Pathways of Change: Grammaticalization in English

PATHWAYS OF CHANGE

PATHWAYS OF CHANGE GRAMMATICALIZATION IN ENGLISH

Edited by OLGA FISCHER University of Amsterdam ANETTE ROSENBACH Heinrich Heine University DIETER STEIN Heinrich Heine University

JOHN BENJAMINS PUBLISHING COMPANY AMSTERDAM/PHILADELPHIA

Table of Contents

Contributors

vii

Preface

ix

Introduction Olga Fischer and Anette Rosenbach

1

A lovely little example: Word order options and category shift in the premodifying string Sylvia Adamson The grammaticalization of the verb ‘pray’ Minoji Akimoto The grammaticalization of concessive markers in Early Modern English Guohua Chen Combining English auxiliaries David Denison

39 67

85 111

Grammaticalisation: Unidirectional, non-reversable? The case of to before the infinitive in English Olga Fischer

149

Remarks on the de-grammaticalisation of infinitival to in present-day American English Susan Fitzmaurice

171

The role of person and position in Old English Elly van Gelderen

187

Remarks on (uni)directionality Roger Lass

207

vi

TABLE OF CONTENTS

Soþlice and witodlice: Discourse markers in Old English Ursula Lenker

229

Onginnan/beginnan with bare and to-infinitive in Ælfric Bettelou Los

251

Some suggestions for explaining the origin and development of the definite article in English Robert McColl Millar

275

Parallelism vs. asymmetry: The case of English counterfactual conditionals Rafał Molencki

311

The grammaticalization of the present perfect in English: Tracks of change and continuity in a linguistic enclave Sali A. Tagliamonte

329

Grammaticalization versus lexicalization: ‘Methinks’ there is some confusion Ilse Wischer

355

Name index

371

Subject index

377

Contributors

Sylvia Adamson Dept of English & American Studies, University of Manchester, Oxford Road, Manchester M13 9PL U.K. [email protected]

Olga Fischer Engels Seminarium Universiteit van Amsterdam Spuistraat 210 1012 VT Amsterdam The Netherlands olga.fi[email protected]

Minoji Akimoto Aoyama Gakuin University 3–12–35 Azamino Aoba-ku Yokohama-shi Kanagawa-ken 225-0011 Japan [email protected]

Susan M. Fitzmaurice Northern Arizona University Dept of English Flagstaff, AZ 86011–6032 USA [email protected]

Guohua Chen Dept. of English Beijing Foreign Studies University 2 Xisanhuan Bei Lu Beijing 100089 China [email protected]

Elly van Gelderen Arizona State University Dept of English PO Box 870302 Tempe, AZ 85287–0302 USA [email protected]

David Denison Dept of English & American Studies University of Manchester Manchester M13 9PL U.K. [email protected]

Roger Lass Dept of Linguistics University of Capetown Rondebosch 7700 South Africa [email protected]

viii Ursula Lenker Institut für Englische Philologie LMU München Schellingstr. 3/RG 80799 München Germany [email protected]

Anette Rosenbach Heinrich-Heine University Anglistik III (Dept of English Linguistics) Universitätsstr. 1 D-40225 Düsseldorf Germany [email protected]

Bettelou Los Dept of English/ATW Vrije Universiteit De Boelelaan 1105 NL-1081 HV Amsterdam The Netherlands [email protected]

Dieter Stein Heinrich-Heine University Anglistik III (Dept of English Linguistics) Universitätsstr. 1 D-40225 Düsseldorf Germany [email protected]

Robert McColl Millar Dept of English King’s College University of Aberdeen Aberdeen AB19 2UB Scotland [email protected]

Sali Tagliamonte Dept of Language and Linguistic Science University of York Heslington York YO1 5DD U.K. [email protected]

Rafał Molencki English Language Institute University of Silesia Zytnia 10 41–205 Sosnowicc, Poland [email protected]

Ilse Wischer Institut für Anglistik/Amerikanistik Universität Potsdam Postfach 601553 D-14415 Potsdam Germany [email protected]

Preface

This volume deals with one of the most-discussed subjects in the theory of linguistic change at present, as manifested in a great number of publications over recent years. It was one of the dominant themes at the XIII International Conference on Historical Linguistics 10–17 August 1997 at Heinrich-Heine University, Düsseldorf. Much of the work in the field was carried out — as is befitting for work on a linguistic concept of general theoretical validity — on a wide variety of languages, and work on English was relatively underrepresented, which produced the idea of creating a volume dedicated to grammaticalization processes in English only, or better, solely in English (cf. § 2.4 of the Introduction). From the papers read at the conference on the intersection of the two criteria, grammaticalization and English, a selection of papers was singled out (the papers by Akimoto, Chen, Fischer, van Gelderen, Lenker, Los, Millar, Molencki, Tagliamonte and Wischer) and subjected to the same refereeing procedures as for the rest of the papers submitted for publication in the main conference volume (Historical Linguistics 1997, eds. Monika Schmid, Jennifer Austin and Dieter Stein, Amsterdam: John Benjamins 1998). In order to yield as full and representative a range of papers on the subject as possible, a number of papers were then solicited from renowned scholars (the balance of the papers). After very extended processes of reviewing, editing and re-writing with no pressure of time, a stage of maturation was reached in the opinion of the editors that certainly warranted publication. In the last stages of the preparation of the volume, John Benjamins, and especially Kees Vaes, Werner Abraham and Michael Noonan, have to be credited for an unusually speedy and efficient processing of the volume. In its present shape, the volume owes a great debt of gratitude to many scholars who have contributed comments to all or some of the papers, especially the reviewers for the volume, and Elizabeth Traugott, who contributed significant comments on an earlier version of the volume. The bulk of the copy-editing task was primarily overseen by one of the editors, Anette Rosenbach, who was

x supported by Sandra Guerrero and Barbara Schulz of the Department of English Language and Linguistics at Heinrich-Heine University, Düsseldorf. Düsseldorf, March 2000 Dieter Stein

Introduction

1.

Olga Fischer

Anette Rosenbach

University of Amsterdam

Heinrich-Heine University

Introduction

The concept of grammaticalization is arguably the most widely discussed concept of linguistic change. As such, it is not surprising that the concept has been a central presence at the recent meetings of ICHL in Düsseldorf, August 1997 and Vancouver, August 1999. Most of the articles in the present volume are a selection of papers presented at the XIII International Conference on Historical Linguistics (Düsseldorf, August 1997), with additionally invited contributions by Sylvia Adamson, David Denison, Susan Fitzmaurice and Roger Lass. The purpose of this volume is to broaden the range of empirical cases of grammaticalization in one particular language, i.e. English, and thereby cast more light on a number of current themes in grammaticalization, which will be highlighted in this introduction. We shall first give a brief description of grammaticalization as an empirical phenomenon (Section 2) with special attention given to the role played by grammaticalization in the English language. We will present an overview of the various approaches to grammaticalization (Section 3), focusing on the different perspectives and objectives in formal and functional accounts of grammaticalization. Next (Section 4), the major mechanisms and causes of grammaticalization will be presented as seen from a functional-diachronic perspective, which is the approach followed by most of the contributors to this volume. This section will pay attention to some controversial issues that are currently being discussed and which are addressed in this volume by some of the contributions, such as the question of unidirectionality in grammaticalization processes (see the studies by Fischer, Fitzmaurice and Lass) and the status of grammaticalization as an explanatory tool (see Fischer and Lass).

2 2.

OLGA FISCHER AND ANETTE ROSENBACH

What is grammaticalization?

2.1 The traditional view Grammaticalization is generally seen as a process whereby a lexical item, with full referential meaning (i.e. an open-class element), develops grammatical meaning (i.e. it becomes a closed-class element); this is accompanied by a reduction in or loss of phonetic substance, loss of syntactic independence and of lexical (referential) meaning. In this sense, grammaticalization is an empirical phenomenon, studied historically; a process which was probably first described under this heading by Meillet (1912) even though the insights date from much earlier (for a succinct history of the development of the idea of grammaticalization, see Hopper and Traugott 1993: 15 ff.). The process of grammaticalization involves changes in both form and meaning. Usually, formal and semantic phenomena go hand in hand. It is important to note, however, that the formal and the semantic do not necessarily go together: there may be formal changes without meaning changes, and meaning changes without formal ones. In addition, not every change is a case of grammaticalization. A crucial question in this connection is: what provides the trigger for grammaticalization? Is it form or meaning? We believe that this is a difficult question to answer in any general sense, but it is a point that should be investigated in each individual analysis of an attested case of grammaticalization. In other words, in each investigation form and meaning developments should be separately discussed. It is clear that the various approaches (within formal and functional theories) to grammaticalization emphasize the roles played by form and meaning differently (see further Section 3). In terms of form (the role played by meaning will be more fully discussed in Section 3), the reduction that takes place when a lexical item grammaticalizes could be described as follows (cf. Hopper and Traugott 1993: 7),   >   >  >   > ₍₎ A well-known illustration of this process is adverb formation in Romance languages, e.g. in French or Italian (cf. Hopper and Traugott 1993: 130–133). We can roughly distinguish the following stages: (1)

a. b.

(Latin) humile mente: ‘with a humble mind’ i. (Old French) humble(-)ment: ‘in a humble(-)way’ ii. lentement: ‘in a slow-way’ iii. humble e doucement: ‘in a humble and gentle-way’

INTRODUCTION

c.

3

humblement: ‘humbly’ humblement et doucement: ‘humbly and gently’

At stage (a), the Latin feminine noun mens (ablative mente) could be used with adjectives to indicate the state of mind in/with which something was done. At a next stage, the phrase acquired a more general meaning (b.i), and mente came to be used also with adjectives not restricted to a psychological sense (b.ii). However, mente retained some of its independence in that, in a conjoined adjectival phrase, the morpheme did not need to be repeated (b.iii). Finally during stage (c), the noun fully developed into a inflectional morpheme, the only remnant of the original construction being the feminine 〈e〉 ending after the adjectival stem, which now serves mainly as a kind of epenthetic vowel to ease pronunciation. Another illustration of a still ongoing grammaticalization process can be given from English (cf. Hopper and Traugott 1993: 2–3), (2)

a. b. c. d. e. f.

I am going (to Haarlem) to visit my aunt I am going to marry (tomorrow) I am going to like it It is going to rain I am going to go there for sure I’m gonna go

In the first example ‘go’ is used as a concrete directional verb (i.e. the verb is still fully lexical), and the infinitive consequently has a purposive function (syntactically it is an adjunct, i.e. it modifies the infinitive). In contexts where the finite verb and the infinitive are adjacent, the directionality of the verb could change from a locative into a temporal one, expressing futurity (b). The meaning of each particular case depends quite heavily on context: e.g., the addition of tomorrow in (b) makes a purely temporal interpretation much more likely. Once this non-directional sense has developed, the verb ‘go’ also begins to be found with infinitives which are incompatible with a purposive meaning as in (c), and from there it may spread to other structures (d–e), more and more losing its concrete directional sense. Syntactic changes seem to go hand in hand with these changes in meaning: in (d–f) the verb ‘go’ has changed from a full verb into a (semi-)auxiliary. As a result of the loss of directional content, the verbal structure also frequently undergoes loss of phonetic substance, which is shown in (f). It is to be noted that this particular grammaticalization process reflects diachronic development as well as synchronic variation. This situation is quite common: the forms reflecting various stages of grammaticalization and the nongrammaticalized forms occur side by side. This phenomenon has been called

4

OLGA FISCHER AND ANETTE ROSENBACH

‘layering’ (cf. Hopper 1991: 22–24; Hopper and Traugott 1993: 123 ff.). When the grammaticalized and non-grammaticalized forms go their own separate ways, Hopper (1991) speaks of ‘divergence’. An example of this would be the indefinite article (a)n and the numeral one, which both go back to the same Old English form an (cf. Hopper and Traugott 1993: 116 ff.; Hopper 1991: 24–25); another instance is the divergence taking place in the verb pray, as described by Akimoto in this volume. Tagliamonte, also in this volume, shows how synchronic layering and diachronic development overlap. She looks at how an isolated dialect of English (Samaná English) expresses the   (i.e. the meaning(s) it has in present-day English) in a layer of different forms (such as the preterite, /been/done + past participle etc.), many of which were used in the history of English. By presenting its synchronic state, she is able to establish which factors cause the appearance of one or other of these forms (factors such as ‘aspect’, ‘temporal distance’, particular collocations etc.); this in turn may deepen our insight into how these forms were actuated and used in the history of English. 2.2 Some more recent developments within grammaticalization With the arrival of structuralism, much less attention was paid to this essentially diachronic phenomenon of grammaticalization. It was only in the seventies, when more and more linguists began to express their dissatisfaction with the strictly dichotomous ‘structural’ model (in terms of the split between diachrony and synchrony) and with the idea of an autonomous syntactic theory, that the phenomenon of grammaticalization gained new interest. Due to this revival and to the spread of functional-cognitive models of language, new perspectives on grammaticalization emerged. In typological work on grammaticalization (see further Section 3), the connection with the historical perspective is still close, but the removal of the strict dividing line between diachrony and synchrony also led to grammaticalization being studied from a more synchronic angle (see especially the work of Elizabeth Traugott [1982, and later studies] and Eve Sweetser 1990). Here grammaticalization is seen as a syntactic, discourse-pragmatic phenomenon, where we witness the semantic development of lexical items from the propositional domain to the textual domain, and from there to the expressive domain; a development whereby the meaning of the lexical item changes from less to more situated in the speaker’s mental attitude. This latter type of grammaticalization, which can also be seen — like the more traditional type discussed above — to operate diachronically, is in this volume discussed synchronically by Lenker with reference to the use of Old

INTRODUCTION

5

English manner adverbs, such as soþlice and witodlice, which are shown to play a role on the discourse level in Old English as well. Akimoto’s contribution (also this volume) addresses this point diachronically. He notes that the phrase I pray thee developed (via reduced forms such as I pray/prithee and pray) from the propositional level into a discourse marker; it skipped the textual level, however, which he attributes to the fact that the phrase retained some of its referential meaning, being used as a marker of politeness rather than a general discourse marker. It is also interesting to observe that Los (this volume) notes, as it were in passing, that discourse-markers need not arise via this particular lexical cline. In her explanation of the grammaticalization of Old English onginnan/beginnan ‘begin’ into inchoative and perfective markers, she shows how both verbs play a role in more or less fixed constructions (i.e. / + to-infinitive and þa + / + bare infinitive) that came to be used as foregrounding devices in discourse, whereby sentence-initial / + to-infinitive functioned as a marker of thematic discontinuity (much like the adverbs witodlice and soþlice discussed by Lenker), while þa + / + bare infinitive is used to continue the smooth flow of narrative events. (More on this development, which often goes under the name of ‘subjectification’, will be found in Section 4.2). 2.3 Grammaticalization versus lexicalization and degrammaticalization Closely linked to grammaticalization is the concept of lexicalization. At present, however, there seems to be no consensus as to what exactly this relation involves. For some linguists, grammaticalization and lexicalization are each other’s opposites. Thus, Ramat (1982) considers lexicalization to be an aspect of degrammaticalization in that “degrammaticalization processes may lead to new lexemes” (p. 550). For instance, in English and also in German, suffixes like -ism and -itis are used (often jocularly, and with pejorative meaning, referring to all the ‘abstract’ ills of present-day society) as full lexical items, with a specialized referential content. For Lehmann (1999) (and see also Traugott 1996, and Chen, this volume), however, lexicalization is an aspect of grammaticalization. He sees both lexicalization and grammaticalization as reduction processes, but taking place on different planes, i.e. in the lexicon and grammar respectively. Lexicalization, according to this view, takes place when a noun, adjective or verb together with a preposition or particle forms a new lexical unit, e.g. in front of, as long as, (to) look after, (to) be going to. This type of lexicalization may constitute a preparatory phase for grammaticalization in that the new, compound, lexical unit may begin to move up the cline of grammatical categories, becoming

6

OLGA FISCHER AND ANETTE ROSENBACH

more and more grammatical on the way, i.e. functioning as a regular preposition (beside, between), conjunction (whilst, because) or auxiliary (to be going to). In this sense, lexicalization is not the opposite of grammaticalization or similar to degrammaticalization, but it is the opposite of folk-etymology, in which language users take an erstwhile lexical item apart and pseudo-transparentize it. The issue of the status of lexicalization in general is addressed in this volume by Wischer. She shows that the ‘lexicalization’ of the Old English impersonal syntactic phrase me/þe/him þynceþ to early Modern English invariant methinks is not an aspect of degrammaticalization (because there is no significant change in the referential meaning of the phrase) but is much closer to the synchronic type of grammaticalization mentioned above in Section 2.2. For a somewhat different case of degrammaticalization, involving not so much lexicalization in the sense of Ramat, but rather a divergent regrammaticalization based on an older lexical sense (a kind of to-and-fro movement), see Fischer, this volume. 2.4 Grammaticalization processes in English: The whys and hows What exactly is the role played by grammaticalization in the English language? Studies on grammaticalization mainly focus on languages with a rich morphology, see for example studies on American languages (e.g. Chafe 1998; Mithun 1998) and the research conducted by Heine and associates on African languages (e.g. Heine and Claudi 1986; Heine and Reh 1984; Heine 1999a and b). Also, the development of creoles presents an ideal field for the study of grammaticalization, since they are typical in developing new morphology fast, using full lexical items to fill the gaps in the pidgin grammar. Creoles, so to speak, represent grammaticalization in statu nascendi. From this point of view, however, the English language does not seem to qualify as the ideal field of activity for the investigation of grammaticalization processes. In the course of the general development from a synthetic to a more analytic character, the English language has lost most of its inflections, and today only meagre traces of morphology are left. This increasing drift towards analyticity has, however, in turn created the need for restructuring the grammatical system. It is in this context that new function words, such as the definite article (see McColl Millar, this volume) and the auxiliaries (see Denison and Tagliamonte, and to some extent also Los and Molencki, this volume) have emerged in processes of grammaticalization. In this respect, the situation in English is comparable to that of a creole. And indeed, there is a huge discussion on whether English should actually be regarded as a

INTRODUCTION

7

creole (e.g. Domingue 1977; Poussa 1982; for a negative conclusion see Thomason and Kaufman 1988: § 9.8; Görlach 1990 [1986] and Allen 1997a). Another advantage of studying grammaticalization processes in English is methodological in nature. In contrast to most African and American languages, English has a well-attested written history and therefore provides a sound empirical basis for diachronic research. Admittedly, the written history of English can only be considered as sketchy and fairly incomplete (or, in Lass’ terms [this volume] it may not be “statistically well-formed”) and is by no means representative of the actual language spoken, but at least some historical evidence is available. Reconstruction, in contrast, relies on synchronic data only to describe a diachronic process and crucially hinges on the assumption that grammaticalization proceeds in one direction (see e.g. Heine 1999b). As the papers by Fischer, Fitzmaurice and Lass in this volume show, however, this may well be not as true and absolute as has usually been assumed (see also Section 4.3 below). In other words, while investigating grammaticalization processes in English may at first sight seem valuable from the perspective of an English historical linguist only, it is also advantageous from a methodological-empirical point of view because of the direct access we have to the diachronic stages of English. This, in addition, makes these investigations an invaluable tool for putting the reconstruction of grammaticalized elements in languages without a long written history on a surer footing too. Interesting in this respect is the contribution by Chen (this volume) on the grammaticalization of concessive markers in English. On the basis of a detailed study of a diachronic corpus, he shows that the general (typological) pathway proposed for concessive markers (as in the work of König) may well need to be rethought. He finds, firstly, that ‘hypothetical concessives’ (also called ‘conditional concessives’) did not always develop out of conditionals, but often out of more general concessive markers, and, secondly, that factual concessive markers are also present at an early stage, and not a later development from hypothetical concessives. This would explain, for example, why (al)though shows no traces of condition in its early (Old English) usage, and why it could express both hypothetical and factual concession from the very beginning. ‘Empirical’ in this volume is used in two ways (see also Figure 1 in Section 4.3). In a strict sense of the term, ‘empirical’ refers to the testing of (potentially falsifiable) hypotheses. It is in this sense that the studies by Fischer and Fitzmaurice on infinitival to have to be seen, both of which challenge the prediction of the hypothesis of unidirectionality by presenting cases of possible degrammaticalization. In a wider (or weaker) sense, ‘empirical’ is simply equivalent to ‘data-based’, which is the approach taken by the remaining articles

8

OLGA FISCHER AND ANETTE ROSENBACH

in this volume (except Lass’ paper, which is theoretical). Within such an approach two different kinds of argumentations can be observed. First, it is possible to argue in terms of language potential. From such a point of view, the fact that a certain form or construction occurs at all is significant in itself, no matter how often. Sometimes it is also argued that the fact that a form or construction does not occur is significant too. Such negative evidence (ex silentio), however, forms a much weaker type of evidence (see also Lass, this volume). Second, within a quantitative analysis not only occurrence versus nonoccurrence counts, but the frequency with which a linguistic form occurs is significant. Such a frequency-based analysis seems particularly fruitful for the analysis of synchronic variation (‘layering’) . This is shown in this volume in the contributions by van Gelderen, Los and Tagliamonte. In the study by Adamson (also this volume), frequency analysis helps establish which of the various senses of a form (‘lovely’) is the more prototypical at a given time, thereby showing how the prototypical meaning of ‘lovely’ changes over time. Note, however, that Lass (this volume) is, in general, fairly sceptical about inductive historical generalizations. In his view, empirical studies often do not define the population on which generalizations are made, or the obligatory contexts of the constructions under investigation . This may, however, be too pessimistic a view. In our opinion, empirical studies do provide a useful tool to reveal the processes involved in the process of grammaticalization, provided that they are conducted in a careful and sensible way, and are not considered definitive.

3.

Approaches to grammaticalization

The term ‘grammaticalization’ is today used in various ways. In a fairly loose sense, ‘grammaticalized’ often simply refers to the fact that a form or construction has become fixed and obligatory, for example when we say that SVO word order has become grammaticalized in English. Similarly, it is often said that certain concepts are — or are not — ‘grammaticalized’ in a language, meaning that they are expressed by grammatical elements. For example, the conceptual distinction between alienable and inalienable possession is ‘grammaticalized’ if it correlates in a systematic way with certain (morpho-)syntactic forms. In these cases, therefore, the term ‘grammaticalization’ is a fairly static concept and simply means ‘fixed’ or ‘codified’. In a stricter sense, however, as introduced above (see Section 2.1), the notion of ‘grammaticalization’ is first and foremost a diachronic process with certain typical mechanisms, a process that can be identified by various diagnostics

INTRODUCTION

9

(see Section 4). The general concept of ‘grammaticalization’ originally comes from Indo-European studies (cf. e.g. Gabelentz 1891) and was given a formal term by Meillet (1912), but, as we mentioned above, the idea was not further pursued within the structuralist framework, because there the focus was on the description of states, and not on processes. Language was not considered as a historical object with a diachronic vector in it, but rather as a succession of synchronic states generated by synchronic grammars. As we said, it was only when such structural axioma were challenged by functionally-oriented approaches that the concept of grammaticalization moved into the limelight again in linguistic research. Recently, however, grammaticalization has also come to figure more prominently in generative accounts of language change, though in a rather different way. In the following we will explore the main differences between functional and generative approaches to grammaticalization (see also discussions in Abraham 1993; Newmeyer 1998 and Haspelmath 1998). 3.1 Formal approaches to grammaticalization The concept of grammaticalization as outlined above (Sections 2.1 and 2.2) is not easily compatible with formal models of language. Following Saussure, the proponents of generative grammar believe in the strict separation of synchrony and diachrony. Even in their diachronic studies the focus is not on language output and the processes of language change, but rather on the description of the synchronic states produced by speakers’ competence before and after a change has occurred. Furthermore, due to the assumption that language in general and syntax in particular are organized in a modular and autonomous way, generative studies are only dealing with syntactic change from a strictly (morpho)-syntactic perspective and they do not take into account the semantic-pragmatic mechanisms that underlie such changes (see below, Section 4). Also, the goal of generative analysis is to find the most appropriate (= maximally constrained) description of the change in terms of the theory of grammar. In other words, an explanation in generative terms means to find a (possibly) universally valid description (which means, in fact, an explanation valid within the current model), which can adequately account for speakers’ internal knowledge of language; it does not attempt to find underlying motivations, which allow the change to occur in the first place. The tool for this description is provided by the theoretical framework of generative grammar — which has undergone several changes in recent years (from Transformational Grammar to Extended Standard Theory, X-Bar Syntax, Principles and Parameters, Government & Binding to Minimalism) — , which explicitly sets out what should, and should not, be possible in

10

OLGA FISCHER AND ANETTE ROSENBACH

language. In this respect, generative grammar is a strong theory, allowing for strong predictions which can be potentially falsified. Thus, while at first sight the concept of grammaticalization seems to be not applicable to generative accounts of language, it is not altogether incompatible with them. It can be said that, strictly speaking, diachronic generative studies only deal with a particular facet of grammaticalization, i.e. the restructuring of the grammatical system by means of re-analysis (cf. Abraham 1993; Haspelmath 1998 and Newmeyer 1998: 292), which is generally seen as one of the main mechanisms of grammaticalization (cf. Hopper and Traugott 1993: 32, but see Haspelmath 1998 for an argument that grammaticalization and re-analysis are two distinct concepts). It seems also clear why this should be so: in so far as reanalysis is involved in grammaticalization, it usually (but not necessarily, cf. below, Section 4.4) only takes place when the process has already been set in motion through semantic-pragmatic factors and has reached momentum at the morphosyntactic level. It is only at this point that generative analysis starts at all. Re-analysis within the generative paradigm is generally accounted for by assigning a structural description both to the old construction and to the new, reanalysed structure, using the principles and constraints of the theory as an ‘explanatory’ tool. In this account, only discrete word-class categories are allowed; gradience of word-class membership (see Haspelmath 1998: 330) is not possible. For this reason, generative studies cannot account for the gradual aspects of grammaticalization processes, but can only capture abrupt, categorical changes. Haspelmath (1998: 330) even argues that “thinking in discrete terms where the phenomena are gradient means that clear instances of grammaticalization are erroneously attributed to reanalysis because grossly oversimplified tree diagrams … do not reflect the gradualness of the change”. Generative models of change also have severe difficulty in dealing with the availability of two structures at one and the same time (as in synchronic variation, or, ‘layering’ phenomena). Can one speaker have access to both the old and the new structure? For a positive conclusion, see Abraham (1993: 21–22), who also refers to Pintzuk (1991) and the possibility that speakers may have access to more than one grammar simultaneously (the so-called double-base hypothesis); for a negative one, see Haspelmath (1998: 341). Language change according to the generative model takes place between successive generations during the process of language acquisition and is manifested either in a change in the structural configuration, a change in movement operations, or in the evolution of or change in functional categories (see also below). Representative for early diachronic generative studies on syntactic re-analysis is the work by Lightfoot (1979) on ‘catastrophic change’ within the English modal auxiliaries.1

INTRODUCTION

11

Recently, with the introduction of functional categories in generative grammar, another kind of reasoning has been introduced into generative accounts of grammaticalization. Elements from functional categories, such as determiners, complementizers or AGR, are taken to serve as heads of constructions (= DP, CP, AGR-P, etc.). Diachronically, functional heads are assumed to evolve out of lexical elements/heads, and it is in this respect that diachronic generative studies can capture grammaticalization phenomena (see e.g. Roberts 1993).2 Only one paper in this volume, by van Gelderen, deals with what could be called grammaticalization phenomena in a generative way. Even though van Gelderen herself does not refer to the term grammaticalization, it could be said that van Gelderen’s study here, on Old English verb morphology, deals with a final stage of a grammaticalization process in that the Old English verbal endings are disappearing and are being replaced (this could become a new cycle of grammaticalization, functionally linked to the earlier one) by personal pronouns and possibly also by a word order becoming more strict (which in itself can be part of a grammaticalization process). Van Gelderen shows that the verbs first reduce their verbal endings when they move to a functional category, such as complementizer position. This is of interest because Abraham (1993) points out that grammaticalization might be captured in formal, generative terms by showing that originally lexically filled nodes (in this case the inflexional morphemes on the verb) may be replaced by functional nodes (here the movement to a functional position). Van Gelderen also indicates that there is a relation between pro-drop (the absence of overt pronouns) and the preservation of verbal endings. This might show a link between the beginning of a new cycle — the use of pronouns to show the function of person, case and number — and the disappearance of the old cycle, in which such features were shown morphologically attached to the verb. Van Gelderen herself does not present the evidence in terms of grammaticalization processes, because she is interested in the consequences this case may have for the theory of grammar. Concentrating on grammar change, she ignores what happens in terms of language change (see also 4.3), which is the level on which grammaticalization works (see also note 1). This study, therefore, shows very nicely how different the objectives are of the generative approach as compared to functional approaches to language change, but it also shows that this different way of looking at the data in question, may unearth further causal factors involved in grammaticalization, which are of a more strictly grammatical nature (see also Fischer, and, somewhat more indirectly, Fitzmaurice, this volume).

12

OLGA FISCHER AND ANETTE ROSENBACH

3.2 Functional approaches to grammaticalization There are a number of fundamental differences between formal and functional models of language in general, which are reflected in the respective approaches to grammaticalization. Although several theoretical frameworks exist for functionalist approaches (e.g. ‘Functional Grammar’ or ‘Cognitive Grammar’), these differ from generative theory by being not that easily falsifiable. The conception of language is holistic and relatively unconstrained; conceptual, pragmatic and language-external factors are believed to have more direct influence on grammatical structure. On the other hand, not being bound to a restrictive, autonomous theory of grammar has the striking advantage of being able to explore how semantic, pragmatic and grammatical factors impinge on one another. Since grammatical elements are not taken as necessarily discrete members of a category but seen rather as more or less prototypical instances of such a category, gradualness can be better accounted for. Diachrony, likewise, is not seen as a succession of discrete synchronic stages, but rather as being inherent in synchrony. In contrast to generative studies, which emphasize mainly the situation before and after grammaticalization, functional approaches may also include aspects of the actuation and implementation of the process, and of the motivations behind the process; in other words, they allow for an explanation in a much wider sense (i.e. outside grammatical competence proper). The subject matter of investigation within functionalist models is primarily the use of language, and not the underlying system. Indeed, in the theory of Emergent Grammar (cf. Hopper 1988 and his later work on this) there is no such thing as a fixed system of grammar at any time, grammar is constantly ‘emerging’ from language being used in discourse. Accordingly, the locus of language change is primarily within language use, i.e. with adults and not children. In Table 1, the basic differences between functional and generative approaches to grammaticalization are summarized. Today, we can today broadly distinguish between more diachronically- and more synchronically-oriented functionalist and typological approaches (for a similar distinction, see also Traugott 1996). Note, that there is a close interdependence between functionalism and language typology: while many functionalists make use of cross-linguistic evidence (see for instance the work of Talmy Givón, e.g. Givón 1979, 1984, 1995), many typologists work within a functional framework, for instance in studies by Martin Haspelmath (e.g. Haspelmath 1990) and Frans Plank (see e.g. the Konstanz project on the Universals Archive), and very often typology and functionalism are not really separable at all. In functional-diachronic approaches (e.g. Lehmann 1982 [1995]; Traugott

INTRODUCTION

13

Table 1. Functional vs. formal approaches to grammaticalization: Basic differences Functional approaches

Formal approaches

• holistic conception of language and grammar • consideration of conceptual, semantic-pragmatic and language-external factors • diachrony in synchrony

• modular conception of language and grammar (→ autonomous subcomponents) • only grammar-internal factors

• subject matter of investigation and locus of change: (mainly) language use • language change = gradual • grammaticalization as the full process from lexical items to grammatical words, including actuation, implementation and motivation • description of the whole process • looking for explanations (inside and outside grammar)

• synchrony vs. diachrony • diachrony = comparison of synchronic stages • subject matter of investigation: competence • locus of change: language acquisition • language change = abrupt • grammaticalization as re-analysis • grammaticalization as the evolution of functional categories/heads out of lexical categories/heads • only description of situation before and after re-analysis • explanation only from the viewpoint of the theory of grammar (e.g. category shifts, changes within functional categories, etc.)

and Heine 1991; Heine, Claudi and Hünnemeyer 1991) the focus lies on the historical development of grammatical constructions, while the main aim of linguists working within functional-synchronic models (e.g. Givón 1979; Hopper and Thompson 1984) is to show the discourse-pragmatic basis of grammatical structure. Positioned somewhere in between are studies on ‘change in progress’, which focus on one particular aspect in the process of grammaticalization, i.e. the fact that in periods of transition old and newly developed linguistic forms may co-exist for some time (‘layering’). Typology explores the concept of grammaticalization by accounting, diachronically, for the evolution of grammatical elements and constructions in general (cf. e.g. Heine 1997; Bybee et al. 1994), and, synchronically, by comparing how certain concepts (e.g. possession) and categories (e.g. mood, tense, aspect) have become grammaticalized in a variety of languages (e.g. Givón 1983; Kemmer 1993).

14

OLGA FISCHER AND ANETTE ROSENBACH

In Table 2, an attempt is made to give a short overview of and distinguish systematically between the various approaches to grammaticalization that are currently on the linguistic market. While typological and functional approaches to grammaticalization, both from a synchronic and diachronic perspective, often go hand in hand and are therefore not mutually exclusive, the most notable contrast is, as has been outlined in this section, between functional models on the one hand, and formal models on the other. Table 2. Approaches to grammaticalization: Short overview Perspective

Synchronic

Approaches to grammaticalization Typological

Functional

cross-linguistic patterns

discourse-pragmatic and not applicable cognitive basis of grammar

‘Change in progress’ Diachronic

• synchronic variation • creoles

Formal

if, at all, only within the double-base hypothesis

evolution of gram- evolution of linguistic re-analysis mar in general forms, Emergent Grammar?

With the exception of van Gelderen’s contribution, who works within the generative paradigm, most of the papers in the present volume come closest to the functional-diachronic approach to grammaticalization, with the articles by Fitzmaurice and Tagliamonte focusing on ongoing developments within American English and Samaná English, respectively.

4.

Mechanisms and/or causes of grammaticalization

4.1 Metaphor and metonymy In the literature on grammaticalization it is generally accepted that the most important semantic mechanisms at work in the process of grammaticalization are metaphorical and metonymic in nature (cf. general studies such as Hopper and Traugott [1993: 77–87] and Diewald [1997: 42–62]).3 Besides these, Traugott and Heine (1991: 7) also mention analogy and re-analysis, which are seen as related to instances of metaphor and metonymy respectively, but then viewed from a

INTRODUCTION

15

structural rather than a semantic/pragmatic point of view. Hopper and Traugott (1993: 87) sum it up as follows, In summary, metonymic and metaphorical inferencing are complementary, not mutually exclusive, processes at the pragmatic level that result from the dual mechanisms of reanalysis linked with the cognitive process of metonymy, and analogy linked with the cognitive process of metaphor. Being a widespread process, broad cross-domain metaphorical analogizing is one of the contexts within which grammaticalization operates, but many actual instances of grammaticalization show that the more local, syntagmatic and structure changing process of metonymy predominates in the early stages.

Since it is quite generally believed that grammaticalization is semantically (or pragmatically) driven, it is not surprising that such essentially pragmatic/semantic factors as metaphor and metonymy are seen as important. It remains to be seen, however, whether the accompanying grammatical changes are a mere appendix to the semantic change or whether they also play a(n) (more) independent role. Here we will briefly consider how these metaphorical and metonymic processes work. We will also discuss in what respect analogy and re-analysis can be said to be similar to metaphor and metonym respectively. According to one school of thought, metaphor is said to play an important part especially in the early stages of grammaticalization. Heine et al. (1991a: 151 ff.) show how only a limited number of basic cognitive structures form the input to grammaticalization; they call these ‘source-concepts’. The fact which makes them eligible is that “they provide ‘concrete’ reference points for human orientation which evoke associations and are therefore exploited to understand ‘less concrete’ concepts” (Heine et al. 1991a: 152). Thus the human body and basic human activities (‘sit’, ‘stand’, ‘go’, ‘leave’, ‘do’, ‘make’ etc) regularly provide source concepts in any language. For instance, in order to express the abstract notion of space, ‘back’ may be used to refer to the space behind, and ‘head’ to refer to space in front. In turn these notions of space may come to be used to express the even more abstract notions of time. Similarly, physical actions like ‘grasp’ may be used to denote mental activities (cf. also the similar etymology of verbs like comprehend, Dutch begrijpen, German fassen etc.). Metaphorical change can be related to analogy. It is a type of paradigmatic change, whereby a word-sign used for a particular object or concept comes to be used for another concept because of some element that these two concepts have in common. It is not surprising that, when this similarity is obvious, often the same metaphorical transfers take place in otherwise totally unrelated languages. Metaphors are of course also an important device in literary language, but there the aspect of similarity is often much less obvious, creating the kind of tension

16

OLGA FISCHER AND ANETTE ROSENBACH

that poets need in order to show well-known objects or concepts in a fresh and unexpected light. Heine et al. (1991b: 50, 60) indeed make a distinction between the type of metaphor that occurs in literary language and in grammaticalization: they call the latter ‘experiental’ or ‘emerging’ metaphors, because they are metaphors that arise in context (i.e. they are metonymic in nature), while the former are termed ‘conceptual’ or ‘creative’ metaphors, which are much more likely to contain conceptual ‘jumps’ and cannot be predicted in any sense. Analogy used as a term in syntactic change is similar to metaphor in that there, too, a form or construction used within a particular paradigm of similar forms or constructions, may replace another one within the paradigm. A clear example of this is the way in which the various noun plurals of Old English (i.e. plural endings in -e, -u, -a, -an, or zero) were almost all replaced by the plural suffix -(e)s (from OE -as, the plural of the masculine strong noun), which had the same function (i.e. the same grammatical meaning) as the disappearing forms within the paradigm or category of ‘number’. Similarly, it can be said that in example (2) above (involving to be going to), a metaphorical change has occurred (cf. also Hopper and Traugott 1993: 88). The change from a concrete, directional verb ‘go’ into a verb referring to the future is semantically a case of metaphor. The physical, ‘bodily’ sense of ‘go’ changes into an abstract temporal concept, a path that is found to be typical in metaphorical change. Heine et al. (1991a: 157) describe this path in a hierarchy (which could be linked to further hierarchies, such as that of case and constituents, see ibid.: 160) as follows,  >  >  >  >  >  Whether this metaphorical change is independent of the metonymic shift taking place in to be going to (see below) is another question. Since the metaphor used is of a contextual type (as indicated above), it may be difficult to draw a distinction, and metonymy may therefore well be the more crucial mechanism. This is indeed the view of Hopper and Traugott (1993: 81), and also Bybee et al. (1994: 289 ff.). The latter distinguish five mechanisms of semantic change that play a role in grammaticalization; at least four of them are essentially metonymic in nature, with metaphor playing only a subsidiary role. Metonymy, like metaphor, is originally a term used in rhetoric but here it is not similarity that causes the association but contiguity, in other words metonymic transfer functions on the syntagmatic plain. So when we speak of ‘the press’ rather than ‘newspapers’, or ‘The White House’ for the US presidency, we use a sign that is indexically related to the substituted one. Both metaphorical and metonymic transfer are cognitive processes, but with metonymy we choose a term from the same field, from the context, whereas with metaphor we

INTRODUCTION

17

substitute a similar cognitive element from a different field or paradigm. What typically happens in grammaticalization processes is what Hopper and Traugott have called “conversational implicatures” (1993: 73) or “pragmatic inferencing” (p. 75). Thus in the above example (2) with to be going to, the change from a directional verb into a verb conveying future time was made possible by the fact that the verb ‘go’ in combination with a purposive infinitive invites the inference that the subject of ‘go’ arrives at a later time at the destination, with the result that the idea of a future plan becomes incorporated into the verb ‘go (to)’ itself. It is clear that the contiguity of the purposive infinitive is essential for the inferencing to happen. Re-analysis,4 which is a term used in syntactic change, is similar to metonymy in that here too the change involves contiguous elements. Thus, the syntactic re-analysis that takes place in the ‘go to’ example in (2) involves a rebracketing of constituents, from [I

NPs[am

going]VP [to visit my aunt]ADV ADJUNCT]

into [INPs [am going to [visit my aunt]]VP] In the case of ‘go to’, there seems to be a relation between the semantic metonymic change and the structural re-analysis (from full verb into semiauxiliary) in that the metonymic shift (which may gradually involve more contexts) can be said to prepare the way for the syntactic re-analysis, which cannot be gradual. The structural change is a result, but it must be noted that this is not a necessary result, as was already indicated in Section 3.1. It is highly likely that the overall structure of the grammar plays a role here too, see further Section 4.5.3 below. 4.2 Semantic bleaching Grammaticalization is one type of macro change, consisting minimally of one process of reanalysis, but frequently involving more than one reanalysis … Grammaticalization is often associated with “semantic bleaching”, and this “bleaching” is the result of reanalysis or, perhaps better said, it is the essence of the reanalysis itself (Harris and Campbell 1995: 92).

Harris and Campbell refer here to ‘semantic bleaching’, which they see as part of the re-analysis itself. In their view, in other words, bleaching is a correlate of the re-analysis, not something that may itself lead to re-analysis. There is also a much more common view (cf. Bybee et al. 1994; Rubba 1994), which regards bleaching as a prerequisite for grammaticalization or even a cause. Fischer

18

OLGA FISCHER AND ANETTE ROSENBACH

(1994), however, shows that bleaching does not necessarily steer the process of grammaticalization. In the case of the grammaticalization of English have to, it was not so much the bleaching of the earlier possessive sense of have that led to the grammaticalization of the verb into an auxiliary, rather it was the change in basic word order from Old English SOV to Middle English SVO, causing have and the to-infinitive to become adjacent in all types of clauses, that set off the re-analysis into an auxiliary. Evidence for this scenario can be found in the fact that the bleached forms of have had been floating around ever since the Old English period for at least six hundred years without causing any further grammaticalization. All other grammaticalization evidence — apart from the bleaching process — such as the development of epistemic meaning, the use of intransitive to-infinitives, double use of have (as in I have to have …) occur only after the word order change. A second type of evidence is the fact that in German and Dutch, which also possessed a bleached form of the cognates of have but where the basic word order remained SOV, the re-analysis did not take place. The French linguist Meillet attributes the process of grammaticalization to the loss of expressivity (which is the same as ‘bleaching’) that occurs in lexical items whenever they occur very frequently (Meillet 1912). The idea that the process of grammaticalization may be caused by the loss of expressivity may indeed explain the continuing cycle of grammaticalization processes, whereby new expressions (Harris and Campbell [1995: 73] refer to these as “exploratory expressions”, which always float around in language but don’t always necessarily get grammaticalized) are constantly used to replace old ones due to a need of speakers to be more expressive.5 However, we must make a distinction between bleaching of one expression that leads to the use of other, new ones (i.e. bleaching at the end of a cline that causes a new cline with a new expression to start), and bleaching within an expression itself (i.e. bleaching within one and the same cline). There is yet another view with respect to the role played by bleaching in grammaticalization, which holds that bleaching occurs only during the last stages of the grammaticalization process (cf. Traugott and König 1991: 190). Traugott and König (and we should also include Sweetser 1990 here) believe that grammaticalization in its early stages involves an increase in meaning, that is, in pragmatic meaning (see also Section 2.2). We have seen that what happens in the early stages of grammaticalization is that a term can come to be used in more senses than one due to pragmatic inferencing; cf. example (2) above, where go comes to indicate both concrete direction and temporal direction (future time). Similarly, mente in (1) comes to be widened to indicate not only ‘mind’, but also

19

INTRODUCTION

‘manner’. This can indeed be interpreted as ‘enrichment’ of meaning because the element now fits into a greater number of contexts. ‘Enrichment’ of meaning also takes place in that meanings that used to be in the extension of an expression move into its intension, i.e. a meaning is added inherently to the defining properties of an expression and not created ad hoc in the context. As argued by Traugott (1995), the process of grammaticalization often (though not necessarily) involves a development towards greater subjectivity, i.e. the tendency of meanings to become increasingly based in the speaker’s subjective attitude towards the proposition. So, in the example of to be going to, the shift in meaning is not only from concrete (lexical) ‘movement’ to more abstract temporal ‘movement’ but also towards a more epistemic meaning in the sense that it expresses the likelihood or intention from the point of view of the speaker. A similar development from deontic to more epistemic can be observed for the English modal auxiliaries, such as must and will (see also Traugott 1995); for further cases of subjectification see the articles in Stein and Wright (1995), which has subjectification as its theme, and the studies by Adamson and Lenker in this volume. Adamson shows on the basis of the historical development of ‘lovely’ how, synchronically, subjective meaning correlates with leftmost position within the NP, and how, diachronically, the meaning change towards subjective meaning goes hand in hand with leftward movement and eventually triggers the syntactic re-analysis of ‘lovely’ as an intensifier.6 She proposes the following grammaticalization pathway from adjectives to intensifiers: Descriptive adjective → Affective adjective →

Intensifier

• referent-oriented

→ • speaker-oriented → (subjective) • increasingly subjective

• 2nd position • within NP

• leftmost position → • within NP →

• leftmost position ↓ • syntactic re-analysis

4.3 The ‘principle’ of unidirectionality Grammaticalization is generally seen as a gradual diachronic process which is characterized as unidirectional, i.e. it always shows the “evolution of substance from the more specific to the more general and abstract” (Bybee et al. 1994: 13). Unidirectionality is said to apply on all levels, the semantic (fully referential > bleached/grammatical meaning; less subjective > more subjective), the syntactic

20

OLGA FISCHER AND ANETTE ROSENBACH

(lexical > grammatical; less bound > more bound) and the phonological (full phonological form > reduced phonological form). Unidirectionality is most strongly defended in Haspelmath (1999), who indeed suggests that it is exceptionless. The emphasis on unidirectionality and on the graduality of the process has led to the idea that the process is mechanistic, that grammaticalization itself is a mechanism or cause for change. Bybee et al. (1994: 298), for instance, write: Thus our view of grammaticalization is much more mechanistic than functional: the relation between grammar and function is indirect and mediated by diachronic process. The processes that lead to grammaticalization occur in language use for their own sakes; it just happens that their cumulative effect is the development of grammar (emphasis added).

It is not at all clear from the literature we have studied what the status of grammaticalization is in theorizing on change. Vincent (1995: 434) for instance writes, even though he is challenging the “pre-eminence [of grammaticalization] as [a] source of new patterns”, that he does not “wish […] to deny the power of grammaticalization as an agent of change” (emphasis added), which seems at least to suggest that he thinks it has explanatory value, that it has independent force. Most students of grammaticalization describe it as a ‘phenomenon’, a ‘process’, an ‘evolution’. However, the fact that for most linguists one of its intrinsic properties is that is is gradual and unidirectional suggests to us that in their view the process must have some independence and that it can be used as an “explanatory parameter” (cf. Heine et al. 1991b: 9, 11) in historical linguistics.7 Roger Lass, in this volume, addresses this very problem. He doubts the validity of the hypothesis of unidirectionality, and questions the way in which it is justified. First, as Lass points out, the criteria for determining the various stages of grammaticalization must be formulated in a clear-cut and explicit way. Lass suggests that we may have preconceived ideas about what ‘lexical’ and ‘grammatical’ is: our definition of ‘lexicality’ and ‘grammaticality’ is more than likely based on some well-investigated languages only, such as English, German or French, and may therefore not function as cross-linguistically valid instruments of description. Second, there is the question of how to deal with possible counter-examples. This is one of the central question raised by Lass and shortly summarized by us in Figure 1 below. According to Lass, if grammaticalization theory aims at being a strong theory, it needs to set out what possible counter-examples should look like. Lass’ position is, we take it, in accordance with the optimal procedure set out for scientific investigation in the sense of Popper (1968). A hypothesis — although it should be formulated in a strong way — is nonetheless always a

21

INTRODUCTION

grammaticalization theory as a strong theory

as a weak theory

allows for explicit predictions as to possible falsification

inductive generalizations only; mere observation

counter-examples count and help in modifying the theory

counter-examples are explained away

no search for counter-examples; only positive data

Figure 1. The role of counter-examples within a theory of grammaticalization

working hypothesis and not a dogma. Given this, the role of counter-examples is to modify the hypothesis in such a way that it can also account for these hitherto unpredicted cases. Another possibility to deal with counter-examples, though, is to simply disregard them, or, in Lass’ terminology, to ‘massage’ them, be it as cases of lexicalization or by simply ignoring them or explaining them away otherwise (as does Haspelmath 1999). A further question is how to find possible counter-examples of grammaticalization? In the Popperian sense of scientific research we should always look for counter-examples and not for cases which conform to our hypotheses. As argued by Lass, this procedure does not seem to apply to grammaticalization research. Here the bulk of research is concerned with finding and reporting prototypical instances of grammaticalization, which, of course, also helps sharpen our understanding of the processes involved. It should, however, not mislead us into thinking that cases of degrammaticalization do not exist. Also, Lass argues, even if there is striking evidence in favour of our theory (in the weak sense), we should not confound ‘commonness’ with absolute truths. Another central problem that Lass addresses is the fact that a strong unidirectional position predicts that all grammatical elements are lexical in origin. Given reconstruction from a uniformitarian perspective, this would predict that there should have been a time when all languages were isolating, i.e. having only lexical and no grammatical material. Lass argues that no such languages are attested, and that therefore such a position is untenable because counter-uniformitarian. If we do not take for granted that the languages of the past looked like today’s languages, how can we, Lass’ argument goes, possibly believe that the principles underlying language change (such as unidirectionality) were the same?

22

OLGA FISCHER AND ANETTE ROSENBACH

At this point, however, a word of caution may be in order: It may well be that Lass is using ‘uniformitarianism’ in two different ways. As recently Deutscher (1999) has pointed out, the original application of uniformitarianism is to diachronic processes only, and not to synchronic states. This, at least, Deutscher argues, is how the notion of uniformitarianism as a methodological tool was originally developed in the natural sciences and from there transferred to linguistics. So, we can only assume that the processes operating in the past were the same (= ‘diachronic uniformitarianism’, in Deutscher’s terms), but it would be wrong to stretch uniformitarianism so as to include the similarity of the languages themselves (= ‘synchronic uniformitarianism’, in Deutscher’s terms). In other words, the fact that we do not have fully isolated languages now, cannot be used to dismiss the ‘principle’ of unidirectionality. According to Lass, we also need to keep the grammaticalization clines and the question of directionality logically apart. As Lass points out, the stages within the clines are causally and ontologically independent of each other: “Information loss processes have no memory”. This is a question also addressed by Fischer in this volume, who concurs with Lightfoot and others that there is no such thing as ‘diachronic grammars’. This point links further to the question of where the locus of change is supposed to be, in ‘language’ i.e. on the performance level, or in ‘grammar’, the abstract system present within each individual speaker? We have argued above (end of Section 3.1) that both must be taken into account to arrive at a full(er) explanation of the phenomenon of grammaticalization. If unidirectionality were indeed a ‘principle’ of language change, the question remains what could possibly motivate it. If a possible explanation turns out to be non-linguistic in nature (e.g. positive feedback as a physico-mathematical principle), then unidirectionality is not a principle of language, i.e. it is not domain-specific, but a general principle. Also, Lass says, the explanation may simply be trivial in the sense that it is highly unlikely to extract anything out of zero.8 Given the importance of the study of counter-examples as advocated by Lass, the studies by Fischer and Fitzmaurice in this volume are especially welcoming for grammaticalization theory. They both set out to explore possible cases of degrammaticalization. Although the development of infinitival to in English cannot be regarded as a case of degrammaticalization back along the macro-level of the cline ‘grammatical > lexical’ — to does not change its grammatical status as an infinitival marker — on a micro-level Fischer shows how the semantic meaning of to moves back to its original semantic meaning of goal or direction, and shows no further phonetic reduction, reduction in scope or increase in bondedness.9 Closely related to Fischer’s paper is the study presented by Fitzmaurice, which looks at infinitival to from a more synchronic perspective,

INTRODUCTION

23

focusing on the negative split infinitive (to not find out) and how it interacts with the grammaticalization of the English semi-auxiliaries (such as have to, want to, be going to). The fact that to within the semi-auxiliaries becomes less bonded with the following VP complement and is therefore indicative of the further degrammaticalization of infinitival to is also mentioned by Fischer. Another indicator for the ongoing degrammaticalization of infinitival to, according to Fitzmaurice is the increasing conventionalization of the negative split infinitive (at least in American English). In the negative split infinitive (to not decide), to not only becomes more detached from the verb, but, according to Fitzmaurice, it also loses its grammatical meaning as an infinitive marker, acquiring a new pragmatic-purposive meaning. Another example for a special case of degrammaticalization, i.e. desubjectification, is pointed out by Adamson (this volume) in the final part of her paper, where she in general draws on the link between word order and subjectivity within the NP. She suggests that there is a pathway from  (e.g. a criminal tyrant) to  (e.g. criminal law), in which the latter stage is less subjective. 4.4 Formal diagnostics of grammaticalization In grammaticalization theory a number of principles or parameters have been distinguished that serve to characterize the process. The clearest discussion of this is to be found in Lehmann (1982 [1995]), whose ‘parameters’ can be used to represent stages in the development. Hopper (1991) presents a number of further generalizations (principles) that can be made regarding the process. Most of these can be subsumed under Lehmann’s parameters. Others, such as ‘divergence’ and ‘layering’, have been mentioned above (see Section 2.1). A final principle mentioned by Hopper, ‘persistence’, points to the fact that traces of the original lexical meaning of the linguistic elements that are grammaticalized, adhere to these elements and that they may be reflected in the way the grammaticalized forms are grammatically constrained. A clear example of persistence is the present-day English auxiliary will, beside the future auxiliary meaning, the old volitional meaning of will lives on, as in, If you will something to happen, you usually succeed. Fischer (this volume) shows how ‘persistence’ may partly explain the divergent route that the infinitival marker to takes in English, compared to its cognates in German and Dutch. Another example of persistence is given by Adamson in this volume, who shows that, today, ‘lovely’ is polysemous in that beside its now prototypical function as an affective adjective or intensifier, it can also still be used as a descriptive adjective (though the different uses correlate with different word order). For Adamson, this synchronic

24

OLGA FISCHER AND ANETTE ROSENBACH

situation reflects the historical development of ‘lovely’ from a descriptive to an affective adjective and an intensifier (see also Section 4.2). Lehmann (1982: 306) presents the following overview (cf. Table 3, slightly adapted in order to indicate the processes taking place). Table 3. Diachronic stages in the process of grammaticalization Parameters

Paradigmatic processes

Syntagmatic processes

Weight

(loss of) integrity

(reduction of) scope

Cohesion

(increase in) paradigmaticity

(increase in) bondedness

Variability

(loss of) paradigmatic variability: increase in oligatoriness

(decrease in) syntagmatic variability

The ‘weight’ or substance of a lexical item involved in a grammaticalization process is reduced (in contrast to similar, but non-grammaticalized items within the same field or paradigm) through both semantic and phonetic erosion. This means that the element becomes syntactically less dominant in the clause, e.g. a full lexical verb such as go in example (2) above dominates the purposive adjunct, whereas the semi-auxiliary go has become part of the VP headed by the infinitive. Similarly, in (1), mente could at first have two coordinated adjectives in its scope (as shown in stage b.iii), but at stage c it needs to be repeated, indicating that its scope has been reduced to the immediately preceding element; it has in fact become a bound morpheme. Concerning ‘cohesion’, the more grammaticalized a linguistic element is, the less choice there is formally, i.e. within the paradigm of forms that have a similar function. Thus, in the expression of a thematic role, a case ending is more paradigmatized than a preposition because usually only one choice exists within the paradigm of case-forms, whereas often more than one preposition can be used to express the same function. Syntagmatically, cohesion is increased in that the grammaticalized item fuses with other linguistic elements, e.g. mente in example (1) becomes a suffix. Paradigmatic variability (in the third row in Table 3) refers to the degree in which a particular linguistic element is obligatory within the clause. Thus, the past tense marker in English is a highly grammaticalized element because it occurs obligatorily within the clause, whereas adverbial markers of time can occur much more freely, their presence being determined not by the grammar but by discourse. Syntagmatically, a grammaticalized element becomes less variable because it takes up a fixed position in the clause. For example, the tense-marker

INTRODUCTION

25

must follow the matrix verb, while the adverbial marker of time can occur in quite a number of positions within the clause. Thus, the parameters in Table 3 indicate the degree to which a particular linguistic item has grammaticalized. It must be noted, however, that these parameters only hold true for the historical or traditional type of grammaticalization, mentioned in Section 2.1. As Wischer indicates in this volume, the discourse-pragmatic type (mentioned in Section 2.2) diverges from these parameters on almost all levels (it undergoes pragmatic enrichment rather than bleaching, increase in scope rather than decrease, there is no ‘obligatorification’ etc.), showing that it is indeed a different type of grammaticalization. Also, Tabor and Traugott (1998) have recently pointed out that one of Lehmann’s parameters, i.e. the reduction of scope, may not be a well-defined and proper diagnostic for grammaticalization. They argue that within a definition of c-command there is rather an increase in scope. Although Lüdtke’s (1980) cyclical theory of language change does not explicitly refer to grammaticalization, it nonetheless links well to the concept. However, where Lehmann’s parameters combine semantic and formal factors in the sense that they occur more or less simultaneously, in Lüdtke’s theory semantic change follows formal change. The basic assumption underlying Lüdtke’s hypothesis is that there is a dualism between sound change and semantic-syntactic change, between reduction and compensation by enrichment. Language change is seen as driven by ‘redundancy management’ (“Redundanzsteuerung”) on the side of both the speaker and the hearer. What sets off language change is phonetic reduction. Too much reduction, however, endangers comprehension for the hearer and therefore needs to be compensated by new lexical material, which then may lead to semantic-syntactic change. This new lexical material will eventually fuse with neighbouring units and become reduced again (since speakers are striving for ease of production), and so the cycle starts again.

sound change (reduction)

lexical enrichment

semantic-syntactic change

26

OLGA FISCHER AND ANETTE ROSENBACH

4.5 Other factors involved 4.5.1 Language contact Another factor that has been mentioned is the use of grammatical material from substrate languages, see for this Traugott and Heine (1991: 7) and more particularly for an in-depth study of this phenomenon, Bruyn (1995). Bruyn (1995) shows that the grammatical processes that take place when a pidgin developes into a creole are often not so much the result of internal developments in the creole (i.e. independent lexical items becoming part of the morphology), as has often been assumed, but that new morphological markers often appear readymade, taken from the substrate languages, which explains perhaps more adequately why the ‘grammaticalization’ in these cases may take place so fast. McColl Millar (in this volume) believes that language contact played an important role in the grammaticalization of the definite article in English. He argues that simply following the typological path that has been suggested for this development, from deictic particle to definite article and further to affixal article, does not explain why languages that started out from the same point, end up in different positions on this cline. Why is Danish typologically most advanced, why is English more in the middle and German still almost at the beginning? He explains the differences between the three languages by showing that the circumstances were different. They all share the decline of inflexions but the difference is that in late Old English there developed a semantic gap due to the specialization in meaning of that, and that this coincided with a time of intensive contact with speakers of Old Norse, who already had a system with separate forms for the article and the distal determiner. This contact, he argues, facilitated the introduction of this system into Old English, using, however, Old English forms. In contrast, Tagliamonte, in this volume, shows that the developments that took place in the expression of the   in Samaná English, was not influenced by the Hispanic context in which this variety of English evolved. 4.5.2 Frequency Yet another factor that plays a crucial role in grammaticalization is frequency. We need to distinguish, however, between frequency as a factor and frequency as an indicator of change. As a factor, frequency matters in that elements eligible for source-concepts are by their very nature frequent, otherwise they would not be source concepts in the first place. It must also be clear that for pragmatic or conversational implicatures to change into conventional implicatures (i.e. for pragmatic inferences to become part of the semantics of a construction), the construction to which they apply must be used frequently. Note, however, that

INTRODUCTION

27

frequency very often is not a necessary precondition for change to occur, but rather a mere consequence of a change, in the sense that a change paves the way for constructions to occur more frequently. For example, if a lexical, open-class item turns into a functional, closed-class item, it is quite obvious that it should be used more often. Functional elements are by definition more frequent than lexical items. In this respect, frequency arguments can be used as an indication of ongoing change and may be used as a diagnostics for the state of the grammaticalization process. This is indeed the approach taken by Adamson (this volume), who shows how the semantic shift towards more subjective meaning in the case of ‘lovely’ and its subsequent re-analysis as an intensifier (see also Section 4.2) leads to more frequent use; that is, increasing frequency follows the change, and does not trigger it in the first place.10 Frequency comes also into play when postulating ‘universals’ or general laws. In this line of argumentation, the fact that certain developments are frequent is taken as a proof that they are universally valid. As pointed out by Lass (this volume), the generally observed tendency that grammaticalization processes proceed from lexical to grammatical elements may be simply due to the fact that they are “statistically commoner, so metaphorically ‘preferred’” (see also Section 4.3). 4.5.3 The current state of the grammar A factor that has been given much less attention, but which is emphasized by Mithun (1991) (and see also Fischer 1997), is the importance of the shape of the current grammar: “the formation of new grammatical categories is motivated or hindered by the contours of the existing grammatical system” (Mithun 1991: 160). This particular point may call into question some of the tenets of grammaticalization theory that have been proposed, such as the belief that grammaticalization processes can be triggered by semantic factors only or the hypothesis of unidirectionality (see also Section 4.3). For instance, in the grammaticalization of to be going to, it is possible that the fact that there was a structural Aux position available in English had a ‘positive’ effect on the rebracketing that has taken place. In addition the semantic and structural function of to (see Fischer, this volume), and the fact that ‘go’ and the infinitive are always adjacent in English, may have played a role, and may indeed explain why this particular verb grammaticalized further in English than for instance in Dutch or German. Fischer, this volume, shows in this respect that the grammaticalization of the infinitival marker in English diverged from the process that the cognate markers underwent in German and Dutch, because the grammatical circumstances in the latter two languages were considerably different. Also, Demske’s (forthcoming) work on the German NP demonstrates nicely that

28

OLGA FISCHER AND ANETTE ROSENBACH

changes can be better accounted for if constructions are not seen in isolation but studied in relation to other constructions within the same structural domain (NP). In particular, Demske argues that in German individual changes within the NP (such as changes in adjective inflection, the use of the definite article, the reanalysis of possessive constructions, and an increasing productivity in nominal compounding) may all be captured by a single change, i.e. a change in the relation between article and NP. 4.5.4 The role played by iconicity11 The concept of iconicity has a long-standing tradition within functional argumentation (see e.g. Haiman 1985a and b; Bybee 1985). The basic idea of iconicity is that the relation between the linguistic sign and the linguistic expression it stands for can be motivated, thereby attacking one of the most basic tenets of structuralism, i.e. the arbitrariness of the linguistic sign.12 The perspective on iconicity can be both synchronic and diachronic. Synchronic studies are basically interested in showing that there is a relation between form and meaning or function. Diachronically, the task is to show which role — if any — iconicity plays in language change in general and in the evolution of grammatical forms in particular, and this is where grammaticalization comes into play. McMahon (1994: § 6.3.5) addresses this question, suggesting that iconicity and grammaticalization take place at different stages, using Bybee’s (1985) work on the ordering of verbal inflections as an example. At a first stage, iconicity ensures that those verbal categories that are conceptually closest to the verb will also occur closest to the verb (according to the iconic principle of conceptual distance, cf. also Haiman 1983: 782). It is only at a second stage that grammaticalization becomes important ensuring subsequent fusion of the inflections. So, while iconicity motivates/initiates the evolution of the form and the order of morphemes, grammaticalization will take over, turning the input structure into more and more grammatical elements in the sense of Lehmann’s (1982) diagnostics as introduced in Section 4.4 above. Another way in which iconicity plays a role in grammaticalization is when a grammatical element that is coming, or has come, to the end of its cline (i.e. when it has become phonologically much reduced), is replaced by a new expression, thus starting a new grammatical cycle. These replacements are generally again iconic (or transparent) with respect to the grammatical function for which they come to be used. For example, in the earliest uses of mente in example (1) above, the choice of mente is motivated by the meaning of the noun mente in other contexts, which means that the word does not need to be learned or stored separately. In the final stage (stage c in [1]), however, mente is no

INTRODUCTION

29

longer motivated by the noun ‘mind’ (indeed in modern French the noun mente has disappeared). It has become a meaningless grammatical attribute that needs to be learned separately. The development of the English s-genitive is another example of how iconicity becomes important when a new cycle starts. Although the s-genitive has not become reduced to zero (though actually in early Modern English, particularly in northern dialects, it used to be increasingly s-less, as in his father boots), it had almost fallen out of use as a productive inflection in Middle English. From late Middle English onwards, however, the s-genitive begins to change from an inflection into a clitic (cf. Allen 1997b). Note, that in this respect we may equally well speak of a genuine case of degrammaticalization (inflection > clitic) rather than the beginning of a new cycle. This change correlates with a highly significant increase in the frequency of the s-genitive (see Rosenbach and Vezzosi forthcoming, 1999). As argued by Rosenbach (forthcoming) the preferred contexts for the use and diachronic spread of the s-genitive point to an iconic motivation for the use — and increase — of the s-genitive, in that, for reasons of efficient language processing, the s-genitive makes it possible for easily accessible possessors, ie. animate and topical possessors, to occur early in the linear order of a possessive construction (note, that in the alternative of-genitive the possessor follows the possessum). Also, the s-genitive represents the more implicit structure to encode close possessive relationships, which is in accordance with the principle of ‘conceptual distance’ proposed by Haiman (1985: § 2.2). This originally strong iconic motivation for the use of the s-genitive seems now to be about to fade. As further shown in Rosenbach (forthcoming), in Modern English the s-genitive, while still gaining ground, is doing so increasingly in non-iconic contexts, particularly with inanimate possessors, thus showing traces of routinization. In both respects, iconicity is closely linked to grammaticalization, and they can be said to occupy two different poles (i.e. an iconic and a symbolic pole respectively) on the axis along which language moves (cf. Plank 1979). The iconic pole stands for creativity and expressivity on the side of the speaker, while the symbolic pole represents the arbitrary and conventional elements of language: through frequent use, originally motivated expressions lose much of their iconic content and gain routine, thereby becoming more economic in terms of processing costs. This may suggest that iconicity and grammaticalization are simply each other’s opposites, and that the pathway is usually from the iconic pole to the symbolic one. Things are not as simple as that, however, as we have tried to illustrate in Figure 2. The opposition is not only between iconic on the one side and symbolic/economic on the other, opposition can also turn up between different, competing iconic motivations (see e.g. Haiman 1985a: ch.6; DuBois 1985).

30

OLGA FISCHER AND ANETTE ROSENBACH

iconic motivation iconic principle 1

economic motivation iconic principle 2

expressivity and creativity (for speaker) not grammaticalized

processing speed, need for clarity (for hearer)

?

grammaticalized

Figure 2. Iconicity and grammaticalization: Competing forces

For example, the principle of placing old information before new information within an utterance can clash with the principle of actuality (Jespersen 1949: 54), i.e. the tendency to express first what is currently most important for the speaker (which is most likely not old information; this is what Tabakowska [1999] has termed ‘experiental iconicity’). In addition, what is often not taken into account is that speakers and hearers may have different needs, which may well clash, too. While the speaker is creative, the hearer may brood over this new expression, trying to figure out what it possibly means. On the other hand, expressions that have already been symbolized to a great extent (i.e. are already near the end of the grammaticalization cycle) may become opaque for the hearer (the speaker always having the advantage that he/she knows what he/she wants to say) and therefore uneconomic in certain situations. As Fischer (1999: 348), referring to Fónagy (1982, 1995) has pointed out, “we are …always at the crossroads of both possibilities”, i.e. the iconic/creative and the symbolic/mechanistic pole. Even the symbolic may become remotivated because, as Fónagy (1999: 3) argues, all linguistic units “are the product of a dual encoding procedure”: when they are generated by the grammar, they “have to pass in live speech through a Distorter (or Modifier) conveying complementary messages, integrated into the original linguistic message”. This ‘dual code’ consists of the arbitrary rules of grammar on the one hand, and the transparent, motivated (by the external world) rules of the ‘Distorter’. For these reasons, the pathway is not necessarily from iconic to symbolic, from less to more grammaticalized, but can potentially also be the other way round (see also Section 4.3). A case in point is the development of infinitival to as shown by Fischer (this volume). In contrast to Dutch, where infinitival te is progressively moving towards the symbolic, i.e. the more grammaticalized pole, the corresponding English to stopped in its grammaticalization process around

INTRODUCTION

31

late Middle English and moved back partly towards the iconic pole, in that to became meaningful again in its grammatical function. Molencki (this volume) shows how iconicity and grammaticalization may intermingle in the most intricate ways. He looks at the expression of counterfactuality in the history of English and finds, not unexpectedly, that the verbal forms used in the protasis and apodosis are being replaced again and again by more expressive forms (the iconic pole) due to the fact that the earlier forms have grammaticalized to (almost) zero. Thus, the Old English preterite subjunctive might be replaced by the pluperfect or by a modal periphrasis. He also finds, which is the actual topic of his paper, that there is a very strong tendency to preserve parallelism between the verbal forms of the apodosis and protasis. When the would-periphrasis first occurs in the apodosis for transparent or iconic reasons (presumably there first, because 1) the earliest uses of would are volitional, and volitionality only plays a role in the apodosis, and 2) because, of the two clauses, the apodosis is the most counterfactual and may therefore be selected for extra marking), it is soon followed by the use of would also in the protasis. This is not only true for English but also for many other related and unrelated languages. Molencki ascribes this further grammaticalized use of would in the protasis (further grammaticalized because it is less motivated in the protasis), to the iconic principle of isomorphism, i.e. the tendency for structures with similar meaning to acquire similar forms (and vice versa). Another case of intermingling can be uncovered in Fitzmaurice’s contribution to this volume. She shows (implicitly) how iconicity may counterbalance the progress of the grammaticalization of the infinitival marker in a number of semiauxiliaries in present-day American English. She shows that there is a strong tendency to place the negator not between to and the infinitive. This placement conveys “an impression of greater negative force”, and could therefore be said to be iconically motivated by the so-called ‘distance principle’ (see Section 4.5.4 above): the closer the negative stands to the activity to be negated, the more forceful the negator is (compare the opposite effect in ‘negative raising’ constructions where the negative force is ‘softened’ by the greater distance between the negator and the verb). The effect of this not-placement is a further degrammaticalization of to, Fitzmaurice argues, which is now no longer bonded to the infinitive: “the investment of purpose force in to [is…] a consequence of the interruption of the infinitive verb sequence by the negator” (p. 178).

32

OLGA FISCHER AND ANETTE ROSENBACH

Acknowledgments We would like to thank Dieter Stein for valuable comments on an earlier version of this introduction.

Notes 1.

In more recent work (1991), Lightfoot tries to incorporate aspects of graduality in his account of language change (according to Harris and Campbell [1995: § 2.3.2.2], this is not very successful). In (1999), Lightfoot ‘solves’ the problem of gradualism by pointing to two different lenses through which we may view change: “languages … change gradually; grammars are a different matter” (p. 83). By concentrating on the purely grammatical and on the individual’s competence, and by following a strictly modular approach to grammar, it is indeed possible to ignore the gradual aspects of change.

2.

For discussions of Robert’s generative account of grammaticalization see Newmeyer (1998: § 5.7.2) and Haspelmath (1998: 341–344).

3.

Recently, Heine (1999c) has stressed the importance of the role of context played in grammaticalization processes. He distinguishes four developmental stages (initial stage → bridging context → switch context → conventionalization), in which different contextual requirements are at work in the evolution of new grammatical meanings without making an appeal to metaphor and metonymy. Heine explicitly states, however, that the contextual mechanisms he proposes and an analysis of meaning transfer in terms of metaphor and metonymy are not incompatible but rather complementary analytical tools in that an understanding of the various kinds of contexts figuring in grammaticalization will help to explain why new meanings evolve out of certain existing meanings.

4.

Note, that recently Haspelmath (1998) has made a case for treating grammaticalization and reanalysis as distinct processes, with analogy being yet another type of change.

5.

Note, that in contrast to Meillet’s view Lüdtke (1980) sees phonetic reduction as the driving force in language change. Only the loss of phonetic content will have effects on expressivity, which will then trigger further semantic-syntactic change, see also Section 4.4 below.

6.

Another example that seems to support the connection between leftward position within the NP and subjective meaning may be the English s-genitive, which seems to have acquired a ‘personalization’ function (see Dabrowska 1998). Note, that the possessor within the s-genitive is realized in left position, which may make this construction especially suitable to express subjective meanings; see Rosenbach, Stein and Vezzosi (2000), who suggest that the English s-genitive has, diachronically, acquired a textual function and may now be undergoing subjectivization.

7.

See also Newmeyer (1998: § 5.3) for a discussion of the nature of grammaticalization as a deterministic process with its own laws or as an epiphenomenon resulting from other processes. He concludes that the latter is the case and he, therefore, argues that there is no need for a separate theory of grammaticalization.

8.

But see also Haspelmath (1998: 318–322) and Haspelmath (1999) for an elaboration of possible motivations for unidirectionality. In Haspelmath (1999), for example, unidirectionality is accounted for in terms of Keller’s (1990) invisible hand account.

9.

Stein (forthcoming: Section 6) discusses how most cases of ‘backward’ development seem to

INTRODUCTION

33

be cases where we have the “re-evaluation of [a] varietal status of a particular form”. What he means is that instances of degrammaticalization seem to be linked to the re-activation of older variant forms, with still fuller referential meaning, which have lain dormant for a while or have survived in other dialects. Accordingly, some cases may look on the surface like cases of degrammaticalization simply because of the fact that they become manifested in the standard (written) language only. If the, supposedly, ‘de-grammaticalized’ form is in fact an older spoken or dialectal variant that simply manages to return into the standard, then this is not degrammaticalization but rather ‘backward divergence’. In this light, it could be said that the cases discussed by Fischer and Fitzmaurice should be called ‘backward divergence’ rather than degrammaticalization. 10.

We owe this observation to Elizabeth Traugott (p.c.).

11.

For a more detailed discussion on the role played by iconicity in grammaticalization we refer to Fischer (1999).

12.

For an overview of the various types of iconicity we refer to Fischer and Nänny 1999.

References Abraham, W. 1993. “Grammatikalisierung und Reanalyse: einander ausschließende oder ergänzende Begriffe?”. Folia Linguistica Historica 13: 7–26. Allen, C. 1997a. “Middle English and the ‘creolization’ hypothesis”. English Language and Linguistics 1: 63–89. Allen, C. 1997b. “The origins of the ‘group genitive’ in English”. Transactions of the Philological Society 95: 111–131. Bruyn, A. 1995.Grammaticalization in Creoles: The Development of Determiners and Relative Clauses in Sranan. Dordrecht: ICG Printing. Bybee, J. 1985. Morphology. Amsterdam: Benjamins. Bybee, J. L., Perkins, R. and Pagliuca, W. (1994). The Evolution of Grammar: Tense, Aspect, Modality in the Languages of the World. Chicago: University of Chicago Press. Chafe, W. 1998. “How a historical linguist and a native speaker understand a complex morphology”. In M. S. Schmid, J. R. Austin and D. Stein (eds). Historical Linguistics 1997. Selected Papers from the 13th International Conference on Historical Linguistics, 101–116. Amsterdam: Benjamins. Dabrowska, E. 1998. “How metaphor affects grammatical coding: The Saxon genitive in computer manuals”. English Language and Linguistics 2: 121–127. Demske, U. Forthcoming. Merkmale und Relationen. Diachrone Studien zur Nominalphrase des Deutschen. Berlin: Mouton de Gruyter. Deutscher, G. 1999. “The different faces of uniformitarianism”. Paper delivered at the XIVth International Conference on Historical Linguistics, August 1999, Vancouver. Diewald, G. 1997. Grammatikalisierung. Eine Einführung in Sein und Werden grammatischer Formen. Tübingen: Niemeyer. Domingue, N. Z. 1977. “Middle English: Another creole?”. Journal of Creole Studies 1: 89–100.

34

OLGA FISCHER AND ANETTE ROSENBACH

DuBois, J. W. 1985. “Competing Motivations”. In John Haiman (ed.), 343–365. Fischer, O. 1994. “The development of quasi-auxiliaries in English and changes in word order”. Neophilologus 78: 137–164. Fischer, O. 1997. “On the status of grammaticalisation and the diachronic dimension in explanation”. Transactions of the Philological Society 95: 149–187. Fischer, O. 1999. “On the role played by iconicity in grammaticalisation processes”. In M. Nänny and O. Fischer (eds), 345–374. Fischer, O. and Nänny, M. 1999. “Introduction: Iconicity as a creative force in language use”. In M. Nänny and O. Fischer (eds), xv-xxxvi. Fónagy, I. 1982. “He is only joking (joke, metaphor and language development)”. In F. Kiefer (ed.). Hungarian Linguistics, 31–108. Amsterdam: Benjamins. Fónagy, I. 1995. “Iconicity of expressive syntactic transformations”. In M. E. Landsberg (ed.). Syntactic Iconicity and Linguistic Freezes. The Human Dimension, 285–304. Berlin: Mouton de Gruyter. Fónagy, I. 1999. “Why iconicity?”. In Nänny, M. and O. Fischer (eds), 3–36. Gabelentz, G. von der. 1891. Die Sprachwissenschaft. Ihre Aufgaben, Methoden und bisherigen Ergebnisse. Leipzig: Weigel Nachf. Givón, T. 1979. On Understanding Grammar. New York: Academic Press. Givón, T. 1983. (ed.). Topic Continuity in Discourse: A Qualitative Cross-Language Study. Amsterdam: Benjamins. Givón, T. 1984. Syntax. A Functional-Typological Introduction, vol. 1. Amsterdam: Benjamins. Givón, T. 1990. Syntax. A Functional-Typological Introduction, vol. 2. Amsterdam: Benjamins. Görlach, M. 1990 [1986]. “Middle English — a creole?”. In M. Görlach (ed.). Studies in the History of the English Language [Anglistische Forschungen 210], 65–78. Heidelberg: Winter. Haiman, J. 1983. “Iconic and economic motivation”. Language 59: 781–819. Haiman, J. 1985a. Natural Syntax. Iconicity and Erosion. Cambridge: Cambridge University Press. Haiman, J. (ed.). 1985b. Iconicity in Syntax. Amsterdam: Benjamins. Harris, A. C. and Campbell, L. 1995. Historical Syntax in Cross-Linguistic Perspective. Cambridge: Cambridge University Press. Haspelmath, M. 1990. “The grammaticalization of passive morphology”. Studies in Language 14: 25–72. Haspelmath, M. 1998. “Does grammaticalization need reanalysis?”. Studies in Language 22: 315–351. Haspelmath, M. 1999. “Why is grammaticalization irreversible?”. Linguistics 37: 1043–1068. Heine, B., Claudi, U. and Hünnemeyer, F. 1991. Grammaticalization: A Conceptual Framework. Chicago: University of Chicago Press. Heine, B. and Claudi, U. 1986. On the Rise of Grammatical Categories: Some Examples from Maa. Berlin: Dieterich Reimer.

INTRODUCTION

35

Heine, B. and Reh, M. 1984. Grammaticalization and Reanalysis in African Languages. Hamburg: Helmut Buske. Heine, B. 1997. Possession: Cognitive Sources, Forces, and Grammaticalization. Cambridge: Cambridge University Press. Heine, B. 1999a. “Polysemy involving reflexive and reciprocal markers in African languages”. In Z. Frajzyngier (ed.). Reciprocals: Forms and Functions. [Typological Studies in Language 41], 1–29, Amsterdam: Benjamins. Heine, B. 1999b. “Historical reconstruction: New evidence from African languages”. Paper presented at XIV ICHL, Vancouver, August 1999. Heine, B. 1999c. “On the role of context in grammaticalization”. Paper given at the international symposium on ‘New Reflections on Grammaticalization’, held in Potsdam, 17–19 June 1999. Hopper, P. J. and Thompson, S. A. 1984. “The discourse basis for lexical categories in universal grammar”. Language 60: 703–752. Hopper, P. J. 1988. “Emergent grammar and the a priori Grammar postulate”. In D.Tannen (ed.). Linguistics in Context. Connecting Observation and Understanding, 117–136, Norwood, N. J.: Ablex. Hopper, P. J. 1991. “On some principles of grammaticalization”. In E. C. Traugott and B. Heine (eds), vol. I, 17–35. Hopper, P. J. and Traugott, E. C. 1993. Grammaticalization. Cambridge: Cambridge University Press. Jespersen, O. 1949. A Modern English Grammar on Historical Principles. Vol. vii. London: Allen & Unwin. Keller, R. 1990. Sprachwandel: Von der unsichtbaren Hand in der Sprache. Tübingen: Francke. Kemmer, S. 1993. The Middle Voice: A Typological and Diachronic Study. Amsterdam: Benjamins. Lass, R. 1980. On Explaining Language Change. Cambridge: Cambridge University Press. Lehmann, Chr. 1982 [1995]. Thoughts on Grammaticalization. A Programmatic Sketch. [Arbeiten des Kölner Universalien-Projekts 48]. Köln Universität. Revised and expanded version appeared in 1995 as Thoughts on Grammaticalization [LINCOM Studies in Theoretical Linguistics 01]. München/Newcastle: Lincom Europa. Lehmann, Chr. 1999. “New reflections on grammaticalization and lexicalization”. Paper given at the international symposium on ‘New Reflections on Grammaticalization’, held in Potsdam, 17–19 June 1999. Lightfoot, D. 1979. Principles of Diachronic Syntax. Cambridge. Cambridge University Press. Lightfoot, D. 1991. How to Set Parameters: Arguments from Language Change. Cambridge, Mass.: MIT Press. Lightfoot, D. 1999. The Development of Language. Acquisition, Change and Evolution. Oxford: Blackwells.

36

OLGA FISCHER AND ANETTE ROSENBACH

Lüdtke, H. 1980. “Auf dem Wege zu einer Theorie des Sprachwandels”. In H. Lüdtke (ed.). Kommunikationstheoretische Grundlagen des Sprachwandels, 182–252. Berlin: de Gruyter. McMahon, A. M. S. 1994. Understanding Language Change. Cambridge: Cambridge University Press. Meillet, A. 1912. “L’évolution des formes grammaticales”. Scientia (Rivista di Scienza) 12, no. 26, 6. (Reprinted 1951. In Linguistique historique et linguistique générale, 130–48. Paris: C. Klingensieck.) Mithun, M. 1991. “The role of motivation in the emergence of grammatical categories: The grammaticization of subjects”. In E. C. Traugott and B. Heine (eds), vol. II, 159–184. Mithun, M. 1998. “The sequencing of grammaticalization effects: A twist from North America”. In M. S. Schmid, J. R. Austin and D. Stein (eds). Historical Linguistics 1997. Selected Papers from the 13th International Conference on Historical Linguistics, 291–314. Amsterdam: Benjamins. Nänny, M. and Fischer, O. (eds) 1999. Form Miming Meaning. Iconicity in Language and Literature. Amsterdam: Benjamins. Newmeyer, F. J. 1998. Language Form and Language Function. Cambridge, Mass.: MIT Press. Pintzuk, S. 1991. Phrase Structures in Competition: Variation and Change in Old English Word Order. Dissertation, University of Pennsylvania, Pittsburg. Plank, F. 1979. “Ikonisierung und De-Ikonisierung als Prinzipien des Sprachwandels”. Sprachwissenschaft 4: 121–158. Popper, K. R. 1968. The Logic of Scientific Discovery. Revised and translated edition. London: Hutchinson. Ramat, P. 1982. “Thoughts on degrammaticalization”. Linguistics 30: 549–560. Roberts, I. 1993. “A formal account of grammaticalization in the history of Romance futures”. Folia Linguistica Historica 13: 219–258. Rosenbach, A. Forthcoming. “The English s-genitive: Animacy, topicality and possessive relationship in a diachronic perspective”. In L. Brinton (ed.). Historical Linguistics 1999. Amsterdam: Benjamins. Rosenbach, A. and Vezzosi, L. 1999. “Was the s-genitive a traveller through England?”. In A. Bergs, M. S. Schmid and D. Stein (eds). LANA-Düsseldorf Working Papers on Linguistics 1: 35–55 (http://ang3–11.phil-fak.uni-duesseldorf.de/˜ang3/LANA/ LANA.html). Rosenbach, A. and Vezzosi, L. Forthcoming. “Genitive constructions in Early Modern English: New evidence from a corpus analysis”. In E. Poppe, A. Shisha-Halevy and R. Sornicola (eds). Stability and Variation in Word-Order Patters over Time. Amsterdam: Benjamins. Rosenbach, A., Stein, D. and Vezzosi, L. 2000. “On the history of the s-genitive”. In R. Bermúdez-Otero, D. Denison, R. M. Hogg and C. B. McCully (eds). Generative Theory and Corpus Study: A Dialogue from 10ICEHL, 183–210. Berlin: Mouton de Gruyter.

INTRODUCTION

37

Rubba, J. 1994. “Grammaticalization as semantic change: A case study of preposition development”. In W. Pagliuca (ed.) Perspectives on Grammaticalization, 81–102. Amsterdam: Benjamins. Stein, D. and Wright, S. (eds). 1995. Subjectivity and Subjectivisation. Cambridge: Cambridge University Press. Stein, D. Forthcoming. “Turning the clock back in English morphosyntax. They never come back (quite the same),-?” In I. Plag and K. P. Schneider. Festschrift in Honor of Rüdiger Zimmermann. Heidelberg: Julius Groos. Sweetser, E. E. 1990. From Etymology to Pragmatics. Metaphorical and Cultural Aspects of Semantic Structure. Cambridge: Cambridge University Press. Tabor, Wh. and Traugott, E. C. 1998. “Structural scope expansion and grammaticalization”. In A. Giacalone Ramat and P. J. Hopper (eds), The Limits of Grammaticalization, 229–272. Amsterdam: Benjamins. Thomason, S. G. and Kaufman, T. 1988. Language Contact, Creolization, and Genetic Linguistics. Berkeley: University of California Press. Traugott, E. C. 1982. “From propositional to textual and expressive meanings: Some semantic-pragmatic aspects of grammaticalization”. In W. P. Lehmann and Y. Malkiel (eds), Perspectives on Historical Linguistics, 245–271. Amsterdam: Benjamins. Traugott, E. C. 1995. “Subjectification in grammaticalization”. In D. Stein and S. Wright (eds). Subjectivity and Subjectivisation, 31–54. Cambridge: Cambridge University Press. Traugott, E. C. 1996. “Grammaticalization and lexicalization”. In K. Brown and J. Miller (eds), Concise Encyclopedia of Syntactic Theories, 181–187. Cambridge: Cambridge University Press. Traugott, E. C. and Heine, B. (eds). 1991. Approaches to Grammaticalization. [Typological Studies in Language 19], 2 vols. Amsterdam: Benjamins. Vincent, N. 1995. “Exaptation and grammaticalization”. In H. Andersen (ed.), Historical Linguistics 1993, 433–445. Amsterdam: Benjamins.

A lovely little example Word order options and category shift in the premodifying string Sylvia Adamson University of Manchester

For Peter Matthews 1.

Manner adverbs v. sentence adverbs

The sentences in (1) and (2) below illustrate a common pattern of alternation in the grammar of modern English, in which (some) manner adverbs have a separate function as sentence adverbs.1 The two variants of such adverbs can be distinguished by position and by meaning. In the canonical word order, as given here, manner adverbs (such as 1a and 2a) occur in post-verb position, whereas sentence adverbs (such as 1b and 2b) occupy clause-initial position, marked off by intonation in speech or punctuation in writing.2 The difference in meaning is commonly described as one of scope: manner adverbs are said to modify the verb or verb phrase while sentence adverbs, as their name implies, include the whole clause in their scope. The contrast is particularly clear in negative sentences, where the manner adverb falls within the scope of the negative (as in 2a) but the corresponding sentence adverb does not (as in 2b). (1)

a. b.

He will answer hopefully. Hopefully, he will answer.

(2)

a. b.

He will not answer frankly. Frankly, he will not answer.

Swan (1988) puts these synchronic options into a diachronic perspective, revising, in the process, the standard account of both word order and scope. She shows that, in the case of those dual-function items for which we have a

40

SYLVIA ADAMSON

sufficient historical record, the manner-adverb function precedes the sentenceadverb function and the development of the latter is associated with a leftward movement into clause-initial position. The canonical position for sentence adverbs in present-day English, illustrated in (1b) and (2b), thus has a special status in the history of the form. As Swan puts it: (3)

[leftward movement] is an important and indeed essential factor in the stabilization of the [sentence adverb] class … Stable [sentence adverbs] may occur anywhere in the sentence with little or no risk of ambiguity, while less stable [members of the class]… are more restricted (Swan 1988: 8)

On the issue of meaning, she notes that sentence adverbs typically act as a comment on the proposition or event encoded in the following clause and suggests that if their meaning is to be described in terms of scope, then the notion of scope has to be defined in terms other than those of syntactic modification: “in a sense the speaker is part of the scope, since they are speaker-oriented” (Swan 1988: 5). This speaker-orientation is well illustrated in the examples above: in (1b) hopefully expresses the speaker’s epistemic or emotional stance (= ‘I am hopeful that he will answer’), in (2b) frankly could be analysed as a manner adverb referring to the current speech act (= ‘I tell you frankly that he will not answer’). By contrast, the manner adverbs in the (a) sentences are descriptive rather than expressive or metalinguistic, referent-oriented rather than speaker-oriented. Viewed in this light, the change of meaning involved in the diachronic category shift that produces sentence adverbs is not so much an enlargement of the syntactic domain that the adverb modifies as an instance of what is now known as subjectivisation.3

2.

Adverbials > discourse particles

A strikingly similar pattern emerges from recent studies of the grammaticalisation of discourse markers (Finell 1996; Traugott 1995). The examples in (4) below (all taken from the Helsinki Corpus) are used by Traugott to illustrate the developmental cline from Clause-internal Adverbial > Sentence Adverbial > Discourse Particle: (4)

a.

al þat þou hauest her bifore i-do, in þohut, all that you have here before done, in thought, in speche, and in dede … ich þe foryeue in speech and in action … I you forgive

(c.1300)

CATEGORY SHIFT IN THE NOUN PHRASE

b.

c.

41

purposely suffring the more noble children to vainquysshe… thoughe in dede the inferiour chyldren haue more lernyng (1531) thereby [the flea is] inabled to walk very securely both on the skin and hair; and indeed this contrivance of the feet is very curious for performing both these requisite motions (1665)

In (4a) in dede is a preposition phrase in which dede still carries its lexical meaning as a noun = ‘action’. The phrase functions as a clause-internal adverbial with narrow scope, the writer drawing a contrast between the different domains in which sins are committed ‘in thought, in word and in deed’. In (4b) — Sir Thomas Elyot’s Machiavellian version of positive discrimination, in which teachers are advised to organise contests in which nobly-born pupils out-perform their more learned low-born class-mates — the adverbial phrase has moved out of the verb phrase into pre-subject position and its syntactic scope is wider, comprising the whole clause, the inferiour chyldren haue more lernyng. Semantically, it is arguably ambiguous between a descriptive meaning close to that of (4a), paraphraseable as ‘in performance’ or ‘in what they do’, and a more subjective meaning, approximating to an epistemic modal adverb such as certainly or truly. In (4c) the ambiguity has pretty much been resolved: the orthography (as is commonly the case after 1600) indicates that the preposition and noun are being treated as a single item, whose main function seems to be metalinguistic or metatextual. Indeed here acts as a signal that the following clause ‘confirms and amplifies a previous statement’ (OED sense 3). The development of indeed in (4) has obvious affinities with the development of frankly in (2). In both cases there is a change of syntactic category accompanied by a semantic shift from referent-oriented meaning to speakeroriented meaning, or from descriptive function to metalinguistic function. And in both cases the establishment of the item in its new function is associated with its appearance in a left-peripheral position. Traugott, noting these similarities, looks for a general explanation in Kiparsky’s account of the clause structure of English, which incorporates a left-peripheral slot for topicalised elements (additional to and preceding a Specifier slot which hosts wh-phrases and focused elements) (Traugott 1995: 9–10; Kiparsky 1995: 140–147). See (5) below.

42

SYLVIA ADAMSON

(5)

CP TOPIC

CP

SPEC

C′ Comp

S

While not disputing this analysis of events at clause level, I want here to explore some data which suggests that the link between subjectivity and the left periphery may be even more general. It is a case-study in which syntactic-semantic change and leftward shift are again combined, but this time in the domain of the noun phrase.

3.

The order of adjectives

In fact, I shall primarily, and certainly initially, be concerned with one subdomain of the noun phrase, the slot occupied by pre-modifying adjectives. In descriptive grammars of English this is commonly located between two other prenominal elements:4 (6)

a. b. c.

pre-adjectival modifiers: e.g. all, next, four … adjectives: e.g. large, white, stiff … post-adjectival modifiers: e.g. leather, dog …

As the nomenclature here implies, if items from different classes are combined in a NP, the canonical order is a + b + c + N, as in four large leather collars. However, that is not to say that there is free ordering within each class. Among pre-adjectivals, the sequence four all [collars] is less acceptable than all four [collars] and among post-adjectivals, dog leather [collars] is less acceptable than leather dog [collars]. In the case of the adjectives proper, ordering seems in some cases constrained and in others not. So fresh crisp [lettuce] and crisp fresh [lettuce] seem equally acceptable, but most native speakers find large green [lettuce] clearly preferable to green large [lettuce]. Dixon (1982), perhaps the most ambitious attempt to account for such patterns of preference, suggests that they are grounded in the division of adjectives into seven semantic sub-classes, as given in (7):

CATEGORY SHIFT IN THE NOUN PHRASE

(7)

1. 2. 3. 4. 5. 6. 7.

43

 — good, nice, excellent, horrible, delicious …  — small, long, thin, large, wide …   — crisp, hard, soft, heavy, smooth …  — fast, quick, slow …   — gracious, kind, proud, generous …  — new, young, old …  — black, green, red …

Adjectives from the same sub-class can be combined in any order, hence the interchangeability of the paired   adjectives crisp fresh/fresh crisp or the paired   adjectives brave clever/clever brave. But when adjectives from different sub-classes are combined, there appears to be a generally preferred left-to-right ordering, reflected in the 1–7 numbering given in (7). In (8) and (9), this ordering is observed in the (a) options and violated in the (b) options: (8) (9)

a. b.

?jealous

horrible jealous old man old horrible man

a. b.

?green

[1-5-6: .-.-]

nice large soft green cushions [1-2-3-7: .-.-.-] soft nice large cushions

It is not that sequences such as green large or soft nice or old jealous cannot occur but that they occur with marked intonation patterns, such as comma disjuncture between the adjectives or contrastive stress on one of them. The sequences in (8a) and (9a) represent what (7) predicts as the semantically unmarked option. Quirk et al., who note a similar pattern of ordering, suggest that “in part the preferences seem to correspond to the ‘natural’ order of recursive qualification” (1972: 924). On this view, adjectival modification is a form of scope-restriction in which the addition of adjectives progressively narrows down the class of referents picked out by, say, cushions in (9a) or man in (8a). But the inverted commas round natural in Quirk et al.’s formulation avoids, rather than provides, an explanation for the particular constraints on ordering that they have observed. Dixon addresses this question by resurrecting and elaborating Whorf’s notion of iconicity. Whorf (1956: 93) had proposed a general principle whereby adjectives encoding inherent properties of the referent are placed closer to the head noun than those encoding non-inherent properties. In Dixon’s elaboration, (7) represents what is essentially a scale of inherence, in which properties of, for instance,  and  are taken to be less inherent than properties of  and . The question of ordering in adjectives is thus in his account divorced

44

SYLVIA ADAMSON

from the question of ‘recursive qualification’. In fact, Dixon explicitly contrasts adjectives in this respect with other classes of premodifiers. Whereas preadjectival modifiers do recursively restrict scope (hence the cleverest two men is not synonymous with the two cleverest men), adjectives, he claims, are not recursive. Rather, each directly qualifies the head noun, so that “a clever brave man has similar cognitive meaning to a brave clever man” (Dixon 1982: 25). He recognises, however, one exception to this generalisation, namely the  class of adjectives: “a  adjective qualifies not the head noun, but some other adjective” (ibid: 25). For example, a good fast car is not a car that is both good and fast but a car that is good because it is fast. Hence the phrase a good fast new car is not cognitively equivalent to a good new fast car (in the first case, the car is good by virtue of its speed, in the second by virtue of its newness). Taking this claim a stage further, Dixon proposes that even if there is no other adjective in the string (e.g. a good car; an atrocious play), the  adjective still does not qualify the head noun. Instead “in a good box, good effectively qualifies some implicit non-value adjective, which itself qualifies box”, the relevant adjective (e.g. strong, roomy) being supplied to the hearer’s mind by the context (Dixon 1982: 26). He sums up his view: (10)

A value adjective cannot directly qualify a noun; if it appears in an NP with a noun and no other adjective, then it has manner function with respect to an implicit non-value adjective (Dixon 1982: 30)

I shall consider the general question of ‘recursive qualification’ in the last section of this paper. My more immediate concern is with two other issues raised by (10). One is its implication of a category-shift in  adjectives. The formulation — ‘manner function’ — seems to point to their kinship with adverbs, and though this is not a suggestion pursued by Dixon himself, it proves particularly illuminating when applied to the historical development of adjectives of this type, as I shall argue in detail below. But first I want to take issue with Dixon’s account of the scope and meaning of such adjectives, in particular with his cheerful positing of invisible syntactic entities for them to modify. What, after all, is the evidence for the presence of an ‘implicit non-value adjective’ in such sequences as a good box or an atrocious play? A much simpler explanation for the data Dixon reviews can be achieved by positing a semantic split within the adjective group, such that adjectives in categories 2–7 are descriptive or referentoriented while adjectives in category 1 — the  class — are affective or speaker-oriented. Their function is not to describe properties in the referent but to express speaker-response to it. So in using a phrase like a good box or a nice book, I’m telling you ‘I approve’ and your knowledge of me then allows you to

CATEGORY SHIFT IN THE NOUN PHRASE

45

infer what properties in the box or book are likely to have evoked that response. In the case of a nice book, for instance, the approval may relate to properties of genre, writing style, binding, illustrations and so on, depending on who occupies the role of I.  adjectives are thus subjective in the same sense as deictic terms: their referential meaning is largely dependent on their speaker’s identity. In fact, the proposal I make here for a split between affective and descriptive adjectives is similar to what Quirk et al. propose as an ordering principle for all premodifiers: (11)

one is tempted to suggest one principle accounting … for … all premodifiers: a subjective/objective polarity. That is, modifiers relating to properties which are (relatively) inherent in the head of the noun phrase, visually observable, objectively recognizable or assessible, will tend to be placed nearer to the head and be preceded by modifiers concerned with what is a matter of opinion, imposed on the head by the observer, not visually observed and only subjectively assessible. (Quirk et al. 1972: 924–925)

I shall return in my final section to the ordering of elements in the premodifying string as a whole. For the moment it’s sufficient to note that the adjective slot, at least, provides grounds for recognising in the NP domain as well as in the clause domain a positional reflex of the division between speaker-comment and referent-description taking the form of a link between leftmost position and the speaker-oriented element.

4.

Change order, change meaning

One complication to the picture presented by (7) is that some adjectives are polysemous and their alternative senses fall into different semantic sub-classes. In particular, the  class includes not only the dedicated evaluative terms that Dixon lists, (nice, excellent, for instance) but many items which can also be used as descriptive adjectives from classes 2–7, e.g. (12)

a. b.

a sweet wine v. a sweet baby ( v.) fair hair v. a fair jury v. a fair performance ( v.  v.)

Here, as in most naturally occurring instances, the relevant sense of the adjective is largely predictable from its head noun. But an implicit claim of Dixon’s model is that word order may also play a role in sense selection, so that adjectives whose semantic range covers both descriptive and affective senses will change

46

SYLVIA ADAMSON

meaning if they change their position in a premodifying string. This does seem to be the case in the pairs below: (13)

a. b. c. d.

short rotten planks v. rotten short planks [- old dark gentleman v. dark old gentleman [- little yellow devils v. yellow little devils [- tall lovely pine-tree v. lovely tall pine-tree [-

v. -] v. -] v. -] v. -]

For each of the italicised adjectives in (13), a position to the right of its neighbour promotes a descriptive reading, a position to the left favours an evaluative reading. In (13a) rotten shifts from ‘in a state of decay’ (OED sense 1–2) to ‘worthless, beastly’ (OED sense 8b); indeed, in this second interpretation, the planks need not be physically decayed at all. In (13b) dark, when following old, is interpreted as a  adjective (referring either to complexion or haircolour), while, preceding old, it is more likely to be construed metaphorically as ‘mysterious, secretive’ (the sense it has in the idiom a dark horse or keep it dark). The  adjective behaves similarly in (13c), this time in relation to a  adjective. Both noun phrases here would be considered evaluative, but the first is more likely to be construed as an expression of anti-Chinese racism, with yellow referring to skin colour, while in the second case, a more widely applicable insult, yellow is construed metaphorically as ‘cowardly’. In (13d), when lovely stands to the right of a  adjective, it is itself construed as a   adjective, synonymous with well-formed, but shifted to the left it is an expression of speaker approval, synonymous with nice or excellent.5 (13d), however, differs from the other examples in this set. It’s noticeable that the pairing with lovely to the left seems to be the more natural order and the affective reading seems to be the more normal interpretation. This suggests that — unlike rotten, dark or yellow — lovely is regarded as predominantly a member of the  class in present-day English. Evidence for this is that the descriptive sense is quite hard to evoke and it may be necessary to violate the unmarked orderings of (7) to achieve it. So, for example, though   adjectives canonically precede  or   adjectives (hence crisp new [banknotes], plump friendly [people]), in the case of lovely a   reading can only be achieved by putting it in the rightmost position of such pairings e.g. old lovely trees; a kind lovely woman (cf. lovely old trees; a lovely kind woman). It is perhaps for this reason that Matthew

CATEGORY SHIFT IN THE NOUN PHRASE

47

Arnold selected the sequence cool lovely country in Empedocles on Etna (1852), to block the reading associated with lovely cool country, which belongs more in the register of tourist enthusiasm than poetic description.6 I say ‘perhaps’ because the date of this example opens up the question of the diachronic status of the synchronic options exemplified in (13a–d). It is this that I want to turn to now. One of the interesting claims made by Dixon is that the model of order-meaning relations expressed in (7) not only represents the unmarked preferences of Modern English but has cross-linguistic validity too. He posits the semantic types as universals and cites studies by Haiman, McElhanon, Hetzron and others as evidence that the relative ordering of the types is very similar across a wide variety of languages (Dixon 1982: 9, 26). If this is the case, it should follow that the ordering of types will also remain constant across different diachronic states of the same language, so that a change of meaning in an adjective can be expected to correlate with a change in its unmarked position in a string. To test this hypothesis, let’s look at the history of lovely, as a token of a whole group of adjectives that have undergone semantic shift from descriptive to affective senses. Classic instances of this kind of change are nice (< ‘fastidious, precise’) and horrid (< ‘bristling’). But in these cases the change is complete, the item has become specialised in its new function and earlier senses have been lost (except in fixed idioms such as a nice distinction). The interest of lovely is that its semantic shift is still in progress.

5.

The development of affective meaning in lovely

At the beginning of the Early Modern English period, lovely has two main senses, both descriptive. In the first of these, illustrated by (14a–d) below, it retains the sense-range of luflic in Old English, which puts it in Dixon’s category 5 as a   adjective, equivalent to Modern English loving or amiable. (14)7 a. b. c. d.

for sheo to him so lovely was and trewe for she to him so loving was and true quhen he wes blyth he was lufly when he was happy he was amiable with much hearty and lovely recommendations being beloved in all companies for his lovely qualities

(1375) (1375) (1533) (1586)

48

SYLVIA ADAMSON

In Middle English, lovely began also to function as a   adjective (Dixon’s category 3) with the sense of ‘physically beautiful’, as illustrated by (15) below. By 1600, this was a popular sense, and in the eighteenth century it became a dominant one. (15)

a.

b. c. d. e. f.

Be he never swa stalworth and wyght, And comly of shap, be he never so sturdy and strong, and beautifully shaped, lufly and fayre (1340) handsome and good-looking the lely, lufely to syghte (1420) the lily, beautiful to behold the tears … like envious floods ore-run her lovely face (1596) Leonora was formerly a celebrated Beauty, and is still a very lovely Woman (1711) the ladies … covered their lovely necks (1751) An amiable disposition, without a lovely person, will render a person beloved. It is distressing to see any one who is lovely in person to be unamiable in character (1816)

The shift from   to   is particularly clear when (15f) is set alongside (14d). In the 1816 example, amiable has taken over the ground covered by lovely in the 1586 example, enabling the writer of (15f) to draw an explicit contrast between the qualities designated by the two terms. But by the time he was writing, the   sense of lovely was itself being challenged by a new sense, one belonging unequivocally in Dixon’s category 1, a  adjective, expressing the speaker’s approval. As the examples in (16) illustrate, the earliest probable instances of this sense belong to the seventeenth century (and the culinary register), its full development to the nineteenth. (16)

a. b. c. d. e.

Come lets to supper … this Trout looks lovely (1653) I ..soon hooked a lovely carp. Play it, play it, said she: I did, and brought it to the bank (1741) Very snug, in my own room, lovely morng, excellent fire (1813) Dear Fred wrote, Directly, such a lovely note (1860) Mr. Lewes had ‘a lovely time’ at Weybridge (1872)

It’s possible, of course, that (16a), like (15b) above, is a comment on the referent’s physical beauty, but much more likely that it and (16b) are registering a fisherman’s approval of the size of the catch or a gourmet’s anticipation of its flavour. In (16c), though the ‘physical beauty’ sense can’t be ruled out, lovely seems to be parallel with the  term excellent and part of a whole series

CATEGORY SHIFT IN THE NOUN PHRASE

49

expressing the speaker’s satisfaction: snug, my own, excellent. (By contrast, the collocational companions of lovely in (15a) are other   adjectives, comly and fayre and in (14) other   adjectives: trewe, blyth, hearty). In (16d) it seems rather unlikely that the physical beauty of the calligraphy is in question and more probable that lovely is registering the speaker’s response to the note’s contents or to Fred’s thoughtfulness in sending it. In (16e), neither   nor   senses can be in play: lovely here reports affect in an experiencer rather than describing any attribute of the stimulus. Read in context (one of George Eliot’s letters) it is not clear whether the quotation marks indicate that the phrase a lovely time is directly quoted from Lewes’s own report or (as I think more likely) that Eliot is flagging its status as a current colloquialism. The context of (16d), too, suggests that in 1860 the ‘speaker approval’ sense was a marked one, associated with female speakers, and particularly those who were regarded as ‘not quite gentlefolk’ (Coventry Patmore: Faithful for Ever, Part 3, Letter 1). Though single examples such as these can help establish the date at which a particular sense was available, they cannot establish the pattern of use for a word or the relative prominence of its various senses in a given period. To try to pin down the historical moment at which  lovely displaced the   or   senses, I turned to a multi-genre representative corpus for evidence of any change in norms of usage. The table below lists all the nouns premodified by lovely in the ARCHER corpus:8 (17)

1650–1700

1700–1750 1750–1800

1800–1850 1850–1900 1900–1950

Proper Noun, youths, face, shade (= Male Human), charmer (= Female Human), creature (= Female Human) i.e. all Human referents Proper Noun (2), boy, face i.e. all Human referents Proper Noun (3), face, lady, girl, girls (2), daughter, maiden, lily (= Female Human), companion (= Female Human), creature (= Female Human), source (= Female Human), statue (= Female Human), innocence (= Female Human) i.e. all Female Human referents Proper Noun, woman, girl, statue (= Female Human). i.e. all Female Human referents face, victim, evening, day’s fishing, ornaments, spot Proper Noun, one (= Male Human), laziness, letter

50

SYLVIA ADAMSON

1950-

cakes, children, creature (= Female Human), family, food, girl (2), hands, house, lady, lampshade, instrument, man, morning, organisations, pageant, place (2), reveling, stay, talk, trip, walks, woman (2)

What we see here is the record of two distinct changes in collocational patterning. In the first two samples, covering the period 1650–1750, all the head nouns carry the semantic feature [+] (as in the case of youths or boy) or, in context, have human referents (as in the case of shade, charmer, creature) or refer to human body parts (as with face). In contrast, the collocations recorded for the following century 1750–1850 show a striking restriction in range, despite the fact that the number of nouns premodified by lovely increases in this period. All of them, either inherently or in context, refer to a female human, a change which suggests that by the second half of the eighteenth century the   sense of lovely has displaced the   sense and that its prototype meaning is ‘female physical beauty’.9 But this restriction in referential range is followed by an even more striking expansion. In the period 1850–1900 the typical early collocates of lovely (here represented by face and victim) are outnumbered by nouns which cannot have a human reference, including one — day’s fishing — to which neither the   or   senses of lovely could plausibly be applied. This pattern is repeated in the remaining time chunks of the corpus, the collocations for the post-1950 period being so various as to suggest that in contemporary English there is hardly a noun with which lovely cannot be collocated, — a situation that points to ‘speaker approval’ being its current prototype sense.

6.

Change meaning, change order

I want now to return to the relation between meaning and word order. One prediction to be drawn from Dixon’s model in (7) is that if the unmarked or default meaning of an adjective changes, so should its default position in a string of adjectives. In the unmarked case, a  adjective will always occur in the leftmost position whereas with   or   senses (category 3 and 5 respectively) there is more chance of an adjective occupying second position in the string. We would expect then that as the prototype sense of lovely shifts progressively from   >   >  it will occur with increasing frequency as leftmost partner. To test this hypothesis, we need to turn to a different kind of

51

CATEGORY SHIFT IN THE NOUN PHRASE

evidence. The representative historical corpora currently available (Helsinki and ARCHER) are limited by the small size of their database. A much larger corpus is provided by the quotation database of the OED and though its corresponding disadvantage, of course, is the relatively arbitrary nature of its text selection, it can reveal gross trends in the frequency of an item and in its patterns of use. In the following Table (18), based on the quotation database of the OED, the first column gives the number of occurrences of lovely per century, the second column the percentage of those occurrences in which it appears with another adjective, and the third column the percentage of those cases in which it occupies leftmost position in such a string.10 (18)

1500–1600 1600–1700 1700–1800 1800–1900 1900-

lovely

% lovely + a.n.other adj.

% lovely leftmost

95 [77] [96] [120] [171]

09 11 05 25 33

34 39 60 92 91

The figures in the first column confirm, on a larger scale, the picture given by ARCHER of an overall rise in lovely’s rate of use, unsurprisingly perhaps given its widening range of reference.11 As for its combinatorial properties, though the numbers here are still relatively small, the pattern is clear enough to have at least diagnostic value. The sixteenth and seventeenth centuries pattern together, with lovely co-occurring with another adjective in about 10% of cases and appearing first in the string in about one third of those instances. The eighteenth century offers an interesting break in this pattern. Although the overall occurrence of lovely remains within the range of the previous two centuries, the percentage of co-occurrence with another adjective is halved (a fact that might repay further investigation) but the percentage of cases in which it occupies the leftmost position doubles, an outcome consistent with the hypothesis that for the eighteenth century the prototype sense is not category 5 ( ) but category 3 ( ), which is likely only to have  or  adjectives preceding it. However, suggestive though the percentages are, the numbers are too small to establish anything with confidence beyond the need for further research into eighteenth-century usage. More unequivocally significant is the contrast between the last two centuries (from 1800) and the first

52

SYLVIA ADAMSON

two: a rise from around 10% to around 30% in the likelihood of lovely occurring with another adjective and from around 30% to over 90% in the likelihood of its appearing leftmost in such constructions. Given the hypothesis emerging from Section 5 that the mid-nineteenth century might be the crucial period of change, it seemed worth breaking the overall figures down again. The result is given below: (19) % lovely + a.n.other adj.

% lovely leftmost

1500–1600 1600–1700 1700–1800

09 11 05

34 39 60

1800–1850 1850–1900

11 32

85 94

1900-

33

91

Comparing the top row of (19) with the bottom row, we can see a tripling in both columns between the sixteenth century and the twentieth (9%–33% and 34%–91%), but the rise takes place at different rates in each. In the second column, there is a major rise in the eighteenth century and another in the period 1800–1850; in first column, however, the figure for 1800–1850 matches those for the sixteenth and seventeenth centuries and the tripling occurs suddenly in a single leap (11%–32%) in the second half of the nineteenth century. One possible explanation for this striking rise in lovely’s rate of co-occurrence with another adjective is to posit a mid-century shift in premodification habits, with double adjectives becoming a favoured stylistic option (in the same way as it was apparently disfavoured in the eighteenth century).12 The other possibility, which is the one I shall pursue here, bears on the ongoing debate about the ordering of syntactic and semantic change in linguistic history. Taken together, the figures of (19) are at least compatible with a scenario in which semantic shift precedes and triggers syntactic reanalysis, such that the figures for 1800–1850 represent the outcome of the semantic shift in lovely charted in the examples of (14)–(16), while the figures for 1850–1900 represent a subsequent — and consequent — category shift in which lovely develops an additional function as intensifier. We may hypothesise a story of change summarised in (20):

CATEGORY SHIFT IN THE NOUN PHRASE

(20)

53

Stage 1 (1500–1700)  /( ) . Stage 2 (1700–1800)  /() . Stage 3 (1800–1850) /( ) . • subjectivisation of meaning leads to: — • left position appropriate for evaluative semantics • increased frequency in left configuration enables reanalysis as: — Stage 4 (1850–1900)  • new syntactic function leads to: — • obligatory co-occurrence with another adj., leading to: — • increased frequency in left configuration

Between Stage 1 and Stage 2, as lovely shifts its prototype sense from category 5 ( ) to category 3 ( ), the ordering constraints of (7) increase the likelihood of its occurring in leftmost position when co-occurring with other adjectives in a string (reflected in the eighteenth century rise to 60% in Table (19)). This in turn increases the likelihood of its being interpreted as a member of the canonically leftmost class, the  adjectives, and in Stage 3, the situation posited for the early nineteenth century, the prototypical sense of lovely shifts from the category of   to that of . This is a more radical shift semantically, since it involves a subjectivisation from descriptive to affective meaning, and, in terms of word order, it makes leftmost position the unmarked option for lovely in any adjectival string (reflected in the rise to 85% for this period in Table (19)). These two factors combine to provide the conditions under which lovely may be reanalysed as an intensifier (Stage 4) and the addition of this syntactic function to its repertoire may account for the leap in its frequency of co-occurrence with another adjective from the later nineteenth century onwards, since as adjective it may co-occur with another adjective but as intensifier it must do so.

7.

Change order, change category: lovely as intensifier

The classic example of category shift from adjective to intensifier is, of course, very, which became established as an intensifier in the seventeenth century, having originally entered the language in the thirteenth century as an adjective <

54

SYLVIA ADAMSON

OF verai (Peters 1994). In this case, the change has proceeded so far that the original adjectival function and meaning have been virtually lost, except in some marginal or ossified usages. But there is a whole group of affective adjectives which in contemporary English retain dual function as adjective and intensifier, as illustrated below: (21)

good nice dirty blood jolly pretty lovely

        

(big) houses

All the items in the lefthand column of (21) can appear directly in front of the noun, in which case they are interpreted as adjectives, but they can also appear in front of an adjective, in which case they may be interpreted as intensifiers, modifying not the head noun but the adjective that follows them in the string. Members of the group show varying degrees of semantic bleaching when used as intensifiers. All lose their original descriptive sense and many also lose their affective sense.13 So dirty and bloody in (21) lose their negative affect just as jolly and pretty lose their positive affect (so that pretty ugly is neither denotatively nor emotionally anomalous). In other cases, such as nice and good, some of the original affect persists, so that the intensifier in nice big and good big usually means something like ‘very + approval’. This is the group to which lovely belongs, in sequences such as: lovely big party, lovely quiet engine, lovely long legs, lovely warm room.14 For all the intensifiers in (21), unlike very, the category shift is synchronically reversible. Changing the word order prompts a reanalysis, reviving either an affective or a descriptive sense and triggering a syntactic re-interpretation of the construction as [adj + adj] rather than [adv + adj]. Consider the contrasts in (22) (22)

a. b. c. d.

good long road v. long good road dirty big box v. big dirty box jolly small woman v. small jolly woman lovely long legs v. long lovely legs

It seems possible, then, that as in the case of the manner adverbs and sentence adverbs of (1), the synchronic alternations of (22) are the residue of a historical process and that the history of other members of the group of intensifiers listed

CATEGORY SHIFT IN THE NOUN PHRASE

55

in (21) is similar to the history of lovely as I have traced it in this paper, with the development from descriptive to affective meaning prompting a shift to the left of the adjective string and the shift to the left triggering a categorial reanalysis from adjective to intensifier. In other words, there may be grounds for adding to the set of recognised grammaticalisation clines the sequence: (23)

8.

Descriptive adjective > Affective adjective > Intensifier15

Word order and subjectivity in the noun phrase

Finally I want to return to the relation between word order and subjectivity. It is clear that affective adjectives are more subjective than descriptive adjectives and intensifiers are more subjective than adjectives. It’s clear too that in both of these cases the more subjective element appears to the left. The question now to be addressed is how far this pattern generalises to the noun phrase as a whole. In the case of a two-element noun phrase, such as the cat, we can argue for a link between subjectivity and the lefthand element on the basis of Lyons’s distinctions between (a) nouns and noun phrases and (b) denotation and reference (Lyons 1977: 208, 1991: 141). In this account, the noun has denotation without reference, its function being to delimit a class (e.g. ), while the determiner’s role is to turn denotation into reference and to locate the referent in relation to a current speech act (e.g. in the utterance have you fed the cat today?, the definite article identifies the referent as a specific member of the given class, known to both speaker and hearer). Other linguists working within the Benveniste tradition (e.g. Teyssier 1968; Milner 1978; Coulomb 1994) have extended this kind of analysis to the expanded noun phrase, in which one or more modifiers are interpolated between determiner and noun. Teyssier, for example, proposes a functional classification of English premodifiers into elements which (i) identify, (ii) characterise and (iii) classify. As shown in (24) below, this tripartite scheme maps readily on to the configurational model of (6) in a way that lends specificity to Quirk et al.’s ‘general principle’ of a subjective-objective polarity, expounded in (11):

56

SYLVIA ADAMSON

A identifying function

(24)

—

pre-adjectival — modifier

B C characterising — classifying function function adjectival modifier

— post-adjectival modifier

subjective

objective

Rightmost in the string is the classifying function, typically associated with postadjectival modifiers and canonically realised by nominal adjuncts (as in dog [biscuit], cheese [omelette], farm [cat]). It is the least subjective of the premodifying elements in that it is assimilated to the function of the noun, working in combination with it to restrict its denotative scope to a smaller sub-category which may not be separately lexicalised; in some cases, where a genre or genus achieves widespread recognition, the modifier-noun collocation may itself become conventionalised or even lexicalised as a compound noun (e.g. houseboat, postman, tom-cat). At the leftmost periphery of the modifying string is the identifying function, canonically associated with such preadjectival modifiers as ordinal and cardinal numbers, superlatives and quantifiers (e.g. first [biscuit], three [omelettes], few [cats]). These items work together with determiners (articles, demonstratives, possessives) to identify more closely the instance relevant to the current speech-act. Their domain is thus reference rather than denotation and they can be seen as extending the functions of the determiner in expanding on or making explicit the notion of definiteness (hence in some accounts they are classed as post-determiners rather than pre-adjectivals). The adjectival modifiers themselves (e.g. crisp [biscuit], large [omelette], black [cat]) are canonically associated with the characterising function, which, both positionally and semantically, occupies the ground between classifying and identifying. Characterisers neither delimit the class denoted by the noun nor identify the class-member being referred to by the speaker; rather, they specify some additional attribute of the referent that the speaker signals as relevant or salient in the given context.16 The examples in (25) instantiate the model of (24), showing all three functional slots occupied by their canonical exponents: (25)

A B C (those) three blind house (mice) (a) few industrious school (girls)

CATEGORY SHIFT IN THE NOUN PHRASE

57

However, the mapping of functions on to syntactic categories is less straightforward than (24) and (25) might suggest. Adjectives in particular are by no means confined to their canonical role. Some items classed as adjectives (e.g. fiscal, dental, lunar) occur almost exclusively in the classifying function associated with nominal modifiers (compare dental surgeon to brain surgeon, fiscal policy to tax policy); others show a clear “kinship to determiners” (Bolinger 1967: 18–20), either by intensifying the sense of ‘already determined’ (same [cat], very [man]) or by relating the referent more explicitly to the time of utterance (former [policy], late [president]) or to the epistemic stance of the speaker (possible [friend], likely [candidate]). Adjectives that have specialised in these functions can be distinguished from the central characterising adjectives by a reduction or loss of syntactic properties: specifically the ability to co-occur with intensifiers or in predicative constructions (contrast: very industrious with *very former, *very monetary; or X is helpful with *X is same, *X is dental) and when they appear in a multi-item premodifying string, they occupy the positions predicted for their function rather than their word-class. Compare (26) below with (24) and (25): (26)

A B C our former stringent fiscal policy the same helpful dental surgeon

What has more seriously muddied the water for grammatical description is that many adjectives can perform more than one function. For example in (*very) sheer madness, sheer is an identifier in (very) sheer cliffs it is a characteriser; in (very) nasal pronunciation, nasal is a characteriser, in (*very) nasal cavity it is a classifier. It is this multifunctionality that produces ambiguities such as those of (27): (27)

a. b. c.

my first disastrous marriage Bloggs is our only criminal lawyer Cat is a common noun

(27a) is discussed by Radford (1988: 222, 1993: 82) as an example of scope ambiguity, the difference in meaning being dependent on whether disastrous is interpreted as a restrictive or a non-restrictive modifier of marriage. But in (27b) (adapted from an example in Bolinger 1967), the adjective is restrictive on either reading (criminal lawyer = (1) not a civil lawyer, = (2) not a law-abiding lawyer). The difference in meaning depends rather on what is being defined: in the first reading, criminal functions as a classifying adjective and restricts the denotative scope of the noun lawyer, whereas in the second reading it functions as a characterising adjective and describes the nature of the specimen Bloggs.17 This is not a peculiarity of the word criminal. (27c) works in exactly the same way.

58

SYLVIA ADAMSON

On one reading, common is a classifying adjective, and common noun delimits a category; on the other reading, common is a characterising adjective and predicates a property. (Hence we can say that, unlike cat, aardvark is a common noun but not a common noun: it belongs to the designated class but does not have the designated characteristic). This analysis could equally be applied to the ambiguity in (27a), the only difference being that the classifier reading of disastrous produces what is (lexically if not socially) a nonce class, disastrous marriage. But as Bolinger points out, it is relatively easy for nonce classes to be established in discourse (Bolinger 1967: 24–27). A passage from Doris Lessing (cited in Coulomb 1994: 8) shows this happening in practice: (28)

Jasmine whispered for a moment to a tall, thin man, the Minister for Native Affairs, who stood up and began to speak … All around Martha people were sitting leaning forward … When the tall thin man sat down, they applauded for a long time.

The string tall thin occurs here twice, but with different status (as indicated by the change in its punctuation). In its first occurrence, it presents information that is new both to the reader and to the central character, Martha, whose point of view is here represented. It is readily interpreted as part of a succession of appositive predications mapping Martha’s developing knowledge of the new character, first his physical appearance (‘tall, thin’), then his social status (‘Minister’), then his subsequent action (‘stood up …’). In the second occurrence, the information about physical appearance is a ‘given’ and the punctuation indicates that the adjective string is here to be read as restrictive modification, or construed together with the noun as a single composite unit (tall-thin-man being, in this context, an equivalent alternative to Minister as a class-designator). Another literary example, this time from P. D. James, (cited in Coulomb 1994: 7) is particularly illuminating: it not only demonstrates the contrast between characterising and classifying functions, but by using the same adjective twice in a premodifying string, it shows the relation between word order and interpretation. (29)

Here was a young, impulsive, over-curious young woman.

Here young occurs as both leftmost and rightmost premodifier. As predicted by (24) and by the general hypothesis of this paper, when it appears on the right, next to the noun, young is interpreted as a classifying modifier, collocating with woman to denote a complex semantic category (+Human +Female −Adult) for which English has no separate lexical item; when it appears on the left, next to the determiner, young is interpreted as a characterising adjective, one of the set

CATEGORY SHIFT IN THE NOUN PHRASE

59

of properties predicated of a specific representative of this class which the speaker signals as relevant in the present context. The diachronic dimension of such shifts — the question of how these word order options relate to pathways of syntactic or semantic change — has been very little studied as yet, either in terms of the history of particular items or of the categories of premodifier that they represent. One context in which such research might be fruitfully pursued is provided by Plank’s (1992) paper, where he argues (on the evidence of the behaviour of possessives in modern German dialects) that the functional distinction between determiners and modifiers is gradient rather than categorial. What further investigation might establish is that this synchronic gradience translates into a historical cline (as seems to have been the case with English possessives, which are now generally classed as determiners but in the sixteenth century still manifested the same variability that Plank finds in German possessives today). Consonant with this hypothesis (though presented within very different conceptual frameworks) is the Carlson/Lightfoot account of the genesis of the quantifier category in Early Modern English (by a reanalysis of certain adjectives) and Spamer’s account of the emergence of the determiner category in Old English (by the reanalysis of the leftmost item in a sequence of adjectival modifiers) (see Carlson 1978; Lightfoot 1979: 168–186; Spamer 1979). But both accounts are controversial, theoretically and empirically, and it is beyond the scope of the present paper to evaluate them here.18 Instead, I’ll conclude with a historical glance at two adjectives whose synchronic description famously bothered Bolinger in his 1967 paper. In Modern English mere is (in the terms I have been using in this section) a specialised identifying modifier. Picked out by Smith (1964: 39) as the token of a type “so different from adjectives as to require separate treatment … effectively developed as part of the determiner”, it well illustrates the kind of categorial indeterminacy that Plank detects in modern German possessives. In an earlier state of the language, however, mere seems to have been used as a characterising adjective, with a similar syntactic and semantic range to Modern English pure (as witness Early Modern English usages such as [the wine] is mere and unmixed). Its transition between the two states allows us to posit a pathway characteriser > identifier in which, in line with the general hypothesis of this paper, leftward movement, category shift and subjectivisation are combined. But the case of criminal suggests that this is not the only pathway of change. As the ambiguity of (27b) shows, criminal now functions as both classifier and characteriser; but it appears (on OED evidence at least) that the characterising function is, historically, the prior one (with a criminal tyrant pre-dating the criminal law). We must

60

SYLVIA ADAMSON

an instance of de-subjectivisation, since the modifier in criminal law no longer expresses speaker comment (‘the law is criminal’) but restricts the noun’s denotative scope (= not civil law). The existence of such a pathway is not a problem for the hypothesis, presented in this paper, of a link between subjectivity and leftness in the noun phrase — since as the item de-subjectivises it moves to the right of the string (compare a criminal old lawyer with an old criminal lawyer). But it poses a challenge to any hypothesis concerning unidirectionality of semantic change, such as the “hypothesis of unidirectional increase in subjectification over time” (Traugott 1995: 45). One way of saving such a hypothesis might be to restrict it to the domain of grammaticalisation and then argue that grammaticalisation is only at work in one of our two cases: so that while the development of mere from characteriser to identifier (mere and unmixed wine > a mere child) is a case of grammaticalisation, the development of a classifying function in criminal (a criminal tyrant > criminal law) may be regarded as a case of lexicalisation (analogous to the change black bird > blackbird). But such an argument has consequences for the synchronic analysis of the syntax of the noun phrase which would take us well beyond the scope of the present paper.

Acknowledgments This paper is gratefully dedicated to Peter Matthews, who first made me aware that adjective ordering was a question of interest and, by quoting from Matthew Arnold, set me thinking about its diachronic dimension. I began to formulate these thoughts for the Workshop on English Historical Syntax, held in Manchester in 1996 (see Denison and Vincent 1997: 6); I am grateful to David Denison and Nigel Vincent for inviting me to speak on that occasion and to the other participants for the helpful discussion that followed. Subsequent revisions have benefited from discussions with audiences at the 9th International Conference of English Historical Linguistics (Posnan, 1996), the 1998 Spring Conference of the University of Northern Arizona, and the University of Cambridge Linguistics Society. I owe particular thanks to those who have taken the time to read and comment on various drafts of the present paper: Keith Brown, Peter Matthews, Elizabeth Traugott, Nigel Vincent and the editors of this volume.

Notes 1.

The difference is so great as to prompt some commentators to treat the two variants of such adverbs as a case of homonymy (see Fraser 1996). More commonly the difference is described as a matter of syntax; in Quirkian terms, for instance, manner adverbs are adjuncts, sentence adverbs are disjuncts (see Quirk et al. 1972: 420–520).

61

CATEGORY SHIFT IN THE NOUN PHRASE 2.

Given the appropriate intonation/punctuation other word orders are, of course, possible for both types of adverb. The limiting case seems to be clause-initial position in negative sentences, where manner adverbs do not normally occur (e.g. *legibly she didn’t write the letter).

3.

Throughout this paper, subjectivity is to be understood as referring to those elements of language that anchor utterance in an originating speech-act, or, more generally, to “the way in which natural languages, in their structure and their normal manner of operation, provide for the locutionary agent’s expression of himself and of his own attitudes and beliefs” (Lyons 1982: 102). Subjectivisation is the diachronic process by which structures and lexemes become imbued with subjectivity, or, in the terms given currency by Traugott (1982, 1990), undergo a semantic change from meanings that may be described as referential or propositional to meanings that may be described as metatextual, expressive or interpersonal. For further discussion and illustrative case-studies, see Stein and Wright (1995). For a cross-linguistic perspective on sentence adverbials, see Ramat and Ricca (1998).

4.

The schema in (6) is taken from Dixon (1982: 24). A similar, though more detailed, account is offered by Huddleston (1984: 227–271).

5.

Similar alternations of meaning have been observed in Romance languages between adjectives in prenominal and postnominal positions. For a helpful summary of the terms in which the opposition has been characterised by Romance linguists, see Vincent (1985: 184–185 and passim). The case of Spanish adjectives is well discussed by Klein-Andreu (1983).

6.

I owe this example to Peter Matthews.

7.

The examples in (14), (15) and (16) are all taken from the OED, either from the entry for lovely itself or from elsewhere in the quotation database.

8.

I am grateful to Merja Kytö for her help in accessing this data. For an account of the design of the ARCHER corpus, see Biber et al. (1994a, 1994b)

9.

As the lovely carp of (16b) shows, it was possible for lovely to modify a [−Human] noun in the eighteenth century, but not with such frequency as to ensure inclusion in a small cross-register sample (in contrast to the position in the twentieth century).

10.

The figures for the number of occurrences have been adjusted to compensate for unevennesses in the OED’s sampling of different centuries. The adjusted figures are given in square brackets. I am grateful to Greer Gilman F.Lib. for the calculations involved.

11.

It is instructive to compare the rate of use for lovely with the figures for three other  adjectives of roughly similar meaning.

1500–1600 1600–1700 1700–1800 1800–1900 1900-

sweet

goodly

lovely

nice

1173 [698] [500] [435] [300]

400 [164] [17] [31] [6]

95 [77] [96] [120] [171]

127 [96] [169] [126] [310]

Lovely and nice, both descriptive adjectives in Early Modern English, have acquired affective meaning and increased their rate of occurrence, while goodly and sweet, two popular Early Modern English affective adjectives, have declined. Goodly, being exclusively a  term (like PDE nice), has become virtually obsolete; sweet retains currency largely by virtue of its

62

SYLVIA ADAMSON descriptive sense (= ‘sugary’) and the synaesthetic extensions of that, but in its purely affective uses (= ‘dear to me’), it has become increasingly restricted, both collocationally and in register.

12.

This explanation cannot be ruled out. It is indeed quite possible that the notable adjectivalism of nineteenth century poetry represents a stylisation of trends in ordinary language use at the time. But there is at present no research that I’m aware of that bears on this question.

13.

The degree of bleaching varies in a number of ways. For instance, it does not always go handin-hand with a widening range of collocational possibilities. Dirty, for example, loses its reference to dirt (e.g. in dirty big, dirty great) but is relatively restricted in collocability. We can’t (I think) say dirty fast.

14.

A more fine-grained analysis of the pragmatics of adjective use might show that such sequences originate as a compensatory mechanism for information deficit: as affective adjectives lose specific descriptive content, speakers may tend to insert an additional adjective to make explicit the grounds of their approval/disapproval, e.g. instead of nice dress preferring nice red dress or nice long dress, instead of good box preferring good strong box or good roomy box. In these cases, the [adj.][adj.] sequence is certainly not equivalent to ‘[adj.] and [adj.]’. In fact some members of this group notably retain their intensifier reading even when in overtly coordinate constructions e.g. it’s lovely and warm in here; the engine was nice and quiet; the water was good and hot.

15.

There is a further development in the history of lovely that deserves mention here. From the late nineteenth century onwards, the OED records usages such as these: a. b. c. d.

‘Is it all right?’ … ‘Lovely … You are duffers not to come in.’ (1889) A few silly languishers flutter and simper, ‘How nice! how lovely!’ (1896) Lovely, loves, I loved it. And Alison here was hysterical, weren’t you, Alison? (1967) ‘Mmmmm, lovely,’ she sighed as she tucked into calorie-laden hors d’œuvres, fattening spaghetti, and an enormous plateful of gooey chocolate gateau. (1971)

Here the sense ‘approved by the speaker’, found in lovely morning (16c) or lovely note (16d), seems to have generalised, so that lovely expresses speaker approval without a specific referent. It seems to be the whole experience rather than an isolable part of it that elicits enthusiasm, and lovely appears not as a premodifier but alone, possibly as an ellipsis of a predicative construction: [it is] lovely. From such uses, it is an easy step for lovely to become conventionalised into what we may call a response particle, equivalent to OK in (e) below or thanks in (f)–(g). Though not yet attested in the OED, this use of lovely has become very frequent in the 1990s, at least in British English. e. f. g.

‘I’ll arrive at six thirty.’ ‘Lovely.’ ‘Three pounds fifty.’ ‘There you are.’ ‘Lovely.’ ‘Gin and tonic?’ ‘Lovely.’

On the basis of (a)–(g) we might posit a grammaticalisation cline of the form: Descriptive adjective > Affective adjective > Pragmatic particle But this then raises the question of how such a cline might relate to the cline posited in (23): Descriptive adjective > Affective adjective > Intensifier

CATEGORY SHIFT IN THE NOUN PHRASE

63

Is there a single cline in which the intensifier stage precedes the particle stage? Or are there two distinct developments, in which the intensifier function develops from the attributive use of affective lovely (via reanalysis of the premodifying string) while the response particle develops from its predicative use (via ellipsis)? Does the intensifier stage either necessarily or empirically precede the particle stage? Or is it the case that pragmaticisation, once begun, proceeds in parallel in different areas? We may note, in this context, a similar multiplicity of outcomes for indeed. Alongside its function as a discourse marker (exemplified in (4c) above), it too can also be used as an intensifier and response particle (see OED senses 5–8). 16.

Given the remit of this paper, I focus here on questions of word order and semantics; but it’s possible that the three classes of modifier are to be distinguished syntactically also, with classifiers being in a lexical relation to the noun, identifiers in a relation of dependency and characterisers in a relation of juxtaposition (as appositive predications or parenthetical comments).

17.

Bolinger himself accounts for the difference by positing a distinction between reference modification (where criminal lawyer = ‘practitioner of criminal law’) and referent modification (where criminal lawyer = ‘criminal practitioner of law’). In this example reference modification (which in Lyons’s more precise terminology would be denotation modification) can readily be equated with the classifying function and referent modification with the characterising function. However Bolinger’s binary schema forces him to conflate classifiers with identifiers (both criminal law and mere law are, for him, examples of reference modification, despite his awareness of the semantic difference between the adjective types involved). He is, of course, engaged in a syntactic debate (combating those who wished to generate adjectives by transformation from relative clauses) and his binarism derives from the syntactic opposition between attributive and predicative constructions, which he wishes to correlate with “two types of generation [for adjectives] … one, termed reference-modification, being in the kernel and allowing for a ‘kind of’ slot among the determiners, the other, termed referent-modification, being by way of a predication which is joined by conjunction rather than by subordination” (Bolinger 1967: 1). Teyssier’s tri-partite functional schema, which appeared the following year in the same journal, is, in effect, if not in intention, a corrective to Bolinger’s account. That it was not so received may be imputed to the prevailing linguistic climate of the late 1960s. It deserves to be better known and, as I suggest in footnote 14 above, its implications for syntax are well worth teasing out.

18.

Objections to Lightfoot’s account of quantifiers are considered by Fischer and van der Leek (1981: 311–317); a critique of Spamer will be included in Fischer (2000). What’s of interest to me in Spamer’s account is that it posits a scenario in which a two-slot Bolinger-like system evolves into a three-slot Teyssier-like system. Translated into the terms I’ve been using here, the scenario runs as follows: stage 1: Old English distinguished between reference modification (associated with weak declension/nominal adjuncts/classifying function/rightmost position/recursive potential) and referent modification (strong declension/adjective + proto-determiner squish/identifying + characterising function/leftmost position/non-recursive); stage 2: the loss of inflectional endings → formal confusion between the two classes of modifier → recursiveness in referent modification → reanalysis of leftmost items as specialised identifiers (ultimately determiners). One of the empirical predictions following from this scenario is that sequences of two strong declension adjectives will not occur in Old English. Testing this claim against a corpus of Old English, Fischer finds two small groups of apparent exceptions. In one group the rightmost elements are ‘denominal adjectives’ (i.e. they are functionally, if not formally, classifiers); the second group she thinks “more difficult to account for”:

64

SYLVIA ADAMSON a. b. c. d.

swa beorht scinende steorra ‘such a bright shining star’ swetum ferscum waeterum ‘with sweet fresh water’ god hluttor eala ‘good clear ale’ mid ofermaete unclaene luste ‘with excessive unclean lust’

But it may not be coincidental that the leftmost adjective in three of these cases is arguably evaluative (swetum, god, ofermaete) while in the other case, the adjective beorht is arguably functioning as an intensifier; certainly in all four examples the rightmost adjective is the more straightforwardly descriptive of the pair and the sequence conforms both to Dixon’s ordering constraints and to the more general correlation I have posited between left-right ordering and subjective-objective semantics. None of Fischer’s counterexamples provides a clear case of the combination of two central characterising adjectives i.e. strings of the type crisp white [collar] or fast new [car].

References Biber, D., Finegan, E. and Atkinson, D. 1994a. “ARCHER and its challenges: Compiling and exploring a representative corpus of historical English registers”. In Creating and Using English Language Corpora. Papers from the Fourteenth International Conference on English Research on Computerized Corpora, Zürich 1993, U. Fries, G. Tottie and P. Schneider (eds), 1–13. Amsterdam and Atlanta GA: Rodopi. Biber, D., Finegan, E., Atkinson, D., Beck, A., Burges, D., and Burges, J. 1994b. “The design and analysis of the ARCHER corpus: A progress report”. In Corpora across the Centuries. Proceedings of the First International Colloquium on English Diachronic Corpora, St Catharine’s College, Cambridge, 25–27 March 1993, M. Kytö, M. Rissanen and S. Wright (eds), 3–6. Amsterdam and Atlanta GA: Rodopi. Bolinger, D. 1967. “Adjectives in English: Attribution and predication”. Lingua 18: 1–34. Carlson, A. 1978. “A diachronic treatment of English quantifiers”. Lingua 46: 295–328. Coulomb, C. 1994. “Les séquences d’adjectifs en position prénominale”. Recherches Anglaises et Nord-Americaines, 27: 1–16. Denison, D. and Vincent, N. (eds). 1997. Papers from the English Historical Syntax Meeting, Manchester, 11–12 May 1996. Issued as a Special Number of Transactions of the Philological Society, 95 (1). Dixon, R. M. W. 1982. Where Have All the Adjectives Gone? Berlin: Mouton de Gruyter. Finell, A. 1996. Discourse markers and topic change: A case-study in Historical Pragmatics. Ph. D. dissertation, University of Cambridge. Fischer, O. 2000. “The position of the adjective in Old English”. In Generative Theory and Corpus Studies: A Dialogue from 10ICEHL, R. Bermúdez-Otero, D. Denison, R. Hogg and C. McCully (eds), 153–181. Berlin: Mouton de Gruyter.

CATEGORY SHIFT IN THE NOUN PHRASE

65

Fischer, O. and van der Leek, F. 1981. “Optional vs radical re-analysis: Mechanisms of syntactic change”. (Review of Lightfoot 1979). Lingua 55: 301–350. Fraser, B. 1996. “Pragmatic markers”. Pragmatics 6: 167–190. Huddleston, R. 1984. Introduction to the Grammar of English. Cambridge: Cambridge University Press. Kiparsky, P. 1995. “Indo-European origins of Germanic syntax”. In Clause Structure and Language Change, A. Battye and I. Roberts (eds), 140–169. Oxford: Oxford University Press. Klein-Andreu, F. 1983. “Grammar in style: Spanish adjective placement”. In Discourse Perspectives on Syntax, F. Klein-Andreu (ed.), 143–179. New York: Academic Press. Lightfoot, D. 1979. Principles of Diachronic Syntax. Cambridge: Cambridge University Press. Lyons, J. 1982. “Deixis and subjectivity. Loquor ergo sum?” In Speech, Place, and Action: Studies in Deixis and Related Topics, R. J. Jarvella and W. Klein (eds), 101–124. New York: John Wiley. Lyons, J. 1991. Natural Language and Universal Grammar: Essays in Linguistic Theory. Cambridge: Cambridge University Press. Milner, J.-C. 1978. De La Syntaxe à L’Interprétation. Paris: Editions du Seuil. Peters, H. 1994. “Degree adverbs in Early Modern English”. In Studies in Early Modern English, D. Kastovsky (ed.), 269–288. Berlin: Mouton de Gruyter. Plank, F. 1992. “Possessives and the distinction between determiners and modifiers (with special reference to German)”. Journal of Linguistics 28: 453–468. Quirk, R., Greenbaum, S., Leech, G. and Svartvik, J. 1972. A Grammar of Contemporary English. Harlow: Longman. Radford, A. 1988. Transformational Grammar. Cambridge: Cambridge University Press. Radford, A. 1993. “Head-hunting: On the trail of the nominal Janus”. In Heads in Grammatical Theory, G. G. Corbett, N. M. Fraser and S. McGlashan (eds), 73–113. Cambridge: Cambridge University Press. Ramat, P. and Ricca, D. 1998. “Sentence adverbs in the languages of Europe”. In Adverbial Constructions in the Languages of Europe, J. van der Auwera and D. P. O. Baoill (eds), 187–276. Berlin: Mouton de Gruyter. Smith, C. 1964. “Determiners and relative clauses in a generative grammar of English”. Language 40 (1): 37–52. Spamer, J. B. 1979. “The development of the definite article in English: A case study of syntactic change”. Glossa 13: 241–250. Stein, D. and Wright, S. 1995. Subjectivity and Subjectivisation. Cambridge: Cambridge University Press. Swan, T. 1988. “The development of sentence adverbs in English”. Studia Linguistica 42 (1): 1–17. Teyssier, J. 1968. “Notes on the syntax of the adjective in Modern English”. Lingua 20: 225–249. Traugott, E. C. 1982. “From propositional to textual and expressive meanings: Some semantic-pragmatic aspects of grammaticalization”. In Perspectives on Historical

66

SYLVIA ADAMSON

Linguistics, W. P. Lehmann and Y. Malkiel (eds), 245–271. Amsterdam: John Benjamins. Traugott, E. C. 1990. “From less to more situated in language: The unidirectionality of semantic change”. In Papers from the 5th International Conference on English Historical Linguistics, S. Adamson, V. Law, N. Vincent and S. Wright (eds), 497–517. Amsterdam: John Benjamins. Traugott, E. C. 1995. “The role of discourse markers in a theory of grammaticalization”. Paper presented at ICHL XII, Manchester, 1995, (http://www.stanford.edu/~traugott/ ect-papersonline.html). Vincent, N. 1985. “La posizione dell’aggettivo in italiano”. In Tema-Rema in Italiano, H. Stammerjohann (ed.), 181–195. Tübingen: Gunter Narr Verlag. Whorf, B. L. 1956. Language, Thought, and Reality: Selected Writings of Benjamin Lee Whorf, ed. J. B. Carroll, Cambridge, Mass.: M. I. T. Press.

The grammaticalization of the verb ‘pray’ Minoji Akimoto Aoyama Gakuin University, Dept. of English

1.

Introduction

The aim of this paper is to discuss the grammaticalization of the verb pray from the fifteenth century to the nineteenth century. The verb pray which was originally used as a full verb came to be used as an interjection or courtesy marker,1 becoming obsolete in present-day English.2 This process of change seems to have taken place in parallel with what Halliday or Traugott describes as the change from propositional to expressive.3 As I discuss later, the change involving pray shows more characteristics of courtesy markers (see Quirk et al. 1985: 569–572) than of discourse markers, although the characteristics of both types partially overlap.4 In this paper, I should like to explain some factors contributing to the grammaticalization process of this verb and then to make some new proposals to explain this process. The following questions are posed: 1. 2. 3. 4.

What made the verb pray turn into a courtesy marker? Why did pray as a courtesy marker fall into disuse? How is politeness related to the development of pray? Does the grammaticalization process of pray follow general tendencies in linguistic change, from the propositional, the textual to the expressive meanings, as Traugott (1982: 247–248) suggests?

2.

Previous studies

Ukaji (1978: 54–55, 131–143) discusses the variant forms of I pray you, and related forms, such as I beseech you and I command you in Early Modern

68

MINOJI AKIMOTO

English, particularly Shakespeare’s works, using transformational rules to account for their imperative structure. Based on Ross’s (1970) proposal, Ukaji argues that the various nuances of their imperative meaning came from the main sentence verbs. (I) pray is one of them. His discussion of the co-occurrence between pray and the imperative has a close connection with the development of pray which I discuss later on. Imperatives are considered to have been derived from base structure performative clauses via processes of inversion and deletion as follows (54–55): (1) Subject-Aux Inversion: opt. SD:

I Pres

pray [+imperative]

SC:

to you

have , X] s be

[NP,

Subjunctive

1

2

3

4

1

3+2

Ø

4

(2) Equi NP Delition: opt SD:

X1

NP

X2

[X3

NP

Subjunctive

X4]s

X5

SC:

1

2

3

4

5

6

7

8

1 2 where 2 = 5

3

4

Ø

6

7

8

He further explains the chronological development of I pray you, I pray, pray you and pray, with the help of data taken from the OED, via a process of deletion. As I discuss later, the sentence I pray you also occurs in the middle of a sentence and therefore is not confined only to performative functions. Brown and Gilman (1989) study polite expressions in Shakespeare’s tragedies, referring to I pray you, pray you and prithee as examples of indirect requests. They also point out the co-occurrence of these indirect request expressions with honorific titles such as your Majesty and your Grace. Barber (1997: 34) also touches on the use of courteous requests by means of I prethe and I prey thee in Early Modern English. All the studies given above concentrate on the Early Modern English period. It seems clear that I pray you and related forms were used for politeness, presumably not only in the Elizabethan period, but before and after the period. As I will argue later on, understanding the growth and development of I pray you and related forms is very important in order to grasp the whole picture of the development

THE GRAMMATICALIZATION OF THE VERB ‘PRAY’

69

of politeness expressions between the fifteenth and nineteenth centuries. As regards historical perspectives on politeness expressions in English, in addition to Brown and Gilman (1989), Kopytko (1995) analyses the plays of Shakespeare in terms of Brown and Levinson’s (1987) strategies of positive and negative politeness. Kopytko concludes that the interactional style or ‘ethos’ of British society has developed from the dominating positive politeness culture in the sixteenth century towards the modern negative politeness culture (1995: 532). Nevalainen and Raumolin-Brunberg (1995) present an analysis of forms of address in letter salutations between 1420 and 1680 on a sociopragmatic basis with special emphasis on the relative power and social distance of the speaker and the hearer, and the ranking of imposition.5 These studies make no reference to pray patterns as polite expressions. On the whole it can be said that there is no comprehensive literature touching specifically on the development of pray.6

3.

Analysis

According to the OED, the verb pray came from OF preier and was introduced around the thirteenth century in the sense of ‘to ask earnestly, humbly, or supplicatingly’. Under heading IV.8., the following idiomatic uses are mentioned: a. b. c. d.

I pray you(thee) Pray you, pray thee I pray Contracted to pray

A 1519 citation is given for use (a) suggesting that it is the oldest, and that the other forms are derived from it. Unlike I gesse in Middle English (see Brinton 1996: 211–263), I pray you(thee) is a main clause followed mostly by subordinate clauses, and is not preceded by so and as, as the argument below shows. A few words are in order regarding the data for analysis. I collected related examples from the fifteenth century to the nineteenth century drawing on various genres of writing, i.e. fiction, drama and letters (see the Texts at the end of this paper). I also make use of copious examples from the OED. Drama and to a less extent, letters represent a colloquial style of English, essays represent more or less a formal style of English, and fiction seems to represent a mixed variety of styles. First of all, I shall give the frequency of the variety of the forms given above in each century from the fifteenth to the nineteenth, and then discuss the

70

MINOJI AKIMOTO

grammaticalization processes which the verb pray and its variants underwent, pointing out some causes for their falling into disuse. 3.1 Fifteenth century The examples were collected from three works, The Paston Letters — approximately 136,000 words, Margery Kempe — approximately 67,000 words and Le Morte D’Arthur — approximately 120,000 words. What is common to these works is that only the I pray you(thee) pattern appears with the exception of six examples of I pray in The Paston Letters. In the pattern I pray you(thee), we find different kinds of complements. As Table 1 shows, the most frequent complement is the imperative, and þat-clauses come next. To-infinitive complements are not so frequent. Table 1. Frequency of complements found with I pray you/thee in the fifteenth century

imperative þat-clause to-infinitive

PL

Kempe

Malory

59 56 09

38 01 02

14 03 04

Example (1) gives representative examples with imperative (1a–c) (but see also below), þat-clause (1d–f) and to-infinitive complements (1g–i). (1)

a. b. c. d. e. f. g. h. i.

I prey you lete it be suyd … (PL 32) I pray Šow beth not dysplesyd wyth me. (Kempe 19) I pray you come and lodge with me here at my place … (Malory 94) I prey yow with al my power þat of yowr wysdom and good discrecion ye wille … (PL 5) I pray Šow þat it be Šowr as Šowr owyn … (Kempe 108) I pray you that at this feast I may be your chamberlain. (Malory 275) I prey you hertily to sette al these matieres in continuaunce vn-to yowr comyng in-to Ingeland … (PL 5) I pray Šow to schewyn me þe occasyon of Šower wepyng … (Kempe 246) I pray you to counsel me. (Malory 150)

THE GRAMMATICALIZATION OF THE VERB ‘PRAY’

71

As regards examples (1a–c), there are two possible analyses: 1. 2.

I pray you [S] I pray [that you …]

Since you (objective) and ye (subjective) were confused around the fifteenth century, it would be possible to consider you to be the object of the matrix verb and the subject in the complement clause; in the singular, there was a distinction between thou (subjective) and thee (objective). Considering the fact that I pray you(thee) and its variants were often used with the imperative in later periods, I prefer analysis (1) over (2). In example (1b), beth has the same form for the indicative plural and the imperative plural, but in this example, I interpret it as an imperative. What makes a difference between these analyses is that in the former case, we have two independent clauses, and in the latter case, we have a main clause (I pray) and a dependent clause. In examples (1e) and (f), there are þat-clauses, possibly because the subjects in the main and subordinate clauses are different. In example (e), the subjunctive appears in the þat-clause, but indicatives may occur in þat-clauses as well. That-deletion took place in the fifteenth century as it does today. Thompson and Mulac’s (1991) study (to be discussed in Section 4) suggests that that-deletion promotes the parenthetical nature of the verb, which seems to reflect a diachronic change. There are two cases in Margery Kempe where I pray you(thee) appears not at the beginning of a sentence, but in the middle and at the end of sentences. In contrast to the overwhelming frequency of I pray you(thee) in head position of sentences, this freedom of position suggests that the form had a parenthetical nature to a certain extent. In (2a), a reversal of superordinate-subordinate structure has occurred. I pray you is parenthetical/subordinate, not main. (2)

a. b.

And þerfor prouydith Šow an-oþer place, I pray Šow. (Kempe 139) A, Lord, I prey þe, for alle goodnes …late þis worthy clerk neuyr (Kempe 169) deyin … 7

3.2 Sixteenth century I examined a work by Nashe (The Unfortunate Traveller and Other Works) — approximately 156,000 words, an Anthology of sixteenth century Elizabethan prose — approximately 69,000 words, a selection of sixteenth century letters — approximately 5,000 words and the plays of Marlowe — approximately 46,000 words. As Table 2 shows, the I pray you(thee) pattern with the imperative form was still dominant in the sixteenth century. This tendency is a continuation of the

72

MINOJI AKIMOTO

previous century (see Table 1). But what differentiates the patterns of these two centuries is the occasional use of wh/how-complements rather than that-complements in the sixteenth century. Both complements also occur with I pray without a second person. Table 2. Frequency of complements in the sixteenth century8

I pray you/thee imperative wh/how-question to-inf direct question I pray wh/how-question imperative

Nashe

Anthology

Letters

Marlowe

5 1

11 01

3

1

2 1 4 2 1

pray you 2

I pray you in mid position I pray in mid position

3

Example (3) gives representative examples with imperative (3a–b), direct question (3c), wh-clause (3d–e), and to-infinitive complements (3f). (3)

a. b. c. d. e. f.

I pray ye, good Monsieur Devil, take some order … (Nashe 65) I pray you be good unto her. (Letters [More] 5) I pray thee, may I ask without offence? (Nashe 178) but I pray you, where is my mistress this morning? (16th Anthology 64) I pray how might I call you. (Nashe 50) I prai you to signifi unto her Majeste that … (Letters [Fry] 9)

I pray you(thee) or I pray occasionally appears in mid position, as in; (4)

a. b.

No, keep thy drink, I pray thee, to thyself. (Nashe l79) Make account, I pray you, of my firm friendship. (Letters [Elizabeth I] ll)

73

THE GRAMMATICALIZATION OF THE VERB ‘PRAY’

(5)

a. b.

Now, I pray, what do you imagine him to be? But, I pray, tell me this…

(Nashe 57) (Nashe 135)

3.3 Seventeenth century As texts, I have selected a number of plays by Behn — approximately 90,000 words, an Anthology of seventeenth century fiction — approximately 242,000 words, and a selection of seventeenth century letters — approximately 25,000 words. As Table 3 shows, prithee appeared frequently in Behn. Pray, which did not appear in the previous centuries, occurred frequently in this century. I pray also occurred with the imperative. I pray you began to decrease. Pray thee appeared only once. Table 3. Frequency of complements in the seventeenth century Behn (1)

(2)

(3)

Anthology

Letters

3 1

4

6 2

1

prithee imperative wh-question

21 13

imperative wh-question question

12 09 02

pray

I pray imperative wh-question

(4)

I pray you

4

(5)

pray thee

1

(6) gives representative examples with imperative (6a, c–d, g–h, j–k) and wh-clause complements (6b, e–f, i). (6)

a. b. c.

Prithee be not so wild. (Behn 6) Prithee, why cam’st thou ashore? (Behn 11) Pray tell me, sir, are not you guilty of the same mercenary crime? (Behn 29)

74

MINOJI AKIMOTO

d. e. f. g. h. i. j. k.

Pray therefore rise, and go into the solitary wood. (17th Anthology 174) Pray, sir, which is Sir Feeble Fainwould’s? (Behn 199) Pray how did he break it? (17th Anthology 463) I pray tell me whether you have lately seen the princ Pamphilia … (17th Anthology 82) I pray present my most humble Service to my good Lady … (Letters [Howell] 19) But I pray, why did you but even now with sighs and tears … (17th Anthology 114) I pray thee advise him as well for the love I bear thee. (17th Anthology 200) … wherefore pray thee be advised … (17th Anthology 119)

Prithee, which is a colloquial form for pray thee, is short-lived and is often found in drama. Although the OED cites examples of prithee from the sixteenth century to the nineteenth century, in my data the form appears mostly in the seventeenth century with fewer occurrences in the eighteenth century. This type of phonological reduction and loss of morphological boundaries is characteristic of grammaticalization and shows the development that pray is undergoing. 3.4 Eighteenth century I have used Richardson’s Pamela — approximately 100,000 words, Farquhar’s plays — approximately 89,000 words and a selection of eighteenth century letters — approximately 58,000 words as texts. Although pray was used in the seventeenth century, it became more frequent in the eighteenth century. In the seventeenth century, 89 tokens of pray and its variants occur (see Table 3), this amounts to one incidence in the corpus approximately every 4,000 words. In the eighteenth century, 146 tokens of pray and its variants occur (see Table 5), which means one incidence approximately every 1,700 words. 31 tokens of pray in the seventeenth century occupy 34% of such variants as prithee, I pray, I pray you and pray thee, and 131 tokens of pray in the eighteenth century occupy 89% of the variants. The position of pray in a sentence was versatile and it appeared in a variety of sentence types. Table 4 shows the positional frequency of pray in Farquhar, where pray was frequently used. Positionally speaking, pray appears mostly at the beginning of sentences. This tendency continues into the nineteenth century. Mid-position was most infrequent. Pray is put at the end of sentences particularly in wh-question sentences. The

75

THE GRAMMATICALIZATION OF THE VERB ‘PRAY’ Table 4. Position of pray in Farquhar’s plays

imperative wh-question question

front

mid

end

33 36 07

2 2 0

01 10 01

frequency of the complements in the works mentioned is shown in Table 5. Table 5. Frequency of complements in the eighteenth century

(1)

Farquhar

Letters

18 03

36 48 08 09

06 02 01 01

1

14

pray imperative wh-question question other

(2)

Richardson

prithee

Example (7) gives representative examples with imperative (7a–c) and wh-clause (7d–h) and question complements (7i). (7)

a. b. c. d. e. f. g. h. i.

Pray, your honour, forgive me! (Richardson 44) But pray, Sir Harry, tell us some news of your travels. (Farquhar 12) So pray remember this … (Letters [Lennox] 115) But pray, what pretty neat damsel was that with you? (Richardson 89) Pray, who is this miraculous Helen? (Farquhar 168) Pray why so? (Letters [Burke] 168) What was it, pray? (Farquhar l0) But whom, pray, sir, have you thought of? (Richardson 119) Well, Sir Harry, and d’ ye like my daughter, pray? (Farquhar 63)

Other structures which do not belong to any of the complement types given above are those where there is no verb, as in j.

Her name, pray, Sir Harry.

(Farquhar 12)

76

MINOJI AKIMOTO

Notice that pray began to be often used with honorific titles such as your honour and Sir Henry. 3.5 Nineteenth century The texts examined were Jane Austen’s Pride and Prejudice — approximately 110,000 words, Oscar Wilde’s plays — approximately l00,000 words and a selection of nineteenth century letters — approximately 58,000 words. The nineteenth century usage of pray is slightly different from that of the eighteenth century in the following respects. Pray ceased to be used in question complements. Prithee, which occurred particularly in Farquhar, disappeared. On the other hand, I pray thee, which did not occur in the previous century, was used in imperatives in initial position by Wilde. The ‘revival’ of I pray thee is an example of Wilde’s conscious use of archaism. The size of the corpus in the respective periods is about the same (247,000 and 268,000 words), and the frequency of pray in the eighteenth century is 131 times, compared to 33 times in the nineteenth century. This shows a notable decline in the use of pray in the nineteenth century. Table 6 shows the frequency of pray and I pray thee. Table 6. Frequency of complements in the nineteenth century

(1)

Wilde

Letters

8 4 1

15

05

pray imperative wh-question other

(2)

Austen

I pray thee

03

Example (8) gives representative examples with imperative (8a–d) and wh-clause complements (8e). (8)

a. b. c. d. e.

Pray tell us your sister that I long to see her. (Austen 41) Pray don’t trouble, Lord Windermere. (Wilde 56) Pray tell me something about Lord and Lady Holland … (Letters [Smith] 230) But I pray thee, Salome, ask of me something else. (Wilde 86) Pray, what is your age? (Austen 148)

THE GRAMMATICALIZATION OF THE VERB ‘PRAY’

77

Based on the description of the frequency of use of these forms given above, the following conclusions can be drawn. a. b.

c.

d. e.

I pray you(thee) was dominantly used with the imperative in the fifteenth and sixteenth centuries, but was less frequent in the seventeenth century. I pray began to appear in the sixteenth and was used until the seventeenth century. It was used with wh-questions, although it was also used with the imperative. Pray began to be frequent from the seventeenth century onwards. It was used both with the imperative and wh-questions, and sometimes with questions. It often appeared in mid-position in sentences. Prithee came to be frequent from the seventeenth century onwards, but seems to have fallen into disuse by the nineteenth century. Pray you(thee) hardly appeared in my data (only once).

In summary, we can depict the general change of the forms discussed above as follows. Notice that the dotted lines denote possible use. Table 7. Change in the forms of pray

15th

16th

17th

18th

19th

I pray you(thee) I pray pray prithee

4.

Discussion

The development of (I) pray you(thee) would appear to be similar in certain respects to that of ‘epistemic’ parentheticals, I guess/think, which have received some attention in the literature. Thompson and Mulac (199l) discuss the relationship between that-deletion and the grammaticalization of ‘epistemic phrases’, as in, (11)

a. b. c.

I think that we’re definitely moving towards being more technological. I think Ø exercise is really beneficial to anybody. It’s just your point of view you know what you like to do in your spare time I think.

78

MINOJI AKIMOTO

Although their study is synchronic, they claim that the grammaticalization process goes from (a) to (c). This process also reflects the diachronic process of grammaticalization of I pray and related forms. That-deletion causes the subordinate clause to cease to behave like one and the consequent free position of the first person + verb strengthens its parenthetical nature. They further suggest that the parenthetical begins to serve the functions of epistemic modals and adverbs. From a diachronic perspective, Brinton (1996: 211–263) discusses suppositional parentheticals, such as I gesse in Middle English, in particular in the works of Chaucer. She refers particularly to their interpersonal functions which serve the purposes of intimacy and politeness. Following Thompson and Mulac’s (1991) method of categorization, Brinton (1996: 248) lists epistemic parentheticals of ‘know’-verbs, such as guess, suppose, trow and woot, which were the most common parentheticals in Middle English. She also calls attention to the development of epistemic parentheticals which are characterized at first by as and so (1996: 249). As and so, which function either as relative pronouns or as subordinators, help those verbs to develop their parenthetical nature. The grammaticalization of pray is different from the development described by Brinton in that this verb does not belong to the ‘know’-class, nor does it begin with as and so in its process of grammaticalization. However, in considering the development of pray, we can refer to two questions posed by Brinton (1996: 50). One concerns the choice of particular items to be grammaticalized. The other concerns the extent to which a word’s original lexical meaning is preserved in the grammaticalized word. We can also consider the evolution of pray in relation to Hopper’s (1991) five principles of grammaticization: (1) layering, (2) divergence, (3) specialization, (4) persistence, and (5) de-categorialization. (3) and (4) in particular relate to the questions posed by Brinton. Applying these principles to the grammaticalization processes of pray, we find the following. Forms such as I pray you(thee)/I pray/pray co-exist (layering). Even after an interjectional pray was in use, the full verb pray continued to be used (divergence). Out of such competing forms as I beseech you and I entreat you, only I pray you turned into a courtesy marker (specialization).9 The courtesy marker pray still retains part of the original meaning of the verb in the sense of supplication (persistence), and it lost its syntactic verbal characteristics being used interjectionally (de-categorialization). Prithee also underwent de-categorialization from the form I pray thee to an interjectional marker. Hopper’s principles, which are useful for the description of the kind of change pray has undergone, are not adequate in explaining the stage-by-stage change of pray from verb form to interjection, becoming obsolete at the final

THE GRAMMATICALIZATION OF THE VERB ‘PRAY’

79

stage in present-day English. Traugott’s theory of semantic change seems to provide the needed explanation. Traugott (1982) argues for unidirectional change as follows: propositional > textual > expressive, taking as an example þa hwile þe (OE) ‘at the time that’ > while (ME) ‘during’ > while (PDE) ‘although’. Traugott (1995: 31) further suggests a unidirectional increase in subjectification, which she defines as “a pragmatic-semantic process whereby ‘meanings become increasingly based in the speaker’s subjective belief/attitude toward the proposition’ ”. The sequential change of pray seems to reflect the tendency towards the expression of an emotive/subjective state. However, during this change, pray does not seem to go through the stage of textual function. Traugott also says little about the disappearance of an item. What happens to the item which acquires the subjective belief state/attitude? Does it continue to exist, or fall into disuse for one reason or another? Theoretically speaking, linguistic change continues for ever, and no stage can be considered the final one. An aspect that studies of grammaticalization have not discussed much is the mitigating/polite function an item may develop in the process of grammaticalization. The development of emotive/subjective meanings involves this function as well. The courtesy marker pray developed this mitigating/polite function most conspicuously. Of the characteristics and functions of pragmatic markers that Brinton (1996: 32–35) mentions, some share those of pray. Of particular interest is the ‘controversial’ suggestion she makes (1996: 35) that pragmatic markers are more characteristic of women’s speech than men’s speech. Is this point applicable to the courtesy marker pray, too? For this purpose, I examined the distribution of occurrences of pray in Farquhar (approximately 89,000 words), because pray appears most frequently here compared with the other works of the eighteenth century, and the dialogues used between the characters are varied. The results are as follows: men’s speech women’s speech

70 31

This suggests that men use pray more often than women, although we must take into account ‘who it is addressed to’.10 The verb pray and its derivative forms were used as devices for expressing politeness. From the nineteenth century onwards, however, interjectional pray seems to have been replaced by please. The appearance of please in this sense seems to have occurred in the nineteenth century in my data.

80

MINOJI AKIMOTO

(12)

Please come and look at the biscuits …

(Letters [Stevenson] 417)

The citations in the OED (‘please’ II.6.c.) endorse my statement. Please was followed by to in eighteenth century English, as in; (13)

Will you please to have your office taken from you …? (Farquhar 234)

Chronologically, the grammaticalization of pray from the verb to the courtesy marker takes place as follows:

I pray you(thee) I pray thee

I pray prithee

pray

Ø Ø

The deletion of you, first of all, is possible because the second person is always the target of address and becomes unnecessary in the imperative; it is clear in the context. Similarly, the first person I becomes redundant because the agent of asking is always presupposed.11 At the same time, semantically, the meaning of pray was diluted, so called ‘bleaching’.12 Syntactically, the pray variants were continuously used with complements which were narrowed down over time. The pray part which was first superordinate gradually became an interjection probably because the following sentence took on more importance in terms of information structure, as a result of the relative decrease in the content of the pray part. Through the repetitive or expected uses of pray with the imperative, it began to become a meaningless interjection. Finally, pray was replaced by please, a new form which has become frequent since the nineteenth century. It should be noted that in the process of grammaticalization, I pray you(thee) had at first more of a propositional content, but with the strengthening of its parenthetical nature involving the change of form, expressive/interjectional uses became more evident and continued until it was replaced and finally disappeared. Reasons for the replacement are various. Firstly, a new form is dynamic, new to the ear, and more expressive. Secondly, pray, because of its religious connotations, may have been narrowed down in its context of use, and finally a long vowel in please may have been more effective in the sense of ‘earnest appeal’. It is, however, interesting that both pray and please came from Old French.

THE GRAMMATICALIZATION OF THE VERB ‘PRAY’

5.

81

Conclusion

I have traced the grammaticalization of the verb pray from a full verb through its use as a courtesy marker during the periods from the fifteenthth to nineteenth centuries to its disappearance. The transformation pray underwent is unique in that it was first used as a verb and then as a courtesy marker, and finally became obsolete in present-day English. Hopper’s and Traugott’s general principles of change give explanations for the grammaticalization processes of pray under general principles of change, but they are not the whole story. Socio-cultural factors, such as politeness, must be taken into consideration for a more convincing explanation of the change involving pray.

Acknowledgments I am grateful to Laurel Brinton, Peter Robinson and the Editors for making detailed comments on earlier versions of the paper.

Notes 1.

Following Quirk et al. (1985: 569–572), I call this interjectional function of pray a courtesy marker, as will be argued later.

2.

The full verb pray in the sense of ‘to ask earnestly, humbly, supplicatingly’ has been in use up to the present. Examples: (1)

… than he ansewereth and prayeth me no more to speke of that matter … (P.L. 161)

(2)

All good people pray heartily to God for these poor sinners … (Letters [Richardson] 106)

The pray to be discussed in what follows [pray2] is related to full verb [pray1], but has split off from it to be used as a courtesy marker. The following diagram may be helpful: pray1 up to the present

pray pray2 3.

courtesy marker

Ø

I shall discuss Traugott’s theory of grammaticalization in due course. Halliday (1978: 117) refers to three functions of language — the observer function (ideational), the intruder function (interpersonal) and the rhetorical function (textual). Although Halliday’s concern is synchronic, the order of functions seems to reflect the historical development of language, a view which

82

MINOJI AKIMOTO Traugott partially shares. She sees the order of change as ideational, textual and expressive (1982).

4.

Schiffrin (1987: 31) operationally defines discourse markers as “sequentially dependent elements which bracket units of talk”, including oh, well, and, but, or, so, because, now, then, I mean, y’know. Crystal (1997: 118–119) defines discourse markers as “sequentially dependent elements which demarcate units of speech, such as oh, well and I mean.” Quirk et al. (1985: 569–572) call please a courtesy subjunct, saying: “When courtesy subjuncts appear in questions, the questions constitute a request”.

5.

There are extensive synchronic studies on politeness, in addition to Brown and Levinson (1987), particularly in the area of pragmatics, such as Leech (1983: 131–151), Mey (1993: 67–74) and Thomas (1995: 149–182). My argument is, however, confined to the expression of I pray you/thee and related forms, and is not directly concerned with these theories. For discussion of the relation between politeness and you and thou, see Wales (1996: 73–76).

6.

Quirk et al. (1985: passim) refer to politeness which includes the preterite forms of modals (could and might ), courtesy subjuncts (kindly and please), and indirect condition (if I may say so).

7.

‘Lord’ functions as a vocative and I prey þe in this sentence may not be an example of middle position with a parenthetical nature. I am grateful to a reviewer for pointing this out to me.

8.

By this time complements had become main clauses since I pray and pray were often used parenthetically. But for convenience sake, I use ‘complements’ throughout.

9.

I thank Laurel Brinton (p.c.) for clarifying this point.

10.

One factor affecting the use of a courtesy marker pray is power/solidarity relationships between speakers. An inferior often uses pray to his superior. For instance, VIZARD SIR HARRY

But pray, Sir Harry, tell us some news of your travels. With all my heart. (Farquhar 12)

Regarding other uses of politeness, such as promoting or maintaining harmonious relations, see Leech (1983: 131–142). For comparison, I examined the uses of I pray you(thee) in The Paston Letters. The result is different from that of Farquhar’s plays. In approximately 7,000 words, the frequencies of I pray you(thee) are as follows: William I Agnes John William II Margaret

10 08 08 08 28

It may be said that I pray you(thee) was a favourite phrase of women. 11.

I say → say seems to undergo a similar process. The OED (‘say’ 12.) assumes the process just given.

12.

Regarding further discussion on ‘bleaching’, see Sweetser (1988) and Hopper and Traugott (1993: 87–93).

THE GRAMMATICALIZATION OF THE VERB ‘PRAY’

83

Texts An Anthology of Elizabethan Prose Fiction. 1573–1588, P. Salzmann (ed.). Oxford: Oxford University Press, 1987. An Anthology of Seventeenth-Century Fiction. 1621–1698, P. Salzmann (ed.). Oxford: Oxford University Press, 1991. Austen, J. 1813. Pride and Prejudice, J. Kinsley (ed.). Oxford: Oxford University Press, 1990. Behn, A. 1677–1688. The Rover and Other Plays, J. Spencer (ed.). Oxford: Oxford University Press, 1995. The Book of Margery Kempe. 1436?, M. S. Brown (ed.) London/New York: Oxford University Press, 1961. Farquhar, G. 1706. The Recruiting Officer and Other Plays, W. Myers (ed.). Oxford: Oxford University Press, 1995. Marlowe, Chr. 1587. The Complete Plays, J. B. Steane (ed.). Harmondsworth: Penguin, 1977. Le Morte D’Arthur. 1485, Ed. Modern Library. New York: Random House Inc., 1994. Nashe, T. 1594. The Unfortunate Traveller and Other Works, J. V. Stean (ed.). Harmondsworth: Penguin, 1972. The Oxford Book of Letters. 1535–1985, F. Kermode and A. Kermode (eds). Oxford: Oxford University Press, 1995. Paston Letters and Papers of the Fifteenth Century. 1429–1489. Vol. 1, N. Davis (ed.). Oxford: Oxford University Press, 1971. Richardson, S. 1740–1741. Pamela, P. Sabor (ed.). Harmondsworth: Penguin, 1985. Wilde, O. 1893–1895. The Importance of Being Earnest and Other Plays. Oxford: Oxford University Press, 1995.

References Barber, C. 1997. Early Modern English. Edinburgh: Edinburgh University Press. Brinton, L. J. 1996. Pragmatic Markers in English. Berlin/New York: Mouton de Gruyter. Brown, P. and Levinson, S. C. 1987. Politeness: Some Universals in Language Usage. Cambridge: Cambridge University Press. Brown, R. and Gilman, A. 1989. “Polite theory and Shakespeare’s four major tragedies”. Language in Society 18: 159–212. Crystal, D. 1997. A Dictionary of Linguistics and Phonetics, 4th edition. Oxford: Blackwell Publishers. Halliday, M. A. K. 1978. Language as Social Semiotic. London: Arnold. Hopper, P. 1991. “On some principles of grammaticization”. In Approaches to Grammaticalization, E. C. Traugott and B. Heine (eds), 17–35. Amsterdam: John Benjamins. Kopytko, R. 1995. “Linguistic politeness strategies in Shakespeare’s plays”. In Historical Pragmatics, A. H. Jucker (ed.), 515–540. Amsterdam: John Benjamins.

84

MINOJI AKIMOTO

Leech, G. 1983. Principles of Pragmatics. London: Longman. Mey, J. L. 1993. Pragmatics: An Introduction. Oxford: Blackwell. Nevalainen, T. and Raumolin-Brunberg, H. 1995. “Constraints on politeness: The pragmatics of address formulae in early English correspondence”. In Historical Pragmatics, A. H. Jucker (ed.), 541–601. Amsterdam: John Benjamins. The Oxford English Dictionary (OED). 1989. (2nd edition) Prepared by J. A. Simpson and E. S. C. Weiner. Oxford: Clarendon Press. Quirk, R., Greenbaum, S., Leech, G. and Svartvik, J. 1985. A Comprehensive Grammar of the English Language. London: Longman. Ross, J. R. 1970. “On declarative sentences”. In Readings in English Transformational Grammar, J. A. Roderick and P. Rosenbaum (eds), 222–272. Waltham, Mass.: Ginn and Company. Schiffrin, D. 1987. Discourse Markers. Cambridge: Cambridge University Press. Sweetser, E. E. 1987. “Grammaticalization and semantic bleaching”. In General Session and Procession on Grammaticalization, S. Axmaker, A. Jaiser, and H. Singmaster (eds), 389–405. Calfornia: Berkeley. Thomas, J. 1995. Meaning in Interaction: An Introduction to Pragmatics. London: Longman. Thompson, S. A. and Mulac, A. 1991. “A quantative perspective on the grammaticalization of epistemic parentheticals in English”. In Approaches to Grammaticalization, E. C. Traugott and B. Heine (eds), 313–329. Amsterdam: John Benjamins. Traugott, E. C. 1982. “From propositional to textual and expressive meanings: Some semantic-pragmatic aspects of grammaticalization”. In Perspectives on Historical Linguistics, W. P. Lehmann and Y. Malkiel (eds), 245–271 Amsterdam: John Benjamins. Traugott, E. C. 1989. “On the rise of epistemic meanings in English: An example of subjectification in semantic change”. Language 65: 31–55. Traugott, E. C. 1995. “Subjectification in grammaticalization”. In Subjectivity and subjectivisation, D. Stein and S. Wright (eds), 31–54. Cambridge: Cambridge University Press. Ukaji, M. 1978. Imperative Sentences in Early Modern English. Tokyo: Kaitakusha. Wales, K. 1995. Personal Pronouns in Present-Day English. London: Longman.

The grammaticalization of concessive markers in Early Modern English Guohua Chen Beijing Foreign Studies University

1.

Introduction

Since “all languages have devices for linking clauses together into what are called complex sentences” (Hopper and Traugott 1993: 167) and these devices form an integral part of the grammar of any language, it is customary to discuss the development of these devices in terms of grammaticalization. Concessive markers, which link the concessive clause with the matrix clause, are one type of such devices. In recent years concessives and what are commonly called concessive conditionals have drawn considerable attention in studies of language change and grammaticalization. Cross-linguistic studies like those made by Haiman (1986), König (1986, 1988, 1991, 1992), Harris (1988), Traugott and König (1991), and Leuschner (1998) have led to some very interesting findings and hypotheses. Apart from König (1985), however, there appears to have been no detailed diachronic study of Modern English concessive markers and, with the exception of Leuschner (1998), which focuses mainly on ‘universal concessive conditionals’ and ‘alternative concessive conditionals’,1 none of the studies is corpus-based. As a result, we still know very little about the actual process in which present-day English concessive markers have evolved. Until detailed diachronic investigation of the changes in individual languages are made and the actual processes of these changes revealed, hypotheses about language change based on cross-linguistic synchronic analyses will always remain hypotheses. The present investigation, which is based mainly on the Early Modern English section of the Helsinki Corpus of English Texts,2 will look into the sources and actual process in which English concessive markers are grammaticalized and explore some of the issues related to the theory of language change and grammaticalization.

86

GUOHUA CHEN

Among scholars working in the field of grammaticalization it is widely accepted that there is “a cline of grammaticality” (Hopper and Traugott 1993: 7) of the following type:   >   >  >   As concessive markers in English mostly originate in content items and are mostly in the form of such grammatical words as conjunctions (e.g. although), prepositions (e.g. despite), prepositional phrases (e.g. in spite of ), and correlative conjuncts (e.g. yet), they are thought to be “in the early stages of grammaticalization” (Traugott and König 1991: 189). However, it is not immediately clear why grammatical words should be inherently less grammatical than clitics or inflectional affixes, nor is it proven that the former tend to develop into the latter. In theory, fully developed grammatical words should be as effective as clitics and inflectional affixes in denoting grammatical relations. In view of the fact that the development of content words into grammatical words is usually a gradual process, it can be assumed that the following cline of grammaticality exists:   > -  >    From this perspective a different set of criteria is adopted in this paper in measuring the grammaticality of concessive markers. Functionally, when a concessive marker can denote the sense of concession without recourse to any contextual cue and when its application is fully generalized, i.e. when it can be used in any situation, it is considered to be fully grammaticalized. Formally, when a phrase develops into a fully grammaticalized concessive marker, it often becomes lexicalized, as in the case of although and albeit.

2.

Definition and properties of concessive sentences

Of all the adverbial clauses, concessives are probably the most difficult to define, both formally and semantically. Formally, although there are a number of easily recognized concessive subordinators like though and although and correlative conjuncts like yet and still, concessive relationships are not always marked by these markers. Subordinators such as if, when, whereas and while, which mark other adverbial relationships, are often used with a concessive implication (see Quirk et al. 1985: 1097–1102). In addition, “asyndetic types of clause linking, absolute constructions, as well as nearly all adverbial clauses may

THE GRAMMATICALIZATION OF CONCESSIVE MARKERS

87

be interpretively enriched and receive a concessive interpretation” (König 1988: 150).3 As a result, the formal features of concessives are not easy to specify categorically. Semantically, the notion of concession is very similar to the notion of contrast and inherently related to the notion of adversity. As Quirk et al. (1985: 1102) point out, “there is often a mixture of contrast and concession.” Compare, for example, the two sentences below: (1)

a. b.

Although John loves Mary, Mary doesn’t love John. Whereas John loves Mary, Mary doesn’t love John.

(1a) is clearly marked as a concessive sentence, but it has an inherent sense of contrast; (1b) may be classified as a contrastive sentence, yet it also allows a concessive interpretation. The close association of the notion of concession with the notion of adversity is indicated by the fact that the concessive marker though is defined as basically “an adversative particle” (OED s.v. though) and that the concessive correlative conjunct yet is pronounced to be “more emphatically adversative than but” (OED s.v. yet, III. 9. a.). In fact it can be argued that a concessive relationship always implies an adversative relationship, though the sense of adversity is sometimes marked by a correlative conjunct, as in (2a), sometimes unmarked, as in (2b):4 (2)

a. b.

Though he is poor, yet he is satisfied with his situation. Though he is poor, he is satisfied with his situation.

When the subordinator in (2a) is omitted, as in (3a), the adversative semantic relationship between the first clause and the second clause still remains the same, but the sentence now looks more like (3b), which is a typical adversative coordinate sentence: (3)

a. b.

He is poor, yet he is satisfied with his situation. He is poor, but he is satisfied with his situation.

Unable to find a clear distinction between concessive and adversative relationships, many investigators speak indiscriminately of the two relationships (see König 1988: 149). In this paper a concessive sentence is defined as a complex structure consisting prototypically of a subordinate clause (the antecedent) and a matrix clause (the consequent), with the subordinate clause conceding or presupposing the existence of an actual or hypothetical adverse situation and with the matrix clause denoting a situation which, contrary to expectation, is not affected by the adverse situation of the subordinate clause. Any linguistic device that serves to mark this relationship is a concessive marker.

88

GUOHUA CHEN

By the above working definition, of the six sentences in (1)–(3), only (1a) and (2a–b) qualify as concessives. Although (1b) allows a concessive as well as contrastive interpretation, it is treated as a contrastive sentence because in an antithetical structure like that of (1b) and with whereas as the subordinator, the sense of contrast is more salient than the sense of concession. The two sentences in (3) are adversatives rather than concessives because a), as Quirk et al. (1985: 644) observe, although their meaning is similar to that of (2), there is a major difference between them: in (2) the man’s poverty is presupposed as a given assumption, whereas in (3) it is stated as a fact; b) the relationship between the two clauses is one of coordination rather than subordination. By the same definition, one can distinguish two kinds of concessives: factual and hypothetical. In this paper I use the term ‘factual concessive’ to refer to what are generally known as simply ‘concessives’ and the term ‘hypothetical concessive’ to refer to what are commonly known as ‘concessive conditionals’ or, less commonly, ‘conditional concessives’. The two new terms not only describe the semantic properties of the sentences concerned more accurately but also help explain the way in which the grammaticalization of concessive markers took place. To those linguists who use the term ‘concessive conditional’, hypothetical concessives are basically conditionals (see König 1986: 231; Leuschner 1998: 161). If we compare the semantic properties of hypothetical concessives with those of conditionals and factual concessives, however, we will find that hypothetical concessives have more in common with factual concessives than with conditionals. Take (4) for example: (4)

a. b. c.

If you dislike ancient monuments, Warwick castle isn’t worth a visit. Even though you dislike ancient monuments, Warwick castle is worth a visit. Even if you dislike ancient monuments, Warwick castle is worth a visit.5

Semantically, a typical conditional sentence like (4a) has two basic properties: (i) the propositions expressed by the relevant clauses are both hypothetical; (ii) the truth of the consequent depends on that of the antecedent. By contrast, a typical factual concessive sentence like (4b) has two opposite properties and a third property: (i) the propositions expressed by the two clauses are both factual; (ii) the truth of the consequent is independent of that of the antecedent; (iii) the situation expressed by the antecedent is assumed to be adversative to that expressed by the consequent. Compared with the other two types of sentences,

THE GRAMMATICALIZATION OF CONCESSIVE MARKERS

89

a typical hypothetical concessive sentence like (4c) shares only one semantic property with a typical conditional sentence — the hypotheticality of the antecedent — whereas it shares all the semantic properties of a typical factual concessive sentence except the factuality of the antecedent.6 Since in present-day English hypothetical concessive clauses are commonly introduced by even if and sometimes by the simple if, it is justifiable on formal grounds to classify them with conditionals. In Early Modern English, however, even was not yet used in collocation with if to mark concessive condition.7 In fact, there was not any concessive conditional marker comparable to even if. The formal distinction between conditionals and concessives was often blurred. Not only was it common for a clause introduced by a conditional subordinator to carry a concessive force without the help of any concessive marker, the concession introduced by the concessive subordinator though or although was often hypothetical, as in the next two sentences: (5)

But, as for flame, our Countreyman Gilbert delivers as his Experiment, That an Electric though duly excited and applied, will not move the flame of the slenderest Candle. (EModE3 Robert Boyle, Electricity & Magnetism 16)8

(6)

Yes, for although he had as many liues, As a thousande widowes, [ …] He shall neuer scape death on my swordes point. (EModE1 Nicholas Udall, Roister Doister L. 1145)

So, even on formal grounds sentences like (5) and (6) deserve to be called hypothetical concessives rather than concessive conditionals.

3.

General trend of grammaticalization of concessive markers

The distinction between hypothetical and factual concessives presupposes perceiving concession as a superordinate notion which is more basic than either the notion of factual or hypothetical concession. As it is a more basic notion, the grammaticalization of its expression ought to have preceded the grammaticalization of the distinction between factual and hypothetical concessives. In other words, the earliest concessive markers of a language were likely those that did not distinguish between factual and hypothetical concessions. This hypothesis is supported by the development of English concessive markers. In Old English the most common concessive marker was þeah with its variants, which were used for the expression of both factual and hypothetical

90

GUOHUA CHEN

concessions. As the verb of the concessive clause was usually in the subjunctive whether the concession was factual or not, the verb form was of little use in distinguishing factual from hypothetical concession. Mitchell (1985: 706–707) observes: The context sometimes enables us to distinguish concessions of fact [ …] from concessions involving possibility or hypothesis […]. But it too frequently fails […]. So little would be gained by attempting to classify OE concessive clauses according to the nature of the concession.

It was not until Middle English that a formal distinction in verb forms began to develop between factual and hypothetical concessives introduced by (al)though (see Fischer 1992: 351–352). The functional specialization of (al)though from a general concessive marker to a factual concessive marker probably also began in this period. A similar development can be observed in many other languages. On the basis of a survey of the concessive markers of 70 languages, König (1988: 152) concludes: All languages seem to have a construction, or at least had at some stage in their historical development, that can be used both as a marker of concessive conditional and of concessive sentences proper.9 Free-choice expressions, in particular, are used as concessive connectives in a wide variety of languages and the differentiation between concessive conditional and concessive interpretation is often left to the context, the mood (subjunctive vs. indicative) or some other marking of the verb. […] In those languages where a clear distinction can be made between concessive conditionals and concessives (e.g. even if vs. even though), this distinction was established fairly late.

However, it seems as if König is not very interested in the development described above, for he neither describes and explains it in sufficient detail nor offers any English example of it.10 What he seems more interested in is, among other things, the development of conditional markers and hypothetical concessive markers into factual concessive markers. He observes (1988: 153–154): In many languages, concessive connectives are composed of a conditional (e.g. G. wenn), originally conditional (e.g. E. though, G. ob) or temporal connective (e.g. Fr. quand; Finn. kun) and/or an additive or emphatic focus particle like E. also, even, too. […] This type of connective, probably the most frequent in the world’s languages, provides further evidence for the assumption that (concessive) conditionals are an important source for the development of concessive constructions.

If, as König believes, factual concessive markers may derive from hypothetical concessive markers, then the grammaticalization of hypothetical concessive markers must have preceded that of factual concessive markers. This is impossible if we accept the earlier hypothesis about the development of general concessive

THE GRAMMATICALIZATION OF CONCESSIVE MARKERS

91

markers into hypothetical and factual concessive markers. According to that hypothesis, before the emergence of any formal distinction between hypothetical and factual concessives, there cannot have been any hypothetical concessive marker. The grammaticalization of factual concessive markers must have begun simultaneously with that of hypothetical concessive markers, one can not proceed without the other. As we examine the case history of the concessive markers in English in the next section, we shall find more evidence in support of the first hypothesis.

4.

The grammaticalization of concessive markers

Since the hypotheticality of hypothetical concessives in Early Modern English could not be effectively marked by the verb form unless the concession was counterfactual, the grammaticalization of hypothetical concessive markers proceeded in the direction of concessive reinforcement of conditional markers, which eventually led to the creation of even if in Late Modern English. At the same time (al)though, which was losing ground in the expression of hypothetical concessions, continued its process of functional specialization as a factual concessive marker in the face of competition from three late-coming concessive subordinators — albeit, for all and notwithstanding. 4.1 Sources of concessive markers Concessive markers may derive from a wide variety of sources (cf. König 1985: 9–11, 1988: 152–156; Harris 1988: 75–87), all related in one way or another to the basic semantic, pragmatic and syntactic properties of concessives. In this paper the sources that contributed to the development of concessive markers in Early Modern English are classified into the following five categories: i.

Expressions that directly concede the existence of an adverse situation, such as admit and grant. ii. Expressions that emphasise the adversity of the conceded situation, such as all and never so. iii. Expressions that assert the ineffectiveness of the conceded situation, such as in (de)spite of and notwithstanding. iv. Expressions that state the concomitance of the two situations expressed by the antecedent and the consequent, such as though. v. Expressions that help affirm the factuality of the situation expressed by the consequent, such as nevertheless, still and yet.

92

GUOHUA CHEN

4.2 Expressions that directly concede the existence of an adverse situation From the point of view of speech act theory, a conditional clause expresses the speech act of supposing (see Van de Auwera 1986: 204). Likewise, a concessive clause expresses the speech act of conceding. Just as supposing can be expressed directly by a suppositional verb or phrasal verb (see Chen 1996: 40–46), conceding can be expressed directly by a concessive verb. Like supposing, conceding can also be expressed directly by an expression containing a verb form denoting optative modality. 4.2.1 Concessive verbs In Early Modern English, admit and grant can serve as a kind of concessive subordinator. The OED (s.v. admit, 2d) notes the use of admit in the sense “allow, concede, grant (either from conviction, or for the sake of argument)” to introduce an object clause. One of the examples it cites clearly functions as a concessive marker: (7)

I know he cannot of himself bring into the field aboue fifty thousand fighting men: But admit he were able to bring an hundred thousand; are not you (if you so please) able to levy a far greater power? (1603 Knolles Hist. Turks (1638) 197)

In the Helsinki Corpus there are altogether five instances of admit (including one in the form of present participle) with an object clause. In three instances, the object clause is concessive: (8)

But admitte it be as you saye, what dothe this proue against me? (EModE1 State Trials, The Trial of Sir Nicholas Throckmorton I, 74.C2)

(9)

“Admitt there were, Sir,” quoth he, “an acte of parliament that all the Realme should take me for kinge. Wold not you, master Moore, take me for kinge?” (EModE1 William Roper, The Lyfe of Sir Thomas Moore 85)

(10)

Admitting I had delivered the same to the Lord Cobham, without allowing or approving, … This I hope is no Treason. (EModE2 State Trials, The Trial of Sir Walter Raleigh I, 214.C1)

The pragmatic difference between (8) and (9) is interesting. Both have a rhetorical question as the consequent, but (8), with the consequent being in the affirmative and therefore not formally adversative to the antecedent, is a

THE GRAMMATICALIZATION OF CONCESSIVE MARKERS

93

conditional sentence with a concessive implication, while (9), with the consequent being in the negative and therefore formally adversative to the antecedent, is a concessive sentence with a conditional implication. The OED (s.v. grant, v. 7 a) also notes a similar use of grant, stating specifically that it often introduces “an adverbial (concessive) clause”. Among the examples it cites are: (11)

Grant they never used drinking and bezling before they came to Sea […] they will soon finde out the art. (1659 D. Pell Impr. Sea 73)

(12)

Granting there were antiently such names […] it remaines doubtfull [etc.]. (1674 tr. Scheffer’s Lapland 4)

In the Helsinki Corpus there is only one instance of granting that: (13)

Granting that Gilbert’s Assertion will constantly hold true, and so, that flame is to be excepted from the general Rule, yet this exception may well comport with the Hypothesis hitherto countenanc’d. (EModE3 Robert Boyle, Electricity & Magnetism 16)

Concerning the use of admit(ing) and grant(ing) in (7)–(13), three points are worth noting: (a) they are at an early stage of grammaticalization; (b) they cannot be labelled as either factual or hypothetical concessive markers, because the nature of the concession is determined by the context; (c) they cannot even be labelled simply as concessive markers, because sometimes they allow the object clause to be purely conditional, as in: (14)

But sometimes the deuill hath power giuen him to plague and doth the harme. Admit he had power giuen him, and did kill the cattle of this man: let vs come nowe to that, who think you, gaue him the power for to strike and kill? (EModE2 George Gifford, A Handbook on Witches and Witchcraft E2R)

It may be argued that admit(ing) and grant(ing) are just like other suppositional verbs and phrasal verbs, such as suppose and put the case, which can function as a kind of conditional subordinator (see Chen 1996: 40–46). In certain contexts these suppositional verbs and phrasal verbs sometimes allow the clause they introduce to have a concessive force, as in:

94

GUOHUA CHEN

(15)

Suppose ten, fifteene, or twenty men and horses come to lodge at their house, the men shall haue flesh, tame and wild-fowle, fish with all varietie of good cheere, good lodging, and welcome. (EModE2 John Taylor, The Pennyless Pilgrimage 138.C2)

(16)

Put the Case, I should come to my Lord Cecil, as I have often done, and find a Stranger with him, with a Packet of Libels, and my Lord should let me have one or two of them to peruse: This I hope is no Treason. (EModE2 State Trials, The Trial of Sir Walter Raleigh I, 214.C1)

If admit(ing) and grant(ing) are essentially conditional markers, it may be fair to say, semantically, that they have a concessive origin. 4.2.2 The optative albeit Optative (volitive) modality and expressions with an optative meaning are found to have contributed to the development of certain conditional markers (cf. Traugott 1985: 290–291; Chen 1996: 46–50). They are also an important source for the development of concessive markers (see Harris 1988; König 1988). In Early Modern English, the optative modality is often realized in the subjunctive mood of the verb. This is the mood shared by the concessive marker all be it, which is usually spelt albeit, and the conditional marker so be it and its variants. Now the question is whether there is a derivational relationship between the two expressions. König(1988: 159) thinks albeit is a typical example of the introduction of the emphatic particle all “into a (concessive) conditional protasis”. In other words it is thought to have a conditional origin. The discussion of the function of all will be postponed till the next section. Here I would like to point out that the chronology of albeit and be it suggests that the former cannot have been derived from the latter. In the OED the first recorded use of al be it so (s.v. credible) and that of its past-tense form al were it so (s.v. albeit) are both dated 1374, and albeit (s.v. albeit) 1385, whereas the first recorded use of so be it (s.v. solidate, pa. pple.) is dated 1542, be it so (s.v. be, v. 3) 1549, and be it (s.v. be, v. 3) 1611. So despite their identity in verb form, it is unlikely that the concessive albeit is a derived form of the conditional be it. The OED gives all be it that (s.v. albeit, conj.) as the ‘proper’ form of albeit and all though it be that its ‘full’ form. It is not clear how the ‘full’ form was reduced to the ‘proper’ form. A distinction is made between albeit that and the simple albeit, but the two forms are given virtually the same factual concessive definition and all the examples it cites seem to allow only a factual concessive interpretation. So albeit appears to have been a factual concessive marker all

THE GRAMMATICALIZATION OF CONCESSIVE MARKERS

95

along. When used to introduce a verbless clause, albeit is defined as “even though, even if [my emphasis], although”. So it may allow the concession to be hypothetical. But this hypothetical concessive use of albeit is a rather late development, with the first example dated 1795. By the OED’s account, then, albeit is essentially a factual concessive subordinator. However, it is by no means impossible for it to introduce a hypothetical concessive. Sentence (17), found in Visser (1972, II: 906) is an early example: (17)

Al be it so the bodi deie, The name of hem schal nevere aweie. (c1390 Gower, C.A. 4, 2393)

The hypothetical concessive use of albeit continued in Early Modern English, though accounting for only a very small minority of instances compared with its factual concessive use. In the Helsinki Corpus there are altogether 27 instances of albeit (including one instance of all be it and two instances of al(l) were it, all in EModE1), of which only the following three introduce a hypothetical concessive: (18)

And that no manne undre the degree of a Baron use in his Apparell of his body or of his Horses eny clothe of golde or clothe of Sylver or tynsyn Satten ne no other Sylke or Clothe myxte or broderd wyth Golde or Sylver uppon payne of forfeyture of the same apparrell, albeit that yt be myxte wyth eny other Sylke or clothe. (EModE1 The Statutes of the Realm III, 8)

(19)

If the parties will at my handes call for iustice, then, al were it my father stood on the one side, and the Divill on the other, his cause being good, the Divill should haue right. (EModE1 William Roper, The Lyfe of Sir Thomas Moore 42)

(20)

The ditches were so anoyed with water, that they were troublesome to passe, albeit no other impediment should have bene offered. (EModE2 John Hayward, Annals of the First Four Years of the Reign of Queen Elizabeth 64)

Judging by its distribution in the Helsinki Corpus, albeit became fully grammaticalized as a concessive conjunction by the end of the sixteenth century, for after that no more instances of its analytic form are found. Unexpectedly, as it became more grammaticalized, its use underwent a sharp decline in frequency. This is shown in Table 1. We shall see later that this sharp decline coincides with a sudden increase in the use of though during the same period.

96

GUOHUA CHEN

Table 1. Distribution of factual concessives (F.C.) and hypothetical concessives (H.C.) introduced by albeit and its variants in the three sub-periods of the Helsinki Corpus Concessive marker

Albeit

EModE1

EModE2

EModE3

F.C.

H.C.

F.C.

H.C.

F.C.

H.C.

18

2

6

1

–

–

4.3 Expressions that emphasise the adversity of the conceded situation It has been found that concessive markers in many languages contain an emphatic universal quantifier like all or an emphatic focus particle like even (see König 1985: 10, 1988: 153; Harris 1988: 75). The motivation for the use of such words seems self-evident. “On any scale,” Harris (1988: 80) explains, “a situation which is depicted as being entirely at one end is clearly ready made to be used concessively, provided that the end specified is that least readily compatible with the main clause, which is nevertheless represented as true”. Expressions of totality or extremity, which emphasise or exaggerate the adversity of the situation, serve to make the concessive relationship clearer. Take (21) for example: (21)

The strength and beauty of this small creature, had it no other relation at all to man, would deserve a description. (EModE3 Robert Hooke, Micrographia 13.5, 210)

The above sentence has the form of a conditional sentence with no concessive marker, yet it is a hypothetical concessive sentence because the supposition that the small creature had no other relation to man is not factual and is adversative to its deserving a description. The adversity of the supposition is enhanced by the phrase at all, which indicates that the concession is unreserved and total. Without it, the concessive relation will be less apparent. Since expressions of totality and extremity can help make the concessive relationship clear, it is not surprising that they have contributed to the formation of concessive markers in so many languages. 4.3.1 The function and uses of all In English all never functions as an independent concessive marker. We have seen in 4.2.2 that it can collocate with the optative be it to form the concessive subordinator albeit. In addition, it can combine with a conditional marker to mark hypothetical concession, or with the concessive marker though for concessive reinforcement, or with the causative marker for to mark factual concession.

THE GRAMMATICALIZATION OF CONCESSIVE MARKERS

97

4.3.1.1 All + a conditional marker. According to the OED (s.v. all, adv. 10 a), “with if and though in if all, though all, all emphasised supposition or concession, = Even if, even though”. It points out that the more common word order was all if, all though. All the examples of if all are from fourteenth century texts. The first instance of all if is dated 1340, the last 1557: (22)

All if he haue power so to do. (1557 Barclay, Jugurtha (Paynell) A ij.)

The OED (s.v. all, adv. 10 c) also notes that “with the subj. mood, though or if, being expressed by the reversed position of vb. and subject (as in be they = if they be), were omitted, leaving all apparently = although. Thus: al be I = all though I be”. The examples it cites are mostly factual concessives, like: (23)

We brought more than ye were able to answer, all were it no Scriptures, nor Councels, nor Doctours. (1560 H. Cole Lett. to Jewel,)

At least (24) is a hypothetical concessive: (24)

All could he further then earths center go. (1599 Bp. Hall. Satires . . ₎

The use of all in combination with a conditional marker does not seem very popular in Early Modern English. No instance of it is found in the Helsinki Corpus. 4.3.1.2 All though. Believing that though was originally a conditional marker, which was reinterpreted as a concessive marker when strengthened by all, König (1985: 13) explains: Recall that conditionals do not entail their antecedents. But, even though they are typically used in situations where the antecedent is not assumed to be true, they are compatible with a factual interpretation. What emphatic elements like all do is to give a factual character to a clause which expresses no commitment of the speaker with respect to its truth or falsity without such a particle.

The origin of though will be discussed later in this paper. The question here is whether all did give a factual character to a clause introduced by though. Suppose though did have a conditional origin. If the reinterpretation of although had been caused by the addition of all, then the bare though should have remained a conditional marker. Yet today though is as much a factual concessive marker as although. Since both though and although were on their way to

98

GUOHUA CHEN

becoming factual concessive markers, the word all cannot have played any significant role in the process. By contrasting a hypothetical concessive introduced by though with a factual concessive introduced by al thoughe, both found in Chaucer’s Canterbury Tales, König (1985: 9) seems to suggest that by Chaucer’s time the reinterpretation of al though had already taken place. Yet in Early Modern English examples of hypothetical concessive use of although can still be easily found, indicating that the concessive reinforcement of though with all did not guarantee that the expression was a factual concessive marker. We have seen an example of counterfactual concessive introduced by although in (6) at the beginning of this paper, where the counterfactuality is signified by the use of backshifted past tense. When the concession is not counterfactual but only hypothetical, the tense of the verb is not backshifted: (25)

If a man with gorgeous apparell come amongst vs, although he bee a theefe or a murtherer (for there are theeues and murtherers in gorgeous apparell) be his heart whatsoever, if his coat be of purple, or velvet, or tissue, every one riseth up, and all the reverent solemnities wee can vse, are too little. (EModE2 Richard Hooker, Two Sermons upon Part of S. Judes Epistle 8)

(26)

Everie p~son and p~sons which after one moneth nexte ensuinge the end of this p~sent Session of Parliament, shall stabbe or thruste any p~son or persons that hathe not then any weapon drawne, […] soe as the person or persons soe stabbed or thruste shall thereof die within the space of sixe moneths then next followinge, although it cannot be proved that the same was done of malice forethoughte, yet the partie soe offendinge […] shall be excluded from the benefit of his or theire Cleargie, and suffer Deathe as in case of Wilfull Murder. (EModE2 The Statutes of the Realm IV, 1026)

If we compare the ratio of factual concessive vs. hypothetical concessive uses of although with that of though, as is shown in Table 2, we will find: (i) the factual concessive use of both expressions is much more common than their hypothetical concessive use; (ii) the percentage of factual concessive use of although is higher than that of though; (iii) there is a dramatic increase in the frequency of though in EModE2 and EModE3. The first finding shows that in Early Modern English both though and although were well on their way of becoming factual concessive markers. The second finding indicates that although was going one step ahead of though in this process. The third finding may explain why the use of albeit sharply decreased over the same period.

99

THE GRAMMATICALIZATION OF CONCESSIVE MARKERS

Table 2. Distribution of factual concessives (F.C.) and hypothetical concessives (H.C.) introduced by although and though in the three sub-periods of the Helsinki Corpus Concessive marker

EModE1 F.C.

although though

H.C.

22 (92) 02 0(8) 37 (76) 12 (24)

EModE2 F.C.

H.C.

33 (83) 07 (17) 65 (77) 19 (23)

EModE3 F.C.

H.C.

018 (95) 01 0(5) 147 (85) 26 (15)

4.3.1.3 The factual concessive marker for all. According to the OED (s.v. for, prep. VII. 23), the use of for in the sense “in spite of, notwithstanding” in relation to “a preventive cause or obstacle” began as a preposition in Old English and was rare without all. The contribution of all to the concessive meaning of for all is apparent. The original meaning of for was “in front of” and “in the presence or sight of” (OED, s.v. for, prep. I. 1). This sense was conducive to for all having the factual concessive sense ‘in spite of’, because when something is in sight it is usually assumed to be factual. It appears that for all was not used as a concessive marker until the beginning of the sixteenth century. Sentence (27) is the first instance of its use cited in the OED (s.v. for, prep. VII. 23): (27)

For all that the frenche kynge sende to hym to delyuer the same castels, yet he refused so to do. (1523 Ld. Berners Froiss. I. clvi. 189)

In the Helsinki Corpus for all that is quite common as a prepositional phrase, but only in (28) is it clearly used as a concessive marker: (28)

How many of this Citie for all that they are Vsurers, yet would be counted honest men, and would faine haue Vsurie esteemed as a trade. (EModE2 Henry Smith, Two Sermons on “Of Usurie” B2V)

4.3.2 SV inversion + never so The phrase never so is defined in the OED (s.v. never, adv. 4) as “in conditional clauses, denoting an unlimited degree or amount”, with the first instance of its use dated as no later than 1122. Interestingly, this emphatic construction has a negative form but a positive meaning. The first three citations in the OED all begin with another negative particle (næfde [= ne + hæfde], ne), which suggests that the construction probably first began as a double negative. By the fourteenth century the single negative form had become the rule and its use continued until

100

GUOHUA CHEN

the end of the seventeenth century when it began to be replaced by the logically correct form ever so (see the OED s.v. ever, adv. 9 d). Although seemingly illogical, never so makes very good sense as a concessive marker, with so denoting the extremity of the adversative situation, and never suggesting the speaker’s contemptuous defiance of the situation. In the Helsinki Corpus altogether six instances of its use are found, all giving only a hypothetical concessive reading, for example: (29)

Were he of hym selfe neuer so feble and faint, nor neuer so lykely to fall, yet the grace of God was sufficient to kepe hym vp and make him stand. (EModE1 Thomas More, The Correspondence of Sir Thomas More 546)

(30)

They can not abide to read any medicine, that is of a long composition, be it neuer so precious. (EModE2 William Clowes, Treatise for the Artificiall Cure of Struma 34)

The use of SV inversion + never so is morpho-syntactically restricted. The subject is restricted to a personal pronoun, the verb to the subjunctive be and were. Because of this, the expression is not a fully grammaticalized concessive marker. 4.4 Expressions that assert the ineffectiveness of the conceded situation The ineffectiveness of the conceded situation is commonly asserted by expressions of contempt, defiance or disregard (cf. König 1988: 152–153). By asserting this ineffectiveness, the speaker affirms the independence of the consequent situation on the conceded situation. The concession expressed in this way tends to be factual. It is natural for such expressions to mark factual concession because they typically bring out a definite noun phrase expressing a presupposition, and when something is presupposed, its existence or factuality is normally assumed. In Early Modern English there are three such expressions — (in) despite (of), in spite of and notwithstanding. 4.4.1 (In) despite (of) and in spite of According to the OED (s.v. despite and spite), both despite and spite were derived directly from OF despit at the end of the thirteenth century, with spite being the aphetic form of despite. From the moment it was borrowed, despite in the phrase in despite of was used in the sense “in contempt or scorn of; in contemptuous defiance of” (OED, s.v. despite, n. 5 a), whereas the use of in spite of did not appear until 1400 (OED, s.v. spite, n. 5 a). König (1988: 157) observes

THE GRAMMATICALIZATION OF CONCESSIVE MARKERS

101

that expressions like in spite of “derive from predicates originally only attributable to arguments denoting human agents or experiencers”. It appears that, when in despite of or in spite of is used in its original sense, not only does the subject of the sentence typically denote a human agent, the object of the phrase also typically denotes a human patient, as in: (31)

And sent all their heddes … to be set upon poles, over the gate of the citie of Yorke in despite of them, and their lignage. (1548 Hall Chron. 183 b)

(32)

In spite both of him and his Legate, they kept company with them that were excommunicate. (1568 Grafton Chron. II. 113)

The reinterpretation of the two expressions as concessive markers seems to have gone hand in hand with the loosening of the collocative restrictions. When the subject of the sentence or the object of the prepositional phrase denotes something non-human, the sense of concession becomes more apparent, for example: (33)

To assaile the entrie of the mouth of Lisbone, in despite of all the fortresses that were there. (1600 E. Blount tr. Conestaggio 132)

(34)

He dyed, Clad in his cote armor paynted all in paper, Al torne and reversed in spyte of his behaver. (1563 Mirr. Mag., Blacke Smyth & Ld. Awdeley lx,)

The OED (s.v. despite, n. 6) notes that, in later use, in despite of in the sense ‘notwithstanding’ often suffered the loss of in and was further shortened to despite, thus eventually becoming a preposition: (35)

If this Bruno … sit in Peters chair, despite of chance. (1590 Marlowe Faust Wks. (Rtldg.) 123/2)

(36)

Il’e … Ransacke the pallace where grim Pluto reignes … Despight his blacke guard. (1613 Heywood Silv. Age iii. Wks. 1874 III. 159)

In spite of sometimes also lost its in (see OED, s.v. spite, n. 6), for example: (37)

Spyte of your enemyes, I shal me so spede, That in short tyme ye may rewarde my mede. (1509 Hawes Past. Pleas. xix. (Percy Soc.) 96)

But somehow it did not go so far as to become a preposition. In the Helsinki Corpus the use of (in) despite (of) and in spite of is very rare. Only one instance of in despite of is found, which does not appear to be concessive:

102

GUOHUA CHEN

(38)

Al though Sicilians had exiled hym, yet in despite of them all he reigned. (EModE1 Thomas Elyot, The Boke Named the Gouernour 22)

Two instances of in spite of are found, both functioning as concessive markers. Sentence (39) is one of them: (39)

The knowledge of the tonges (in spite of some that therein had florished) was manifestly contemned. (EModE1 Roger Ascham, The Schoolmaster 281)

4.4.2 Notwithstanding The formal make-up of notwithstanding (not + withstanding) is transparent, and its etymology is clear. According to the OED (s.v. withstanding), it was coined in the late fourteenth century in imitation of the medieval Latin phrase non obstante or OF non obstant. Initially a preposition used in the sense ‘in spite of’, in the mid-fifteenth century it began to be used as a conjunction in the sense ‘although’. This use continued throughout Early Modern English, for example: (40)

Notwithstonding your grace had commaundid us to retorne, yet we … wold be content to make here abode. (1502 Lett. Rich. III & Hen. VII (Rolls) II. 111)

(41)

Dick, I thank God, continueth free from any infectious disease notwithstanding since Michalmas there hath dyed in the town seaven score and the greatest part of the small pox. (EModE3 Elizabeth Oxinden, The Oxinden and Peyton Letters 308)

From the late fifteenth century onwards, notwithstanding began to be positioned after the noun phrase. Following is an Early Modern English example: (42)

But, her worthiness notwithstanding, and that he had a fair daughter by her, he … was wonderfully tormented in conscience. (1555 Harpsfield Divorce Hen. VIII (Camden Soc.) 179)

In terms of grammaticalization, this seems to be a retrogression, because prepositions are more grammatical than participles and this postpositional notwithstanding can in effect be interpreted as a present participle, and the whole phrase as an absolute clause. The motivation for this change in word order probably lay in a desire to smooth out the apparent incongruity between the literal meaning of notwithstanding ‘not standing against’ and the sense of concession it is meant to express. Consider the prepositional use of notwithstanding in (43):

THE GRAMMATICALIZATION OF CONCESSIVE MARKERS

(43)

103

He dothe wronge, nothwithstondynge the said lawe declared by the prophete. (1460 Fortescue Abs. & Lim. Mon. iv. (1885) 117)

If nothwithstondynge the said lawe is paraphrased as ‘not standing against the said law’, then the intended concessive meaning will be lost. By contrast, (42) presents no such problem, because there the noun phrase expressing the conceded situation is the subject, and it is this situation that is not standing against the consequent situation, so everything is in place. In the Helsinki Corpus, I have identified only four instances of conjunctional notwithstanding, so its use is not very common. There are 27 instances of postpositional notwithstanding as compared with 21 instances of its prepositional use, but the overwhelming majority of its postpositional use is concentrated in just one genre — Law. In this type of text the sense of concession expressed by the postpositional notwithstanding is often not very strong, for example: (44)

All Clothmakers within the same Countie may make cloth and use clothmakyng as they before the makyng of this acte have usyd and accustomed, this acte or any thyng therin conteyned notwithstondyng. (EModE1 The Statutes of the Realm III, 29)

4.5 Expressions that state the concomitance of the two situations expressed by the antecedent and the consequent As we have seen in Section 2, an essential semantic property of concessives is that the adverse situation expressed by the antecedent does not affect or prevent the existence or occurrence of the situation expressed by the consequent. Since both situations coexist or co-occur, expressions such as while and though, which denote concomitance, become a natural source of concessive markers. 4.5.1 While According to Traugott and König (1991: 201), the temporal conjunction while probably began to have a concessive inference in the early seventeenth century. The example they cite is from the OED (s.v. while, conj. 2.b): (45)

Whill others aime at greatnes boght with blood, Not to bee great thou stryves, bot to bee good. (1617 Sir W. Mure Misc. Poems xxi. 23)

But by the definition adopted in this paper, (45) and the other seventeenth century example sound more like sentences of contrast. No genuine concessive use of while is found in the Helsinki Corpus.

104

GUOHUA CHEN

4.5.2 Though Both König (1985: 10–11, 1988: 153–154) and Harris (1988: 75) regard conditional markers as one of the most important sources of factual concessive markers. The English example they cite is the OE þeah. According to König (1985: 7), “OE þeah is related to Gothic þau + h ‘if + also/and’” and “the earlier Gothic cognate shows that this form has a clear conditional origin and was originally used to express a concessive conditional relationship”. According to the OED (s.v. though, adv. and conj.), however, the þau of Gothic þauh means ‘in that case’. Since ‘in that case’ can be interpreted both hypothetically and factually and since in its earliest days þeah was actually used to introduce both factual as well as hypothetical concessions, it was more likely that it started out as a general concessive marker rather than a conditional marker. If though had a conditional origin, we should be able to find more traces of it in its early use. What I find instead is just the opposite. According to the OED (s.v. though, conj. 1 and 4), its concessive use can be traced back to c888 while its use in the form as though in the sense ‘as if’ did not appear until c1200. On the basis of an exhaustive study of concessive sentences in Old English poetry, Quirk concludes that the purely conditional function of þeah is “extremely rare in OE and probably not in evidence at all in the poetry” (Quirk 1954: 39). Quirk’s verdict is confirmed by Campbell, who categorically declares: “There is no evidence that þeah can be used for gif ” (Campbell 1956: 66 fn. 2). If þeah did not begin as a conditional marker, what kind of expression was it initially? I think the answer lies in the sense ‘also/and’ denoted by h. Crosslinguistically, “additive focus particles” like the English also and too are frequently a component of concessive connectives (see König 1985: 10–11). This is motivated by the fact that such particles signify the notion of concomitance. In the case of though, because of its semantic make-up, from the very beginning it could be used as a concessive adverb (see OED, though, I. a. adv.). In Mitchell’s view, which is drawn on Burnham (1911: 14), “it is probable that the adverbial use came first and that the conjunction þeah developed from it through the intermediate stage þeah þe” (Mitchell 1985, II: 709). We have seen in Table 2 that in Early Modern English though and its emphatic form although were used predominantly as factual concessive subordinators. It had taken them several centuries to travel thus far along the road of functional specialization, a process which we know little about. Although had nearly reached its goal by the end of the Early Modern English period, but it would take though another several centuries to complete the process, for its hypothetical concessive use can be found as late as in 1900 (see Visser 1972, II: 905). This process seems to have little, if anything, to do with verb forms. In

THE GRAMMATICALIZATION OF CONCESSIVE MARKERS

105

Early Modern English it was perfectly normal for the subjunctive to be used to express factual as well as hypothetical concession. Although there was a general tendency to use the indicative for factual concession, it was possible to use it to express hypothetical concession as well, for example: (46)

O Lord! I’ll go and put on my Lac’d Smock, tho’ I am whipt till the Blood run down my Heels for’t. (EModE3 John Vanbrugh, The Complete Works I, 60)

Of course, when the concession is counterfactual, as in (47) below, it is usually marked by backshifted past or past perfect tense. This feature is shared by both conditionals and concessives. (47)

He shall neuer scape death on my swordes point, Though I shoulde be torne therfore ioynt by ioynt. (EModE1 Nicholas Udall, Roister Doister L. 1145)

4.6 Expressions that help affirm the factuality of the consequent situation The concessive markers we have examined so far all concern some aspect of the antecedent. Expressions of similarity, continuance and immutability, which help affirm the factuality of the consequent situation in one way or another, also provide an important source for concessive markers in the form of correlative conjuncts such as yet and nevertheless. 4.6.1 Yet Mätzner (1874, III: 360) believes that yet was originally a temporal expression; it became an adversative particle by “denoting that the following thought is still (notwithstanding) valid, alongside of the preceding one”. The OED (s.v. yet, adv. (adj.) and conj. 1 and 2) seems to think otherwise. Finding it to be “of obscure origin”, the OED identifies two basic meanings of yet: (i) “in addition, or in continuation; besides, also; further, furthermore, moreover”, (ii) “implying continuance from a previous time up to and at the present (or some stated) time”, and believes that its adversative meaning “in spite of that, for all that, nevertheless, notwithstanding” developed from the non-temporal meaning when yet was used to introduce “an additional fact or circumstance which is adverse to, or the contrary of what would naturally be expected from, that just mentioned” (s.v. yet, adv. (adj.) and conj. 9a). It is worth noting that both meanings (i) and (ii) proposed by the OED have the sense of continuation or continuance. As continuation or continuance often implies resistance to interference or adverse circumstances,

106

GUOHUA CHEN

perhaps it is from the sense of continuance that yet developed into a concessive consequent marker. When the antecedent is already marked by a concessive marker, the use of yet in the consequent is an optional extra. Its role as a concessive marker is more significant, sometimes vital, when the antecedent is not marked. Very often the sense of concession, though inherent in the sentence, is not immediately apparent. Without yet, the sentence still allows only a concessive-conditional reading, but the use of yet immensely facilitates this reading, for example: (48)

Yff your Grace wer eligible and undir th’empire, yet ye coud not be chosen Emperor, by cause ye were never Kinge of Romains. (EModE1 Cuthbert Tunstall, Original Letters I, 137)

In sentences like (49) below, the use of yet is vital to the concessive reading of the condition. Without it, such instances are likely to lose their concessive force and to be interpreted as simple conditionals: (49)

This is the end of the Vsurer and his money, if they stay together till death, yet at last there shall bee a diuision. (EModE2 Henry Smith, Two Sermons on “Of Usurie” E2R)

In the Helsinki Corpus the use of yet as a concessive correlative conjunct is not as common as we might expect. Of the 68 hypothetical concessives introduced by if, only 10 have yet in the consequent. 4.6.2 Nevertheless The motivation of the use of nevertheless or its variant forms natheless and nonetheless as a concessive marker is transparent. By affirming that the truth of the consequent is in no way less valid because of the adverse situation expressed in the antecedent, it stresses the concessive relation. Its use as a concessive marker in a conditional sentence is not specifically noted in the OED. In the Helsinki Corpus only one instance is found and in this one instance it is used together with yet: (50)

And be it further Inacted, That if anye Horse Mare Geldinge Coulte or Fillie, after Twentie Dayes next ensuynge thende of this Session of Parliament, shalbe stolen, and after shalbe soulde in open Fayre or Markett, and the same Sale shalbe used in all Poynt~ and Circumstaunc~ as aforesaide, that yet nevertheles the Sale of any suche Horse Mare Geldinge Coulte or Fillie, within Sixe Monethes next after the Fellonye done, shall not take awaye the P~pertie of the Owner from whom the same was stolen. (EModE2 The Statutes of the Realm IV, 811)

THE GRAMMATICALIZATION OF CONCESSIVE MARKERS

5.

107

Clause order and hypothetical concession

While accepting that an if-clause preceding the consequent may have a concessive reading, Haiman (1986: 221–222) thinks that such concessive use of if is comparatively rare. He observes that “in English, at least, concessive conditionals are usually doubly marked: either by the presence of the focus marker even, itself a diacritic, or by clause inversion”, that is, by inverting the order of antecedent and consequent. Although he does not go so far as to allege that clause inversion is a marker of concession, he believes that parenthetical conditionals like “Greetings from your affectionate, if absentminded, son” are invariably concessive and “owe this meaning to their marked position rather than their superficial morphology”. Haiman, of course, is talking about present-day English. In Early Modern English the predominant clause order for hypothetical concession is antecedent + consequent. But I have noticed a notable shift toward clause-order inversion. As in present-day English, there are three clause orders in Early Modern English hypothetical concessives: (i) antecedent + consequent (A + C); (ii) consequent + antecedent (C + A); and (iii) antecedent embedded in consequent (C + A + c). When the antecedent is embedded in the consequent, it is the so-called parenthetical conditional clause, for example: (51)

I made the Experiment also at differing times, and with some months, if not rather years, of interval, but with the like success. (EModE3 Robert Boyle, Electricity & Magnetism 34)

From Table 3 we can see that in the Helsinki Corpus A + C is the predominant clause order for hypothetical concessives, accounting for 60%; however, there is a visible shift from A + C to non-A + C order. In EModE3 non-A + C order even outnumbers A + C order by 2 to 1. This is due mainly to the surge of parenthetical conditionals.

Table 3. Distribution of different clause orders of hypothetical concessives in the three subperiods of the Helsinki Corpus Clause order

EModE1

EModE2

EModE3

Total

A+C C+A C+A+c

18 06 –

33 14 02

09 08 10

060 028 012

Total

24

49

27

100

108 6.

GUOHUA CHEN

Conclusion

The grammaticalization of concessive markers is motivated by the need to emphasise or express explicitly some aspect of the semantic and pragmatic properties of the concessive sentence. Concessive markers derive from a wide variety of sources. Although conditional markers are one of the sources for hypothetical concessive markers, there is no evidence in English to testify that some of them have developed into factual concessive markers. In Early Modern English the majority of the concessive markers did not distinguish between factual and hypothetical concessions, though some of them, such as (al)though and albeit, were used predominantly as factual concessive markers and some others, such as if, were used predominantly as hypothetical concessive markers. If need be, factual and hypothetical concessions could be explicitly marked. To mark factual concession, one could use for all that and notwithstanding, etc.; to mark hypothetical concessive one could use SV inversion + never so or a concessive conjunct like yet in the consequent of a conditional sentence. But these expressions, being not fully grammaticalized, were not used very frequently in such functions. In the great majority of cases people would rather use the more ambiguous (al)though for the expression of factual concession and the more ambiguous bare if for the expression of hypothetical concession, feeling content that what they meant was sufficiently clear from the context. What caused them to change their mind remains a mystery.

Acknowledgments I am indebted to Merja Kytö for kindly arranging for me to use the Early Modern English section of the Helsinki Corpus of English Texts for this study and to Olga Fischer, Dieter Stein and the three anonymous reviewers for their comments on the earlier versions of this paper.

Notes 1.

The two terms are adapted from Quirk et al. (1985: 1099–1102), where the original terms are universal conditional-concessive clause and alternative conditional-concessive clause, the former referring to clauses typically marked by however, whatever and so on, the latter to clauses typically marked by whether … or. For lack of space the present paper does not cover these two types of concessives.

2.

This section of the corpus consists of three sub-sections — EModE1 (1500–1570), EModE2 (1570–1640) and EModE3 (1640–1710). The three sub-sections are comparable in size and genre make-up except that EModE3, unlike EModE1 and EModE2, contains no Bible text. To

THE GRAMMATICALIZATION OF CONCESSIVE MARKERS

109

ensure full comparability among the three sub-sections, I have deleted the Bible texts from EModE1 and EModE2 in the version of the corpus I use for this study. In addition, for a reason that has nothing to do with the purpose of the present investigation, I have replaced samples of Shakespeare’s The Merry Wives of Windsor in EModE2 with excerpts of roughly the same length from Ben Jonson’s Every Man in his Humour. For a detailed description of the corpus, see Kytö (1991). 3.

Haiman (1986: 220–224) maintains, however, that it is impossible for a paratactic structure to have a concessive reading.

4.

The sentences in (2)–(3) are based on Quirk et al. (1985: 644).

5.

The sentences in (4) are based on Quirk et al. (1985: 1099).

6.

When the focusing particle even of a hypothetical concessive sentence focuses on only one part of the antecedent the truth of the consequent depends on that of the antecedent. See König (1986: 232).

7.

The earliest instance of even if recorded in the OED is s.v. poetess: 1748 Lady Luxborough, Let. to Shenstone 28 Apr., “I am no Poetess; which reproachful name I would avoid, even if I were capable of acquiring it.”; its earliest use cited s.v. even, adv. is dated 1824.

8.

All examples with references beginning with EModE are from the Helsinki Corpus (for full reference of the sources, see Kytö 1991), others beginning with dates are from the OED.

9.

In König’s terminology, concessive sentence proper means “factual concessive sentence” and concessive connective means “factual concessive connective”.

10.

The only example König (1988: 152) offers is the Norwegian selv om.

References Burnham, J. M. 1911. Concessive Constructions in Old English Prose [Yale Studies in English 39]. New York: Henry Holt and Company. Campbell, A. 1956. “Review of Quirk 1954”. Review of English Studies 7: 64–68. Chen, G. 1996. Conditional sentences in early Modern English: A study in grammaticalization. Ph.D dissertation, Faculty of English, University of Cambridge. Comrie, B. 1986. “Conditionals: A typology”. In On Conditionals, E. C. Traugott, A. ter Meulen, J. Snitzer Reilly and Ch. A. Ferguson (eds), 77–99. Cambridge: Cambridge University Press. Fischer, O. 1992. “Syntax”. In The Cambridge History of the English Language, 1066–1476, vol. II, N. Blake (ed.), 207–408. Cambridge: Cambridge University Press. Haiman, J. 1986. “Constraints on the form and meaning of the protasis”. In On Conditionals, E. C. Traugott, A. ter Meulen, J. Snitzer Reilly and Ch. A. Ferguson (eds), 215–227. Cambridge: Cambridge University Press. Harris, M. 1988. “Concessive clause in English and Romance”. In Clause Combining in Grammar and Discourse, J. Haiman and S. A. Thompson (eds), 71–99. Amsterdam: John Benjamins.

110

GUOHUA CHEN

Hopper, P. J. and Traugott, E. C. 1993. Grammaticalization. Cambridge: Cambridge University Press. König, E. 1985. “On the history of concessive connectives in English: Diachronic and synchronic evidence”. Lingua 66: 1–19. König, E. 1986. “Conditionals, concessive conditionals and concessives: Areas of contrast, overlap and neutralization”. In On Conditionals, E. C. Traugott, A. ter Meulen, J. Snitzer Reilly and Ch. A. Ferguson (eds), 229–246. Cambridge: Cambridge University Press. König, E. 1988. “Concessive connectives and concessive sentences: Cross-linguistic regularities and pragmatic principles”. In Explaining language universals, J. A. Hawkins (ed.), 145–166. Oxford: Basil Blackwell. König, E. 1991. “Concessive relation as the dual of causal relations”. In Semantic Universals and Universal Semantics, D. Zaefferer (ed.), 190–209. Berlin: Foris Publications. König, E. 1992. “From discourse to syntax: The case of concessive conditionals”. In Who Climbs the Grammar Tree, R. Tracy (ed.), 423–434. Tübingen: Niemeyer. Kytö, M. 1991. Manual to the Diachronic Part of the Helsinki Corpus of English Texts. Helsinki: Department of English, University of Helsinki. Leuschner, T. 1998. “At the boundaries of grammaticalization: What interrogatives are doing in concessive conditionals”. In The Limits of Grammaticalization, A. Giacalone Ramat and P. J. Hopper (eds), 159–187. Amsterdam: John Benjamins. Mätzner, E. 1874. An English Grammar: Methodical, Analytical, and Historical, trans. from the German by Clair James Grece, 3 vols. London: John Murray. Mitchell, B. 1985. Old English Syntax, vol. II. Oxford: Clarendon Press. Poutsma, H. 1960. A Grammar of Late Modern English, 5 vols. Tokyo: Senjo [1928]. Quirk, R. 1954. The Concessive Relation in Old English Poetry [Yale Studies in English 124]. New Haven: Yale University Press. Quirk, R., Greenbaum, S., Leech, G. and Svartvik, J. 1985. A Comprehensive Grammar of the English Language. London: Longman. Thompson, S. A. and Longacre, R. E. 1985. “Adverbial clauses”. In Language Typology and Syntactic Description, vol. II, T. Shopen (ed.), 171–234. Cambridge: Cambridge University Press. Traugott, E. C. 1985. “Conditional markers”. In Iconicity in Syntax, J. Haiman (ed.), 289–307. Amsterdam: John Benjamins. Traugott, E. C. and König, E. 1991. “The semantics-pragmatics of grammaticalization revisited”. In Approaches to Grammaticalization, vol.1, E. C. Traugott and B. Heine (eds), 189–218. Amsterdam: John Benjamins. Van de Auwera, J. 1986. “Conditionals and speech acts”. In On Conditionals, E. C. Traugott, A. ter Meulen, J. Snitzer Reilly and Ch. A. Ferguson (eds), 197–214. Cambridge: Cambridge University Press. Visser, F. T. 1972. An Historical Syntax of the English Language, vol. II. Leiden: E. J. Brill.

Combining English auxiliaries David Denison University of Manchester

The whole entirely depends, added my father, in a low voice, upon the auxiliary verbs, Mr Yorick. Laurence Sterne, Tristram Shandy V.xlii

1.

Introduction

The processes of grammaticalisation by which lexical verbs turned into auxiliaries in English raise a number of interesting questions, many of them already studied in depth. Grammaticalisation involves semantic and syntactic and sometimes morphological change, as well as changes of distribution, and the changes need not be simultaneous. In this paper I wish to explore the dating and significance of grammaticalisation of certain English auxiliaries by looking at their combinatory possibilities. The sequencing and co-occurrence constraints on the present-day auxiliaries are among the most systematic areas of English syntax, if we allow ourselves to leave aside marginal and dialectal forms. Some combinations once possible have become ungrammatical, while other combinations have come into existence during the period of recorded history, not always as soon as might have been predicted from the behaviour of the individual auxiliaries. These facts allow us to infer dates for grammaticalisation certainly of the progressive and perhaps of other auxiliaries too. The auxiliary system of present-day English in tensed, finite clauses can be represented as follows: (1)

(Modal) (Perfect) (Progressive) (Passive)

112

DAVID DENISON

The four slots come in a fixed relative order. Each slot can be filled or not, independently. If none is filled, then the dummy auxiliary  may appear as sole auxiliary; whether  is necessary, optional, or forbidden depends on the clause type and the lexical verb. All of this is familiar and of course oversimplified, but it will serve the purpose adequately for now. There are pros and cons in using a grammatical label like ‘perfect’ rather than a lexical item like . The former allows us to generalise across different possible exponents of a slot, such as ,  or Old English  for the perfect, the latter to generalise across different uses of a particular verb — notably . I shall use both (and see also § 4.3 below). With the possible exception of Passive, all of these auxiliaries form periphrases with a lexical verb in the sense that the combination Aux + V commutes with the simple verb V. Cross-linguistically it is probably justifiable to treat passive Aux + V as a periphrasis too, but for English — especially after the demise of the unique inflectional passive of  ‘be called’ — the label is less than ideal. For the reason that follows, however, it will be convenient to retain passives in the definition. These auxiliaries form probably the most orderly and systematic area of English syntax. It is a truism that each of the items which can serve as an auxiliary is a development — historically speaking — out of some full-verb use, and it is reasonable to call all of them ‘grammaticalised’. (I shall ignore the main-verb analyses of present-day English auxiliaries.) There are various grounds for this: the facts that they are closed-class items, virtually without restriction as to the lexical verbs they can collocate with (though this is less true of  than of other auxiliaries), mostly without argument structure of their own, morphologically odd, semantically general in having senses to do with tense or aspect or at least with sentence modification (epistemic meaning, etc.), and so on. But grammaticalisation is alternatively a diachronic process or a synchronic gradient, and it is a moot point how far into the process or how far along the gradient a verb has to go before it can be said to be ‘grammaticalised’. Consider perfect  in diachronic terms. The English perfect is generally regarded as a development by reanalysis of structures involving possessive . Contrast (2) and (3), for example: (2)

OE Bede 4 23.328.6 ðonne hæbbe we begen fet gescode then have we both feet shod(..) suiðe untællice very blamelessly

COMBINING ENGLISH AUXILIARIES

(3)

113

OE Or 132.17 Nu ic hæbbe gesæd … hu … now I have said … how …

In rare examples of the older type like (2),  can be a transitive lexical verb meaning ‘possess’, the word order may involve a sentence brace in which the NP — which is object of  — precedes the participle, and the participle is an object predicative which may carry adjectival inflection. In (3), however,  cannot mean ‘possess’, and in this particular case there is no object and no adjectival agreement on the participle. Various stages of development of the perfect are potentially relevant: (A) when the  perfect became available for any lexical verb which did not conjugate with  (late Old English?) (B) when it had come to be a pure tense equivalent (late Old English?) (C) when it had developed approximately its present-day meaning (seventeenth century?) — which would have involved the loss of B (D) when it became available for every non-auxiliary verb (late Modern English) I am content to regard A as indicating the stage when perfect  had become an auxiliary verb, since it suggests that  was being used transparently, i.e. without an argument structure or selectional restrictions of its own. It had been reached when perfect  occurred with transitive participles meaning ‘distributed’, ‘lost’, ‘eaten’ — meanings incompatible with possession — or with intransitive verbs, or with less than fully transitive verbs, as in (3). As discussed in Denison (1993: 346–348), word order and participial inflection are not in themselves reliable indicators of the syntactic status of possible  perfects. For other auxiliaries all sorts of evidence of grammaticalisation may be used. In this paper I shall concentrate on one seemingly simple source of evidence: the combinatory possibilities of (potential) auxiliaries. As we shall see, combinations of such verbs may provide evidence that one of them — usually the first but exceptionally the second — has been grammaticalised with auxiliary status, or conversely that the second remains ungrammaticalised. The analysis of each case is different. Exhaustive coverage is not aimed at. I begin with what from a present-day English point of view looks like the repetition of a slot, then look at some other combinations.

114 2.

DAVID DENISON

Doubling of auxiliaries

2.1 Double  Late in the Middle English period, and especially in the works of Caxton, frequent use was made of a double  construction: (4)

1490 Caxton Prol.Eneydos 108.14 And also my lorde abbot of westmynster and also my lord Abbot of Westminster ded do shewe to me late certayn euydences “did do” show to me recently certain pieces-of-evidence wryton in olde englysshe written in old English ‘and also my lord, the Abbot of Westminster, had me shown recently certain pieces of evidence written in ancient English’

(In some examples — e.g. (60) below — either the first or the second verb is not  but  or .) I follow Ellegård (1953: 110–115) in reading did do examples as an attempt to mark causative meaning at a time when the periphrasis was on the increase and simple causative use of  open to misunderstanding. If, as I believe, periphrastic  was a development of causative  (Denison 1985a, 1993), examples like (4) provide evidence of the grammaticalisation of the -periphrasis as a transparent auxiliary, since two consecutive causatives with empty argument slots would be both redundant and highly opaque. Thus the first  is periphrastic; the second, untensed, is causative. 2.2 Double modals Double modals are a much-tilled field and I shall not spend long on them here. Untensed forms of modals are dealt with by Visser in his (1963–1973: §§ 1649–1651, 1684–1687, 1722–1723, 1839, 2042, 2134),1 with further Middle English examples in Ogura (1993, 1998) and discussion in Nagle (1993, 1995). Some examples: (5)

c1180 Orm. 2958 … Þatt I shall cunnenn cwemenn Godd … that I shall have-ability() please() God ‘ … that I shall have the ability to please God’

COMBINING ENGLISH AUXILIARIES

115

(6)

c1450 Pilgr.LM(Cmb) 1.467 And whan ye wole go withoute me and when you will go without me [sc. Reason] ye shul wel mown avaunte yow you shall well be-able-to be-boastful ‘and when you wish to go without me you shall certainly be able to be boastful’

(7)

(c1463) Paston 66.16 … and wythowte I knowe þe serteynté … and unless I know the truth I chal not conne answere hym. I shall not be-able-to answer him ‘… and without knowing the truth I shall not be able to answer him.’

(8)

c1483(?a1480) Caxton, Dialogues 3.37 Who this booke shall wylle lerne … he-who this book shall wish learn … ‘He who wishes to master this book …’

(9)

1532 Cranmer Let. in Misc.Writ.(Parker Soc.) II.233 I fear that the emperor will depart thence, before my letters shall may come unto your grace’s hands.

A double modal implies that the second modal is in the infinitive, which is also the case in the non-finite clause of (10): (10)

1533 More, Wks. IX 84.4 [885 C1] some waye yt [= þat = that: D.D.] appered at ye [= the] firste to mow stande the realme in great stede

The modal in the infinitive is non-epistemic (only examples (6), (9) and (10) above show any possibility of an epistemic interpretation). This is consistent with Plank’s observation (1984: 310, 314) that non-modal syntax and morphology in modal verbs (taking of direct objects, untensed forms) has always been associated with non-modal semantics. The double modal construction in historical texts therefore suggests that the second modal was not grammaticalised; I am unable to deduce anything about the first modal. In later English double modals are confined to dialects of northern and Scots English, plus (later still) certain southeastern American dialects (Montgomery 1989; Montgomery and Nagle 1994 [for 1993]; de la Cruz 1995). Again the second modal is generally root rather than epistemic (Nagle 1994: 205–206), though futurity, which does occur as second modal, is not so obviously — to me at least — a root meaning. On the relation-

116

DAVID DENISON

ship between future meaning and (other) kinds of modality see Bybee, Pagliuca and Perkins (1991, esp. 24–25). 2.3 Double perfect There are several ways in which a double perfect can be formed. One has been common in English for centuries, though it is unclear whether it has ever attained the level of standard usage. It depends on the notion of unrealised action or unreality. 2.3.1 Unreality A correlation has developed between unrealised action and the use of the  perfect in certain contexts. Some examples are unreal conditionals, where  may appear in the protasis, the apodosis, or both, but the usage is not confined to conditionals: (11)

c1230 (?a1200) Ancr. 13b.24 hwa se hefde iseid to eue … ‘A eue anyone who had said to Eve … O Eve went te awei …’, hwet hefde ha iondsweret? turn yourself away … what had she answered ‘If anyone had said to Eve … “O Eve, turn away” … What would she have answered?’

(12)

(1448) Paston 128.21 … and told here þat Še had sergyd to a fownd … and told her that you had searched to have found wrytyng þer-of and Še kwd non fynd in non wyse. writing thereof and you could none find in no way ‘… and told her you had gone searching to find written evidence of it, and you could not find any anywhere’

(13)

(1478) Let.Cely 34.5 and thay spake to me, and desyryd to haue had iij sarpelers …2 and they spoke to me and desired to have had 3 bales (of wool) … ‘and they spoke to me and asked to have three bales …’

(14)

1660 Pepys, Diary I 102.18 (3 Apr) This day came the Lieutenant of the Swiftsure (who was sent by my Lord to Hastings, one of the Cinque ports, to have got Mr. Edw. Mountagu to have been one of their burgesses); but could not, for they were all promised before.

COMBINING ENGLISH AUXILIARIES

117

(15)

1660 Pepys, Diary I 216.27 (7 Aug) Here I endeavoured to have looked out Jane that formerly lived at Dr Williams at Cambrige, whom I had long thought to live at present here; but I found myself in an errour, meeting one in the place where I expected to have found her, but she proves not she, though very like her.

(16)

1667 Pepys, Diary VIII 446.5 (23 Sep) the glass was so clear that she thought it had been open, and so run [: D.D.] her head through the glass and cut all her forehead.

Many of Visser’s examples of  + past participle fall into the category of unrealised action; see especially his (1963–1973: §§ 2030–2050, 2154–2156, 2188). A few of his types begin with isolated late Old English instances, but most of them did not appear until Middle English. The prescriptive tradition frowns upon some of the patterns with double use of  (e.g. would have liked to have gone), even though each pair of adjacent verbs conforms to the morphosyntax of verbal groups in standard English. However, the tendency to use  may even be strong enough to produce two adjacent instances (ignoring non-verbs) of perfect , a pairing which certainly contravenes what is now a clear rule in the standard language: (17)

a1425 Dial. Reason & A. 9.2 Hadde neuere infirmite haue asailed Job & had never infirmity have assailed Job and Tobye: here holinesse hadde not Šit be fully opened. Tobias their holiness had not yet been fully revealed ‘If infirmity had never assailed Job and Tobias, their holiness would not yet have been fully revealed.’

(18)

(1446) Paston 16.21 … the valew of the heye, yf … the value of the hay if it it [sc. a meadow not cultivated in good time] had well a be dite. had well have been cultivated ‘… the value of the hay, if the meadow had been properly cultivated’

(19)

1837 Dickens, Pickwick xxvi.393 “Well, I raly would not ha’ believed it, unless I had ha’ happened to ha’ been here!” said Mrs. Sanders.

118

DAVID DENISON

(20)

1848 Dickens, Dombey xxxii.445 Little Dombey was my friend at old Blimber’s, and would have been now, if he’d have lived.

(21)

1987 Wolfe, Bonfire xix.409 I wish we hadn’ta moved so fast with the sonofabitch.

The syntagm seen in the last clause of (20) is variously expanded as had have Ved and would have Ved, both by syntacticians and in attested instances, though it is commonest with contracted ’d for the first verb. See Visser (1963–1973: § 2157), Wekker (1987), and also some comments in Denison (1992, 1998: 140–142). Suppose we take the first option and treat the construction as involving double  (certainly correct for (21)). One analysis would treat the first  as modal, since it appears to be followed by an infinitive. Example (22) shows how modal  normally behaves: (22)

Before an X-ray they have to have gone without food for a whole day.

Examples like (21) would therefore be anomalous in lacking an obligation sense and in not requiring to. An alternative analysis of (20)–(21), which I prefer, takes both s as perfect, the first marking anteriority (central use of the perfect) and the second unreality (secondary use): each function is separately realised by a grammaticalised form of . The morphological oddity then consists in the fact that the second auxiliary is an infinitive rather than a past participle despite being in the  perfect, rather as Dutch auxiliaries followed by an infinitive behave when they themselves have a perfect auxiliary (Geerts et al. 1984: 523–525): (23)

Ik had het moeten zien. I had it must( for .) see() ‘I ought to have seen it.’

(24)

Ik ben wezen kijken. I am be( 3 for .) look() ‘I have been to have a look.’

In the English double  pattern we might expect to find forms where the second auxiliary is a past participle: (20′) *If he’d had lived … /*If he’d’d lived (21′) *I wish we hadn’t had moved so fast …

COMBINING ENGLISH AUXILIARIES

119

The usual handbooks do not mention any, but Visser (1946–1956: § 710b) has two somewhat similar examples: (25)

1442 Let.Bekynton II. 213 He might never have had escaped.

(26)

1535 Joye, Apol. Tindale(Arb.) 30 He wold … neuer haue had so farre he would … never have had so far swaruen from his principal, as … swerved from his principle as … ‘He would never have strayed so far from his principle, as …’

But (25)–(26) are clearly most unusual (and (25), with its modern spelling, suspicious). Why so rare? Among the reasons may be: (A) phonetic awkwardness and/or avoidance of repetition (B) residue of historical confusion in ’d of had and would, the latter of which would have collocated with an infinitive, not a past participle (C) possible association of infinitival rather than other forms of  with unreality Factor B could only be relevant for examples from around 1600 onwards, when contracted auxiliaries are first recorded. Factor C might be a consequence of the frequent use of a modal in unreal clauses, especially with the obsolescence of the non-modal type seen in (11). It is convenient to mention here — rather than in the sections on combinations of different auxiliaries — two further pieces of evidence that suggest that infinitival have is undergoing further grammaticalisation. One is word order evidence that suggests that strings like would have or would’ve are increasingly being treated as uninterruptable ‘chunks’. The first example below is among many gathered by Boyland (1998: 3): (27)

What would’ve you done?

(28)

1961 Brown Corpus, Belles Lettres G65:85 a sentiment he would have probably denied

(29)

1992 A. Maupin, Maybe the Moon xv.225 ‘I should’ve never went on a stupid blind date. They never work out.’

Then there is the stressed form of created from the unstressed enclitic ’ve:

120

DAVID DENISON

(30)

1819 Keats, Letters 149 p. 380 (5 Sep.) Had I known of your illness I should not of written in such fiery phrase in my first Letter.

(31)

1992 D. Tartt, Secret History ii.57 If I’d of been the bartender at the Oak Room he wouldn’t have noticed.

In (30), the earliest example I have found, the of spelling for have occurs in the apodosis of a conditional which is otherwise standard; (31) draws attention to double  in a protasis that accompanies a wholly standard apodosis. Many speakers thus apparently fail to see any connection between a noninitial, infinitival occurrence of  in a verbal group and the normal auxiliary. 2.3.2 Perfect  + perfect  There is actually an occasional  perfect of the  perfect (Visser 1963–1973: § 2162)! Rydén and Brorström (1987: 25) find it obvious that “this variant emerged to satisfy a need for stressing the resultative aspect more emphatically than the be + P[ast ]P[articiple] construction was capable of at the time”: (32)

a1400(a1325) Cursor 7074 Bot als þe tan als but as-if the one [sc. half] as be þat toþer | Of al þis werld had risen bene against the other of all this world had risen been ‘but as if one half of this world “had been risen” against the other’

(33)

(modern) She’s been gone a long time.

Visser (1946–1956: § 682) suggests that the -perfect “gradually got the character of” copula  + adjective, “especially when the collocation was not accompanied by verbal adjuncts”. This would then have allowed the normal conjugation of (copula)  to operate. Visser’s suggested reanalysis would be the exact converse of grammaticalisation, and it is simpler to assume that the -perfect never was fully grammaticalised. 2.4 Double progressive? Actually I have no examples of the double progressive, that is, doubling of progressive , but perhaps it is relevant to mention the doubling of -ing, even though the first -ing should not be regarded as progressive, since verbs like ,  which resist the progressive have -ing forms in non-finite clauses.

COMBINING ENGLISH AUXILIARIES

121

The syntagm being + Ving should occur when a finite progressive is turned into a gerundial or present participial construction, as in (34)

1660 Pepys, Diary I 302.21 (26 Nov) … I being now making my new door into the entry, …

The being + Ving pattern had some currency at least from the mid-sixteenth century to the early nineteenth (Denison 1985b, 1993: 394–395, 411 n. 8). Modern grammars claim it to be impossible in present-day English. The gap is an odd one. Consider the following pairs, where a finite clause in the (a) sentence is turned into a non-finite clause in (b) by altering the first verb to an -ing: (35)

a. b.

Jim teaches/taught five new courses. Teaching five new courses makes it easier.

(36)

a. b.

Jim has/had taught these courses before. Having taught these courses before makes it easier.

(37)

a. b.

Jim has/had been teaching these courses for some time. Having been teaching these courses for some time makes it easier.

(38)

a. Jim is/was teaching five new courses. b. *Being teaching five new courses makes it easier.

There is now a systemic gap at (38b), a gap, furthermore, which has actually opened up where previously the paradigm was complete.4 There are some glorious examples which combine being + Ving with passival usage (see § 3.3.2 below): (39)

1676 Prideaux, Letters 50.4 a great deal of mony beeing now expendeing on St. Mary’s …

(40)

1774 Woodforde, Diary I 125.12 (13 Mar) I talked with him pretty home [‘directly’] about matters being so long doing — .

See Denison (1993: 394–395, 440–443) for more detail, where an attempt is made to link the loss of the being + Ving pattern to a reanalysis and grammaticalisation of progressive . 2.5 Double passive? The following examples appear to combine passive  and passive  in a single syntagm:

122

DAVID DENISON

(41)

1736 Butler Anal. i. iii These hopes and fears … cannot be got rid of by great part of the world.

(42)

1810 Syd. Smith Wks. (1850) 183 Nor is this conceit very easily and speedily gotten rid of.

Indeed the construction of (41)–(42) is still wholly grammatical. I assume that these apparent double passives were formed in the following steps: (43)

A neat solution rids us of the problem.

(44)

a. b.

(45)

The problem is got rid of.

We are rid of the problem. We get rid of the problem.

That is, (44a, b) were possible passives of (43), but either could be interpreted as containing a statal AP rid of the problem, with / taken as ergative and the whole clause as active. (It is impossible to assign historical priority between active and passive readings.) Then (44b) — which would also have had an overall dynamic meaning in the active reading and even perhaps an agentive role for we — was open to reanalysis as containing a group-verb  rid of. This is lexicalisation but not grammaticalisation. Only then could a new prepositional passive be formed from it, in the same way as from  care of or  paid to. In this way, and because the semantic role of the problem is Patient, we can explain how (45) is possible but the double passives (46)–(48) are not: (46) *The problem is been rid of.

(⇐ (44a))

(47) *A ride is got taken for.

(⇐ He gets taken for a ride.)

(48) *Free tickets were got given us.

(⇐ We got given free tickets.)

If this account is correct it suggests that the apparent -passive (44b) was not (exclusively) perceived as the passive turn of (43).5 Compare the dual reading of has been gone, with  as auxiliary or as copula (§ 2.3.2 above).

COMBINING ENGLISH AUXILIARIES

3.

123

Other combinations of auxiliaries

3.1 Auxiliary +  In most accounts of standard present-day English, the dummy auxiliary  is presented as incompatible with any other auxiliary verb, though there may be mention of its occurrence with marginal auxiliaries, as for instance in didn’t used to V. We now look at some constructions, older and newer, where  appears to collocate with a central auxiliary. First we look at instances where  precedes . Then we investigate cases where  follows another verb, where it will necessarily be untensed. 3.1.1 Tensed  +  The dummy auxiliary  does not co-occur with  except in imperatives, and imperative do and don’t are anyway arguably different from the usual dummy auxiliary. Sporadically, however, one does find the passive auxiliary  — in the infinitive, be — after tensed , usually interrupted by other material: (49)

1713 J. Swift, L-19, Vol. II, Letter LXI, p. 634 Parvisol has sent me a Bill of 50ll. as I orderd him; wch I hope will serve me & bring me over, pray Gd […]6 does not be delayd for it; but I have had very little from him this long time.

(50)

1998 letter, Oldham Evening Chronicle p. 4 (23 Nov.) Did littering the streets not once be considered breaking the law?

(51)

1998 Oldham Evening Chronicle p. 4 (16 Dec.) Does everyone get cards from others and be obliged to say, “No idea who they are”?

(52)

1995 att. D.D., Gerald Hammond We agree that particular students do be flogged.

One might include here examples like: (53)

1865 Arnold, L-33 pp. 258–9 I do not feel quite certain that little Tom will not be more reconciled to school by the end of the week. If he does not, however, I suppose you cannot come to Italy.

involving substitute  rather than periphrastic auxiliary . Mistakes or anacolutha such examples may be, but they seem to suggest the beginnings of an

124

DAVID DENISON

invariant passive auxiliary closely tied to its participle. Related to this is a combination of  with lexical : (54)

If you don’t be careful, you’ll …

(55)

1888 A. T. Ritchie, L-67, letter, p. 207 I read your letter to-day and I could have cried to think you sometimes feel so far away, but one thing you need never feel, that you don’t live and talk and be here just as much as if you were.

(56)

1995 letter, Oldham Evening Chronicle 26/6 (24 Jan.) If the taxi driver …was having a dig at me … why didn’t he stop and be a witness?

which may not yet be generally acceptable but which occur in speech with reasonable frequency. The quasi-imperative Why don’t you be … ? construction is now fully accepted in standard: (57)

1920 Wharton, Age of Innocence xiv.118 “Who’s ‘they’? Why don’t you all get together and be ‘they’ yourselves?”

The typical syntactic contexts for tensed  +  are non-assertive, frequently both negative and either interrogative, quasi-imperative, or conditional. If  is lexical, it usually forms a nonstative group-verb with its complement. 3.1.2 Other auxiliaries + untensed  Periphrastic  in present-day English has no untensed forms, but in the period when the periphrasis was being grammaticalised there were occasional exceptions. Here are some reasonably likely examples of infinitive and past participle periphrastic  prior to 1500: (58)

(c1300) Havelok 1747 He … bad him … Hauelok wel yemen and he … bade him … Havelok well look-after and his wif, | And wel do wayten al the nith his wife and well “do” watch-over all the night ‘He … asked him … to look after Havelok and his wife well and to guard them well all night.’

(59)

a1400(a1325) Cursor 2818 Þe angls badd loth do him flee. the angels bade Lot “do” him flee

COMBINING ENGLISH AUXILIARIES

125

(60)

(c1395) Chaucer CT.Sq. V.45 He leet the feeste of his nativitee | Doon he had the feast of his nativity “do” cryen thurghout Sarray his citee. cry throughout Tzarev his city ‘He had the feast of his birthday announced throughout his city of Tzarev.’

(61)

?c1425(?c1400) Loll.Serm. 2.592 … þat resenable men … schul þanne do … that rational men … shall then “do” make hem redy aŠen þe comynge of þe Lord. make themselves ready for the coming of the Lord

(62)

(?1456) Paston 558.12 The parson wyth yow shall do well the person with you shall do() well sort my maister evidenses sort my master’s pieces-of-evidence ‘The person with you will certainly sort my master’s evidence for him’

(63)

?a1475 Ludus C. 283.339 and þis Še knowe now All and haue don and this you know now all and have “done” here | þat it stant in þe lond of galelye. hear that it stands in the land of Galilee

(64)

a1500 Partenay 2367 behold | ho shall doo gouerne And rule this contre behold who shall “do” govern and rule this country

See further Visser (1963–1973: §§ 1414a, 2022, 2133). (Examples (58)–(60) involve infinitival  as complement to a causative or other non-auxiliary verb.) This material provides some weak evidence for differentiating phases of grammaticalisation of  (Denison 1985a: § 4.6), as examples like (59) and (64) seem to show uses of  + infinitive which are contextually like the periphrasis in excluding an intermediary (subject of  = subject of lexical verb), and yet which precede the restriction to tensed forms which periphrastic  and the modals later underwent in standard English. The past participle construction gained some vogue in sixteenth-century Scottish poetry, and we find examples like (65) as well as examples of the present participle and infinitive of periphrastic :

126

DAVID DENISON

(65)

1568(1500–20) Dunbar Poems 55.13 Thow that hes lang done you that have long “done” Venus lawis teiche … Venus’s laws teach …

Some Scots dialects even now allow untensed forms of modal verbs like  (§ 2.2 above), and it seems likely that the untensed use of  must be tied in with this. If some Scots dialects have been able to use periphrastic  with fewer restrictions than other varieties of English, it seems almost paradoxical that “ … the Scotch language used periphrastic do much more sparingly than the dialects South of the Humber even in the 16th and 17th centuries” (Ellegård 1953: 46, and cf. 164, 200n., 207n.). But perhaps there is no paradox: if these are the same dialects which allow untensed modals, then they rely less on the operator~non-operator distinction which goes hand-in-glove with periphrastic . Arguably they show a less advanced stage of grammaticalisation than that reached by  in standard Modern English. 3.2 Perfect  + passive  Visser claims that perfect + passive  can be traced back to Old English (1963–1973: § 2161), citing the following examples: (66)

Lk(WSCp) 12.50 Ic hæbbe on fulluhte beon gefullod. I have in baptism been/(to) be baptised Lat.: ‘baptisma autem habeo baptizari’

(67)

a1225(?a1200) Trin.Hom. 59.14 feren it is þat we and ure heldrene afar it is that we and our ancestors habbæð ben turnd fro him. have been turned from him ‘It is long since that we and our ancestors have been turned away from him.’

But Mitchell (1985: § 753 n. 188) rejects (66) as containing infinitival, not participial, beon, and (67) as being Middle English. And indeed Visser has a respectable collection of examples from early Middle English onwards:

COMBINING ENGLISH AUXILIARIES

(68)

c1180 Orm. 18232 & forr ðatt Crist ær haffde ben and because Christ earlier had been Fullhtnedd att teŠŠre maŠŠstre baptised by their master

(69)

c1230(?a1200) Ancr. 86b.28 … Šef ich hefde ibeon akeast wið strengðe. … if I had been overthrown by force

127

The perfect auxiliary is always , which is interesting.7 Three reasons suggest themselves: (A) passive  was not yet grammaticalised (B) the -perfect was already recessive (C) some restriction on double  Of these, A seems most cogent for early Middle English, giving a terminus a quo for the grammaticalisation of the passive. There is no independent evidence that passive  was grammaticalised at that time. Suppose to the contrary that it already had been. A syntagm consisting of grammaticalised passive  + past participle would arguably have been a mutative intransitive, precisely the sort of syntagm liable to form its perfect with . Since that did not happen, the likelihood is that passive  was still an ungrammaticalised main verb. By the time it did become grammaticalised, the productive way to form a perfect was with , perfect  being by then recessive — factor B. Further evidence that passive  was indistinguishable from main verb  in early English is that main verb  has likewise never formed a perfect with auxiliary , but rather — since very late Old English — always with . Instead of  + been + past participle or indeed the non-occurring * + been + past participle, one often finds just  + past participle: (70)

(a1387) Trev.Higd. I 1.xxiv.235.20 And whan þe ymage was made, and when the image was made hem semede þat þe legges were to feble … them() seemed that the legs were too weak …

(71)

1623(1606) Shakespeare, Mac IV.iii.204 Your Castle is surpriz’d; your Wife, and Babes | Sauagely slaughter’d:

(72)

1853 Dickens, Bleak House lv.814 That the visitors … have been here this morning to make money of it. And that the money is made, or making.

128

DAVID DENISON

For discussion, references and further examples see Visser (1963–1973: § 1909), Denison (1998: 183–184). It is possible that the increased tendency in recent years to use  + been + past participle (had been made, has been surprised) for the notion of current relevance in such sentences may follow from the loss of the  perfect as a simple tense equivalent (= stage C in the grammaticalisation of the perfect in § 1 above). The last clause of example (72) actually contrasts two forms which would both now be inappropriate here, the non-perfect passive and the passival (§ 3.3.2 below); in present-day English it might read And that the money has been, or is being, made. And in example (73) it is unclear whether were … made should be modernised as present-day English had been made or were being made: (73)

1848 Cottle, Reminiscences of … Coleridge and … Southey p. 434 Independently of which, an idea had become prevalent amongst the crowd of afflicted, that they were merely made the subjects of experiment, which thinned the ranks of the old applicants, and intimidated new.

I have noticed one example of perfect + passive , (74) — Visser has some Old English examples too in his § 2166 — and here the perfect auxiliary is indeed , as it is with all main-verb uses of  too, e.g. (75)–(76): (74)

c1180 Orm. 19559 Šæn himm, þatt wass att Sannt Johan Bapptisste against him that was at Saint John Baptist wurrþenn fullhtnedd become(.) baptised ‘against him that had been baptised by John the Baptist’

(75)

c1180 Orm. 3914 Annd Godess enngless wærenn þa Well swiþe glade and God’s angels were then well very glad wurrþenn | Off þatt, tatt Godd wass wurrþenn mann become(.) concerning that that God was become man ‘and the angels of God had then become very glad of the fact that God had become man’

(76)

c1180 Orm. 2272 Forr þatt nass næfrær wurrþenn, for that not-was never-before happened ‘For it had never happened before that …’

COMBINING ENGLISH AUXILIARIES

129

There is one context in present-day English, the “strange existential” of (77) mentioned by Lakoff (1987: 562–565), where the auxiliary before a passive seems to hover between  and : (77)

There’s a man been shot.

(78)

A man has been shot.

(79)

a. b. c.

There’s a man in the garden. There’s some men in the garden. There’s been an accident.

In Lakoff’s analysis, ’s in (77) is a contraction of perfect has (cf. (78)), not is, but cannot be used in uncontracted form — a “rational property” which depends on phonological identity with its “ancestor” element, the ’s = copula is of normal existentials like (79a). The invariant form there’s appears to have been grammaticalised, a claim corroborated by its well-known colloquial use with plural NPs, as in (79b). However, the strange existential also provides evidence for the status of the second auxiliary. In normal existentials, all uses of  behave alike, auxiliary and non-auxiliary, and has been would be treated as a form of  around which the true subject could be moved under there-insertion, as in (79c). Pattern (77) rather suggests that the  of been shot has been grammaticalised as an auxiliary of . 3.3 Progressive +  In this section I consider combinations of the progressive,  + Ving, with a second use of , the most important of which is the combination of progressive  and passive  (§ 3.3.1). In order to put its appearance in context we need to mention alternatives (§ 3.3.2), precursors (§ 3.3.3), and analogues (§ 3.3.4), before considering an analysis (§ 3.3.5). 3.3.1 Progressive  + passive  According to Mossé (1938: §§ 263–264) and Visser (1963–1973: § 2158), finite progressive passive constructions only began to be used in the late eighteenth century. Mossé and Visser show that progressive passives were at first stigmatised in print and heavily condemned. To Visser’s 28 examples prior to 1872, we can add quite a few more, all of them three-verb syntagms of the type, tensed form of  (+ …) + being + past participle, for example:8

130

DAVID DENISON

(80)

1772 J. Harris, in Lett. 1st Ld. Malmesbury (1870) I.264 I have received the speech and address of the House of Lords; probably, that of the House of Commons was being debated when the post went out.

(81)

1779 Mrs. Harris in ibid. I.430 The inhabitants of Plymouth are under arms, and everything is being done that can be.

(82)

1801 tr. Gabrielli’s Myst. Husb. I.125 ‘It [sc. a bill] is being made out, I am informed, Sir.’

(83)

1829 Landor Imag. Conv., Odysseus, etc., While the goats are being milked, and such other refreshments are preparing for us as the place affords.

Since then the construction has become generally acceptable. For an account of its spread via a social network see now Pratt and Denison (2000). 3.3.2 Passival  + Ving (= ‘with passive sense’) Until the progressive passive (84)

The house is being built.

entered the language, it was necessary either to do without explicit progressive marking, as in (85) and the last clause of (86): (85)

1662 Pepys, Diary III 51.25 (24 Mar) I went to see if any play was acted9

(86)

1838–9 Dickens, Nickleby v.52 he found that the coach had sunk greatly on one side, though it was still dragged forward by the horses;

or to do without explicit passive marking, as in the curious construction of (87): (87)

The house is building.

This is not formally a passive (there is no  + past participle), but its subject NP is the argument which would be subject in a true passive and object in a normal active, which is why Strang (1982: 441) calls (87) a “covert passive” and Visser (1963–1973: §§ 1872–1881) calls it “passival”. Passival (87) seems to have fulfilled the function of the missing (84); see Denison (1993: 389–391) for details of its early history. Note also (39)–(40) above. The alternatives remained in use even after the progressive passive began to be possible. Visser (1963–73: §§ 1879–81) suggests that the retreat of the passival in the

COMBINING ENGLISH AUXILIARIES

131

face of the advancing progressive passive did not begin until the twentieth century — though Nakamura’s (1991: 126–129) statistics on usage in diaries and letters show a steep decline from mid-nineteenth century. I shall suggest that the replacing construction, progressive + passive, is evidence of the grammaticalisation of the progressive. 3.3.3 Precursors of progressive  + passive  A precursor of the progressive + passive construction involved the participial or gerundial phrase being + past participle used absolutely10 or separated from a tensed . The gerundial pattern appeared in the fifteenth century; for discussion see Denison (1993: 431–433): (88)

(1456) Paston 562.4 … we being enformed þat the matiere is … we being informed that the affair is pitevous praie you hertly þat ye wul … lamentable pray you earnestly that you will …

(89)

1769 Mrs. Harris in Lett. 1st Ld. Malmesbury (1870) I.180 There is a good opera of Pugniani’s now being acted

(90)

1779 J. Harris in ibid. (1870) I.410 Sir Guy Carlton was four hours being examined.

(91)

1798 Woodforde, Diary V 137.19 (14 Sep) … that the French … had been defeated, and that the Irish were in a fair Way, of being made quiet.

Sentences like (89) cannot be confidently separated from the progressive passive, given that (89) looks like a normal there-transform of (89′)

A good opera of Pugniani’s is now being acted.

with only a light adverb interrupting the verbal syntagm. These non-finite constructions probably do not contain a combination of auxiliaries. While Visser regards none of them as true progressive passives, Nehls (1974: 158 n.149) singles out (90) as the first certain example of the construction; see also my comment on (89). I have found a seventeenth-century example which looks exactly like the progressive passive: (92)

1667 Pepys, Diary VIII 249.28 (3 Jun) … thinking to see some cockfighting, but it was just being done; and therefore back again …

132

DAVID DENISON

I believe, however, that (92) probably belongs with the (88) type as another kind of precursor of the progressive passive, because being done seems to mean ‘just finished’ or ‘becoming finished’ for Pepys, like the non-finite (93): (93)

1667 Pepys, Diary VIII 250.30 (4 Jun) and that being done, …

For clear evidence that being could mean ‘becoming’ in Pepys see Denison (1993: 433). I find one of Mossé’s (1938: § 262) alleged near-progressives, not as it happens repeated by Visser, particularly interesting: (94)

1766 Goldsmith, Vicar xxv.141 I … immediately complied with the demand, though the little money I had was very near being all exhausted.

The all confirms that semantically, (94) is no progressive. The analysis must be something like: (95)

[VP was [AP [DEG very near ] [AP being all exhausted ]]]

Compare my discussion of is being wicked in § 3.3.4 below, where I recognise a non-progressive structure like (95) for some early examples. There are also prepositional patterns which seem to resemble the progressive passive in the same way that  + P + Ving resembles the ordinary progressive: (96)

1669 Pepys, Diary IX 475.1 (8 Mar) He tells me that Mr. Sheply is upon being turned away from my Lord’s family, and another sent down.

3.3.4 Progressive of  The construction  + lexical being is interesting for the light it throws on the relation between the progressive and stative verbs — here the archetypal stative verb,  itself. It may also be relevant to the history of the progressive passive, which begins with an identical sequence of verb forms. For possible early examples see Denison (1993: 395–396). For Mossé (1938: § 266) and Visser (1963–73: §§ 1834–1835) the first late Modern English example is: (97)

1761 Johnston, Chrysal II 1.x.65 but this is being wicked, for wickedness sake.

They ignore the fact that (97) and examples from Fanny Burney and Jane Austen over the next sixty years11 do not appear to contain a progressive verbal group

COMBINING ENGLISH AUXILIARIES

133

is being at all: rather the verb is just equative is, which links an inanimate pronoun subject (it, this, there) to a gerundial phrase being AP. The surface subject is not an argument of the AP. The pattern may have helped prepare the ground for the introduction of a progressive of , but it is difficult to think of a sentence like (97) which could actually have been reanalysed as a true progressive, since the function of the subject NP would have to change so radically. The first modern-looking example in Visser’s collection is Jespersen’s first (1909–1949: IV 225): (98)

1819 Keats, Letters 137 357.4 (11 Jul) You will be glad to hear … how diligent I have been, and am being.

Here I is underlyingly an argument of being diligent. Visser explicitly (1963–1973: 2426 n.1) — but, I would say, wrongly — accuses Jespersen of getting the date of introduction too late. For examples with NP rather than AP as complement, the one late-seventeenth-century example, (99), is better analysed as a non-progressive, just like (97). For good examples we must wait until the nineteenth, (100): (99)

1697 Vanbrugh, Provok’d Wife III.i.198 That’s being a spunger, sir, which is scarce honest:

(100) 1834 R. H. Froude Rem. (1838) I. 378 I really think this illness is being a good thing for me. Mossé and Visser do not distinguish two possible structures — non-progressive (99) and progressive (100) — for  + being + AP/NP. Nor do they do so for  + being + .. Mossé (1938: § 266) merely observes that they are analogous constructions which appeared at about the same time, but that the former remained rare until the end of the nineteenth century. Visser (1963–1973: §§ 1834–1835, 2158), however, who claims a much earlier date, suggests that  + being + AP may have been another subsidiary cause of the use of the progressive passive. Note also this apparent example with PP as adjunct or complement: (101) c1515 Rastell, Interlude 376 Yet the eclyps generally is alwaye | In the hole worlde as [sc. at] one tyme beynge;

134

DAVID DENISON

3.3.5 Reanalysis of progressive  I take it that progressive , like other auxiliaries, is developed out of a lexical verb by reanalysis. However, of all the auxiliaries, progressive  is the one where the semantic difference between a full-verb use and auxiliary use is least perceptible, giving us wide latitude in dating a reanalysis. I hypothesise that it occurred comparatively late. Incidentally, in my conception of syntax there is no need to assume unique, black-and-white analyses everywhere. A recently dominant but now less salient analysis can still play a part in the behaviour of a construction, and not only in non-productive relics. Compare the concept of ‘persistence’ in grammaticalization theory (Hopper 1991). If there has been a reanalysis of the progressive, what are the consequences of locating (the most rapid phase of) the changeover in the late Modern English period? I have sketched out a scenario in previous publications (e.g. Denison 1993: 441–443; 1998: 155–157) and will be briefer here. The crucial points are that before the reanalysis a putative progressive passive: (102) *The house was being built. would have had to be analysed as containing the progressive of , but it could not have been supported by pattern (103), progressive  + predicative adjective, since that was not in use before the nineteenth century: (103) Jim was being stupid. Here I follow Jespersen (1909–1949: IV 225) and Strang (1970: 99) against Visser (1963–1973: § 2158); see § 3.3.4 above. And the semantics of syntagms like being built would not generally have been durative: see the discussion of (92)–(93) above. Hence the semantic and syntactic oddity of the progressive passive would explain its non-appearance until near the end of the eighteenth century and the fierceness of some people’s reactions to it when it did finally begin to appear in print in the nineteenth century. After the reanalysis, the progressive passive (102) became possible for those speakers with the new grammar, since it was the progressive not of passive  but of the lexical verb. Warner (1986: 164–165) has also proposed a reanalysis of constructions involving tensed forms of , giving 1700 and 1850 as extreme limits. All uses of  belong together in Warner’s account. In subsequent publications (1993, 1995, 1997) he has constructed an explanation of how auxiliary verbs came to differ from full verbs by having a series of forms with independent syntactic properties, rather than belonging to a paradigm with a single subcategorisation. In this explanation the loss of the being + Ving pattern, (34), is another symptom of the same change.

COMBINING ENGLISH AUXILIARIES

135

3.4 Combinations involving passive  We take up a recent addition to the roster of possible English auxiliaries, one that is not fully grammaticalised even now: . For a valuable recent study see Gronemeyer (1999). 3.4.1 First appearance of passive  Most authorities follow OED in giving the mid-seventeenth-century (104) as the first recorded passive with : (104) 1652 Gaule Magastrom. 361 A certain Spanish pretending Alchymist … got acquainted with foure rich Spanish merchants. Strang cautiously — and rightly — describes acquainted as a “predicative which could be taken as a participle” (1970: 150–151). A better example is: (105) 1693 Powell, A very good wife II.i p. 10 [ARCHER] I am resolv’d to get introduced to Mrs. Annabella; There is then something of a gap. In Jespersen’s collection (1909–1949: IV 108–109) the next examples chronologically are: (106) 1731 Fielding, Letter Writers II.ix.20 so you may not only save your life, but get rewarded for your roguery (107) 1759 Sterne, Tristram Shandy III.ii.126 he should by no means have suffered his right hand to have got engaged (108) 1766 Goldsmith, Vicar xvii.90 where they give good advice to young nymphs and swains to get married as fast as they can. For some reason Visser’s collection (1963–1973: § 1893) misses (106)–(108) and continues with OED’s next examples, dated around 1800: (109) 1793 Smeaton, Edystone L. § 266 We had got (as we thought) compleatly moored upon the 13th of May. (110) 1814 D. H. O’Brien, Captiv. & Escape 113 I got supplied with bread, cheese and a pint of wine.

136

DAVID DENISON

Strang, too, seems to be unaware of (106) (and indeed (105)) when she writes that “unmistakably passive structures are not found till late in the 18c” (1970: 151). 3.4.2 Modal and/or  / + passive  Soon after passive  entered the language it began to occur with preceding auxiliary verbs, including modals and , as in (106) and: (111) 1816 ‘Quiz’ Grand Master viii.213 Or else they wou’d Get most confoundedly bamboo’d. (112) 1819 Southey Lett. (1856) III.150 I shall get plentifully bespattered with abuse. (113) 1901 Shaw, Cæsar and Cleopatra II 272b CÆSAR. No man goes to battle to be killed. — CLEOPATRA. But they do get killed. (114) 1989 Gurganus, Confederate Widow III.i.2 328 If I do get killed, I’ll only be dead. When preceded by  it seems reasonable to speak of perfect  + passive , as in (107), (109) and: (115) 1950- Survey of English Usage N2 If they don’t offer it this time, I won’t drag it away once somebody mentioned it but it hasn’t got mentioned very much. (116) 1989 Gurganus, Confederate Widow II.i.2 164 … he settled near his company’s bonfire. It’d got built one mile from the meadow where … When preceded by  there is room for doubt, as we shall see, as to the status of  (perfect or passive auxiliary?) and/or  (passive auxiliary or causative?): (117) 1837 Carlyle Fr. Rev. I VII.x.281 the first sky-lambent blaze of Insurrection is got damped down; (118) 1837 Carlyle Fr. Rev. I III.ii.69 An expedient … has been propounded; and … has been got adopted (119) 1870 Alford in Life (1873) 457 I only hope the Master’s work may be got done by bedtime. (120) 1662 J. Davies, Olearius’ Voy. Ambass. 220 They were both gotten sufficiently Drunk.

COMBINING ENGLISH AUXILIARIES

137

(121) 1701 W. Wotton, Hist. Rome, Alexander, iii. 510 Maximus was got as far as Ravenna. (122) 1888 Berksh. Gloss. s.v. Veatish I be got rid o’ the doctor, an’ be a-veelin’ quite veatish [‘fairly well in health’] like now. Examples (117)–(118), (120)–(121) come from Haegeman (1985: 55–56, 71). In my opinion it is highly doubtful whether any of (117)–(122) contain  as passive auxiliary. If Carlyle is not to be charged with using a double perfect,  in (118) should mark passive, not perfect, but in fact from a present-day English point of view (117)–(119) look like passives of the pattern  + NP + ., and Jespersen (1909–1949: V 16, 36) seems to agree; he asserts that they correspond to the mainly American type (I give a later example): (123) 1945 Coast to Coast 1944 103 Well, he’s got me beat. In that case the  of (117)–(119) and (123) is no passive auxiliary. Nor is the  of (120)–(121) either, since there is no lexical past participle. For the sake of completeness we may note that the sequence Modal +  +  + . is attested too, though my example is recent: (124) 1950- Survey of English Usage T1 If you, in fact, cleared that cupboard out to put offprints in it, it might have got cleared out then. 3.4.3 Progressive  + passive  This is a predictable combination, though Visser (1963–1973: § 2160), perhaps surprisingly, has no examples before the very end of the nineteenth century:12 (125) 1819 Scott Let. in Lockhart (1837) IV.viii.253 My stomach is now getting confirmed, and I have great hopes the bout is over. (126) 1837 Carlyle Fr. Rev. I VII.viii.268 One learns also that the royal Carriages are getting yoked (127) 1837 Dickens, Pickwick xxxii.479 Extraordinary place that city. We know a most astonishing number of men who always are getting disappointed there. The gerundial use of passive  is earlier still:

138

DAVID DENISON

(128) 1776 G. Semple Building in Water 46 Our Coffer-dam … which we began to despair of ever getting made even tolerably stanch [‘water-tight’], However, at present I have no examples of progressive  + passive  preceded by another auxiliary, whether modal or perfect , though they are clearly grammatical in present-day English. Examples like (129) 1931 — Big Money xiii.309 … even if he had been getting steadily plastered [‘drunk’] all the afternoon. are not convincingly passive.

4.

Multiple auxiliary combinations

We are now in a position to look for generalisations about how certain combinations of auxiliaries came about, and when. 4.1 Two auxiliaries A modal can be followed by an auxiliary of the perfect, the progressive or the passive, and this has been the case since Old English for all such pairs except perhaps modal + perfect , which is said to date from the fourteenth century but has a couple of possible Old English examples. Perfect  can be followed by an auxiliary of the progressive (from a1325) or the passive (c1180), and perfect  is followed by passive  (Old English). For perfect  + passive  I have examples from 1759/1793 (§§ 3.4.1–2). Progressive  can be followed by an auxiliary of the passive. Tensed progressive  + passive  is found from 1772 (§ 3.3.1), tensed progressive  + passive  is found from 1819 (§ 3.4.3) — both rather earlier when the first verb is untensed . 4.2 Three or four auxiliaries If we treat  as the only significant auxiliary of the progressive, the following four-verb combinations should be possible, with dates of their earliest occurrence where known:

COMBINING ENGLISH AUXILIARIES

(A) (B) (C) (D)

modal + perfect  + progressive  + V: modal + perfect  + passive  + V: perfect  + progressive  + passive  + V: modal + progressive  + passive  + V:

139

?a1425 c1300 1886/1929 1915

Patterns A to C require a past participle of . Patterns C and D combine progressive  and passive  in a single syntagm. Including passive  brings the following additional possibilities, all grammatical now, though data on first occurrences are not readily available:13 (E) modal + perfect  + passive  + V: (F) modal + progressive  + passive  + V: (G) perfect  + progressive  + passive  + V:

1950- (§ 3.4.2) present-day English present-day English

The following table arranges the information given above so that dates of first occurrence of three-auxiliary (four-verb) patterns can be compared with the dates of first occurrence of each adjacent pair of auxiliaries they contain.14 To clarify what is being tabulated, the first line claims that pattern A (modal + perfect + progressive might have been singing) is found from ?a1425, whereas the adjacent pairings that make it up (modal + perfect might have sung and perfect + progressive has been singing) are found from Old English and a1325, respectively. Table 1. Earliest combinations of auxiliaries pattern

first pair

second pair

three auxiliaries

A B C D E F G

Old English Old English a1325 Old English Old English Old English a1325

a1325 c1180 1772 1772 1832 1819 1819

?a1425 c1300 1886/1929 1915 1950– present-day English present-day English

I have not discussed or tabulated the maximal, four-auxiliary (five-verb) sequences, on the assumption that nothing of great significance will be lost by the omission. 4.3 The process of combining Interestingly, Table 1 shows that it is always the first pairing which occurs earliest, then the second pairing, and finally the three-auxiliary pattern. The table

140

DAVID DENISON

would appear to support the conclusion that auxiliaries are added on at the left, at the tensed end of the verbal group, in a development like (130): (130) has been being sung

←

is(was) being sung

In this hypothesised development, there is an easily motivated substitution of perfect has been for simple tensed is/was, and after only a modest time-lag. That is a much more satisfactory hypothesis than an imaginable (131): (131) has been being sung

←

has been sung

with the rather opaque and very long-delayed substitution of progressive participle been being for simple past participle been. However, the evidence for some of the dates is too skimpy to justify any weighty conclusions. One might compare also (132) (=(77))

There’s a man been shot.

There too we seem to have a grammaticalised item, ’s or rather there’s, added on at the left of a pre-existing been shot syntagm, though the process is a rather more complex one involving blending. Kossuth (1982: 291) presents a theory “that the order of appearance in cooccurrences parallels that of the original auxiliarization, but with a lag of a good century”. That last figure seems about right, to judge from my Table 1. However, I assume that passive  was grammaticalised before progressive  (see below), which is not her assumption. In Kossuth’s view, finite clauses in English have always been subject to what she calls a Once-per-Clause Constraint (Kossuth 1982: 290). This states that each optional auxiliary can appear at most once, but the basis of the rule has undergone a significant change in the last two hundred years. Formerly it had to be stated in terms of lexical items like , latterly in terms of grammatical categories like Progressive. I have given a detailed critique in Denison (1993: 454–455). The crucial dating problem is the progressive passive, as in the present-day English example: (133) Max was being serenaded. From a present-day perspective its introduction in the late eighteenth century is completely unmysterious, representing as it does the syntactic combination of two long-established periphrases in a semantically compositional way. The question then arises, why it took so long to appear, and why its early use met with such a torrent of abuse. In the light of the discussion of how auxiliaries combine, we can say that the combination of progressive and passive had to

COMBINING ENGLISH AUXILIARIES

141

await the full grammaticalisation of the progressive. In § 3.3.5 above I suggested that the reanalysis which produced the fully grammaticalised progressive did not take place until around the late eighteenth century.15

5.

Conclusion

We have returned to the question of grammaticalisation of individual auxiliaries. The process of grammaticalisation of an originally lexical verb — which is a matter of both semantics and syntax — can be long-drawn-out and hard to assign dates to. In semantics grammaticalisation probably involves generalisation and perhaps bleaching of meaning (but cf. Brinton 1988), while in syntax the (pre-)auxiliary changes from being head of its phrase to a modifier of the lexical head. (The latter characterisation will not apply to abstract formal analyses which stack present-day English auxiliaries, like catenatives, in a nest of left-headed phrases, so that apart from the first, tensed verb, each verb is part of the complement of the one preceding.) In the course of this paper on combinations of auxiliaries I have given specific pieces of evidence for certain datings. Summing up, I suppose that the auxiliaries were grammaticalised in the following order:16 Table 2. Dates of grammaticalisation auxiliary verb

grammaticalisation

modals and  perfect  periphrastic  passive  progressive  passive 

already in Old English already in Old English fourteenth-fifteenth centuries fourteenth-eighteenth centuries? late eighteenth century? twentieth century and continuing

I have concentrated on the central auxiliaries (plus some brief observations on ), but the many verbs which are, or have been, marginal to the auxiliary system would also repay investigation from this point of view, as with syntagms like gonna go, imperative don’t let’s V/let’s don’t V, and so on. I discuss a number of marginal auxiliaries in Denison (1998). The history of all verbal periphrases in English is a much larger topic than can be dealt with in a single paper. The pathways of development of each periphrasis and the relationships

142

DAVID DENISON

between periphrasis and simple form are intricate matters, some of which are gone into in Denison (1993). Even where these matters are well understood it can be difficult to decide where on the scale from full verb to auxiliary a particular example falls — in other words, to pin down the degree of grammaticalisation involved. All I have attempted here is to gather one particular sort of evidence, in the belief that it may shed light on the processes of grammaticalisation.

Acknowledgment To construct a coherent account I have built on material scattered over six long chapters of Denison (1993), where a full list of primary sources can be found. Early versions of some parts were presented in Helsinki, Amsterdam and Durham in 1990–91; the paper then fell victim to the vicissitudes of publishing. The present paper adds some new data and analysis and benefits from revisions suggested by the editors.

Notes 1.

Visser observes that with three-verb clusters of modal + modal + V, the first modal is almost always  (1963–73: § 2134). The only exceptions Visser gives for the Middle English and early Modern English periods are the following: a.

b.

c.

a1400 Lanfranc 17.2 Also he muste kunne evacuener him þat is also he must know-how-to free him that is ful of yuel humouris. full of evil humours (c1443) Pecock Rule 375.2 infantis mowe receive … þi sacrament of infants may receive … your sacrament of baptym eer þei mowe kunne worschipe þee. baptism before they may know-how-to worship you c1454 Pecock Fol. 129.5 if y se my neiŠbour goyng … forto drenche if I see my neighbour going … to drown him silf, y oughte … forto wille defende him fro drenching … him self I ought … to wish prevent him from drowning …

References are given to may can V, etc., in current American English dialects. Elsewhere, however (1963–73: § 1357), he notes infinitival modal , as seen for instance in: d.

1871 Macduff Mem. Patmos xi. 153 We cannot dare read the times and seasons of prophecy.

COMBINING ENGLISH AUXILIARIES

143

Ogura (1998: 232–3) cites wolde mot in one MS of Cursor Mundi, though mot seems to me an obvious error for not, the negative particle used in that MS., and (1434) Misyn ML may will with root will. 2.

The context of (13) makes clear that a definite order for the wool had not yet been placed.

3.

Wezen is a special infinitive form — differing from the normal infinitive zijn — used colloquially to replace the past participle geweest in this construction (Geerts et al. 1984: 578).

4.

The judgements remain the same if the subject NP Jim is retained in the non-finite versions. A similar exercise with passive examples reveals the possibly unexpected absence of being + being + Ving: a. b.

The courses are/were being taught by the same tutor. *Being being taught by the same tutor makes it easier.

This is of less interest, however, since here the (a) pattern is relatively new. 5.

It is perhaps not surprising that OED should be inconsistent in its analysis of what are surely parallel constructions, the idioms  quit of and  rid of, which provide many of the possible early examples of passive  (to judge from a computer search of OED citations). It defines the former s.v. quit, quite a., therefore not as a -passive, but the latter s.v. transitive rid v. 3d.

6.

The editor marks [….] as “quite undecipherable” but notes that previous editors of the journal had read “MD” [= Swift’s monogram for Stella, and sometimes Stella and Dingley]. I am grateful to Dr Fujio Nakamura (p.c.) for examples (49), (53) and (55) and the information on (49).

7.

Here English is like, say, French and unlike, say, Italian or Dutch.

8.

Further examples up to c1830 in OED2 (found by means of an early test release of the CD-ROM version) are dated 1826 s.v. new a. A.5b, a1834 s.v. preconception, 1828 s.v. ring v.2 B.6a.

9.

The sense of (85) is ‘… if any play was being acted (later that evening)/was to be acted’, so that it would not be a substitute for the central sense of the progressive.

10.

Non-finite being + passive participle should not be called ‘progressive’, since verbs like ,  which resist the progressive have non-finite -ing forms.

11.

It is worth pointing out that the two examples from Jane Austen — Mossé cites Pride & Prejudice II.iii[xxvi].144 and Visser Emma II.xiv[xxxii].280 — are dialogue by Eliza Bennet and Mr Woodhouse, respectively. It is highly unlikely that Austen, even with her general predilection for the progressive, would have put such a novel construction into the mouths of ‘careful’ speakers, especially the fussy, old, prim Mr Woodhouse. Phillipps (1970: 117) cites an example outside dialogue: a.

1816 Austen, Emma III.xv[li].444 She was so happy herself, that there was no being severe.

By such gerundial usage, he suggests, “Jane Austen does approach the modern construction”. And OED2 has an example from 1679 s.v. idiotical a. 1. 12.

Note that his earliest examples of progressives of  and  + AP ( old,  impolite, etc.) are also from the turn of the twentieth century, though he has much older citations with the verbs  and  (Visser 1963–1973: § 1840). Earlier instances of

144

DAVID DENISON progressive  + AP include (a), (b) and (c) below and (129) above, while participial  + AP is much older still, (d) below: a. b. c. d.

1802 Woodforde, Diary V 403.19 (29 Aug) My Throat is daily getting better he says. 1834 T. Medwin Angler in Wales I.21 The race of our bull-dogs is getting fast extinct, … 1839 Dickens, Ol. Twist (1850) 60/1 ‘You’re getting too proud to own me afore company, are you?’ 1624 Saunderson 12 Serm. (1637) 172 The Morter getting wet dissolveth …

13.

If we included passive  it would bring the additional possibility of modal + perfect ⁄ + passive  + V, but according to Mitchell (1985: §§ 753, 1095) this did not occur in Old English, and I have not located any examples in Middle English. The theoretical combinations involving progressive  + passive  did not apparently occur.

14.

I tabulate the order as idealised to the present-day English norm. Compare now Warner (1997: 182–184); I have modified one date — c1300 rather than a1325 for pattern B — in the light of Warner’s useful discussion.

15.

The likelihood of differential dates of adoption of the new grammar by different groups of speakers is discussed in Pratt and Denison (2000), and for other evidence on the dating of grammaticalisation of the progressive see Denison (1998: 143–146).

16.

I cannot justify all the various datings in Table 2, and I have not mentioned  ‘begin’ at all: see Denison (1985a, 1993) for some discussion. There are, incidentally, significant differences from the dates given in Kossuth’s Table 2 (1982: 294). Quite a bit of the data in her interesting sketch can in fact be antedated (as she herself foresaw). In the light of the argument developed here I have modified some datings from the table originally given in Denison (1993: 440).

References Boyland, J. T. 1998. “A corpus study of would + have + past-participle”. In Historical Linguistics 1995: Selected Papers from the 12th International Conference on Historical Linguistics, Manchester, August. 1995, vol. 2, Germanic [Current Issues in Linguistic Theory 162], R. M. Hogg and L. van Bergen (eds), 1–17. Amsterdam and Philadelphia: John Benjamins. Brinton, L. J. 1988. The Development of English Aspectual Systems: Aspectualizers and Post-verbal Particles [Cambridge Studies in Linguistics 49]. Cambridge: Cambridge University Press. Bybee, J. L., Pagliuca, W. and Perkins, R. D. 1991. “Back to the future”. In Approaches to Grammaticalization, vol. 2, Focus on Types of Grammatical Markers [Typological Studies in Language 19], E. C. Traugott and B. Heine (eds), 17–58. Amsterdam and Philadelphia: John Benjamins.

COMBINING ENGLISH AUXILIARIES

145

de la Cruz, J. 1995. “The geography and history of double modals in English: A new proposal”. Folia Linguistica Historica 16: 75–96. Denison, D. 1985a. “The origins of periphrastic : Ellegård and Visser reconsidered”. In Papers from the 4th International Conference on English Historical Linguistics: Amsterdam, 10–13 April 1985 [Current Issues in Linguistic Theory 41], R. Eaton, O. Fischer, W. Koopman and F. van der Leek (eds), 45–60. Amsterdam and Philadelphia: John Benjamins. Denison, D. 1985b. “Some observations on being teaching”. Studia Neophilologica 57: 157–159. Denison, D. 1992. “Counterfactual may have”. In Internal and External Factors in Syntactic Change [Trends in Linguistics / Studies and Monographs 61], M. Gerritsen and D. Stein (eds), 229–256. Berlin and New York: Mouton de Gruyter. Denison, D. 1993. English Historical Syntax: Verbal Constructions [Longman Linguistics Library]. London and New York: Longman. Denison, D. 1998. “Syntax”. In The Cambridge History of the English Language, vol. 4, 1776–1997, S. Romaine (ed.), 92–329. Cambridge: Cambridge University Press. Ellegård, A. 1953. The Auxiliary ‘do’: The Establishment and Regulation of its Growth in English [Gothenburg Studies in English 2]. Stockholm: Almqvist & Wiksell. Geerts, G., Haeseryn, W., de Rooij, J. and van den Toorn, M. C. 1984. Algemene Nederlandse Spraakkunst. Groningen: Wolters-Noordhoff. Gronemeyer, C. 1999. “On deriving complex polysemy: The grammaticalization of get”. English Language and Linguistics 3: 1–39. Haegeman, L. 1985. “The get-passive and Burzio’s generalization”. Lingua 66: 53–77. Hopper, P. J. 1991. “On some principles of grammaticization”. In Approaches to Grammaticalization, vol. 1, Focus on Theoretical and Methodological Issues, E. C. Traugott and B. Heine (eds), 17–35. [Typological Studies in Language 19]. Amsterdam and Philadelphia PA: John Benjamins. Jespersen, O. 1909–49. A Modern English Grammar on Historical Principles, 7 vols. Heidelberg: Carl Winters Universitätsbuchhandlung. Kossuth, K. C. 1982. “Historical implications of the co-occurrence constraints on auxiliaries”. Lingua 56: 283–295. Lakoff, G. 1987. Women, Fire, and Dangerous Things: What Categories Reveal About the Mind. Chicago and London: University of Chicago Press. Mitchell, B. 1985. Old English Syntax, 2 vols. Oxford: Clarendon Press. Montgomery, M. B. 1989. “Exploring the roots of Appalachian English”. English WorldWide 10: 227–278. Montgomery, M. B. and Nagle, S. J. 1994 (for 1993). “Double modals in Scotland and the southern United States: Trans-Atlantic inheritance or independent development?” Folia Linguistica Historica 14: 91–107. Mossé, F. 1938. Histoire de la forme périphrastique ‘être + participe présent’ en germanique, vol. 2, Moyen-anglais et anglais moderne [Collection Linguistique, La Société Linguistique de Paris 43]. Paris: C. Klincksieck.

146

DAVID DENISON

Murray, J. A. H., Bradley, H., Craigie, W. A. and Onions, C. T. (eds). 1933 (and CD-ROM version. 1988). The Oxford English Dictionary, 13 vols. Oxford: Clarendon Press. Nagle, S. J. 1993. “Double modals in early English”. In Historical Linguistics 1989: Papers from the 9th International Conference on Historical Linguistics, Rutgers University, 14–18 August. 1989 [Current Issues in Linguistic Theory 106], H. Aertsen and R. J. Jeffers (eds), 363–370. Amsterdam and Philadelphia PA: John Benjamins. Nagle, S. J. 1994. “The English double modal conspiracy”. Diachronica 11: 199–212. Nagle, S. J. 1995. “The English double modals: Internal or external change?” In Linguistic Change under Contact Conditions [Trends in Linguistics / Studies and Monographs 81], J. Fisiak (ed.), 207–215. Berlin and New York: Mouton de Gruyter. Nakamura, F. 1991. “On the historical development of the activo-passive progressive: ‘the house is building’”. In Aspects of English Philology and Linguistics (Festschrift offered to Dr Masatomo Ukaji on his sixtieth birthday), S. Chiba (ed.), 121–143. Tokyo: Kaitakusha. Nehls, D. 1974. Synchron-Diachrone Untersuchungen zur Expanded Form im Englischen: eine Struktural-Funktionale Analyse [Linguistische Reihe 19]. Munich: Max Hueber Verlag. OED. See Murray et al. (1933,1988) and Simpson and Weiner (1992 [1991]). Ogura, M. 1993. “Schal (not) mowe, or double auxiliary constructions in Middle English”. The Review of English Studies 44: 539–548. Ogura, M. 1998. “On double auxiliary constructions in medieval English”. In English Historical Linguistics and Philology in Japan, J. Fisiak and A. Oizumi (eds), 229–236. Berlin and New York: Mouton de Gruyter. Phillipps, K. C. 1970. Jane Austen’s English [The Language Library]. London: André Deutsch. Plank, F. 1984. “The modals story retold”. Studies in Language 8: 305–364. Pratt, L. and Denison, D. 2000. “The language of the Southey-Coleridge circle”. Language Sciences 22: 401–422. Rydén, M. and Brorström, S. 1987. The ‘be/have’ Variation with Intransitives in English: With Special Reference to the Late Modern Period [Stockholm Studies in English 70]. Stockholm: Almqvist & Wiksell. Simpson, J. A. and Weiner, E. S. C. 1992 (beta test version. 1991). The Oxford English Dictionary: CD-ROM version (2nd edition). Oxford: Oxford University Press. Strang, B. M. H. 1970. A History of English. London: Methuen. Visser, F. Th. 1946–56. A Syntax of the English Language of St. Thomas More [Materials for the Study of the Old English Drama n.s. 19, 24, 26.], 3 vols. Louvain: Librairie Universitaire. Visser, F. Th. 1963–73. An Historical Syntax of the English Language, 4 vols. Leiden: E. J. Brill. Warner, A. R. 1986. “Ellipsis conditions and the status of the English copula”. York Papers in Linguistics 12: 153–172.

COMBINING ENGLISH AUXILIARIES

147

Warner, A. R. 1993. English Auxiliaries: Structure and History [Cambridge Studies in Linguistics 66]. Cambridge: Cambridge University Press. Warner, A. R. 1995. “Predicting the progressive passive: Parametric change within a lexicalist framework”. Language 71: 533–557. Warner, A. R. 1997. “Extending the paradigm: An interpretation of the historical development of auxiliary sequences in English”. English Studies 78: 162–189. Wekker, H. Chr. 1987. “Points of Modern English syntax LXIX”. English Studies 68: 456–463.

Grammaticalisation: Unidirectional, non-reversable? The case of to before the infinitive in English Olga Fischer University of Amsterdam

1.

Introduction

In the literature on grammaticalisation it is quite generally assumed that this process is unidirectional and non-reversable, and also that it is essentially a process driven by semantic or pragmatic factors with grammatical and phonetic changes as it were automatically following (cf. Brinton 1988; Lehmann 1991; Traugott and Heine 1991; Hopper and Traugott 1993; Bybee, Perkins and Pagliuca 1994). I have tried to show on two other occasions (cf. Fischer 1994b, 1997b) that the syntactic and semantic changes do not necessarily go in tandem, and that the assumption of grammaticalisation as a diachronic process clashes with models of change that accept that change takes place within the grammar of each speaker, i.e. that each speaker builds up his grammar afresh and does not take account of processes that have started long before he was there. I agree with Harris and Campbell (1995) that it is probably methodologically more accurate to consider grammaticalisation not as a separate mechanism or cause of change, but as a process built up out of a series of reanalyses and extensions.1 A process that may proceed all the way, from lexical item to affix or even zero, but as often as not may stop halfway, or may even regress to some extent. Each station in the series or the chain is dependent on synchronic circumstances and determines the direction of the change afresh. In another paper (Fischer 1999), I also tried to show that many of the processes at work during grammaticalisation are motivated by iconicity, such as analogy, metaphor, isomorphism and persistence. I do not wish to go very deeply into the iconic aspects of grammaticalisation here (I refer the interested reader to Fischer 1999), but I would like to look in

150

OLGA FISCHER

detail at what happened in one particular case of grammaticalisation, and determine the nature of the process, and what may have caused it, in order to get a clearer picture of what grammaticalisation is. But first, before turning to this case study, I want to briefly show what I understand by iconicity, since this plays an important background role in the discussion that follows. As I mentioned above, I believe that iconicity plays an important role in grammaticalisation processes, and may even upset some of its presumed unidirectionality. Since iconicity is used by linguists in different ways, a description of what I understand by iconicity is in order. Iconicity as a semiotic notion refers to a natural resemblance or analogy between the form of a sign (‘the signifier’, be it a letter or sound, a word, a structure of words, or even the absence of a sign) and the object or concept (‘the signified’) it refers to in the world or rather in our perception of the world. The similarity between sign and object may be due to common features inherent in both: by direct inspection of the iconic sign we may glean true information about its object. In this case we could speak of ‘imagic’ iconicity (as in a portrait or in onomatopoeia, e.g. ‘cuckoo’) and the sign is called an ‘iconic image’. When we have a plurality of signs, the analogy may be more abstract: we then have to do with ‘diagrammatic’ iconicity, which is based on a relationship between signs that mirrors a similar relation between objects/concepts or actions (e.g. a temporal sequence of actions is reflected in the sequence of the three verbs in Caesar’s famous dictum veni, vidi, vici: in this instance, the sign — here the syntactic structure of three verbs — is an ‘iconic diagram’). Obviously, it is primarily diagrammatic iconicity that is of great relevance to language. Both imagic and diagrammatic iconicity are not clean-cut categories but form a continuum, on which the iconic instances run from almost perfect mirroring (i.e. a semiotic relationship that is virtually independent of any individual language) to a relationship that becomes more and more suggestive and also more and more language-dependent (for a convenient overview of various types of iconicity see Fischer and Nänny 1999). Isomorphism, a term also used in the discussion that follows, is an example of diagrammatic iconicity of a rather abstract kind. It is used here in the sense used by Haiman (1980: 515–516), i.e. the principle of one form (signifier) corresponding to one meaning (signified). Iconicity also plays a fundamental role in the so-called naturalness approach to language theory (e.g. in the work of Wolfgang Dressler, Willi Mayerthaler, David Stampe, Wolfgang Wurzel and others), and in cognitive approaches to language (as in the work of Talmy Givón, Ronald Langacker, Paul Hopper, Sandra Thompson and others). Basic to both these approaches is (in the words of Dressler 1995: 22) that “it does not assume an autonomous module of grammar, but attempts to find cognitive and other extralinguistic bases … for

UNIDIRECTIONALITY: TO BEFORE THE INFINITIVE

151

grammatical principles and preferences”, that “questions of the relationship between language and the mind can be approached only by considering language in its natural functional context” (Hopper and Thompson 1984: 747–748), and that the cognitive strategies underlying language systems have a strong perceptual basis. After this brief excursion, I now wish to consider what exactly the nature of grammaticalisation is. One of the problems I have with the way grammaticalisation has been dealt with in the literature is that the mechanistic side of it has been overemphasised, with the result, I think, that the mechanism has become too powerful as an explanatory tool or as a description of a diachronic process of linguistic change. Thus, the following quote from Bybee et al. (1994: 298) suggests that grammaticalisation is seen as an independent process with independent explanatory value: “Thus our view of grammaticization is much more mechanistic than functional: the relation between grammar and function is indirect and mediated by diachronic process. The processes that lead to grammaticization occur in language use for their own sakes; it just happens that their cumulative effect is the development of grammar” (italics added). Similarly, Vincent (1995: 434) talks about “the power of grammaticalisation as an agent of change” (italics added), which suggests that grammaticalisation has explanatory value, that it has independent force. Finally, Heine et al. (1991b: 9) write that “Meillet followed Bopp rather than Humboldt in using grammaticalization as an explanatory parameter in historical linguistics” (italics added), and the authors themselves seem to follow this line too (see 1991b: 11). Let us first look at the way the process has been described in the literature. Grammaticalisation is generally seen as a gradual diachronic process which is characterised as unidirectional, i.e. it always shows the “evolution of substance from the more specific to the more general and abstract” (Bybee et al. 1994: 13). The unidirectionality applies on all levels, the semantic, the syntactic and the phonological. Almost without exception, the process is seen as semantically driven, with bleaching of meaning playing a primary role. (This is not true for all linguists, notably Hopper and Traugott (1993) believe that bleaching only plays a role in the later stages of grammaticalisation.) Rubba (1994: 81), for instance, describes it as primarily a process of semantic change. Bybee et al. (1994: 17–18) even suggest that we can reconstruct the path of grammaticalisation with the help of the “hypothesis that semantic change is predictable”. The notion of graduality implies that grammaticalisation is seen as “an evolutional continuum. Any attempt at segmenting it into discrete units must remain arbitrary to some extent” (Heine and Reh 1984: 15, and see also Heine et al. (1991b: 68, 165 and passim). In this light it is not surprising to read that the mechanisms at work in and the causes of grammaticalisation are also seen as basically semantic/pragmatic in

152

OLGA FISCHER

nature. For most linguists writing on grammaticalisation, the main mechanisms involved are metaphoric and metonymic in nature.2 Metaphoric change can be related to analogy, it is a type of paradigmatic change whereby a word-sign used for a concrete object (i.e. the word back as part of the body) can be reinterpreted on a more abstract level as an indication of ‘location’ (because of some element that these concepts have in common), and then further interpreted along the metaphorical axis as an indication of ‘time’. Metonymic change can be related to re-analysis, and functions on the syntagmatic plane. It takes place mainly via the “semanticization of conversational inferences” (Hopper and Traugott 1993: 84). A good example of metonymic change is the case of to be going to discussed in the introduction (this volume). Another example discussed by Hopper and Traugott (1993: 85) is the change in OE hwile ‘while’; the meaning of temporal simultaneity that it has, may change into a causal meaning because in many cases “the conditions specified in the subordinate clause [i.e. the clause introduced by while] serve not only as the temporal frame of reference for those in the main clause, but also as the grounds for the situation”. As far as the cause of grammaticalisation is concerned, this is usually seen as being pragmatic in nature. Bybee et al. (1994: 300) write: “the push for grammaticization … originates … in the tendency to infer as much as possible from the input, and in the necessity of interpreting items in context”. They show that grammaticalisation occurs in cycles or is self-propelling. The process of grammaticalisation (loss of concrete form) itself leads to a search for new expressive means to indicate the same function, and when the new expression has again grammaticalised, the search for a new concrete expression begins again. Likewise Hopper and Traugott (1993: 86) concur with Heine et al. (1991a: 150–151) that “grammaticalization can be interpreted as the result of a process which has problem-solving as its main goal”. It is the result of a “search for ways to regulate communication and negotiate speaker-hearer interaction” (1993: 86). Although I would agree with the views just now discussed, i.e. that reanalysis and analogy, or metonymic and metaphorical processes, are important in language change, and also that grammaticalisation may be caused by the need for expressivity and routinisation, I still cannot see that there is room for a separate or ‘independent’ process of grammaticalisation. Where most linguists see a unidirectional process from concrete to abstract, a process that cannot be cut up into segments, I can only see a more or less accidental concurrence. The processes underlying grammaticalisation may lead one way as well as another, i.e. there is no necessary link between one segment of the chain of grammaticalisation and another.3

UNIDIRECTIONALITY: TO BEFORE THE INFINITIVE

153

I also think that grammaticalisation processes themselves can only be discovered with hindsight, which means that if we have a preconceived notion of what grammaticalisation is, we will indeed discover mainly those processes that have run a full or ‘fullish’ course, and we will not realise that there may be many cases where the path of grammaticalisation proceeded differently. So it may only seem that grammaticalisation usually follows the same channel. Aborted and reversed processes are very difficult to find when one looks backwards in this way.4 So the similarities in known cases of grammaticalisation may have led to an overemphasis on a common core, and through that the idea may have arisen that grammaticalisation is an explanatory parameter in itself. To my mind it is the subprocesses that explain the change. I agree with linguists such as Lightfoot (1979, 1991, 1999) and Joseph (1992) that, logically, diachronic processes cannot exist because diachronic grammars do not exist. Each speaker makes up his own grammar afresh on the basis of data surrounding him, and on the basis of his general cognitive abilities or strategies (or, so one wishes, on the basis of some innate Language Acquisition Device). So why should a grammaticalisation process necessarily run from a to b, to c etc.? Why should there be unidirectionality?5 With Harris and Campbell (1995: 20, 336 ff.) (and see also Fischer 1997b) I would tend to accept that grammaticalisation has no independent status, no explanatory value in itself. It was when I looked at the so-called grammaticalisation of have to in English from a possessive verb to a modal auxiliary (see Fischer 1994b) that I began to realise that not all grammaticalisation change is driven semantically, and that unexpected, language specific factors may play a role. I think it pays to look at any hypothetical grammaticalisation process in detail, next to taking the wider, typological bird’s eye view. Here, therefore, I would like to consider the case of infinitival to in English.

2.

The case of infinitival to

There is a widespread belief that the development of the original preposition to before the infinitive into a meaningless infinitival marker follows a well-known grammaticalisation channel. This is clear from Haspelmath’s (1989) study, the essence of which is expressed in his title, “From purposive to infinitive — a universal path of grammaticization”. Haspelmath shows that in many languages in the world the allative preposition — to in English — which expresses location, or rather the goal of motion, also comes to express goal or purpose more abstractly; and that in combination with the infinitive, the preposition begins to lose its original purposive function, ending up as a purely grammatical

154

OLGA FISCHER

element to indicate that the verbal form is an infinitive. This interpretation of the development is already seen in Jespersen (and cf. also Mustanoja 1960: 514), In … the to-infinitive, to had at first its ordinary prepositional meaning of direction, as still in “he goes to fetch it”; … But gradually an enormous extension of the application of this to-infinitive has taken place: the meaning of the preposition has been weakened and in some cases totally extinguished, so that now the to-infinitive must be considered the normal English infinitive, the naked infinitive being reserved for comparatively few employments, which are the solitary survivals of the old use of the infinitive. This development is not confined to English: we find it more or less in all the Gothonic languages, though with this preposition only in the West Gothonic branch (G. zu, Dutch te), while Gothic has du, and Scandinavian at ( …) (Jespersen 1927: 10–11)

It seems to me that the expectations raised by the fact that this seems to be a frequent grammaticalisation pattern, has led us too much to see the English case as following the well-trodden path. I think it pays to look more closely at the linguistic details. I have compared the development of the infinitive marker in Dutch and English6 and come to the conclusion that to and cognate te have not grammaticalised in the same way. On the contrary, it looks as if to has been stopped early in its development and has even regressed in some respect. I think this could be characterised as a process of what Frans Plank (1979) has called Ikonisierung, a moving away from the symbolic pole back to the iconic one. I will first briefly explain what I mean by these two poles. It is well-known that in language there is competition between iconic and economic motivation (cf. Haiman 1983) or between the need for clarity and the need for processing speed. General erosion leads to the loss of expressivity and consequently to a constant need for new linguistic expressions. One could say therefore that language moves or is situated along an axis with two poles: an iconic, concrete pole at one end, and a symbolic (or perhaps ‘arbitrary’ or ‘conventional’ is a less confusing term here), abstract one at the other. In grammaticalisation, elements move along this axis, from concrete to abstract. One could also refer to the iconic pole as original and creative and to the symbolic as derivative and mechanistic. I believe indeed that the forces behind the development of to have been to a large extent iconic (with ‘persistence’ and analogy or isomorphism playing an important role, see for more detail Fischer 1999), although there were some syntactic factors too, which I will come back to below. The main iconic factor indeed is isomorphism. I am using the term here, as used by Haiman (1980), meaning the existence of a one-to-one relation between signans and signatum, similar to von Humboldt’s principle of ‘one form-one meaning’. One can see that

UNIDIRECTIONALITY: TO BEFORE THE INFINITIVE

155

through the grammaticalisation of to, the original isomorphic or one-to-one relation between the signans and the signatum (as given in (1a)), (1)

Structural stages in the grammaticalisation of to a.
b.
c.


x

xy

x y

(α = the signans to; β = the reduced signans of to; x = signatum ‘goal’; y = signatum ‘infinitival marker’) is disturbed (as shown in (1b)). Through grammaticalisation, the sign to acquires two signata: the first is the original prepositional purposive ‘to,’ and the second the semantically empty, infinitive marking element ‘to’. The result is then an asymmetric, non-isomorphic situation as shown in (1b). This lack of isomorphism can be amended in two ways. The usual way according to the grammaticalisation hypothesis is for the new signatum to acquire its own distinctive linguistic form. This may be obtained through the phonetic reduction of to (this is another iconic principle, the ‘quantity principle’, see Givón 1995: 49), which would then coexist with the full form to (stage (1c)). This development is most clear in Dutch, which has infinitival te, next to the earlier particle toe. But in Middle English, too, we find occasional te spellings or other spellings indicating the phonetic reduction of to.7 So with stage (1c) we have a new stable isomorphic relation. The other solution for the asymmetry of (1b) is to go back to the earlier symmetry (i.e. 1a). This also makes the relation isomorphic again, and it is more strongly iconic than (1c) because here the sign to is linked back up with its original meaning, i.e. it is re-iconicised, going back to the iconic pole (in Fischer 1999, I argue that this process, usually called persistence, is also iconic in nature). So my suggestion is that diacronically English to moved back to stage (1a), while Dutch te moved on to stage (1c). In what follows, I will have a look at the (comparative) facts, and also offer some suggestions as to why English to re-iconicised, both in terms of isomorphism and persistence. 2.1 The grammaticalisation of to in its early stages It seems that at first, in the late Old English, early Middle English period, to developed very much like Dutch te. Evidence for this can be found in the following facts: (2)

a. b. c. d.

strengthening of to by for phonetic reduction of to loss of semantic integrity occurrence of to-infinitive after prepositions other than for

156

OLGA FISCHER

2.1.1 To strengthened by for First we find the need for an additional preposition (for) to emphasise the goal function of the to-infinitive. This use of for is attested from 1066 onwards (see Mustanoja 1960: 514) and steadily increases in the Middle English period until 1500 (see Table 1). A similar development can be seen in Middle Dutch, where om(me) te begins to occur quite frequently (see Stoett 1909: § 283) and becomes more and more regular for the expression of purpose (see Gerritsen 1987: 143–147), becoming obligatory in many positions in Modern Dutch and remaining there whenever purpose or direction is intended. 2.1.2 Phonetic reduction of to The phonetic reduction of to to te can be found in Middle English, as shown in Table 1. In Middle Dutch we already only find the reduced form, but this can be reduced even further to a single phoneme t attached to the infinitive (Stoett § 283 gives the form tsine for te sine ‘to be’). I have found a few bound forms in the Helsinki corpus too, all from the later Middle English period (examples are given in note 7). Table 1. The frequency of for to in the Middle English and early Modern English periods, based on the Helsinki corpus (taken from Fischer 1997a)

forto forte for to for te (te, t’, to-

1150–1250

–1350

–1420

–1500

–1570

–1640

–1710

01 87 14 03 36

29 15 91 01 03

46 01 3230 00 01

46 00 2510 00 02

01 00 41 00 01

00 00 07 00 00

0 0 5 0 (0)

2.1.3 Loss of semantic integrity We see the occasional use of the to-infinitive in Middle English in structures where it cannot possibly be goal-oriented, i.e. in positions where the plain infinitive and the present participle (which express simultaneity rather than purpose) had been the rule in Old English (cf. Fischer 1996: 119–121). The following is an example from a fourteenth century text, where to wepe clearly expresses a state not a purpose, (3)

And in my barm ther lith to wepe / Thi child and myn … and in my bosom there lies weeping thy child and mine (Macauley 1900–1901, Gower, Conf.Am. III 302)

UNIDIRECTIONALITY: TO BEFORE THE INFINITIVE

157

In Middle Dutch, too, the usual forms in these constructions were the plain infinitive and the present participle (but a coordinated construction is also quite often found, cf. Stoett 1909: §§ 10, 281). But here, too, the te-infinitive, which becomes the rule in later Dutch (cf. the examples in (4)), begins to make headway, (4)

a.

b.

Hij lag te slapen he lay to sleep ‘He lay sleeping’ Zij stond te wachten she stood to wait ‘She stood waiting’

2.1.4 Occurrence of to-infinitives after prepositions other than for We see the occasional occurrence in Middle English of a to-infinitive preceded by another preposition which also governs the infinitive, making clear that to can no longer be prepositional. According to Visser (1969: § 976), this structure does not occur in Old English, and is very rare again in later English. Most of his examples are from the period 1200 to 1500. Some illustrations are given in (5), (5)

a.

b.

bliss of herte that comþ of God to lovie the happiness of heart that comes from God to love ‘… from loving God’ (Morris 1965, Ayenbite 93)  himm birþþ Šeornenn aŠŠ þatt an, / Hiss Drihhtinn wel to cwemenn / … Wiþþ messess  wiþþ beness /  wiþþ to letenn swingenn himm and it behoves him to always desire that one (thing), to please his Lord well … with masses and prayers and with to let scourge himself … ‘and by having himself scourged …’ (Holt 1878, Orm. 6358–62)

In Middle Dutch, and more frequently in early Modern Dutch, the te-infinitive begins to occur too after other prepositions, such as van ‘of’, met ‘with’, na ‘after’, and especially sonder ‘without’, and these constructions can still be found in present-day Dutch, especially in colloquial speech (see Stoett 1909: §§ 282–283; Overdiep 1935: §§ 354–358). Some seventeenth century instances are given in (6),

158

OLGA FISCHER

(6)

a.

b.

Hy starf, niet sonder seer beclaeght te wesen, he died not without deeply lamented to be (‘being lamented’) den 8sten April the 8th of April (van Mander, Overdiep 1935: 420) … sal ick eindigen naer mijn groetenisse aen alle de vrinden ghedaen te hebben I will end after to have (‘having’) given my greetings to all my friends (Reig.77/16, Overdiep 1935: 421)

2.2 New developments involving to So the initial stages in Middle English look like a regular grammaticalisation process. However, towards the end of the Middle English period the trend seems to reverse. All the structures discussed in Section 2.1 above seem to disappear. Table 1 makes quite clear that the strengthening of the to-infinitive with for, disappears quite suddenly — at least from the Standard language — in the early Modern period. I believe that the reason for this is that to went back to its original meaning, again strongly expressing goal or direction (there is some difference with Old English usage, I will come back to that below). But apart from the disappearance of the grammaticalisation characteristics enumerated in (2), there are also new developments that indicate the renewed, semantic independence of to before the infinitive:8 (7)

a. b. c.

appearance of split infinitives absence of ‘reduction of scope’ no loss of semantic integrity

2.2.1 The appearance of split infinitives The first split infinitives are attested in the fourteenth century (see Mustanoja 1960: 515; Visser § 977; Fischer 1992a: 329–330).9 (8)

Blessid be þou lord off hevyn … / That suche grace hath sent to his / Synfull men for to þus lede / In paradice (Cursor Mundi, Morris 1876, Laud Ms 18440–44)

This shows that the grammaticalisation of to is disturbed in that the usual process would have been for grammaticalised to to become more and more ‘bonded’ to the infinitive, in accordance with one of the grammaticalisation parameters distinguished by Lehmann (1985).

UNIDIRECTIONALITY: TO BEFORE THE INFINITIVE

159

2.2.2 Absence of the ‘reduction of scope’ Another phenomenon showing ongoing grammaticalisation, also mentioned by Lehmann, is the reduction of scope, which is absent in English. In Dutch, on the other hand, there is reduction of scope: when two infinitives are coordinated in Modern Dutch, it is the rule for both infinitives to be marked by te, if the first one is so marked (for some idiomatic exceptions see Fischer 1996: 112–113). In other words the scope of te has been reduced to its immediate constituent. This is not the case in English where the first to can have both infinitives as its scope, as the literal English translation of the non-acceptable Dutch example in (9) shows, (9)

a. *Je kunt deze shampoo gebruiken om je haar mee te wassen en je kleren schoonmaken b. You can use this shampoo to wash your hair and clean your clothes

2.2.3 No loss of semantic integrity Another of Lehmann’s parameters, ‘the loss of integrity’, is relevant here too. It is clear that in Dutch, te has gradually lost its semantic integrity, i.e. it has become de-iconicised, and no longer expresses ‘goal’ or ‘direction’; this is now expressed obligatorily by om te. One result of this semantic loss in Dutch was already mentioned in Section 2.1.3 above. Another one is the appearance of the te-infinitive with a future auxiliary in Dutch. Overdiep (1935: § 336) shows that they are quite regular already in early Modern Dutch. This again is a clear contrast with English where such a future infinitive simply never develops, neither in Middle English when shall and will could still be used in infinitival form, nor later with the new future auxiliary to be going to. Overdiep also mentions that zullen is especially common when the matrix verb itself is not inherently future directed, so after a verb like say. The reason for this difference may be clear by now. To itself expressed future and therefore had no need for a future auxiliary, whereas Dutch te no longer carried future meaning; it had become empty of referential meaning, and therefore the Dutch infinitive may need reinforcement. The loss of the purposive meaning of te has also widened the possibility of using non-agentive subjects with a te-infinitive in Dutch (showing grammaticalisation along the ‘animacy’ hierarchy, cf. Hopper and Traugott 1993: 157). With a verb like dreigen ‘threaten’, the use of a non-agentive or an expletive it subject (i.e. with the verb being used epistemically) is quite common in Dutch, while it is more awkward in English, because of the stronger purpose meaning of to,

160

OLGA FISCHER

(10)

a.

Het dreigde te gaan regenen, toen ik het huis verliet threatened to rain, when I left the house.’

?‘It

Traugott (1993) notes that occurrences with expletive it are very rare in her corpora but that they do occur. It is interesting to note that all ten informants that I questioned, except one, do not like this sentence. They would much prefer, It threatened rain or Rain threatened or It looked like rain, and they find it also more acceptable with a progressive verbal form. Obviously it is not a construction they would comfortably use themselves. Traugott also notes (1993: 187) that although inanimate subjects with threaten occur, there is usually “something about the subject that leads to an expectation of the proposition coming into being”; in other words, there is a strong tendency still present to ascribe some agentive function to the subject. Similarly (10b) can be non-agentive in Dutch, (10)

b.

Hij dreigde van zijn fiets te vallen *‘He threatened to fall of his bike.’

All my informants except one agree that (10b) is only possible in English if the subject wanted to injure himself and thus inflict pain on someone who cared for him. The epistemic meaning is the usual interpretation of this sentence in Dutch. When the subject has to be interpreted as agentive, the preferred construction would be with a finite clause: Hij dreigde dat hij …. The same situation holds when dreigen is followed by a passive infinitive, making an agentive function of the matrix subject, which is also the subject of the infinitive, impossible (cf. Traugott 1995: 34: “the passive demotes the inference that the subject … is volitional or responsible with respect to the purposive clause”). (11) then, is a perfectly possible sentence in Dutch, but unacceptable for most English speakers, (11)

Hij dreigde ontslagen te worden threatened to be fired.’

?*‘He

And a construction like (12) is ambiguous in Dutch, but not usually in English, (12)

Hij dreigde haar te doden ‘There was a danger that he would kill her.’ ‘He threatened to kill her’

English speakers strongly prefer the second, agentive interpretation. The reason for these differences is the fact, as I mentioned above, that to in English is still more strongly purposeful and therefore by default as it were one expects a controlling agent. It should be mentioned here that threaten/dreigen is not the

UNIDIRECTIONALITY: TO BEFORE THE INFINITIVE

161

only verb that shows this difference in usage between Dutch and English. Traugott (1993) also discusses the behaviour of the verb to promise, which verb again in Dutch can be used non-agentively more easily than in English. In Fischer (1997a: 271–273), I also point out that in English to-infinitives regularly occur with the categories of verbs that Haspelmath (1989) has described as ‘irrealis directive’ and ‘irrealis potential’, but not with the categories ‘realis nonfactive’ and ‘realis-factive’. The latter two categories contain clearly nondirectional verbs, and it is interesting that in Dutch and German these last two categories do take te/zu infinities much more easily than in English. Thus a verb like affirm does not take a to-infinitive in English, but its Dutch semantic equivalent verzekeren does (for more details see Fischer 1997a). A final difference between Dutch and English is the formation of new modal auxiliaries in English consisting of a matrix verb that has semantically inherent future reference and the to element that belongs to the infinitive following the verb, as in to be going to/gonna, to want to/wanna, to have (got) to/gotta etc. Plank (1984: 338–339) notes that these verbs are unlike auxiliaries in that they occur with to, but notes at the same time that these same auxiliaries “allow the conjunction [i.e. to] to be reduced and contracted in informal speech”, even when this is not fast speech, and before pauses, indicating that this to has grammaticalised and become as it were affixed to the matrix verb. This amalgamation is possible because both to and the matrix verb express future modality. (So it seems to can become further grammaticalised in English only when it coincides with another future-bearing element.10) In Dutch, however, this development has not taken place, because there was no meaningful ‘future’ or purposeful te for the matrix verb to attach to. In fact, whenever we do get a (semi-)auxiliary followed by a te-infinitive, it is clear that te goes with the infinitive. This is shown in the position of the adverb in examples such as the following, (13)

a.

b.

Ik zit nu te denken /*Ik zit te nu denken I sit now to think /I sit to now think ‘I am thinking now’ Het dreigt thans te mislukken /*Het dreigt te thans mislukken It threatens now to fail /It threatens to now fail ‘there is a possibility that it will fail’

In linguistically similar cases in English, the adverb can occur between to and the infinitive, showing that to and the infinitive do not form a cluster. That to in fact forms a cluster with the matrix verb is shown by cases in which matrix verb and to can be contracted as in the second example of (14),

162

OLGA FISCHER

(14)

3.

I want to immediately go there I wanna go there immediately

The role played by the grammar

Now the question must be asked, what has caused the reversal in the grammaticalisation of to? I believe this is due to the grammatical circumstances under which to developed. In one respect English came to differ radically from Dutch, and this influenced the use and interpretation of to. In early Middle English the infinitive became much more strongly verbal than in Dutch (for instance, Dutch infinitives can be preceded by a possessive pronoun or an article, which is impossible in English (for more details see Fischer and van der Leek 1981: 319). This verbal nature of the infinitive was strengthened by the fact that to-infinitives started to replace that-clauses on a grand scale in the Middle English period (cf. the rough statistics in Manabe 1989, and more specifically Los 1998); that is, they replaced clauses which have a tense-domain separate from the tense expressed in the matrix clause (cf. Fischer 1997c). This caused the element to, which originally expressed ‘goal’ or direction, to function as a kind of shift-oftense element. What I mean is, to came to express a ‘break’ in time, a movement away from the time of the main clause; i.e. it again expressed ‘direction’. It is indeed only in English that we later (the first examples date from the late Middle English period) see the development of two different kinds of infinitival complements after perception verbs, where to becomes crucial in expressing a shift in tense, (15)

a.

b.

it thoghte hem gret pite / To se so worthi on as sche, / With such a child as ther was bore, / So sodeinly to be forlore ‘it seemed to them a great pity to see so worthy a woman as she was to be destroyed together with the child that was born to her’ (Macauley 1900–1901, Gower, Conf.Am. II, 1239–42) ‘for certeynly, this wot I wel,’ he seyde, / ‘That forsight of divine purveyaunce / Hath seyn alwey me to forgon Criseyde,’ ‘for certainly, this I know well, he said, that the foresight of divine providence has always seen that I would lose Criseyde’11 (Benson 1988, Chaucer T&C IV, 960–62)

In both cases the to-infinitive refers to something happening in the future. The construction contrasts with the usual complement structure of physical perception

UNIDIRECTIONALITY: TO BEFORE THE INFINITIVE

163

verbs, which until then had only allowed a bare infinitive, expressing the simultaneous occurrence of what had been seen, heard or felt, as in I saw her cross(ing) the street. In present-day English, this to-infinitive after perception verbs now no longer expresses future time, but it still expresses a shift in tense, making the experience indirect, as in (16), (16)

Alex saw Julia to have been in a hurry when she dressed (because she was wearing her T-shirt inside out) (the example is from van der Leek 1992: 13)

The type of construction shown under (15) was further strengthened by the influx of Latin-type accusative and infinitive constructions (aci) (as in (17)), which again appear in the late Middle English period (i.e. when to ‘reverted’) showing similar ‘breaks’ in tense between matrix verb and infinitive, (17)

I expect him to be home on time

These Latin-type aci constructions always have a to-infinitive in English. (For more details on this development, see Fischer 1992b, 1994a.). It seems that we can conclude that special syntactic circumstances as it were forced infinitival to to become more isomorphic again with the preposition to.

4.

A brief conclusion

I have tried to show that grammaticalisation processes do not always run the same course, that there may be differences between similar languages, that the process may indeed be reverted, and that this relates to the specific grammatical circumstances that a language finds itself in. In other words, grammaticalisation need not be a process driven purely semantically, whereby the grammatical changes are the result solely of semantic and/or pragmatic change. The way the process developed in the case of to was (co-)determined by syntactic factors, specific to English, and by universal iconic constraints or patterns such as persistence and isomorphism.

Acknowledgments I would like to thank audiences at Zürich and Vienna University, the editors of Viewz (in which an earlier version of this paper was prepublished [Viewz 7.1, 1998]), Elizabeth Traugott and Dieter Stein, and especially my colleagues Frederike van der Leek at Amsterdam University and Bettelou Los at the Free University of Amsterdam for their valuable criticisms and comments, and for their

164

OLGA FISCHER

willingness to discuss this paper with me. The paper has prospered from it, but of course those errors and weaknesses that have no doubt remained are all my own responsibility.

Notes 1.

For a general discussion of the necessity to make a distinction between grammar change and aggregate language change, see also Lightfoot 1999: 220 ff. Even though logically this distinction (i.e. between the process of language acquisition and the historical development of language) is clear-cut, the question remains as to what provides a proper explanation for language change. It must be clear that explanations are to be found both in language use (the triggering experience) and in the theory of grammar. The importance of each for explanation all depends on how ‘deep’ one’s model of grammar is, the shallower one’s grammar is, the more an explanation will have to be found in the changing circumstances of language use (see also Lass 1997 for a discussion of explanation and grammar-depth).

2.

Bybee et al. (1994: 289 ff.) recognise three other mechanisms of semantic change that play a role in grammaticalisation (it is quite clear that for them the mechanisms of semantic change are more or less equivalent to the mechanisms found in grammaticalisation, see p. 282), i.e. (3) generalisation, (4) harmony and (5) absorption of contextual meaning. It is clear from their description that all three mechanisms are essentially metonymic in nature, with metaphor playing a subsidiary role. Indeed they conclude (p. 297): “The most important point that can be made from the discussion of mechanisms of change is that context is all-important”.

3.

Heine et al. (1991a) indeed refer to the process as a ‘chain’.

4.

It is interesting to note that Bruyn (1995), who was looking at the developments taking place in a pidgin becoming a creole (where it is believed that grammaticalisation plays an important role), and so, as it were, looking for grammaticalisation evidence from another perspective (not on the basis of selected cases from many languages as is usually done, but on the basis of full data from one language where the process might be expected to apply according to the hypothesis), found very few cases where grammaticalisation ran its full course. She found that language contact (especially substratum influence) often caused divergence (p. 241 ff.), or early abortion (p. 53 ff.), or that sometimes a development was much more abrupt than is usual in grammaticalisation cases (pp. 237–239). Her investigation shows that it is important at each stage to take into account the synchronic circumstances, which will ultimately (and freshly) decide what will happen.

5.

The principle of unidirectionality has been much in the limelight recently. The strongest adherent of the principle must be Martin Haspelmath (1999), who indeed believes it is without exceptions. This view came under strong attack at the recent (1999) conference on grammaticalisation held at the university of Potsdam, in papers delivered by among others, Johan van der Auwera (“Degrammaticalization”), Aidan Doyle (“Yesterday’s affixes as today’s clitics: functional heads and grammaticalisation in Irish”), Muriel Norde (“The final stages of grammaticalization: affixhood and beyond”), Günter Rohdenburg (“Degrammaticalization phenomena with ‘seeing’ and other adverbial conjunctions in English”), who all argued, to my mind convincingly, that true cases of degrammaticalization exist, even though they are more rare. Roger Lass (this volume) goes into the philosophical, methodological and empirical problems that this notion entails.

6.

For more details, also on the comparative development of zu in German, see Fischer (1997a). This article takes a different approach in that it considers the degrees of grammaticalisation of

UNIDIRECTIONALITY: TO BEFORE THE INFINITIVE

165

to, zu and te only from the point of view of the parameters of grammaticalisation distinguished by Lehmann (1985). 7.

See e.g. King Horn (Hall 1901: 25), te lyue, and signs of reduction in forms such as tobinde (Havelok, Smithers 1987: 56), tobe (Rolle, Psalter, Bramley 1884: 380), and tavenge (Caxton, Reynard, Blake 1970: 54). All these instances are from late Middle English, none have been found in earlier or later periods in the corpus I used (the Helsinki Corpus).

8.

Some of the editors of Viewz (in which this paper was prepublished, see Vienna English Working Papers 7.1, 5–24) raised the question here whether the developments mentioned under (7) are not “independent syntactic developments which only happen to affect the way in which the infinitive marker can be placed more or less accidently”. This is of course a question of interpretation. I do not believe they are independent because they take place at more or less the same time, and because the same developments did not take place in either Dutch or German. These developments are indeed closely tied up with the process of grammaticalisation (as described in more detail in Fischer 1997a). The same editors mention as a possible counterexample the case of the ‘group genitive’ as in the teacher of music’s room, which according to them is not a case of de-grammaticalisation. I think it is interesting to observe in this connection that Janda (1980) has indeed suggested that it is a case of de-grammaticalisation (the genitive–s being reinterpreted as a possessive pronoun).

9.

There are also cases of split infinitives found in early Middle English but these are of a different type, they involve infinitives preceded by the negative particle or by a personal pronoun. Presumably (cf. van Kemenade 1987) these particles/pronouns are still clitics, which explains their position next to the infinitive which was becoming more verbal around this time.

10.

The importance of to for the development of these new modals is also emphasised by Hopper and Traugott (1993: 81), where they write, “[t]he contiguity with to in the purposive sense must have been a major factor in the development of the future meaning in be going to as an auxiliary”. A full discussion follows on pp. 81–83. And see also Fitzmaurice (this volume) on the link between the degrammaticalization of infinitival to and the grammaticalization of to in combination with semi-auxiliaries.

11.

For more details on how this rather difficult example (of which there are three in Chaucer) should be interpreted, see Fischer 1995: 10–11.

Texts Benson, L. D. (gen.ed.). 1988. The Riverside Chaucer [3rd ed.]. Oxford: Oxford University Press. Blake, N. F. (ed.). 1970. The History of Reynard the Fox, Translated from the Dutch Original by William Caxton [EETS 263]. London: Oxford University Press. Bramley, H. R. (ed.). 1884. The Psalter or Psalms of David and Certain Canticles with a Translation and Exposition in English by Richard Rolle of Hampole. Oxford: Clarendon Press. Hall, J. 1901. King Horn. A Middle English Romance. Oxford: Clarendon. Holt, R. (ed.). 1878. The Ormulum. Vols. I and II. Oxford: Clarendon Press. Macauley, G. C. 1900–1901. The English Works of John Gower. Vols I & II. [EETS e.s. 81,82]. London: Oxford University Press.

166

OLGA FISCHER

Morris, R. 1876. Cursor Mundi. Four Versions. Part III. [EETS o.s. 62]. London: Oxford University Press. Morris R. 1965, with Pamela Gradon (eds), Dan Michel’s Ayenbite of Inwyt or Remorse of Conscience. Vol.I [re-issue of 1866 text]. [EETS o.s. 23]. London: Oxford University Press. Smithers, G. V. 1987. Havelok. Oxford: Clarendon Press.

References Brinton, L. J. 1988. The Development of English Aspectual Systems. Aspectualizers and Post-Verbal Particles. Cambridge: Cambridge University Press. Bruyn, A. 1995. Grammaticalization in Creoles: The Development of Determiners and Relative Clauses in Sranan [Studies in Language and Language Use 21]. Amsterdam: IFOTT and Dordrecht: ICG Printing. Bybee, J., Perkins, R. and Pagliuca, W. 1994. The Evolution of Grammar. Tense, Aspect, and Modality in the Languages of the World. Chicago: University of Chicago Press. Dressler, W. 1995. “Interactions between iconicity and other semiotic parameters in language”. In Iconicity in Language, R. Simone (ed.), 21–37. Amsterdam: Benjamins. Fischer, O. C. M. 1992a. “Syntax”. In The Cambridge History of the English Language. Vol. II. 1066–1476, N. Blake (ed.), 207–408. Cambridge: Cambridge University Press. Fischer, O. C. M. 1992b. “Syntactic change and borrowing: The case of the accusative and infinitive construction in English”. In Internal and External Factors in Syntactic Change, D. Stein and M. Gerritsen (eds), 17–88. Berlin: Mouton de Gruyter. Fischer, O. C. M. 1994a. “The fortunes of the Latin-type accusative and infinitive construction in Dutch and English compared”. In Language Change and Language Structure. Older Germanic Languages in a Comparative Perspective, T. Swan et al. (eds), 91–133. Berlin: Mouton de Gruyter. Fischer, O. C. M. 1994b. “The development of quasi-auxiliaries in English and changes in word order”. Neophilologus 78: 137–164. Fischer, O. C. M. 1995. “The distinction between to and bare infinitival complements in late Middle English”. Diachronica 12: 1–30. Fischer, O. C. M. 1996. “The status of to in Old English to-infinitives: A reply to Kageyama”. Lingua 99: 107–133. Fischer, O. C. M. 1997a. “The grammaticalisation of infinitival to in English compared with German and Dutch”. In Language History and Linguistic Modelling. A Festschrift for Jacek Fisiak on his 60th Birthday, R. Hickey and S. Puppel (eds), 265–280. Berlin: Mouton de Gruyter. Fischer, O. C. M. 1997b. “On the status of grammaticalisation and the diachronic dimension in explanation”. Transactions of the Philological Society 95: 149–187.

UNIDIRECTIONALITY: TO BEFORE THE INFINITIVE

167

Fischer, O. C. M. 1997c. “Verbal complementation in Early Middle English: How do the infinitives fit in?”. In English Historical Linguistics 1994, D. Britton (ed.), 247–270. Amsterdam: Benjamins. Fischer, O. C. M. 1999. “On the role played by iconicity in grammaticalisation processes”. In M. Nänny and O. C. M. Fischer (eds), 345–373. Fischer, O. C. M. and Leek, F. C. van der. 1981. “Optional vs radical re-analysis: Mechanisms of syntactic change”. Lingua 55: 301–350. Fischer, O. C. M. and Nänny, M. 1999. “Introduction: Iconicity as a creative force in language use”. In M. Nänny and O. C. M. Fischer (eds), xv-xxxvi. Gerritsen, M. 1987. Syntaktische Verandering in Kontrolezinnen, een Sociolinguistische Studie van het Brugs van de 13e tot de 17e Eeuw. PhD diss. Univ. of Leiden. Dordrecht: ICG Printing. Givón, T. 1995. “Isomorphism in the grammatical code: Cognitive and biological considerations”. In Iconicity in Language, R. Simone (ed.), 47–76. Amsterdam: Benjamins. Haiman, J. 1980. “The iconicity of grammar: Isomorphism and motivation”. Language 56: 515–540. Haiman, J. 1983. “Iconic and economic motivation”. Language 59: 781–819. Harris, A. and Campbell, L. 1995. Historical Syntax in Cross-Linguistic Perspective. Cambridge: Cambridge University Press. Haspelmath, M. 1989. “From purposive to infinitive — a universal path of grammaticization”. Folia Linguistica Historica 10: 287–310. Haspelmath, M. 1999. “Why is grammaticalization irreversible?”. Linguistics 37: 1043–1068. Heine, B., Claudi, U. and Hünnemeyer, F. 1991a. “From cognition to grammar: Evidence from African languages”. In E. C. Traugott and B. Heine (eds), vol. I, 149–188. Heine, B., Claudi, U. and Hünnemeyer, F. 1991b. Grammaticalization. A Conceptual Framework. Chicago: University of Chicago Press. Heine, B. and Reh, M. 1984. Grammaticalization and Reanalysis in African Languages. Hamburg: Helmut Buske. Holyoak, K. J. and Thagard, P. 1995. Mental Leaps. Analogy in Creative Thought. Cambridge, Mass.: MIT Press. Hopper, P. J. and Thompson, S. A. 1984 “The discourse basis for lexical categories in universal grammar”. Language 60: 703–750. Hopper, P. J. and Traugott, E. C. 1993. Grammaticalization. Cambridge: Cambridge University Press. Janda, R. D. 1980. “On the decline of declensional systems: The overall loss of OE nominal case inflections and the ME reanalysis of — es as his. In Papers from the 4th International Conference on Historical Linguistics, E. C. Traugott et al. (eds), 243–253. Amsterdam: Benjamins. Jespersen, O. 1927. A Modern English Grammar on Historical Principles. Part III Copenhagen: Munksgaard.

168

OLGA FISCHER

Joseph, B. D. 1992. “Diachronic explanation: Putting speakers back into the picture”. In Explanation in Historical Linguistics, G. W. Davis and G. K. Iverson (eds), 123–144. Amsterdam: Benjamins. Kemenade, A. van. 1987. Syntactic Case and Morphological Case in the History of English. Dordrecht: Foris. Kytö, M. 1991. Manual to the Diachronic Part of ‘The Helsinki Corpus of English Texts’: Coding Conventions and Lists of Source Texts. University of Helsinki, Helsinki. Lass, R. 1997. Historical Linguistics and Language Change. Cambridge: Cambridge University Press. Leek, F. C. van der. 1992. “Significant syntax: The case of exceptional passives”. Dutch Working Papers in English Language and Linguistics 27: 1–28. Lehmann, C. 1985. “Grammaticalization: Synchronic variation and diachronic change”. Lingua e Stile 20: 303–318. Lehmann, C. 1991. “Grammaticalization and related changes in contemporary German”. In Traugott and Heine (eds), vol. II, 493–535. Lightfoot, D. W. 1979. Principles of Diachronic Syntax. Cambridge: Cambridge University Press. Lightfoot, D. W. 1991. How to Set Parameters: Arguments from Language Change. Cambridge Mass.: MIT Press. Lightfoot, D. W. 1999. The Development of Language. Acquisition, Change, and Evolution. Oxford. Blackwell. Los, B. 1998. “The rise of the to-infinitive as verb complement”. English Language and Linguistics 2: 1–36. Manabe, K. 1989. The Syntactic and Stylistic Development of the Infinitive in Middle English. Fukuoka: Kyushu University Press. Mustanoja, T. F. 1960. A Middle English Syntax. Part I. Helsinki: Société Néophilologique. Nänny, M. and Fischer, O. C. M. (eds). 1999. Form Miming Meaning: Iconicity in Language and Literature. Amsterdam: Benjamins. Overdiep, G. S. 1935. Zeventiende Eeuwse Syntaxis. Groningen: Wolters Plank, F. 1979. “Ikonisierung und De-Ikonisierung als Prinzipien des Sprachwandels”. Sprachwissenschaft 4: 121–158. Plank, F. 1984. “The modals story retold”. Studies in Language 8: 305–364. Rubba, J. 1994. “Grammaticalization as semantic change: A case study of preposition development”. In Perspectives on Grammaticalization, W. Pagliuca (ed.), 81–102. Amsterdam: Benjamins. Stoett, F. A. 1909. [2nd ed.]. Middelnederlandsche Spraakkunst. Syntaxis. The Hague: Nijhoff. Traugott, E. C. 1993. “Subjectification and the development of epistemic meaning: The case of promise and threaten”. In Modality in Germanic Languages, T. Swan and O. Jansen Westwik (eds), 185–210. Berlin: Mouton de Gruyter. Traugott, E. C. 1995. “Subjectification in grammaticalization”. In Subjectivity and Subjectivisation in Language, D. Stein and S. Wright (eds), 31–54. Cambridge: Cambridge University Press.

UNIDIRECTIONALITY: TO BEFORE THE INFINITIVE

169

Traugott, E. C. and Heine, B. 1991. (eds). Approaches to Grammaticalization. Vols I and II, Amsterdam: Benjamins. Vincent, N. 1995. “Exaptation and grammaticalization”. In Historical Linguistics 1993, H. Andersen (ed.), 433–445. Amsterdam: Benjamins. Visser, F. Th. 1963–73. An Historical Syntax of the English Language. Vols. I-IIIb. Leiden: E. J. Brill.

Remarks on the de-grammaticalisation of infinitival to in present-day American English Susan Fitzmaurice Northern Arizona University

1.

Introduction

Infinitives split by negatives are becoming more noticeable in American English, as exemplified by the following utterances transcribed from television broadcasts: (1)

We will send enough troops to not let Macedonia shut down its borders. (William Cohen, NBC Today, April 5, 1999)

(2)

You have to learn to not let it start. (The Puzzle Place, PBS TV, March 16, 1999)

In addition, the coalescence of the infinitive to into the quasi-auxiliaries going to and have to seems to be reinforced by the occurrence in children’s language of utterances like the following: (3)

You have to not say that word.

(4)

I’m going to not eat strawberries.

In this brief essay, I will examine the patterns of occurrence and pragmatic motivations for these constructions in present-day American spoken English. I will argue that the gradual regularisation of quasi-modals as auxiliary verbs on the one hand, and the pragmatic utility of the negative split infinitive on the other hand provide support for the claim that infinitival to is in the process of de-grammaticalising. I will discuss plausible semantic-pragmatic motivations for the de-grammaticalisation of infinitive to, and comment on the extent to which these motivations are reflected in speakers’ behaviour. Finally, I attempt to connect the processes of de-grammaticalisation and the ongoing development of auxiliaries such as have to and be going to.1

172 2.

SUSAN FITZMAURICE

Background

2.1 Reversal of grammaticalisation of infinitival to: Fischer (1997, this volume) argues that the infinitive marker to exhibited the following signs of grammaticalisation in late Old English and early Middle English: – – –

–

strengthening of to by for (e.g. for to 1420–1500); phonetic reduction of to; loss of semantic integrity (e.g. use of to in non-purposive or goal-oriented constructions, such as the stative: And in my barm ther lith to wepe / Thi child and myn ‘And in my bosom there lies weeping thy child and mine’. (Gower, Confessio. Am. III 302); occurrence of to-infinitive after prepositions other than for.

However, by the end of the Middle English period, the trend seemed to be beginning to reverse itself: – –

– –

to ceases to co-occur with for in the standard language; split infinitives begin to appear, e.g. Blessid be þou lord off hevyn. / That suche grace hath sent to his /Synfull men for to þus lede / In paradice (Cursor Mundi, Laud Ms. 18440–44) — 14th c.; absence of ‘reduction of scope’ you can use this shampoo to wash your hair and clean your clothes; no loss of semantic integrity: to is still strongly purposeful (He threatened to kill her — agentive interpretation as default).

For comparison, Fischer (1997) examines the fortunes of the infinitive marker te in Dutch, observing that by contrast with English, Dutch te exhibits typical features of grammaticalisation, for example, increased bondedness with the base form of the verb, loss of semantic integrity, and the reduction of the scope of the infinitive. It is striking that Fischer identifies as a key construction in the degrammaticalisation of to, the split infinitive. Of course, the split infinitive most commonly referred to in this context is the infinitive split by adverbial elements, as exemplified in ‘to boldly go’. This construction is a traditional bugbear of traditional grammarians, and therefore its high profile militates against an objective assessment of the progress of the de-grammaticalisation of the infinitive marker.2 By contrast, the negative split infinitive — the construction split by a negative operator — has received rather less attention from prescriptivists and has thus remained less obtrusive in speakers’ conscious linguistic behaviour. This

DEGRAMMATICALISATION OF TO IN AMERICAN ENGLISH

173

type of split infinitive therefore seems worth further investigation. It seems reasonable to argue that the development of quasi-auxiliaries with to such as want to, be going to, be supposed to, be to, and have to is licensed by the progressive de-grammaticalisation of the infinitive marker to. That is, it seems that to is allowed to coalesce with selected verbs to create periphrastic auxiliaries, thus becoming part of the grammaticalisation process of another construction, by virtue of its release from the infinitive construction. One way of connecting the two processes — grammaticalisation of the to-auxiliaries and degrammaticalisation of the infinitive to — is to examine the effects and consequences of the isolation of to. Research on the history of English quasi-auxiliaries from the perspective of grammaticalisation has been growing steadily (for example, Sweetser 1990; Hopper and Traugott 1993; Fischer 1994; Brinton 1988, 1991). Fischer (this volume) and Traugott and Hopper (1993: 81) underline the importance of the purposive function of to in the development of the new modal auxiliaries with ‘future’ meaning, such as be going to. Traugott and Hopper (1993: 82) note, “we hypothesize that the future meaning of be going to was derived by the semanticisation of the dual inferences of later time indexed by go and purposive to, not from go alone”. They offer a full discussion of the history of be going to. Plank (1984: 338–339) offers a different perspective on the coalescence into modals of to and the preceding verb (want, have, be going). He notes that these verbs are unlike auxiliaries in that they occur with to, but that they “allow the conjunction [i.e. to ] to be reduced and contracted in informal speech” even when this is not fast speech, and before pauses, indicating that this to has grammaticalised and become, as it were, affixed to the matrix verb. Presumably this amalgamation is possible because both to and the matrix verb express future modality. This amalgamation of verb with the future/purposive to involves a process by which each element itself undergoes the semantic bleaching associated with grammaticalisation. Olga Fischer’s work on the de-grammaticalisation of English infinitive to (for example, Fischer 1997, this volume) explores the interdependence of the history of the infinitive and the semi-auxiliaries; I seek to contribute to the literature by exploring the ways in which the regularisation of the English quasiauxiliaries with to interacts with the conventionalisation of the negative split infinitive, and their impact on the process of de-grammaticalisation of the infinitive marker to.

174

SUSAN FITZMAURICE

2.2 Systems of polarity and scope in negative VPs The nature of the interaction of the quasi-auxiliaries with to and the negative split infinitive constructions centres on their relation to negation, and to the systems of polarity and scope in the VP. The formal structural description of each construction is dependent on the assignment of syntactic boundaries. In other words, the regularisation of to quasi-auxiliaries involves the dissolution of the boundary between the governing verb (have, want) and the complement VP signalled by to. That is, the grammaticalisation of a quasi-auxiliary such as have to in part consists in the shift from a structural description illustrated in (5a) to that illustrated in (5b): (5)

a. b.

[X have [to VP]COMP ]S [X have to V]S

The conventionalisation of the negative split infinitive is similar; that is, the placement of the negator after to, directly before the main verb in the subordinate complement isolates the infinitive to from its verb, arguably weakening the traditional boundary between main clause and subordinate complement, even shifting it. Compare the bracketing in (6a) with that in (6b): (6)

a. b.

She decided not [to deliver the plan] She decided to [not deliver the plan]

In (6a), the negator not has scope over the infinitive VP and the infinitive marker to which marks the VP boundary. By contrast, in (6b) the negator not falls within the scope of to, and within the VP proper. It is possible that the marker to loses its grammatical infinitive meaning in structures like (6b), and assumes a pragmatic, purposive, meaning that is adverbial in flavour and force. It contributes to the positively negative meaning which is conveyed when not is dominated in this way by to. In this section, I examine the role of the negator in both constructions, with the aim of demonstrating the contrary pressures placed on the infinitive marker as a consequence of the polarity and scope of the negative. Because the item that splits the quasi-auxiliary with to and negative infinitive constructions under examination is a negator, and because the grammar of negation involves semantic issues like scope and polarity which also affect verb grammar, I will review ways in which negative scope and polarity may be affected by negative placement with respect to the infinitive to. Huddleston (1984: 135) considers polarity a VP system as it specifies the structures in the VP that express negation. Positive polarity is structurally

DEGRAMMATICALISATION OF TO IN AMERICAN ENGLISH

175

unmarked, whereas negative polarity is marked, either analytically, by “the word not functioning as dependent in the structure of the phrase, as in is not working”, or inflectionally, by “a negative form of the verb, as in isn’t working”. Negative polarity is relevant to this discussion because the placement of the infinitive marker to relative to the boundary of the VP may have particular consequences for the degree of negativeness, or the place of the negative on the polar continuum. Polarity has most commonly been explored with respect to synthetic and analytic negation, so that synthetic negation conveys positive negation of a proposition, e.g. she’s unkind, analytic negation conveys neutral negation of a proposition, e.g. she’s not kind, and a combination of the two conveys negation of a negative proposition, e.g. she’s not unkind. However, the force of the latter expression is neither an unequivocal denial of the negative proposition, nor an affirmation of the positive proposition (see Matthews 1981; Adamson 1990 for examples, and Horn 1989: 273–308 for extensive discussion). Polarity is often affected by the scope of a negator, and therefore I will combine their discussion in what follows. 2.2.1 Negation of semi-auxiliaries with to Scope and polarity are relevant in the construal of negation in semi-auxiliaries with to. Semi-auxiliaries like going to, want to and supposed to accommodate a choice of placement of the negator because of their periphrastic composition.3 The examples in (7) illustrate the different force borne by the negator depending upon its placement: (7)

a. b.

She’s not/isn’t going to make friends She’s going to not make friends

— neutral negation — positive negation

The positioning of the negator after the first element of the quasi-auxiliary (7a) places the whole VP within its scope, and is construed as straightforwardly negative. By contrast, in (7b) the positioning of the negator after the quasi-modal and adjacent to the main verb make removes be going to from the scope of the negative. Note that this is the normal position for a negator in an auxiliary–verb construction (does not make).4 The very proximity of not to the proposition to be negated appears to offer an opportunity for greater emphasis, and thus strengthens the force of the negation. By contrast, the opposite effect is reached in the construction marked by negative raising (7a). The pattern introduced above in (7) applies equally to quasi-auxiliaries such as be supposed to and be to with similar emphatic effect.5

176

SUSAN FITZMAURICE

(8)

a. b.

She’s not/isn’t supposed to answer the question. She’s supposed to not answer the question.

(9)

a. b.

She’s not to answer the question. She’s to not answer the question.

Have to and need to appear less flexible than these quasi-auxiliaries with respect to the placement of negators and the resulting scope of negation. Firstly, not may not appear immediately before the semi-auxiliary without being ungrammatical ((10a), (11a)). Placing not between the two elements (as in (10b) and (11b)) creates an expression that does not participate in the same system of polarity; that is, it appears to imply that the subject has a choice of whether or not to answer the question.6 This construction as well as the negative constructed with do-support ((10b), (11b)) cannot be interpreted as conveying positive or unequivocal negation. The only negative construction with need to and have to which conveys definite positive negative force is one in which the negator occurs after to, immediately before the main verb, as in (10c) and (11c): (10)

a. *She not has to answer the question. b. She has not to/does not have to answer the question. c. She has to not answer the question.

(11)

a. *She not needs to answer the question. b. She needs not to/does not need to answer the question. c. She needs to not answer the question.

The semi-auxiliary have to presents problems of analysis in the negative; I leave for now the question of how need (to) acts in the negative.7 In semantic-pragmatic terms, have to, like other modals of necessity, bears subjective meaning, notably in its speaker-centred obligative force. But as illustrated in (10c), it does not behave syntactically like other modals of necessity such as must. Specifically, the negative sentence (10c) says what it means; neither the modal nor its modality is negated. By contrast, in the similar sentence (You must not say that word) with must, while the modal is negated the modality is not (Palmer 1995). The equivocal status of have to as a modal is highlighted by the fact that when negated using do, and when placed into a question using do, the string have to changes meaning. The semantic-pragmatic effect of negative placement using do is to negate both the verb and its modal force, replacing the emphatic imposition of obligation with a choice of whether to oblige or not. The meaning change thus consists of a loss of force.

DEGRAMMATICALISATION OF TO IN AMERICAN ENGLISH

177

2.2.2 Negation and infinitive VPs In present-day English the usual position for not in non-tensed VPs is at the beginning of the phrase (not to be working, not having seen) (Huddleston 1984: 136). As we shall see, negation associated with this position appears less markedly negative than negation which places not after the infinitive marker in infinitive VPs because of the pragmatic strength lent by the proximity of not to the VP in its scope. Scope is a semantic feature that has been shown to be important in the construal of modal meaning (Nordlinger and Traugott 1997). It is relevant to the relation of the negator to the proposition expressed in the VP, as it concerns the (variable) extent of negation in a construction, i.e. what falls within the domain or influence of the negator. In the following pair of sentences, the placement of the negative does not change the scope of the negative: (12)

a. b.

She decided not to identify the culprit but to ignore him (>> she did not identify him) She decided to not identify the culprit but to punish him (>> she did not identify him)

In the first example (12a) not falls outside the domain of the matrix verb, and within that of the complement VP. The result is that the infinitive complement falls within the scope of not to yield subclausal negation, and the interpretation that the proposition in the complement is negated. That is, it has the entailment: ‘she did not identify the culprit’. In the second example (12b), although not falls inside the infinitive complement rather than being placed at the boundary of the infinitive VP, the scope of the negative is no different from that in the first example; it shares the entailment, ‘she did not identify the culprit’. However, it might be argued that the greater proximity of the negator to the subordinate VP by virtue of its placement after the infinitive marker to gives an impression of greater negative force. In addition, the negator falls within the scope of to, the latter no longer infinitive because of its isolation from the verb, but purposive. So not only is the negative polarity of the second example (12b) greater than the negative polarity of the first (12a), it is pragmatically more forceful. This force might be somewhat elaborately paraphrased thus: ‘she decided to make no effort in order to come to some identification of the culprit’.8 In the following infinitive constructions governed by the predicative adjective careful, the semantic scope of the negative does not vary with the placement of not. However, the fact that not is included in the scope of to in (13b) alters the force of the negative meaning inferred. In this example, the purposive force of to seems quite clear; the negative infinitive VP indicates why ‘she was careful’:

178

SUSAN FITZMAURICE

(13)

a. b.

She was careful not to identify the culprit She was careful to not identify the culprit

The positioning of not at the boundary of the infinitive complement of the predicative adjective specifies the scope of the negative as subclausal (i.e. she was careful to not X). In (13b) the scope of the negative is restricted to the domain of the infinitive complement only. Despite the fact that the entailment is the same for both examples, namely, she did not identify the culprit, the negative force conveyed by (13b) is stronger than the force conveyed by (13a). Thus, in addition to altering the scope, the positioning of the negator may have some influence on the strength of the negative and on the expression of purposeful action. The adjacency of the negator to the verb in the sentence in (13b) appears to encourage a more firmly negative reading than the position of the negator on the infinitive clause boundary. One likely contributor to this force is the investment of purposive force in to as a consequence of the interruption of the infinitive verb sequence by the negator.

3.

Evidence for the de-grammaticalisation of infinitive to in present-day American English

3.1 Negative split infinitive and purposive to In order to examine the strength or weakness of to as an infinitive marker, it may be useful to examine the occurrence of the split negative infinitive in American English. If a range of verbs and predicates accommodate the split negative infinitive for a range of speakers in a range of different conversational situations, this may indicate the independence of to from the (bare) infinitive verb.9 A search of the American Conversation Register in the Longman Spoken and Written English Corpus (Biber, Johansson, Leech, Conrad and Finegan 1999), yields a number of split negative infinitive constructions, including the following:10 (14)

a. b. c. d.

I can’t afford to not do math again trust me to not lock the keys in the car It is safer for her to not deal with the quilt I would really like to not work when I have kids

The first example (14a) combines two negatives for extra emphasis; a reasonable, but somewhat weaker paraphrase might be, ‘I must do math again’. In the second example (14b) the infinitive VP is a complement of a rhetorical imperative. This

DEGRAMMATICALISATION OF TO IN AMERICAN ENGLISH

179

is not a directive so much as an exclamation, and its function easily accommodates the extra pragmatic strength offered by the placement of not. The positive negative polarity of (14c) lends emphasis to the conversational implicature that ‘the subject should not be allowed to deal with the quilt’, not ‘that it would benefit her not to deal with the object’. (14d) is strongly self-expressive, by virtue of both the intensifier adverb really and the positioning of not. This example is further strengthened by not having the negator raised into the matrix VP, which would be possible without changing the basic meaning: ‘I wouldn’t really like to work when I have kids’. The semantic-pragmatic difference between the raised construction just given and the corpus token in (14d) lies in the greater certainty associated with the negation in the latter example. Doug Biber’s frequency count of the distribution of not relative to the infinitive marker to reveals that the split infinitive occurs more frequently in American conversation than in British conversation, and less frequently than the unsplit (not to V) in American conversation (Fitzmaurice, Biber and Reppen 1998). The figures are reproduced in Table 1. Table 1. Distribution of not relative to the infinitive to (approximate frequency counts per one million words)

British Conversation (c. 4.2 million words) American Conversation (c. 5 million words)

not to

to not

07 80

01 10

The figures in Table 1 suggest that while the split infinitive is infrequent in American conversation, it does occur. Although the proportion of to not constructions is small, at a ratio of 1 : 8, I would argue that if examined in the light of change in progress, these data are not negligible. In addition to occurring in conversation, the construction is appearing regularly in the speech of middleclass, educated adult Americans, from teachers and academics to announcers, critics, and politicians who are regularly given voice in the American broadcast media. (15) and (16) below contain samples of examples which I myself collected recently from these contexts: (15)

a.

Some departments seem to not be very interested in developing a writing intensive course. (male university professor of Linguistics I, department meeting, Jan. 13, 1999)

180

SUSAN FITZMAURICE

b.

c. d.

It’s only superficial to not have objectives as outcomes for this [writing] course if we’re focusing on the process [of writing].’ (female university professor of Sociology, Jan. 7, 1999) The students hope to not get calls at that office. (male university professor of Linguistics II, Jan. 14, 1999) It forces them to not run large classes and let them do journals instead. (male university professor of Linguistics I, Jan. 14, 1999)

These examples represent rather different kinds of structures in the matrix clause, from raising verbs like seem in (15a), predicative adjectives such as superficial in (15b), epistemic verbs like hope (15c), to verbs like force (15d) which require raising to subject. This variation does not seem to alter the pragmatic strength of the negative’s place in splitting the infinitive. The occurrence of the construction in the expressive, pragmatically fluctuating speech settings of public radio news magazine programs should not be surprising. Placement of not immediately before the clausal complement arguably focuses the negator with respect to the adjacent verb, with the rhetorical effect of emphasis on the connection of negator and phrase governed by the verb. Because the variable placement of the negator may result in a variation of emphasis and accentuation, increasing use of the to not construction seems pragmatically motivated. In these expressions, not has scope over the complement it precedes, its adjacency to the embedded verb lending it extra emphatic force. As in the examples in (15), the self-collected examples in (16) exhibit the range of matrix verb and predicate styles: (16)

a.

b. c.

d.

You should also be briefed to not give him the keys to the cabinet. (male interviewee, (National Public Radio NPR) ‘All things considered’, Sep. 28 1998) It’s going to be hard to not take advice. (male interviewee, (NPR) ‘All things considered’, Sep. 28 1998) He tended to not like people who refused to be subservient to him. (male Sinatra biographer (NPR) ‘Morning Edition’, Dec. 9 1998) We didn’t expect Newt Gingrich to not be speaker of the House. (Cokie Roberts, ‘A Presidency at the Crossroads’, ABC TV, Dec. 18, 1998)

DEGRAMMATICALISATION OF TO IN AMERICAN ENGLISH

e.

181

People like myself will vote to not remove the president from office. (Senator Jay Rockefeller, W. Virginia (D), ‘Meet the Press’, NBC TV, Feb. 7, 1999)

The construction also appears in written reports of spoken language. The examples in (17) are quotations of speech, reported in my local paper, the Arizona Daily Sun. Both illustrate the greater rhetorical force of placing the negator immediately adjacent to the proposition being negated. The choice of lexical predicate (be entitled) in (17a) enhances the emphatic, modal nature of the negative expression. The positioning of not in second place in the infinitive VP as a subject nominal in (17b) gives it the heavy stress in an iambic foot, making the negative prominent in the utterance. (17)

a.

b.

‘ “Disabled students are entitled to not be subject to discrimination. They are entitled to reasonable accommodations to give them every opportunity that a student without disabilities has,” said Jane Jarrow, a nationally renowned expert on the ADA in higher education.’ (‘Conference considers reasonable accommodation for disabled students’, Arizona Daily Sun, Sat., Nov. 7, 1998) ‘This is 1999, we’re about to go into the year 2000. To not have videotape available for senators, and maybe even beyond that, is not … defensible’, (attributed to Trent Lott, Senate Republican leader. ‘Sex, Lies — and more videotape’, Arizona Daily Sun, Fri., Jan. 29, 1999)

Finally, I have encountered the construction in the language of student papers. The following, final example is from an essay written by a female undergraduate English major: (18)

‘Humankind is constructed to detest suffering, to the extreme of actually ending a life just to not have to witness the suffering of that creature for a moment longer than necessary.’ (final essay for English 253: World Authors. Female English student, NAU, May 12, 1999)

This rather odd example is a sample of energetic opinionating. Note that the infinitive VP is adverbial in function; it is itself modified by the scalar adverb just, and the infinitive VP consists of semi-auxiliary have to. The appearance of the split negative infinitive, albeit in a vigorous rhetorical style, in student

182

SUSAN FITZMAURICE

academic writing suggests that it may be used in increasingly formal situational contexts and registers, including written registers. 3.2 Semi-auxiliaries with to Although semi-auxiliaries like have to, want to, be going to, etc. are well-attested in speech and writing in positive declarative constructions, and are thus routinely treated, in semantic terms at least, as more modal-like than verb-like (e.g. Sweetser 1990; Kolln and Funk 1998), it is not clear how well they meet grammatical criteria, such as the so-called NICE criteria for operators/auxiliaries (see Denison 1993, Chapter 10 for details and caveats). Let us briefly review the extent to which these semi-auxiliaries in American English do satisfy these criteria for auxiliary status. Have to, want to, be going to and supposed to appear to act like auxiliaries with respect to Emphasis and Code (post-verbal ellipsis). They may convey emphasis without do support (I have to/want to/am supposed to/need to leave), and they can appear in a reduced or elided form (Do I have to leave? Yes you do/yes you have to, etc.). However, few of these semi-auxiliaries can participate in full Inversion with the subject to form interrogatives (*Have to you leave? *Want to you leave? but Are you supposed to leave? Are you going to leave?). Instead have to and want to require do support (Do you have to leave? Don’t you have to leave? Do you want to leave? Don’t you want to leave?).11 If the semi-auxiliaries with to could be negated directly (Negation), without do support, this ability would provide support for the observation that the semi-auxiliaries had moved another step further toward auxiliary status, as the placement of the negator would further separate to from the (bare) infinitive verb. This would arguably underline the coalescence of the verb and particle elements, and thus provide support for claiming the ongoing grammaticalisation of the form. Although there is little readily attested evidence of have to with direct negation in American English speech, there is good evidence for the use of supposed to, want to and used to with following not. The American Conversation Register in the Longman Spoken and Written English Corpus (Biber, Johansson, Leech, Conrad and Finegan 1999), has examples such as the following: (19)

a. b. c. d.

Both wanted to not find out about the other I know that I’m supposed to not allow him enough time to talk They used to not be like that I don’t think you ought to not eat

Other examples that I collected recently include the following:

DEGRAMMATICALISATION OF TO IN AMERICAN ENGLISH

(20)

a.

b.

183

This was my nervous breakdown last year, but you’ve got to not talk about it. (white woman, about 30 years old. Overheard at MLA Conference, Dec. 29 1998) You want to not be confused about whose mother is who at kids’ birthday parties. (female junior high school English teacher, Flagstaff, Jan. 9, 1999)

These real-life examples illustrate how the elements in the periphrastic semiauxiliaries coalesce, isolating not. The consequence of such patterns of coalescence is a possible reanalysis of the structure along the lines described in Section 2 above, so that the clause boundary is removed by the coalescence of to with the preceding adjacent verb. In this scenario, the following verb is no longer a bare infinitive, but a main verb supported by an auxiliary containing to. Examples such as these offer some evidence for the ongoing grammaticalisation of the semi-auxiliaries at the same time as the de-grammaticalisation of the infinitive marker to.

4.

Concluding squib

I have offered evidence from present-day American speech (and some writing) to support the ongoing de-grammaticalisation of the infinitive marker to. This evidence consists of attested occurrences in adult speech in a range of situational contexts of split negative infinitives, as well as the direct negation of semiauxiliaries. Further work on the de-grammaticalisation of ‘infinitival’ to might include the investigation of the occurrence and variation of infinitive VPs which have adverbial functions, in order to explore the extent to which the purposive force that Traugott and Hopper (1993) and this paper claim is acquired by postinfinitive to is connected with the infinitive’s capacity for adverbial meaning.

Notes 1.

Nichols (1986) offers the English infinitival marker to, which attaches to the main verb (wanna, gonna), and the development of the English split infinitive to illustrate the typological ‘migration’ of affixes from dependent to head. Thus the process of de-grammaticalisation of one grammatical function (infinitive) that is involved in the grammaticalisation of another (auxiliary verbs) may occur in accordance with general typological tendencies. The discussion of this issue is beyond the scope of the present paper, but the connection appears to support

184

SUSAN FITZMAURICE challenges to the strength of unidirectionality as a feature of grammaticalisation. I am grateful to Anette Rosenbach for bringing Nichols’ work to my attention.

2.

For an entertaining digest of the ‘split infinitive story’, see David Crystal (1995: 195).

3.

Although there are restrictions. In particular, not cannot be inserted between going and to without turning go into a main verb and to into an infinitive marker proper: *She’s going not to make friends (but to influence people) The placement of the negator in this position appears to violate the semantic integrity of the periphrastic modal.

4.

It is also possible to add a negator to this example, to create a hedge; an equivocal denial of positive negation: She’s not/isn’t going to not make friends. This double negative and the fact that it may be construed quite unproblematically makes use of the plasticity of the be going to construction as one that is not fully grammaticalised as an auxiliary.

5.

Note, however, that the periphrastic quasi-auxiliary be supposed to may accommodate a third alternative for positioning the negator, with different results: She’s supposed not to make friends. The force of the negative in this example seems to be positive, but less positive than the example of positive negation in the text. In British English, this sentence may be ambiguous, such that it could be interpreted as an agentless passive: ‘They suppose that she does not make friends’. In the same way that two negators may be added to the be going to construction, two may be added to both be supposed to and be to, with the following results and consequences for polarity: She’s not supposed to not answer the question. She’s not to not answer the question. In both cases, the addition of a negator creates the negation of the second negator, so that each is construed as a hedge; an equivocation. Both might be paraphrased as follows: ‘Although she might not answer the question directly, she must respond in some fashion’.

6.

An alternative treatment is to construe the have not to construction as part of a very conservative variety of British English, as ‘she does not have to’. Thus old-fashioned British English has the following possible negative declarative with have to, in which dummy auxiliary do is not used: You have not to go home now (= ‘you need not go home now’). This is related to the use of possessive have without a dummy auxiliary do in the interrogative, Have you time to look at this paper?

7.

Need (to) has a long, complex history, and so is not directly comparable with the case of have to.

8.

In speech, the placement of not as in (12b) allows the negative to fall in a position in which the nuclear pitch and an accompanying rising tone convey the force of utterance.

9.

The split negative infinitive types examined here may well be part of a broader group of constructions in which the negator not interacts with to, whether infinitive marker or preposition. For example, the following utterance makes use of the same strategy of placing the negative for maximum emphasis: She may have reconciled herself to not getting everything. The placement of the negator ensures that it bounds the gerund headed by getting. It is possible that the regular placement of the negator after infinitive to is influenced by this non-infinitive construction.

DEGRAMMATICALISATION OF TO IN AMERICAN ENGLISH

185

10.

I am grateful to Doug Biber for the information. The sampling for this sub-corpus was demographically-based, so that informants were selected to represent a range of age, sex, social group and regional differences. The informants tape-recorded all their conversational interactions over a week. The American sub-corpus has the conversations of 491 speakers, 292 female and 199 male. There were 112 speakers under 20 years old, 114 between 21 and 30 years old, and 265 31 years and older (Biber, Johansson, Leech, Conrad and Finegan 2000).

11.

In British English, have to does not need do support to form interrogatives (Have you to leave?), but it is arguable that this presents a counter-example to the argument for greater grammaticalisation because the integrity of the coalesced have to is challenged by the fact that the construction can be divided.

References Adamson, S. 1990. “The what of the language”. In The State of the Language, C. Ricks and L. Michaels (eds), 503–514. London: Faber and Faber. Biber, D., Johansson, S., Leech, G., Conrad, S. and Finegan, E. 1999. The Longman Grammar of Spoken and Written English. London: Longman. Brinton, L. 1988. The Development of English Aspectual Systems: Aspectualizers and PostVerbal Particles. Cambridge: Cambridge University Press. Brinton, L. 1991. The origin and development of quasimodal have to in English. Unpublished Ms, University of British Columbia. Crystal, D. 1995. The Cambridge Encyclopedia of the English Language. Cambridge: Cambridge University Press. Denison, D. 1993. English Historical Syntax: Verbal Constructions. London: Longman. Fischer, O. This volume. “Grammaticalisation: Unidirectional, non-reversible? The case of to before the infinitive in English”. Fischer, O. 1977. “The grammaticalisation of infinitival to in English compared with German and Dutch”. In Language History and Linguistic Modelling. A Festschrift for Jacek Fisiak on his 60th Birthday, R. Hickey and S. Puppel (eds), 265–80. Berlin/New York: Mouton de Gruyter. Fischer, O. 1994. “The development of quasi-auxiliaries in English and changes in word order”. Neophilologus 78: 137–164. Fitzmaurice, S., Biber, D. and Reppen R. 1998. “Subjectivity, scope and the interaction of negation and modality in English: Perspectives from grammaticalisation, corpus linguistics and child language”. Panel discussion presented at LASSO, Tempe, AZ. October 11, 1998. Hopper, P. J. and Traugott, E. C. 1993. Grammaticalisation. Cambridge: Cambridge University Press. Horn, L. R. 1989. A Natural History of Negation. Chicago/London: The University of Chicago Press. Huddleston, R. 1984. Introduction to the Grammar of English. Cambridge: Cambridge University Press.

186

SUSAN FITZMAURICE

Kolln, M. and Funk, R. 1998. Understanding English Grammar, 5th edition. New York: Allen & Unwin. Matthews, P. H. 1981. Syntax. Cambridge: Cambridge University Press. Nichols, J. 1986. “Head-marking and dependent-marking grammar”. Language 62: 56–119. Nordlinger, R. and Traugott, E. C. 1997. “Scope and development of epistemic modality: Evidence from ought to”. English Language and Linguistics 1 (2): 295–317. Palmer, F. R. 1995. “Negation and the modals of possibility and necessity”. In Modality in Grammar and Discourse, J. Bybee and S. Fleischman (eds), 453–471. Amsterdam: John Benjamins. Plank, F. 1984. “The modals story retold”. Studies in Language 8: 305–364. Sweetser, E. 1990. From Etymology to Pragmatics: Metaphorical and Cultural Aspects of Semantic Structure. Cambridge: Cambridge University Press.

The role of person and position in Old English Elly van Gelderen Arizona State University

1.

Introduction

Throughout the history of English, agreement endings disappear (as well as Case) and word order becomes more fixed. This change from synthetic to analytic can be accounted for in different ways. In a Minimalist framework (Chomsky 1995), one could argue that more Functional Categories are activated in the later stages. It has been claimed (van Gelderen 1997a, b) that in constructions where verbs move to a Functional Category such as the Complementizer position, there is less agreement on the verb. In this paper, I examine these constructions in more detail and examine whether there are differences between the agreement on first, second and third person verbs. I show that verbs with first and second person subjects are more likely to lose their endings than those connected with third person subjects. I also discuss pro-drop, which is more common with third than with first and second person. This fits if third person features are somehow stronger on the verb (Safir 1985; Jaeggli and Safir 1989); or, alternatively, if third person endings are still pronouns. The Old English texts I examine are Beowulf, The Junius Manuscript, The Exeter Book, The Lindisfarne Gospel of St. Matthew (which is a gloss to the Latin text), some of Alfred’s works and Aelfric’s Homilies. Thus, poetry as well as prose is included; Northumbrian as well as West-Saxon; translation from Latin as well as original work.1 In Section 2, I outline some assumptions I make and provide some background on features and inflection. In Section 3, I show that first and second person verbs lose agreement before third person ones. Section 4 discusses the relationship between null subjects and verbal inflection.

188 2.

ELLY VAN GELDEREN

Rich inflection and movement

I assume that Old English is a language with movement of the verb to the complementizer position when this position is available (e.g. in main clause questions, cf. van Kemenade 1987). This accounts for the Verb-second nature of Old English. Thus, constructions where a verb precedes a subject are ones where the verb is in the Complementizer position. In a Minimalist framework (cf. Chomsky 1995), verbs (and nouns) move to the I(inflection) ) and C(omplementizer) positions in order to check certain features. If the features are ‘strong’, the movement has to be overt, i.e. visible. Within this framework, Bobaljik and Jonas (1993) argue that movement is in fact triggered by overt inflection. My data indicate another scenario. I show that verbs lose inflection with first and second person pronouns when the latter follow the verb, i.e. when the verb is in C. To account for the data, I argue that first and second person features are less specified and that this is especially noticeable (to be expanded on below) in C. This is also the reason why null subjects occur with third person but not with first and second. In Old English, the present and past (preterite) tenses are distinguished in that, in the past tense, the weak verbs have a suffix containing a dental, the strong verbs have a stem change and the irregular verbs have suppletive forms. For instance, a first person present is ic lufie ‘I love’ and a first person past is ic lufode ‘I loved’. Person and number are generally distinguished for the singular present tense, for instance, ic lufie ‘I love’, þu lufast ‘you love’, he/heo/hit lufað ‘he/she/it loves’, but not for the plural: we lufiað ‘we love’, ge lufiað ‘you- love’, hi lufiað ‘they love’ (cf. Quirk and Wrenn 1955 [1977]: 43; also Campbell 1959: 295–351, and Mitchell 1985, who follows Campbell). The subjunctive has a singular -e ending and a plural -en. There are of course many variants and many different verb classes (as will be shown below) but they all make similar distinctions. For weak verbs in the past tense, the second person singular (-est) is different from the first and third persons singular (-e) and from the plural (-on). Strong verbs in the past tense distinguish first and third from second singular as well. In the imperative and subjunctive moods, only singular and plural are distinguished. Thus, number is consistently distinguished but person is only sometimes. In constructions where the subject follows the verb, verbal agreement is often reduced. Jespersen (1942: 15) writes “[i]n OE a difference is made in the plural, according as the verb precedes we or ge or not” and Quirk and Wrenn (1955 [1977]: 42) remark that “[t]here are alternative 1 and 2 forms of all tenses and moods in -e when the pronouns […] immediately follow” the verb.

THE ROLE OF PERSON AND POSITION IN OLD ENGLISH

189

Campbell (1959: 296) says: “[w]hen a pronoun of the 1st or 2nd pers. follows, the  endings -þ, -on, -en can be reduced to -e”. To put these observations in terms of verb-movement, it appears that, when the verb moves to C as in (1), there is less inflection (-e) than when it does not move as in (2) (-aþ). Sentence (1) is an imperative, but the appropriate ending would be -að since ge is plural. If it were a subjunctive, the ending should have been -en (the verb endings are in bold):2 (1)

Beowulf, 2529 Gebide ge on beorge ‘Wait (you) on the hill’.

(2)

Beowulf, 1340 ge feor hafað fæhðe gestæled you far have a feud inflicted ‘You have gone far to inflict a feud’.

The same phenomenon can be found in other languages, e.g. Dutch. In Dutch, Verb-second is a main clause phenomenon and does not occur when an overt complementizer is present. It has therefore been argued that the verb in a verbsecond construction moves to C (cf. den Besten 1983). The reason behind this movement might be that certain features in C must be lexicalized. If this is correct, one might expect more morphology on verbs in C than on those not moved to second position. However, even though there is a difference in morphology in a number of cases, the morphology of verbs in C as in Dutch (3) is no richer than of those not in second position as in (4). Under certain circumstances, as in (5), it is even weaker, i.e. zie ‘see’ rather than the regular second person ziet ‘see-2S’ (cf. Abraham 1995 for similar evidence in German): (3)

Vandaag ziet zij een javelina today sees she a javelina

(4)

dat zij vandaag een javelina ziet that she today a javelina sees

(5)

Waarom zie jij die javelina altijd en ik niet? why see you that javelina always and I not

The data in (1), (2) and (5) are unexpected in a system where overt inflection is linked to overt movement, as in e.g. Bobaljik and Jonas (1993) and Platzack and Holmberg (1989). In the next section, I examine the factors that determine which verb forms followed by subjects lack inflection.

190 3.

ELLY VAN GELDEREN

Inflection and person

First, I focus on the relationship between first and second plural inflection and overt movement. Then, I examine first and second singular verbal inflection. I will show that in the plural, lack of agreement is extremely general in Verb Subject (hence, VS) constructions by Late Old English, but that also in the first person singular, the loss has already occurred in most of the dialects, especially the Southern (i.e. West Saxon) ones. The second person singular is complicated but the pattern is that here too inflection is lost. With third persons, there is no evidence in Old English that the inflection is lost. 3.1 First and second plural In this section, I show that there is little evidence of reduced inflection in Beowulf and Junius, the earlier poetic texts; but there is some in The Paris Psalter and in the later prose writings ascribed to Alfred (ninth century). There is quite a lot of reduction in the Northumbrian Lindisfarne Gospels but this is not related to person or word order. In texts by Aelfric, there is a lot of reduction of first and second person inflection in VS constructions. In Beowulf, there are 17 instances of ge ‘you-’ as in (6). In two of these, (7) and (1) above, the verb precedes the subject. One of the two, namely (7), has full inflection for plural3 but the other one, namely (1), has reduced inflection. First person we occurs 25 times but in the two instances where the verb precedes we, the inflection is full as in (8): (6)

Beowulf, 245–6 ne ge leafnesword guðfremmendra gearwe ne wisson not you password warriors completely not knew ‘You had no knowledge of the warriors’ password’.

(7)

Beowulf, 237 Hwæt syndon ge searohæbbendra ‘What are you warriors’.

(8)

Beowulf, 270 Habbaþ we to þæm mæran micel ærende ‘We have for the celebrated a great message’.

In Junius, 75 instances of we occur of which nine are ‘inverted’ with the inflected verb. None of these are reduced. In the same text, there are 19 ge forms of which two are inverted without loss of inflection. In Exeter, of the 124

THE ROLE OF PERSON AND POSITION IN OLD ENGLISH

191

instances of we, 10 follow the verb and the only ‘reduction’ is a change from -on to -un and -an (e.g. in l. 1895). This ‘reduction’ also occurs when the verb follows the subject (e.g. in ll. 1834 and 2086) and seems therefore a regular sound change. There are 91 ge pronouns with eight inverted. None of these have reduced inflection. The metrical (West-Saxon) Paris Psalter has a number of reduced verb forms when ge follows as (9) shows, but this is not consistent as (10) shows. Some of these might be imperatives, but even then the expected ending would be -að: (9)

Paris Ps 74.4 Nelle ge unriht ænig fremman not-want you unjust any advance ‘Do not wish to advance evil’.

(10)

Paris Ps 61.10 Nellað ge gewenan welan unrihte not-want you imagine riches unjust ‘Do not imagine unjust riches’.

Herold (1968: 52) gives a number of instances of reduced inflection after we and ge in Alfred’s translation of Orosius. In the Meters of Boethius, ascribed to Alfred, there are several instances of reduced verbal inflection in VS constructions such as (11), and (13) to (15). Checking the Helsinki Corpus (the pre 950 part), there are five willað ‘want-’ endings connected with we and four of those have we preceding whereas only one has we following the verb. The reduced form wille never follows we but always precedes it as in (13) and (14). There are three instances of habbað and in all, we precedes the verb; there are two instances of habbe/hæbbe and we follows these as in (15): (11)

Alfred, Meters 2.16–7 Forhwam wolde ge, weoruldfrynd mine, secgan ‘Why do you want, my friends-of-the-world, to say’.

(12)

Alfred, Meters 19.15 Hwæðer ge willen … secan ‘Do you want to seek’.

(13)

Alfred, Boethius 85.8 Hwæt, wille we þonne secgan ‘What, do we want to say …’.

(14)

Alfred, Boethius 22.23 Ac hwæt wille we cweðan

192

ELLY VAN GELDEREN

(15)

Alfred, Pastoral Care, 45.12 ðonne hæbbe we begen fet gescode ‘then have we both feet shod’.

A substantial change in inflection can be seen in the tenth century Northumbrian Lindisfarne Gospels. For instance in Matthew, there are 134 instances of ge and 56 of we but the present tense inflection is changed, as in (16), in many of the cases. In fact, Holmqvist (1922: 2) mentions that, apart from solitary earlier instances, Lindisfarne, The Durham Ritual and the Northumbrian part of Rushworth (not discussed here) are the first texts where -s appears this way (and in third person singular and plural). The word order seems irrelevant as the two forms in (16) show. The imperative ending too is often changed as in (17), especially when the pronoun is present. I find 11 instances of the -(a)ð ending in both present indicative and imperative, and subjunctive. The West-Saxon version of (17) provided in (18) is interesting since the ending is present but the pronoun is not. This is also true in the Mercian section of The Rushworth Gospels as in (19): (16)

Lindisfarne, Matthew 3.9  nællas ge cuoeða bitiuh iuih fader we habbas abraham and not-want you say between you father we have abraham ‘And think not to say within yourselves, we have Abraham as our father’.

(17)

Lindisfarne, Matthew 7.7 gebiddas ge  gesald bið iuh soecað ge  ge infindes ask you and given be you seek you and you find and ge begeattas cnysaþ and cnyllas ge and un-tyned bið iuh and you get knock and strike you and opened be you ‘Ask and it shall be given you; seek and you shall find; knock and it shall be opened for you’.

(18)

West Saxon, Matthew 7.7 Biddaþ  eow bið geseald. seceað  ge hit findaþ cnuciað  eow biþ ontyned.

(19)

biddaþ  eow bið sald. soecaþ  ge gemoetaþ cnyssaþ  eow biþ ontyned.

Since the loss of inflection is one across the paradigm and not related to word order, I disregard it here. Checking Aelfric’s Catholic Homilies (first and second series, abbreviated as ‘Hom I’ and ‘Hom II’, from the Dictionary of Old English version), all but two VS constructions in the first series have reduced inflection with ge as in (20)

THE ROLE OF PERSON AND POSITION IN OLD ENGLISH

193

and (21), and all but two with we as in (22) to (24). There is a suppletive form sint that does precede the subject. In the second series, there are 384 instances of ge, of which 69 are VS. All but three of these lack inflection on the verb (but seven involve the suppletive sint). In the same text, there are 677 instances of we, of which 52 are inverted. All but one of these lack inflection (three involve sint). Some of these might be imperatives, but that means the ending is still reduced: (20)

Hom I, 280.4 Nu hæbbe ge gehyred … now have you heard

(21)

Hom I, 286.15 Ac wite ge ðæt nan man but know you that no man

(22)

Hom I, 88.32 Nelle we ðæs race na leng teon nor want we that argument not long pull

(23)

Hom I, 154.22–3 Nu hæbbe we þæt leoht on urum mode … and we habbað … now have we that light on our mind and we have-3 …

(24)

Hom I, 158.25 Nu bidde we … now ask we …

For convenience, the results for the two series of Homilies are summarized in Tables 1 and 2. To give an idea of the number of VS constructions, Table 1 gives the total number of VS constructions involving first and second person plural. Since the verbal endings on SV are not lost, I only provide the data on inflection in VS constructions in Table 2. As full inflection I have included both present -að endings as well as preterite -on endings (subjunctive -en endings do not occur): Concluding, in early Old English, there is some loss of first and second plural endings. However, by the time of Aelfric, practically all verbs followed by we or ge lack inflection. 3.2 First and second singular Compared to Gothic -a and Old Saxon -u, the Old English inflection with first person present singular is reduced to -e, especially in the Southern dialects. In these dialects, the -e occurs early on. The older Germanic forms are still found

194

ELLY VAN GELDEREN

Table 1. First and second person plural VS word order in Aelfric’s Homilies VS

non-VS

total

Hom I 1 P we 2 P ge

54 0(8%).0 51 (12%).0

627 373

681 424

Hom II 1 P we 2 P ge

52 0(7.7%) 72 (18.8%)

625 310

677 382

Table 2. First and second person plural inflection and VS order Inflection:

-e/Ø

full

suppletive

total

Hom I 1P 2P

49 46

2 2

3 3

54 51

Hom II 1P 2P

49 61

1 4

2 7

52 72

in the heavy (Anglian) endings -o and -u. Sievers/Brunner (1942: 282–283) take -u to be the older and -o to be the younger. These and an occasional -a occur in Mercian and Northumbrian dialects. Beowulf, is a Southern text in this respect since, out of 181 instances of ic in Beowulf, there are only three forms with a full ending as in (25) to (27). They all occur in SV constructions. There are two instances of possibly further reduced inflection with wen(e) as in (28) and (29); both VS constructions; when the word order is SV, the ending is as in (30). They might also be subjunctives, but then their ending would be expected to be -e: (25)

Beowulf 2150 ic lyt hafo ‘I have little’.

(26)

Beowulf 2523 forðon ic me on hafu bord ond byrnan ‘Therefore I have on me a shield and a coat of mail’.

THE ROLE OF PERSON AND POSITION IN OLD ENGLISH

(27)

Beowulf 3000 ðæs ðe ic wen hafo ‘that of which I hope have’

(28)

Beowulf 442 Wen ic þæt he wille ‘Hope I that … .’

(29)

Beowulf 338 wen ic þæt ge for … ‘Hope I that you … .’

(30)

Beowulf 279 þæs ic wene ‘That I hope … .’

195

This is all the evidence I find in Beowulf. I disregard the past tense since with strong verbs, there is no ending and with weak verbs, there is only an -e. In the Junius Manuscript, there are no unreduced endings and in the Exeter Book, there are three instances of hafu, one of which occurs in a VS structure, and two occur in an SV structure as in (31): (31)

Exeter, Riddle 35, 5 Wundene me ne beoð wefle, ne ic wearp hafu wound me not are woofs nor I warp have ‘Woofs are not wound for me, nor have I a warp’.

Checking the Helsinki Corpus (OEI-II), there are 54 first person endings in -u and 14 in -o, and except for three, all of these are SV. All are from the Vespasian Psalter, which is Mercian, presumably of the ninth century (cf. Kuhn 1965). Sievers/Brunner (1942: 283) claim “von den merc. Texten hat Vesp. Ps. ganz überwiegend -u, bzw. bei den schwachen Verbis der II. Kl. -iu, seltener -o bzw. -io, daneben je ein Šebidda, seŠcŠa und einige e-Formen”. In the Northumbrian Lindisfarne Gospels (I have only examined Matthew), the -o ending is strong. In texts such as Alfred’s Pastoral Care and Aelfric’s Homilies, the special endings have disappeared and only -e occurs. For instance in the two volumes of Supplementary Homilies, there are 293 instances of ic but none are connected to verbs other than ones ending in -e. Thus, the first person -o ending has disappeared by Late Old English, but the -e ending is common. A solid link between VS and inflection cannot be found. Comparing the Old English second person singular with its Gothic and other Germanic counterparts, one notices the addition of a -t to the -s ending. Most Old English grammarians argue that it “arose in inverted forms, e.g. ritstu <

196

ELLY VAN GELDEREN

rides þu” (Campbell 1959: 297), and that the “earlier ending is found in some texts” (Quirk and Wrenn 1955: 42). Jespersen (1942: 6) agrees that -st “is no doubt due to the frequent use of combinations of verb and the pronoun”. The development is said to have occurred independently in Old High German where, up to the present day, the second person present tense is -st. Later on, in English, the -t is lost again, starting in the North. This scenario seems unlikely for the following reasons: (a) if forms such as hafest originate in hafestu, one expects tu to replace þu but tu is very infrequent; (b) it is unexpected to have the same development in different languages, i.e. in Old English and Old High German; (c) forms such as hafestu are infrequent (15 times in the Old English part of the Helsinki Corpus; once in Junius; once in the Paris Psalter). For these reasons, I assume that in Old English the regular second person ending is (already) -st but that this ending may be weakened to -s. There are 69 instances of second person singular inflection in Beowulf and even though 11 of these are ‘inverted’, the inflection is not reduced as, for instance, (32) shows: (32)

Beowulf 1221 Hafast þu gefered þæt … ‘You have achieved that’.

In Junius, there are 201 instances of the second person singular and six of the 41 ‘inverted’ ones have less inflection as in (33), i.e. -es rather than -st. In Exeter, there are 318 second person singular pronouns. Many are inverted but the ones with zero or -e agreement, as in (34) to (36), are mainly imperative or subjunctive, and thus not reduced. Sentence (34) may be an imperative and then the ending is expected. Sentence (35) could be a subjunctive in which case the -e is expected and (36) is an imperative. The pronoun is never tu, which would be expected if the reduction to -s originates in the contraction of the V and pronoun: (33)

Junius, Christ and Satan 59 Wendes ðu ðurh wuldor ðæt þu woruld ahtest believed you through splendor that you the world own

(34)

Exeter, Juliana 87–8 Dem þu hi to deaþe … swa to life læt judge you her to death … or to life let

(35)

Exeter, Christ 1487 For hwon ahenge þu mec why did you hang me

THE ROLE OF PERSON AND POSITION IN OLD ENGLISH

(36)

197

Exeter, Resignation 59 Forstond þu mec ond gestyr him defend you me and restrain them

In Alfred, there are a number of reductions. Sweet (1871: xxxiv) says “[t]he ending of the 2nd pers. sing. occasionally appears without the final t ” as in (37). In the Pastoral Care, even though verbs in VS constructions generally have no ending, one -st precedes ðu. Thus, instances of full inflection usually occur with SV order, but not always as (38) shows: (37)

Pastoral Care, 193.3 ðonne hafas ðu ðin we[d] geseald then have you your pledge given

(38)

Pastoral Care, 207.11 ðe ðu on iuguðe worhtes which you in youth wrought

Checking the Helsinki Corpus (pre 950), some more instances from Alfred can be found: (39)

Boethius, 118.29 For hwylcum oðrum ðingum woldes þu þæt sprecan for which other things wanted you that say

(40)

Boethius, 122.28–9 hwæðerne woldes þu deman which wanted you to judge

In the Lindisfarne Gospels (again just Matthew is examined), some -st endings occur when the subject precedes the verb, in addition to -s ones. The ratio is 12:9. When the verb precedes, there are no -st endings (there are two suppletive wilt forms), but 20 -e(s) ones. Thus, word order is relevant here. In Aelfric’s Homilies, there are many instances of reduced inflection in VS structures, which is expected if the inflection on a second person singular verb is gradually lost. There is less reduction than with first and second person plural. Thus, in Aelfric’s second series of the Homilies, there are 454 instances of the second person singular pronoun, but at least half of these do not lose their inflection. The pronoun is never tu, again indicating that the verbal ending is not reduced because of the contracted form.

198

ELLY VAN GELDEREN

3.3 Third person and ‘feature weakness’ As to third person inflections in inverted constructions, there is little evidence of a lessening in the Old English period (at least where pronominal subjects are concerned). Later on, the ending changes from -þ to -s, first in the North. The Northern paradigm has -s throughout (except for first person singular). Jespersen’s (1942: 17) explanation for the change is that s is more ‘efficient’ than -þ. According to Sievers/Brunner (1942: 284), the reason is “ungeklärt”, but the replacement occurs more frequently in the plural than in the singular. This is the same as with first and second person, where the plural changes before the singular. No instances of reduced inflection are found with third person in Beowulf. No evidence of reduced inflection when the third person pronoun follows the verb can be found in Exeter. The Northumbrian Lindisfarne Gospels has both -s and -ð but in both SV and VS order. In Aelfric’s two volumes of Supplementary Homilies, he occurs 1443 times and even though there are at least 40 instances where the verb is followed by a subject, none indicate reduction. So far, I have examined the relationship between inflection and movement and have concluded (a) that agreement occurs somewhat less often when the verb precedes the subject, i.e. when it is in C, and (b) that agreement is lost especially in first and second persons. Third person verbs do not show reduction. Conclusion (a) presents problems for an approach that links rich morphology and overt movement, as is the case in Bobaljik and Jonas (1993) and Platzack and Holmberg (1989). That link must therefore be questioned, and the person split of (b) must be accounted for. Regarding (a), in van Gelderen (1997a), I suggest that agreement checking in C does not check all features and that it may not check certain un(der)specified features. Regarding (b), it may be possible that third person features are cliticized pronouns, but that first and second ones are not. Pronouns such as he would be adjuncts, i.e. dislocated elements. I will not pursue this option, but the availabity of third person pro-drop (as outlined in the next section) may point in that direction. Another possibility is that person features may have different specifications and that this makes it possible for some endings to disappear. If third person is more specified, it becomes understandable that these features are not lost. First and second person features on verbs are less specified and are lost early on. If true, this would explain the person split. Additional evidence for the ‘strength’ of third person features is that third person pronouns are frequently left out as will be shown in the next section.

THE ROLE OF PERSON AND POSITION IN OLD ENGLISH

4.

199

Pro-drop

Evidence for the differences in strength between the different pronouns also comes from null subject (or pro-drop) data in Old English. In Modern English, as is well-known, an empty subject is possible with infinitives as in (41). This subject is referred to as PRO and has the features of its antecedent, i.e. first person singular in (41); a tensed clause cannot have a null-subject as (42) shows: (41)

I like PRO to talk to myself.

(42) *pro am talking to myself. Languages such as Spanish and Italian have null subjects as in (42). In order to distinguish between (41) and (42), the terms ‘big PRO’ and ‘little pro’ were coined. There has been a debate at least since Chomsky (1982) as to how PRO and pro are licensed (e.g. Jaeggli and Safir 1989). The problems with ‘big PRO’ are that it is both a pronominal and an anaphor, has no governing category and, even though it is an argument, it does not have Case. PRO is not relevant here and hence, I will not go into it deeper. Regarding ‘little pro’, some (e.g. Taraldsen 1978; Huang 1984) have argued that strong agreement is responsible for the appearance of pro; Ura (1994) argues that pro is licensed by Case or phifeatures. I assume these accounts are correct and that pro can be licensed by specified person features. There is a phenomenon that resembles pro-drop, namely Topic-drop, that occurs in Modern English, as in (43). It does not take place when an adverb or complementizer occupy CP, as (44) shows. The structure of these is not related to feature licensing: (43)

Hope to see you soon.

(44) *Tomorrow hope to see you. As to pro-drop in Old English, one of the better known instances is found in some of the many versions of Caedmon’s Hymn: (45)

Nu scylun hergan hefaenricaes uard now must praise heavenly-kingdom guard

(46)

Alfred’s translation of Bede’s Eccl. Hist (Sweet’s AS Reader) Nu we sculan herian heofonrices Weard

This poem exists in many versions and in some, as in (46), we appears. (45) is the Northumbrian version, probably from the eighth century and (46) is a WestSaxon one, from the tenth. Visser (1963: 4ff) lists many others in Beowulf. Many

200

ELLY VAN GELDEREN

of these, e.g. as in (47) to (49), are in embedded finite clauses and can therefore neither be PRO or Topic-drop: (47)

Beowulf 567–9 þæt syðþan na that since-then never ymb brotne ford brimliðende on broad water-way seafarers lade ne letton passage not let ‘that they after that never kept people from passing that water.’

(48)

Beowulf 67 þæt healreced hatan wolde that palace command would medoærn micel men gewyrcean meadhall large men to-build ‘that he would order his men to build a palace, a big meadhall.’

(49)

Beowulf 300 þæt þone hilderæs hal gedigeð that that battle-storm unhurt endure ‘that they will withstand unhurt the heat of the battle.’

Thus, early texts such as Beowulf display pro-drop. Visser also lists instances from The Exeter Book (Juliana) as in (50), Junius (Exodus, Daniel), Judith, Alfred as in (51) and Aelfric as in (52). So does Schrader (1887) for Aelfric. In (51), þa could be taken as a subject rather than as an adverbial. However, there is a second pro-drop after þonne: (50)

Juliana 142 Het hi þa swingan ordered her then beat ‘He ordered her beaten.’

(51)

Alfred, Orosius, 86.7 sume men secgað þæt þa beteran wæren þonne nu sien some men say that then better were than now are ‘Some men say that the tides were better then than they are now.’

(52)

Hom I, 316.23 namon þa to rede, … ‘They took then to council ….’

THE ROLE OF PERSON AND POSITION IN OLD ENGLISH

201

Discussing the use of subjects in Aelfric, Ropers (1918: 1) argues that the subject pronoun is not always expressed but mostly it is added to the Latin where the subject need not be expressed. He does not mention a person split. In the Lindisfarne (Northumbrian, i.e. Northern) interlinear translation of Matthew, pro-drop occurs frequently. This may be due to the attempt to translate the Latin text (where pro-drop is frequent) word by word. As I show below, however, there is a systematicity to pro-drop which can only be explained if prodrop is part of the syntax. Instances of pro-drop are (53) to (56). Skeat’s edition also includes a Mercian and two West Saxon texts and the West Saxon versions have less pro-drop, e.g. none in these examples: (53)

Matthew 8.26 et dicit eis quid timidi estis modicæ fidei  cueð to him huæt frohtende aron gie lytlo geleafa and says/said to them what fearing are you little faith ‘and he says to them why do you fear, you of little faith.’

(54)

Matthew 9.37 Tunc dicit discipulis suis messis quidem … ða cueð ðegnum his hripes soðlice … then says/said disciples his harvest truely ‘Then he says to his disciples about the harvest truely ….’

(55)

Matthew 12.44 tunc dicit … ða cueoeð … then says/said ‘Then he says ….’

(56)

Matthew 17.20 dicit illis propter incredulitatem uestram cueð him fore ungeleafulnise iurre says/said them for unbelief yours ‘he says to them because of your unbelief ….’

If third person is in fact more specified as I argue is the case in Old English, one would expect pro-drop with third but not with first and second because an empty subject would be licensed through the strength of features. This is indeed the case. In the first 20 lines of Beowulf, there are five third person instances of pro-drop but none with first person. In (57), an instance of third person pro-drop is given:

202

ELLY VAN GELDEREN

(57)

Beowulf 7–11 He þæs frofre gebad he that- consolation- waited weox under wolcnum weorðmyndum þah [He] grew under clouds- honors- grew oð þæt him æghwylc þara ymbsittendra until him every of-the people-around- ofer hronrade hyran scolde across sea obey should gomban gyldan. tribute pay ‘He was consoled for that. He grew up. His honor grew until everyone of the neighboring people on the other side of the sea had to obey him; had to pay him tribute.’

Representative instances of a lack of pro-drop with first person are (58) and (59): (58)

Beowulf 290–3 Ic þæt gehyre þæt þis is hold weorod I that hear that this is friendly army frean Scyldinga. Gewitaþ forð beran to the lord of the Scyldings Go on carrying wæpen ond gewædu. Ic eow wisige your arms and armor I you will lead swylce ic maguþegnas mine hate … just as I will command my men (to …)

(59)

Beowulf 335–8 Ic eom Hroðgares I am Hrothgar’s ar one ombiht. Ne seah ic elþeodige messenger and officer Never saw I foreign þus manige men midiglicran so many men more courageous. Wen ic þæt ge … I hope that you …

The vast majority of Visser’s examples is with third person. Berndt (1956) argues that pro-drop in late Old English occurs more with third person than with first and second (as opposed to Modern English). In his examination of late Old English verbal inflection, Berndt also tabulates the increased use of personal

THE ROLE OF PERSON AND POSITION IN OLD ENGLISH

203

pronouns. His tables indicate a clear first/second versus third person split. For instance, in the early tenth century Durham Ritual, which shows fewer pronouns than the other texts examined, 87% of the first person singular pronouns appear; 78% second person singular; 7% of the third singular; 98% of first person plural; 93 of second person plural and 17% of third person plural. Berndt divides The Lindisfarne Gospels and The Rushworth Glosses in two parts each because one part of the latter is from a different dialect area than the other. The figures for indicative constructions for the presence of first person singular are 96%, 99%, 97% and 96%; for second person singular 87%, 93%, 88%, 90%; for third singular 21%, 15%, 54%, 16%; for first plural 100%, 99%, 98%, 98%; for second plural 95%, 95%, 89%, 83%; third plural 29%, 20%, 52%, 19%. Berndt does not discuss the different kinds of pro-drop but looking at some verb forms, the split becomes obvious. In the first 30 occurrences of Latin dico ‘I say’, seven are preceded by ego ‘I’ whereas in the Northumbrian 29 ics are present, i.e. 22 have been added. Of the six occurrences of faciam ‘I make’, none have a subject pronoun in Latin but all have ic added in the Northumbrian version. So pro-drop with first person is not frequent. In (53) to (56) above, prodrop with third person is much more frequent. Of the four occurrences of dicit ‘he says’, the Northumbrian version adds he only once; of the 14 occurrences of facit ‘he does’, 13 have a specified NP or relative subject but one has pro-drop (Matt 5.45) in the Latin and the Northumbrian. As mentioned, Skeat’s edition includes a Mercian and two West Saxon texts and the latter generally have less pro-drop. Even though in general pro-drop is less frequent with first and second person, there is no direct relationship that texts with full verbal inflection will drop the subject pronoun. For instance, in Rushworth 2, the first person -o endings are very present (cf. Berndt 1956: 97) but pro-drop is not. First and second person features are weaker but this comes out differently in different texts. Thus, the occurrence of pro-drop with third person subjects provides evidence that third person features are more specified and that the pronoun can therefore appear as pro. Cross-linguistically, there is evidence that first and second person features are weaker. For instance, Solá (1996: 236) presents evidence from Italian dialects where first and second person object pronouns as in (61) need not trigger agreement on the participle whereas third person ones as in (60) do: (60)

Le ha viste/*visto them has seen-./seen-. ‘S/he has seen them..’

204

ELLY VAN GELDEREN

(61)

Ci ha viste/visto Us has seen-./seen-. ‘She has seen us..’

Pro-drop in the Old Dutch Wachtendonckse Psalmen also follows the Old English pattern. For instance, van der Wal (1992 [1994]: 98) claims that first and second person pronouns were present in the majority of cases.

5.

Conclusion

In this paper, I show that fronted verbs (situated in C) have less agreement. This is especially true with first and second person subjects. I account for this by saying that these features are less specified on the verb. If this is true, it accounts for the fact that pro-drop is least common with first and second person. Third person features are specified, i.e. pronominal and third person pronouns can be left out.

Notes 1.

I have used computer-readable versions provided by the Oxford Text Archive and the Dictionary of Old English project together with TACT.

2.

Two kinds of glosses are provided. The one is a word-by-word gloss and the other, enclosed in single quotation marks, provides a freer translation. I do not always provide both since the meaning is often clear from the word-by-word gloss; and sometimes a word-by-word gloss is redundant.

3.

With most verbs, the -on ending is a preterite. In the case of the so-called present-preterite verbs, where the past tense forms are used for the present, such as witan ‘know’, sculan ‘have to’ and cunnon ‘know’, the -on ending indicates a present tense. This is also the case in the suppletive form syndon, which is a present.

Texts Godden, M. (ed.). 1979. Ælfric’s Catholic Homilies, the second series. London: Oxford University Press. Klaeber, F. (ed.). 1950 Beowulf. Boston: D. C. Heath [1922]. Krapp, G. (ed.). 1931. The Junius Manuscript. New York: Columbia University Press. Krapp, G. and Dobbie, K. E. V. (eds). 1936. The Exeter Book. New York: Columbia University Press.

THE ROLE OF PERSON AND POSITION IN OLD ENGLISH

205

Kuhn, S. (ed.). 1965. The Vespasian Psalter. Ann Arbor: University of Michigan Press. Pope, J. (ed.). 1967–8. Homilies of Aelfric: A Supplementary Collection, volumes I and II. London: Oxford University Press. Sedgefield, W. (ed.). 1899. King Alfred’s Old English Version of Boethius. Oxford: Clarendon. Skeat, W. (ed.). 1887. The Gospel according to Saint Matthew. Cambridge: University Press. Sweet, H. (ed.). 1934. King Alfred’s West-Saxon Version of Gregory’s Pastoral Care. London: Oxford University Press [1871].

References Abraham, W. 1997. Deutsche Syntax im Sprachenvergleich. Tübingen: Gunter Narr. Berndt, R. 1956. Form und Funktion des Verbums im nördlichen Spätaltenglischen. Halle: Max Niemeyer. Besten, H. den. 1983. “On the interaction of root transformations and lexical deletive rules”. In On the Formal Syntax of the Westgermania, W. Abraham (ed.), 47–131. Amsterdam: John Benjamins. Bobaljik, J. and Jonas, D. 1996. “Subject positions and the roles of TP”. Linguistic Inquiry 27 (2): 195–236. Campbell, A. 1959. Old English Grammar. Oxford: Oxford University Press. Chomsky, N. 1982. Some Concepts and Consequences of the Theory of Government and Binding. Cambridge: MIT Press. Chomsky, N. 1995. The Minimalist Program. Cambridge: MIT Press. Gelderen, E. van. 1997a. Verbal Agreement and the Grammar of its ‘Breakdown’. Tübingen: Niemeyer. Gelderen, E. van. 1997b. “Inflection and movement in Old English”. In Problemsyntax, W. Abraham and E. van Gelderen (eds), 71–82. Tübingen: Niemeyer. Herold, C. 1968. The Morphology of King Alfred’s Translation of the ‘Orosius’. Den Haag: Mouton. Holmqvist, E. 1922. On the History of the English Present Inflections particularly -th and -s. Heidelberg: Carl Winter. Huang, J. 1984. “On the distribution and reference of empty pronouns”. Linguistic Inquiry 15: 554–561. Jaeggli, O. and Safir, K. (eds). 1989. Introduction. The Null Subject Parameter. Dordrecht: Kluwer. Jespersen, O. 1942. Modern English Grammar VI. London: Allen & Unwin. Kemenade, A. van. 1987. Syntactic Case and Morphological Case in the History of English. Dordrecht: Foris. Mitchell, B. 1985. Old English Grammar. Oxford: Oxford University Press. Platzack, C. and Holmberg, A. 1989. “The role of AGR and finiteness”. Working Papers in Scandinavian Syntax 43: 51–76.

206

ELLY VAN GELDEREN

Quirk, R. and Wrenn, C. L. 1977. An Old English Grammar. London: Methuen [1955]. Ropers, K. 1918. Zur Syntax und Stilistik des Pronominalgebrauchs bei Ælfric. Kiel: Schmidt & Klaunig. Safir, K. 1985. Syntactic Chains. Cambridge: Cambridge University Press. Schrader, B. 1887. Studien zur Aelfricschen Syntax. Jena dissertation. Sievers, E. 1942. Altenglische Grammatik nach der Angelsächsischen Grammatik, neubearbeitet von Karl Brunner. Tübingen: Niemeyer. Solá, J. 1996. “Morphology and word order in Germanic languages”. In Minimal Ideas, W. Abraham et al. (eds), 217–251. Amsterdam: John Benjamins. Taraldsen, T. 1978. “On the NIC, vacuous application and the That-trace filter”. Indiana University Linguistics Club. Ura, H. 1994. “Varieties of raising and the feature-based bare phrase structure theory”. MIT Occasional Working papers in Linguistics 7. Visser, F. 1963. An Historical Syntax of the English Grammar. Leiden: Brill. Wal, M. van der. 1992. Geschiedenis van het Nederlands. Utrecht: Aula.

Remarks on (uni)directionality Roger Lass University of Cape Town

If it cannot be shown that a pet hypothesis has better support than one or more alternative explanations, then (in my opinion) we should either propose none of the possible explanations or all of them … to be responsible historical linguists, we will sometimes have to say that we don’t know and can’t guess what happened in some particular historical situation. My subtext is that it is a mistake to believe (as a considerable number of historical linguists apparently do) that a weakly-supported explanation is always better than no explanation at all. (Sara Grey Thomason 1993: 485f)

1.

Scene-setting

The ‘hypothesis of unidirectionality’ (UD) is a central pillar of the theory of grammaticalisation (GR).1 In its basic form (see e.g. the important, and now for students canonical, textbook exposition in Hopper and Traugott 1993), it consists of at least these defining claims: 1.

There exist in processes of (morphosyntactic) change pathways or clines along which constructions and lexical items proceed during their evolution. One major pathway is {lexical > grammatical} (including {less grammatical > more grammatical}), with lexical or ‘content’ items typically ending up in their most grammaticalized form as inflectional morphs. Other pathways include {free > bound}, {objective > subjective}, etc.

2.

Movement on any cline is unidirectional; degrammaticalisation is at the very least extremely rare, so rare in fact that even clear cases are not counterexamples to (1).2

208 3.

ROGER LASS

The apparent strength and empirical support for (1)–(2) enable a universal claim: that all grammatical items in natural languages ultimately derive from lexical items (see below).

The UD hypothesis is important and influential; it also raises such interesting philosophical, methodological, and empirical problems that it deserves some detailed unpacking and critical examination. My concern is not with the general question of whether there are indeed very common cross-linguistic phenomena that show what has come to be seen as ‘characteristic’ or by now ‘traditional’ UD: there are, and that’s a fact (or perhaps factoid: we will see). But what kind it might be is a less straightforward matter. For instance: i.

Is UD a ‘principle of language change’, or simply a very common phenomenon?

In other words (see §§ 7–8 below) is it a ‘thing’ or a reification? If the latter, what is it that is being reified? A number of other questions, largely concerned with empirical accountability, follow from this: ii.

If the answer to (i) is (reasonably) positive, is UD a strong enough principle to be used as a basis for reconstruction, rather than simply a phenomenon to be (more or less) expected, recognizable post hoc, but not predictable? iii. Are the canonical grammaticalisation pathways or ‘clines’ well-defined enough to be uncontroversially recognizable? I.e. are oppositions (or more weakly, cluster-points on clines) like lexical/grammatical, free/bound, major/minor category, objective/subjective, etc. clear enough so that researchers not predisposed to find unidirectional GR lurking behind most morphosyntactic change can recognize them and agree where on any cline a particular item is? (See § 2) iv. Does a GR theory with the standard foundations (including UD) risk becoming impregnable? (A theoretical bag that contains everything in the world might just as well be empty as far as empirical accountability goes.) Is GR — at least with respect to directionality claims — an invulnerable orthodoxy rather than an empirical theory? v. Following from (iv), is GR sufficiently articulated as an empirical theory so that it can specify its potential falsifiers — either for the theory globally, or for particular exemplifications? vi. What is the statistical structure of the universe over which GR phenomena are defined, and is it well-formed enough to be an empirically satisfactory domain? In other words, what is the role of counterexamples, and can a sufficiently large number of them be massaged so that they cannot be turned into potential

REMARKS ON (UNI)DIRECTIONALITY

209

falsifiers? Does GR theory contain protective ‘blocking devices’? Of all of these issues, the most central and philosophically important is the ‘universal’ claim (3) about the source of grammatical items. Hopper and Traugott (1993: 128f) formulate it this way: To date there is no evidence that grammatical items arise full-fledged, that is, can be innovated without a prior lexical history in a remote (or less remote) past. Some grammatical items show enormous longevity, and we cannot look back into their pre-history. Among the highly stable grammatical items with no known lexical origin is the Indo-European demonstrative to-. Given the unidirectionality hypothesis, we must hypothesize [emphasis mine] that tooriginated in some currently unknown lexical item. We do not … know what that item was. But neither do we know that there was none, or indeed that there might theoretically have been none.

Before getting down to details (like the question of whether the available evidence actually supports such a claim), it would be worth unpacking the rather odd argument in the last two sentences. It boils down to an interesting and complex case of what I like to call the ‘Grandmother’s Balls’ argument, after a wise old Yiddish proverb: az di bobe volt gehat bejcim, volt zi geven majn zejde (‘if my grandmother had balls she’d be my grandfather’).3 The particular argument here is a kind of circuitous derivation, which seems to be deconstructable as follows: a. b. c. d.

If my grandmother had balls she’d be my grandfather. However, she apparently doesn’t, at least on the surface. But since at the moment she’s fully dressed, there’s no reason why she couldn’t. Therefore my grandmother is my grandfather.

If this is a fair (if lightly parodied) construal, we have an argument that is pretty poor even for reconstructive history, which is not, as normally practised, known for the sharpness of its argumentation (see Lass 1997: passim). But in addition to this metalinguistic failure, there are more significant empirical problems as well. For instance, as I’ll show below, in at least two major and historically wellknown language families (Indo-European and Uralic), it is impossible to make a case for lexical origin with respect to most of the older grammatical material. Could there perhaps be language types that simply have always had grammatical items that were just that? If so, forcing the strong form of UD on all data may simply leave an empty promissory note on the table, or get us from linguistics proper into the murky and ill-defined world of hypothetical palaeolinguistics (§§ 5–6 below).

210 2.

ROGER LASS

On ‘lexicality’: Preliminary confusions

How far are the notions ‘lexical’ and ‘grammatical’ determinate or publicly definable? Even though in the literature they’re defined as cluster-points on clines rather than (necessarily) ‘absolute’ categories, they are still assumed to be something like ‘natural kinds’ (GR thinking is fundamentally essentialist): otherwise there would be no point in having clines at all, and placing objects on them in relative positions, and no theory. There are probably cases where just about nobody would disagree: stone is lexical and -s in stone-s is grammatical (an affix), and so is -s in stone-’s (a clitic: as it happens, a clitic that historically derives from an inflection, and thus goes up the hierarchy the wrong way: see note 2). This would be true both in terms of content (plural and genitive are prototypically ‘grammatical’), and categoriality (so are affixes and clitics). Similarly L lapis is lexical, or at least lap- is, and -i-, -s are grammatical but in somewhat different ways: -i- is merely a theme or declension marker (where do these fit in the grammaticality hierarchy?), and -s is nominative , a marker of a grammatical relation. Similarly accusative sg lapidem: lexical lap-, thematic -i-d-, and affixal -em. To put it another way, slightly hocus-pocussy but within the bounds of conventional praxis: a morph in a complex word belongs to a lexical category if in a bracketing according to one set of traditional protocols we would want to put two brackets around it, and to a grammatical (thematic, affixal or clitic) one if we would not: so [[stone]s], [[[lap]id]em], etc. But sometimes obviously grammatical items (e.g. case-markers) carry so much lexical, ontological or pragmatic weight, in short ‘semantics’ that they seem just about as lexical in everything except form and/or degree of boundness as other uncontroversially lexical items in the same construction. Take for instance: (4)

The prince became a frog

Aside from uncontroversially lexical prince and frog, there seems no doubt that became is a full lexical verb (if maybe more ‘abstract’ than ate or kissed). And there would be just as much ‘lexicality’ in the predicators in semi-paraphrases like: (5)

The prince turned into a frog

(6)

The prince underwent a change of state such that he became a frog

(7)

The prince shifted from the natural kind ‘human’ to the natural kind ‘frog’

But precisely this degree of semantic loading is carried in some languages by items that are clearly, by all the normal criteria, grammatical — e.g. case

REMARKS ON (UNI)DIRECTIONALITY

211

markers. Consider for instance the Balto-Finnic translative case. The normal translation of any of the above in Finnish would be4 (8)

Prinssi muuttu-i sammako-ksi prince change- frog-

The element -ksi is clearly just as much a case marker as the accusative or genitive  -n in sammako-n, the inessive -ssa in sammako-ssa ‘in the frog’, the ablative -lta in sammako-lta ‘from the (exterior of) the frog’, etc. Formally, there is no difference in boundness between the final elements in [[sammako]n] and [[sammako]ksi]. There are even phonological markers of this boundness: the translative triggers medial consonant gradation like any other suffix producing a non-initial closed syllable within the (non-compound) word: kukka ‘flower’, with degemination in the inessive kukassa /ku.kas.sa/, translative kukaksi /ku.kak.si/ according to the standard syllabification, similarly joki ‘river’, joessa, joeksi, etc. (one of the gradations of /k/ is zero). Some of the uses of the translative may seem a bit less ‘lexical’, at least in the sense that they can translate fairly obviously with rather more ‘grammatical’ classes of elements (glosses for translative in bold): (9)

Hän ei sovi talo-n-poja-ksi he not.be.3. suitable house--boy- ‘he isn’t suited to be a farmer’

(10)

Sa-i-n kirja-n lahja-ksi receive--1 book- present- ‘I received (a/the) book as a present’

(11)

Mitä sano-n X Suome-ksi how say-1 X Finnish- ‘How do I/does one say X in Finnish?’

But the ‘true’ paraphrases, given the overall use of the translative, would be more like: (12)

He isn’t suitable to be in a state such that he is a farmer

(13)

I received the book it having been transformed into a present

(14)

How do you change the state of X so that it becomes (from whatever else it was) Finnish?

This may be pushing things a bit, but it does raise a point worth discussing: how much of what we call ‘lexical’ and ‘grammatical’ is based on particular canonical instances in particular canonical languages (like English, German, French)

212

ROGER LASS

that the bulk of workers on grammaticalisation habitually conduct their discourse in? Are our judgements of prototypicality as ‘objectively’ cross-linguistic as we seem to think? Or do we (unconsciously) often take translation-equivalence as one of our main criteria? So perhaps mere formal categoriality is not in itself an index of grammaticalness. Finnish can in fact furnish a reverse instance. Presumably, in any commonsensical hierarchy, a negating operator would be more grammatical than lexical. In Finnish, however, the most common negator is a verb, which normally occurs as the finite verb of a predication, with any following verb in the infinitive. It is moreover fully inflected: (15)

e-n menn-ä tal-oon -1 go-1 house- 5 ‘I don’t go into the house’

(16)

e-i menn-ä tal-oon -3 go-1 house- ‘he/she/it doesn’t go into the house’

And so on. This prototypically ‘grammatical’ operator assumes the full panoply of ‘lexical’ characters, in particular serving as a stem that takes inflections. Would we want to say that e- is grammatical by virtue of its semantics (surely a lexical property), but lexical by virtue of its morphology (which is certainly under any interpretation more grammatical)?

3.

What is unidirectionality a theory of?

This is not quite as inane a question as it sounds. It is not entirely clear, at least to me, what UD really is, or what it’s about. Is it a claim about languagehistories as landscapes viewed by historians? If so, it’s patently wrong if it attempts to be exclusivist, but it hasn’t yet developed a way to cope with counterexamples except to say that they’re ‘rare’. So are pandas, flightless birds and bacteria that metabolize sulfuric acid. The problem of what a ‘theory of rarity’ really means is a difficult one: it certainly bedevils all thinking about markedness, naturalness, and the like. This is the problem with ‘explanation by universals’ in general (see Lass 1975 on markedness): the ‘odder’ a type is, the fewer languages we should expect to show it. Does this mean anything? Is it anything more than a tautology? Is ‘rarity’ a property of (a) language? Is there a distinction between the properties of a universe and the properties of its members? At the moment, all we can say about the extant examples of reverse

REMARKS ON (UNI)DIRECTIONALITY

213

grammaticalization, etc. is that they’re surprising (see Campbell 1991; Harris and Campbell 1995). Another way of putting the question: does GR so far say more than that if we find five counterexamples we ought to be five times more surprised than if we find none, but that we have no further responsibilities? (See the next section for discussion.)6 UD is also problematical in another way, if it attempts to project (see the next section) beyond a certain kind of typological barrier (again, the discourse at the moment is restricted to UD as a kind of ‘observational’ claim about a domain). There appears to be something impermeable ‘before’ GR in its modern sense got started, and this barrier may be methodologically destructive (§§ 5–6). On the other hand, is UD a theory about human cognition? Obviously in some hands it attempts to be, but in fact even then it is still really only a set of claims about frequently-observed pathways: except that these then projected — primarily on the grounds of their observed frequency, but with a certain amount of bolstering from (possibly erroneous) ideas about ‘preferred’ types of human action — to a timeless universality. There is a larger problem here too: an implicit claim, as in all views of ‘universal tendencies’, that there is only one ‘human mind’, trans-cultural and uniform, and capable of being penetrated ‘hermeneutically’ by any other human being. All such universalist theories in essence deny the role that culture and history may play in human mental activity (Lass 1997: ch. 7). Of course there must be a residue of penetrability, since despite differences we do all share the same evolved cognitive capacities and ‘metaculture’ (Tooby and Cosmides 1994). But how far can this be taken? The ‘observational’ interpretation is ontologically less specific, and so safer; the ‘universalist’ one is more daring, and apparently vulnerable. But is it vulnerable enough to have empirical content? Or don’t we expect linguistic theories (as opposed to those in other domains) to be cast in such a way that they specify (or could specify in principle) their potential falsifiers? At least it’s clear that no argumentative framework in which positive examples count and negative ones don’t can really be more than a set of inductive generalizations. It is certainly not a theory in the strong sense.

4.

The nature of the UD universe

One problem hinted at in the last section is that it appears that nobody seems to have done the kind of sampling of really big corpora that would show us if UD-believers’ intuitions correspond with reality to any respectable degree of statistical significance. (This may be a lack of knowledge of the literature on my

214

ROGER LASS

part, but the sort of thing we’re after just doesn’t, and as we’ll see maybe can’t, exist.) What’s worrying about virtually all directionality, markedness, etc. claims is that they are generally not based on properly constituted corpora, or operate over ill-defined universes. Say in the course of your work you have found 542 changes that confirm a direction, and none that don’t. Question is, 542 out of what? Does a UD-believer’s inability to find the counterexamples, and/or the observed frequency of the confirming instances, reflect a ‘real’ property of the domain or merely the accidental tendentiousness of a chosen data base? Note that not finding things is an argumentum ex silentio, which is not at the top of anybody’s hierarchy of epistemic goodness. That is, there’s a big worry about whether any inductive historical generalizations are safe (or at least theoretically interesting), because the field over which they’re being defined is not (and may in principle not be able to be) statistically well formed. There seems to me to be, in much GR work, no proper antecedent definition of what constitutes an ‘individual’ or even a ‘population’. And at least the latter must be defined, because ultimately any empirical claim over populations must be able (at least in principle) to meet some criteria of significance.7 That is, a lot (how much we don’t know) depends on what one happens (contingently) to know, and what one is looking for. What is obscure or absent in the literature is a clear definition of the universe over which directionality predictions are being made. Is there really one at all? Has anybody worked as hard to find counterexamples as UD proponents have to find confirming cases? To take a rather priggish but not outlandish position, until you’ve stipulated a well-formed universe and prior definitions of ‘significance’, and said what will count as a refutation, you have not actually made an empirical claim.8 That is, say some UD believer finds those 542 changes going in the right direction, and I (whose exposure to languages and their histories is of course partly different from his) find 3 going the wrong way. All this says (so far) is that this is the way the numbers happen to come out. But without stipulating a corpus size and composition, we do not and cannot know whether this is a freak distribution (maybe the next 500 examples will go in the wrong direction too, or maybe the next after the next …). That is, the ratio 542:3 may be an artifact; perhaps if we looked at 3000 more changes the figures would come out no better than chance.)9 In other words, are we simply on to a crazy run (50 heads in a row), or something ‘lawful’? The methodological point is that no matter how many instances of X are put forward, they don’t really count (even though they suggest confirmation) until the believer can stand up and say at least roughly (a) how

REMARKS ON (UNI)DIRECTIONALITY

215

many non-X will make him recant, and (b) how many out of what number in the first place. We must remember that the (putative) domain is actually all the language changes that have occurred and will occur, bolstered or quasi-justified by the implicit (inductive) belief that the sample conventionally being worked with is a fair one. And of course nobody seems to have done more than scratch the surface of the available data. As a community of linguists, historians or not, we do tend to find the sheer number of cited UD-favourable examples suggestive, even impressive; but the problem remains. The actual corpus is still rather like the corpus of all species: nobody knows how many there are, and it’s not even clear how to find out. I have seen biodiversity estimates running from our knowing 1% of the existing species to something closer to 60% or so, and since the remaining 99% or 40% are undiscovered, this is not very satisfactory. In our discipline the problem is made worse by the fact that even within what is conventionally called a ‘language’, there may be so many varieties that just using (as is normally the case) one or two can be dangerously misleading. For instance, one talks about ‘English’: but the Survey of English Dialects clearly shows (and this is undercounted) 314 quite distinct varieties in England and the Isle of Man alone; and nearly all of these lack detailed histories. In most published descriptions of ‘English’, for instance, it would be made clear that unlike German and most other Germanic languages, ‘it’ does not have front rounded vowels. But in fact this is really only the case for some very archaic rural dialects in England, some US dialects, and RP and its relatives; more English dialects than not lack anything like /u:/, and have front rounded (or at least very advanced central rounded, which might count for the same thing) vowels in the historical categories represented by e.g. words like goose, nurse. The point is not this little factoid, but what its implications might be. A survey of ‘languages’ with and without front rounded vowels might include ‘English’ and ‘Yiddish’ as the two Germanic languages without any; but if in a larger if still coarse sample one construed Scots, standard South African English, New Zealand English, ‘Sloane Ranger’ English and RP as distinct ‘languages’ (which is phonologically reasonable), then four out of five ‘Englishes’ in fact do have them, so the numbers are skewed in a completely different way. The only language none of whose varieties seem to have any turns out to be Yiddish (probably). And if we remember that each of these varieties has an at least partly independent history, we could be in trouble just on the basis of our way of sampling.

216 5.

ROGER LASS

UD and uniformitarianism … relaxing methodological criteria in order to make an explanatory hypothesis possible does not advance our knowledge of language or its history. (Thomason 1993: 492)

As I noted in § 1, a major aspect of UD — to some writers its natural conclusion — is the claim that all grammatical material is ultimately lexical in origin. Let’s call this ‘strong UD’, as opposed to the more temperate ‘weak UD’, which is an inductive statement of commonly observed tendencies, pathways, etc.10 Strong UD is methodologically problematical, and probably untenable. The argument goes like this: i.

All sound historical theories are ‘uniformitarian’ or ‘actualist’; the laws and boundary-conditions governing their domains must be isotropic in time and space.

That is, the ‘laws of nature’ do not change. This has been the foundation-stone of rational historical science since Lyell (1830–33): a time when ‘everything was different’ is in principle unavailable to history. ii.

The claim that all grammatical material is ultimately lexical means that there was a time when all human languages were ‘isolating’ (in the days of Homo erectus or whatever everybody spoke Vietnamese). iii. No period for which we have attested linguistic data, nor even any period reconstructable by standard methods (given the corpus of all languagefamilies that allow investigation at reasonable time-depths), shows anything other than roughly the current distribution of isolating, agglutinative and inflectional/fusional languages. iv. Among these languages are many the bulk of whose morphology (inflectional and/or derivational) shows no evidence of earlier lexical sources, and no possibility of reconstructing any. v. Therefore positing a period when there was a ‘law’ that says ‘all languages are isolating, and all their material is lexical’ (or similarly, following Givón 1979, that all languages were once SOV) is counter-uniformitarian, and so methodologically inadmissible.

Note that this contrasts sharply with the organic fossil record, which shows long periods of minimum diversity: it was once a law that ‘all living organisms are cyanobacteria’, it was once a law that ‘all plants are non-flowering’, etc. But this presupposes an theory (such as exists in biology) that accounts for increase in typological or phyletic diversity over time, and there is no such theory in linguistics. The source of this problem is unclear: either languages have always been

REMARKS ON (UNI)DIRECTIONALITY

217

this diverse, or we face the equivalent of encountering the fossil record for the first time at the end of the Cretaceous. Either way, the issue is the same, except that language gives us no evidence for earlier ‘pre-modern’ states, whereas both the post-Cretaceous fossil record and the existing biota do: we still have cyanobacteria and ferns. vi. Therefore a warrant for reconstructing a very non-diverse domain must be stronger than one for reconstructing a period when all the current typological varieties existed — if in different proportions; this is a matter of statistical tact, not substance.11 None of this would be particularly troublesome, if, not for the following: vii. There are very well studied language families (e.g. Indo-European, Uralic, Dravidian, Semitic) in which the bulk of the inflectional and derivational material in the (well-justified) protolanguage as well as in the surviving daughters is not traceable to any lexical source. As far back as you can get by standard reconstructive methodology there’s only morphology. (See the following section.) ix. Therefore any universally more uniform state (e.g. a world in which IndoEuropean is isolating and all its morphs are lexical) is not reconstructable by comparative method, internal reconstruction, or any other known projective technique. x. Therefore any claim that Indo-European or its putative ancestor(s), from ‘Proto-World’ down to ‘Nostratic’,12 must have been (lexical-only) isolating at some point is a non-payable promissory note based on projection back to an (a) in principle unobservable, and (b) untestable, because unreconstructable, domain. It is palaeolinguistics, not linguistics. xi. Since uniformitarianism has failed (or been abandoned) here, there is no guarantee that it would succeed elsewhere. Who is to say that in a linguistic world so different from ours that it has no morphology but only lexis the {lexical > grammatical} or {free > bound} or any other pathways didn’t run the other direction? (Induction fails in universes nonisomorphic to the one in which the induction is made.) By giving up one set of constraints we lose our motivation for retaining any others, since all our constraints are based on inductive generalizations over the set of existing languages. I think the discussion so far allows these preliminary conclusions: I.

No UD claims can be made for any domain that is different typologically (particularly in terms of allowed diversity of type) from the set of all observed languages (past and present).

218

ROGER LASS

II.

By definition, any UD claim that derives all nonlexical morphs from lexical ones is then made over a domain like that defined in I: hence it cannot hold for ‘language-in-general’ in the sense in which it is the subject-matter for modern linguistics, since the claim is not being made about the subject matter of modern linguistics. III. Hence strong UD cannot form part of linguistic theory. Now it’s true that a lot of hard work has reduced, or should we say ‘lexified’, the amount of what may once have been thought to be ‘original’ morphology. E.g. the -b- suffix in Latin imperfects and futures can be plausibly traced to grammaticalisation of *bhu- ‘be, remain’, and the Germanic weak verb suffix to *dhe:-/dho:- ‘place, put, do’ (though this is more controversial). But this is only one side of the problem.

6.

The ‘grammatical in principio’

Here is a small list of items, which, along with an argument to be given a bit later, seems to show that Hopper and Traugott’s IE *to- example might be a bit disingenuous, or at best misleading. It is not in fact a strange isolated case, but rather closer to the norm. To take some further Indo-European examples, nobody has ever found a lexical source for the first-person pronominal and athematic verbal *-m-, the deictic *kw-, the feminine marker *-a:, the neuter *-d, the acc sg *-m, the thematic verb present 1 sg -ò, and so on. Even when some of these have been traced back to different categories (e.g. the feminine marker *-a: to an old collective), these are still affixal. And given current (and probably even conceivable) reconstructive techniques, and what has survived from older IndoEuropean, it seems unlikely that anybody will ever find lexical ancestors. There is simply no evidence for any more substance ever being attached to these items. It looks rather as if (say in the Latin copula with its various grades) it is words that are made (historically, in real time) out of lexical roots and bits of original morphology. E.g. if we take *es- ‘be’, and *-m- ‘1’, we can combine them into the word sum; similarly with *es-t, etc. The only lexical element is *es-; all the rest is morphology as far back as you can go.13 The case of Balto-Finnic is similar. According to the standard accounts,14 the canonical lexical root structure in Proto-Uralic, as still in modern BaltoFinnic languages like Finnish and Karelian, was *(C)VC(C)V-. Monosyllabic lexical items (in fact monosyllables in general) are rare in these languages; virtually the only free monosyllables in Finnish for instance are conjunctions like

REMARKS ON (UNI)DIRECTIONALITY

219

ja ‘and’ (probably a Gothic borrowing), jos ‘if’, kuin ‘when’, etc.15 Even the few monosyllabic roots are usually not fully lexical, but represent relatively grammatical categories, like deictics (tä- ‘this’, ku- ‘who’), or the negative verb e-. Or they are derivational morphs, like -ja ‘agentive’. Many important formatives can be reconstructed only as consonants (e.g. *-c´- ‘living being, object’, *-l- ‘location, movement’, *-kk- ‘diminutive’: Collinder 1960: §§ 773ff). To turn to inflection, none of the cases are reconstructable as looking anything like lexical items: all are monosyllabic, and some are combinations of other nonlexical bits: so Proto-Uralic *-m ‘accusative’, *-n ‘genitive’, *-nA ‘locative’,16 *-tA ‘partitive’. In Balto-Finnic these were combined with other nonlexical material to form new cases, e.g. *-s- ‘lative’ + *-nA > -sna > -ssa ‘inessive’, *-l- ‘lative’ + *-tA > -lta ‘ablative’. These items are not only unwordlike in syllabic structure, but also contain illegal initial geminates or clusters not reconstructable for any ProtoUralic or Proto-Balto-Finnic words or lexical roots. As far as any known evidence tells us, these elements were always what they are now, and still function the same way. It will not do to say that on theoretical grounds that there must have been a (now eroded) lexical item, only it happens to have left no trace (cf. the ‘Grandmother’s Balls’ discussion in § 1). Languages with this kind of morphology typically do not have much in the way of ‘words’ anyhow, but rather build themselves out of roots and unanalysable sub-word elements. In fact, given such languages, one could conceive an argument (here only a bit of kite-flying), which if pursued properly could turn UD on its ear. Why should there not be a type of language in which typically lexical words are made out of grammatical morphs? One could imagine a case where a prefix meaning ‘motional’ and one meaning ‘directional’ (surely ‘grammatical items’) could fuse into a verb meaning ‘turn’ or ‘go’. Why not? This is not to say either that such languages actually exist (I don’t know), or that if they do they had no lexical morphemes originally; only that there are both kinds, and perhaps a preponderance of the latter, so that a lot of things that are expressed in non-agglutinative languages by word-like elements are handled by concreted affixes. This cannot be ruled out in principle, if on no other grounds than the frequent occurrence of word-stems that contain bits of derivational or inflectional morphology that have fused onto stems (‘phonogenesis’). Thus G fressen (cf. Gothic fra-itan), glauben < OHG gi-lauban, etc.; or best of all, the relict accusative in the now unanalysable French rien < re-m. This may be the right place for some exemplification, by means of a kind of infinite regress reconstruction. This is a particularly neat and clear instance of a solid counterexample to the claim for lexical origin of inflectional material as a universal. We might call this the case of the ‘Proto-Proto Barrier’. Since the

220

ROGER LASS

end of the nineteenth century at least (Anderson 1879; Sweet 1900), it has been clear to many writers that there is some kind of ‘affinity’ or distant relation between Uralic and Indo-European, though it’s not clear precisely what kind it is. The most radical claim is that there is a common protolanguage, conservatively ‘Indo-Uralic’, more radically the larger ‘Nostratic’, which contains not only these families but perhaps Afro-Asiatic, Kartvelian, etc. (depending whose account you read: see the discussion in Lass 1997: ch. 5). One of the classic markers of this ‘affinity’ is the so called ‘Mitian’ or ‘Me-Thee’ phenomenon: the presence in both Indo-European and Uralic of a marker in /-m-/ for the first person (pronominal and verbal), and /-t-/ for the second person. A few examples: (17)

Indo-European personal endings of athematic verb: Skr, Gr 1 -m-i, -t-i, L 1 -m-us, -t-is; pronouns: 1 L m-e¯, OE m-e¯, 2 L t-e¯, OE þ-e¯.

By standard comparative reconstruction, the Proto-IE forms of these markers must be *-m-,*-t- respectively. (18)

Balto-Finnic verb: Finnish, Karelian, Estonian 1 -n < *-m, 2 -t (Est -d), 1 -mm-, 2 -tt-; pronoun: 1 Finn min-ä, Kar mie, Est m-(in)a; 2 Finn sin-ä Kar šie, Est sin-a < *t-in-V.

By the same argument as for (17), the markers must be reconstructed as ProtoFinnic *-m-, *-t-; and the same would hold for Proto-Uralic. Given the results of (17–18), since there is only identity of endings, the ancestral maximal forms (‘Indo-Uralic’, ‘Nostratic’, whatever) must of course also be *-m-, *-t-. A lot more data would be needed to make a really convincing case, but these examples at least suggest that UD is simply an observable but by no means indefeasible tendency. The picture seems rather like this: out of two possible directions, lexical > grammatical, and grammatical > lexical, the first is statistically commoner, so metaphorically ‘preferred’. But there is a further ‘turtles all the way down’ possibility: much of the grammatical morphology in the world’s languages has never been anything else. There are good methodological arguments that this must have been the case; the ‘reverse’ direction is not generally discussed, and the evidence is weak, but some well-known phenomena (hinted at in rien above) suggest that there are processes by which some irreducible morphology can become part of some lexis. The point is that there is no grounding for the claim that because (now) we

REMARKS ON (UNI)DIRECTIONALITY

221

observe that lexical > grammatical is a very strong (if not indefeasible) pathway, we can therefore say that all today’s morphology is yesterday’s lexis. As it stands, this is simply a non sequitur, not a ‘strong’ empirical (or any other respectable) kind of claim. Now of course the reconstructive barrier suggested by the *-m-, *-t- and other examples may be an artifact of weak technique: if we really understood comparative method better, we’d be able to get behind the 10,000-year or so veil that seems to block us so often. But this is not just a peculiarity of Indo-European, as we’ve seen (and the same demonstration would be available for Dravidian and Semitic). If it were just an artifact, it would be surprising that it was so common in languages that have just the kind of structure that seems to favour it, and not in languages that appear to have been isolating forever. This reinforces the suggestion that there are just different kinds of languages, and some of them have morphology all the way down (or up or back), and nothing we might want to believe can provide evidence that this was not the case. Except of course a desire to believe; but if we claim that all grammatical material has a lexical source regardless of the evidence, this is a confessio fidei, not linguistics. In linguistics or other rational subjects matters of faith and works are distinct, and works win.

7.

Possible reasons for (some) unidirectionality

Anyway, the real problem, or one of them, seems to be: if there are indeed unidirections, why are there? Are they due to something interesting, or if they exist, merely to rather boring facts like if you lose enough information during an information-losing process, you end up with zero, which is kind of a black hole out of which you can’t extract anything? In that case, at least some posited clines are tautologies. If you reduce a lexical item of the shape CVCVCV to an affix of the shape C, there’s precious little left to tell you what it once was, so of course you can’t get enough information to reconstitute it. Even worse for deletion. OK then, why if there are clines (and let’s forget for the moment the question of the ill-definedness of terms like ‘lexical’, ‘grammatical’, ‘bound’, ‘bonding’ … ) do they look the way they do? One answer is something that characterises innumerable natural systemspositive feedback or autocatalysis. There are all kinds of processes in the world in which if you do X, this generates a propensity to do more X. E.g. think of the way a ball gathers speed rolling down an inclined plane, or more generally the way an object accelerates under gravity: the behaviour is a consequence of the maths, not of whoever or whatever

222

ROGER LASS

is doing the behaving. If the acceleration of a falling body is a function of its mass times a gravitational constant, then there’s nothing about the body itself that determines how fast it (ideally) falls: whatever it’s all ‘about’ is the nature of falling, under a regime of (physico-mathematical) ‘law’, not properties of the bodies that happen to come under the influence of the law. In other words, how much of what seems to happen in GR is really GR or linguistics at all, and how much might be general autocatalytic or feedback behaviour, to be expected in any similar circumstances, and therefore less interesting, at least from a strictly linguistic point of view? Unidirectionality, insofar as it does exist (and that’s another question) might be simply the kind of behaviour you expect in an evolutionary landscape where flow-paths converge on attractors. (Virtually, in fact, a dynamical definition of what an ‘evolution’ is.) Which is to say, unidirectional pathways are a part of the mathematical furniture of the world in all domains, and if you find them in language change then it’s not surprising that they are unidirectional, because that’s how they’re defined. Or if that sounds weird, if X exists and you look for X and find it, then X exists. There’s a parsimony issue tied up in all this, which is perhaps the most important point of all. Let there be two accounts for some phenomenon, one domain-specific and the other operative in many domains (under a sufficiently abstract construal, anyhow). In this case the less domain-bound one is more parsimonious, because it invokes less in the way of ontological specificity. UD then may not be a property of languages except derivatively; it is a property of epigenetic landscapes of particular kinds, and similar phenomena can be found in any domain meeting the requisite conditions, i.e. the possibility of structural loss.17

8.

More simplification: why clines may be conceptual artifacts

I have been (carefully) failing to separate two issues here. One is that of directionality, and the other the existence of clines in the ‘classical’ Hopper and Traugott sense. It would seem at first that there’s really no distinction: if there’s a known pathway that goes {lexical word > grammatical word > clitic > affix}, isn’t this pathway a cline in the sense that of being a graded series of categories that moves in a single direction? Actually the answer is no, or at least not necessarily yes. Consider for instance a familiar type of cline in another linguistic module: a lenition pathway. Everyone would agree that the following are very typical pathways, which show all the properties of clines (step-by-step movement, at least tendential irreversibility, etc.):

REMARKS ON (UNI)DIRECTIONALITY

(19)

p > pf > ph > f > h > Ø

(20)

b>v>w>u>Ø

223

Formally these look exactly like: (21)

lexical word > grammatical word > clitic > affix > Ø

The ‘cline’ interpretation of these examples is ‘progressive weakening’ (loss of oral occlusion or lexical substance and positional freedom) via a series of prespecified steps, which map onto the cline’s ‘positions’ or ‘cluster-points’. But there is an alternative and more parsimonious interpretation. Processes apparently involving ‘progressive’ loss of articulatory strength or lexicality are common (the pathways are inductively derived), and according to our chosen primitives and formalisation, ‘quantal’ rather than continuous (guaranteed by our selection of named cluster-points). ‘Squishes’ are presumably less common than objects that sit on the named points, even if some phenomena may be ambiguous (e.g. /t/ going to a voiceless alveolar tap, or an item varying between inflection and clitic while the clitic is becoming an inflection via lexical diffusion). But if we pursue the basic quantal notion, we have the possibility of a different kind of interpretation. I suggest something more or less along the following lines: i.

Let any phonological segment or lexical (or other) morph be a set of matrices of hierarchically ranked, unary features. These are the quanta defining categorial information, and the units of change. ii. Processes involving loss of ‘articulatory strength’ or ‘lexicality’, are commonplace, and quantal. That is, they lop off ‘one quantum’ on a given scale. iii. Such processes, for inertial or whatever other reasons, have a habit of returning over time, either with a defined period or not; and the same item may recycle through the same process (an ‘instruction’ to lose one information-quantum, say) again and again. Therefore, when such a change arises, and an item is susceptible, it will undergo the process. But note that its being affected is independent of its prior state. A stop will become a fricative, a clitic will become an affix simply because it is what it is at the time the (generalized) change applies. The change of one category into a ‘weaker’ one is ontologically and processually independent of its prior state, what changes happen to have come before, or where the item began n changes ago. That is, under this construal the ‘stages’ in a cline are causally independent, like the tosses of a coin. Information-loss processes have no memory. We could argue then that ‘clines’ or ‘pathways’ have no independent existence or explanatory value; they are post-facto reifications, pure epiphenomena of our notations.

224

ROGER LASS

So in principle all we need for the examples discussed is one kind of process, and no complications like predefined pathways with content. This process may be content-free and memory free, and ‘directions’ in such a framework are artifacts of presentation, not properties of the objects or the processes that change them. Perhaps a graphic illustration can sum this up more clearly: (22)

Let ☞ = any periodic change, ✘, —, ✡, ★ arbitrary quanta of information: T1

T2

T3

T4

T5

✘

✘

✘

✘

✘

—

—

—

☞—

✡

✡

☞✡

★

☞★

There is a lot of work to do, of course, in refining a proposal like this (if indeed it’s worth proceeding). In particular, the kinds of elements that make up the quantal sets have to be defined. In this introductory polemic I will shirk that responsibility; but it does not seem impossible, and does seem worth trying.

Acknowledgments Thanks to Claire Cowie, Ana Deumert, Olga Fischer, Don Ringe, and Elizabeth Traugott for helpful comments on earlier drafts; and to discussants at oral presentations in Oxford and Edinburgh, especially Anna Morpurgo Davies, JC Smith, Jim Hurford and John Anderson. The present paper is certainly better than earlier drafts because of these interventions; I’ve surely ignored some important ones, and of course made all my own errors.

Notes 1.

Whether there is a ‘theory’ of GR is debatable, or at least debated. For a negative conclusion see Newmeyer (1998: ch. 5). Newmeyer’s ‘deconstruction’ of GR was brought to my attention after this paper was written, thanks to Jim Hurford. Interestingly Newmeyer and I largely agree, and come to similar conclusions on the basis of quite different arguments. This is either a supporting ‘consilience of inductions’ (Whewell 1837), or a folie à deux.

2.

E.g. it is impossible, from the literature, to determine the status of well-known changes like the ‘upgrading’ of the Old English masculine/neuter a -stem genitive singular affix -(e)s, which derives from an Indo-European affix, into a clitic in modern English: note the early type [the

REMARKS ON (UNI)DIRECTIONALITY

225

king]’s nose of England (where the -s is only an affix on an N) vs. the later ‘group-genitive’ [the king of England]’s nose, where it is a clitic to a phrase. It can even be a clitic in many dialects to a phrase with one or more embedded sentences: [the man who I used to know]’s daughter. It is at least arguable that the short-lived English ‘his-genitive’, e.g. John his book, shows upgrading of an affix (or at that stage probably a clitic) to a free word, which can serve as member of a paradigm: thus after John his book we get the types Mary her book, John and Mary their book (see Lass 1999: § 3.8.1). 3.

I was reminded of this childhood treasure by Steven Pinker’s quotation of it in a slightly different context (1995: 66). I use a different transliteration, but the proverb is the same.

4.

Thanks to Minna Andersen for Finnish data and detailed and enlightening discussion.

5.

‘Inf1’ = ‘first infinitive’, the traditional name for this form. There are other ‘infinitives’ that are more like what non-Finnish traditions would call ‘participles’.

6.

Note that it would be perverse to make a universal claim that “all mammals with long canines, claws, short guts and cutting cheek-teeth are carnivores”, and then have to deal with the ‘problem’ of how to treat the herbivorous ‘counterexample’ of the panda. The question of how to interpret apparent counterexamples can sometimes be a function of an ill-considered (if beloved) hypothesis, formulated in such a way that it ignores potential counterexamples ab initio.

7.

Cf. Don Ringe’s devastating 1995 paper, where he shows that the standard ‘cognate-sets’ and correspondences supposedly supporting the existence of ‘Nostratic’ are no better than what would be expected by chance, and that their profile can be shown to be mathematically so different from that of ‘real’ families like Indo-European that we have to reject Nostratic as a pure invention.

8.

This is not to say that the UD tradition hasn’t produced a lot of exquisite exegesis, populated the universe of discourse with new ideas of great fruitfulness, increased our understanding of particular changes or types of change, etc. But in the present connection this is not the main issue.

9.

This is just an intuition: it certainly seems to be the case in some markedness examples that have been subject to statistical analysis (see Lass 1975).

10.

It should be clear that my target here is the strongest possible Strong UD. I don’t know what percentage of writers on GR actually hold this position, but it has been enunciated in print, as well as by personal communication from various people (notably Martin Haspelmath who I thank for taking such a strong position a couple of years ago). But I think it’s good strategy to attack the strongest version of any position, since the weaker ones have already yielded something.

11.

That is, uniformitarian theories must have sufficient latitude to allow for lawful emergence. The likelihood of a resident of New York speaking Yiddish or having a PC in 1828 is not the same as now; but there are ‘legal’ phenomena like migration and technological development that allow for this. For detailed discussion see Lass (1997: ch. 1).

12.

The scare-quotes indicate my doubt that any such things existed, or at least are recoverable by normal techniques; but let’s assume them or something of the sort anyway; since it makes no difference.

13.

This appears to be the consensus in the handbooks, both old and new (e.g. Meillet 1964; Krahe 1962; Szemerényi 1989; Beekes 1990). The issue does not even appear interesting enough to discuss.

226

ROGER LASS

14.

For the following material, see for example Collinder (1960) on Uralic in general, and Laanest (1982) on Balto-Finnic.

15.

This is not the case for many more easterly Uralic languages belonging to other subgroups, and some innovative Balto-Finnic ones like Estonian and Vepsian.

16.

Notations like *- nA, etc. indicate a low vowel subject in later dialects to vowel harmony. So the locative survives as the Finnish essive, with two allomorphs: kala ‘fish’, kala-na, hiiri 〈mouse〉, hiiri-nä, where 〈a, ä〉 = [!, æ].

17.

The common phenomenon that Meillet (1912) called renouvellement, the recreation of lost structures out of others, is not to the point here. Even if you lose something and make a new one in some other way, the old one is still gone, and the movement toward its loss is directional. De novo creation is not reversal of directionality.

References Anderson, N. 1879. Studien zur Vergleichung der Indogermanischen und FinnischUgrischen Sprachen. Leipzig: K. F. Koehler. Beekes, R. S. P. 1990. Vergelijkende taalwetenschap. Tussen Sanskrit en Nederlands. Utrecht: Het Spectrum BV. Campbell, L. 1991. “Some grammaticalization changes in Estonian and their implications”. In Approaches to Grammaticalization, E. Traugott and B. Heine (eds), 285–299. Amsterdam: Benjamins. Collinder, B. 1960. Comparative Grammar of the Uralic Languages. Stockholm: Almqvist & Wiksell. Givón, T. 1979. On Understanding Grammar. New York: Academic Press. Harris, A. C. and Campbell, L. 1995. Historical Syntax in Cross-Linguistic Perspective. Cambridge: Cambridge University Press. Hopper, P. J. and Traugott, E. C. 1993. Grammaticalization. Cambridge: Cambridge University Press. Krahe, H. 1962. Indogermanische Sprachwissenschaft. 2 vols. Berlin: de Gruyter. Laanest, A. 1982. Einführung in die ostseefinnischen Sprachen. Hamburg: Helmut Buske Verlag. Lass, R. 1975. “How intrinsic is content? Markedness, sound change, and ‘family universals’”. In Essays on the Sound Pattern of English, D. Goyvaerts and G. K. Pullum (eds), 475–504. Ghent: E. Story-Scientia. Lass, R. 1997. Historical Linguistics and Language Change. Cambridge: Cambridge University Press. Lass, R. 1999. “Phonology and morphology”. In The Cambridge History of the English Language, III, 1476–1776, R. Lass (ed.), 56–186. Cambridge: Cambridge University Press. Lyell, C. 1830–33. Principles of Geology; Being an Attempt to Explain the Former Changes to the Earth’s Surface by Reference to Causes Now in Operation. 3 vols. London: John Murray.

REMARKS ON (UNI)DIRECTIONALITY

227

Meillet, A. 1912. “L’évolution des formes grammaticales”. In Linguistique historique et linguistique générale, 131–148. Paris: Champion. Meillet, A. 1964. Introduction à l’étude comparative des langues indo-européennes. Reprint. University Alabama: University of Alabama Press. Newmeyer, F. 1998. Language Form and Language Function. Cambridge, MA: MIT Press. Pinker, S. 1995. The Language Instinct. New York: Harper Collins. Ringe, D. 1995. “‘Nostratic’ and the factor of chance”. Diachronica XII:1.55–74. Sweet, H. 1900. The History of Languages. London: J. M. Dent. Szemerényi, O. 1989. Einführung in die Vergleichende Sprachwissenschaft. Darmstadt: Wissenschaftliche Buchgesellschaft. Thomason, S. G. 1993. “Coping with partial information in historical linguistics”. In Historical Linguistics 1989. Papers from the 9th International Conference on Historical Linguistics, H. Aertsen and R. J. Jeffers (eds), 485–496. Amsterdam: Benjamins. Tooby, J. and Cosmides, L. 1994. “The psychological foundations of culture”. In The Adapted Mind. Evolutionary Psychology and the Generation of Culture, J. H. Barkow, L. Cosmides and J. Tooby (eds), 19–136. New York: Oxford University Press. Whewell, W. 1837. History of the Inductive Sciences from the Earliest to the Present Time. 3 vols. London: John W. Parker.

Soþlice and witodlice Discourse markers in Old English Ursula Lenker University of Munich

1.

Introduction

Those who study grammaticalization processes from a semantic-pragmatic perspective widely agree that the early stages of grammaticalization are characterised not only by semantic weakening but, more importantly, by pragmatic strengthening and increased subjectification (Hopper and Traugott 1993: 63–93; Traugott 1995; Brinton 1996). In an application of this approach Traugott (1995b) has recently examined the role of the development of discourse markers in a theory of grammaticalization. She argues — on the basis of an analysis of the development of Modern English indeed, in fact and besides — that the cline “clause-internal adverbial > sentence adverbial > discourse particle” should be added to the inventory of clines that are the subject of grammaticalization. In this paper I want to show that the Old English adverbs soþlice and witodlice can serve as examples for such a cline. In Old English we find these adverbs used in several coexisting functions: from original manner adverbs and sentential adverbs (disjuncts) they develop to boundary markers, i.e. discourse particles marking thematic discontinuity. Such a ‘layering’ of functions is a characteristic of all language change and in particular a property of the early stages of grammaticalization processes (Hopper and Traugott 1993: 124–126).1

2.

Soþlice and witodlice: The traditional approach

Adverbs such as soþlice and witodlice are notoriously difficult for linguists:

230

URSULA LENKER

(1)

The lexicographers have much work to do with eornostlice, soðlice, witodlice, and the like. When I consider the great variety of Latin words they translate (see BT(S), s.vv.) and when I read Ælfric De Coniunctione (…), I do not envy them (Mitchell 1985: § 3168).

The validity of Mitchell’s statement can be illustrated by the entry for eornostlice in the most recent of the dictionaries of Old English, the Dictionary of Old English (DOE; Cameron et al. 1987):2 (2)

eornostlice (ca. 325 occ.) A.1 strictly, solemnly A.2 steadfastly, stalwartly, resolutely B. used as an introductory or conjunctive adverb, especially to render a variety of Latin conjunctions; when used in this way, the word is usually placed initially in the sentence or clause, but is sometimes postponed (often, but not always, reflecting the position of the Lat. conjunction translated) B.1 therefore, then, so, accordingly (without any definite expression of consequence or result) B.1a rendering ergo ‘therefore, then, so, accordingly’ B.1a.i rendering ergo, used in interrogative constructions B.1b rendering igitur ‘then, therefore, accordingly, consequently’ B.1c rendering itaque ‘and thus, accordingly, therefore’ B.2 indeed, in fact, truly; for in fact B.2a rendering quippe ‘indeed, for, for in fact’ B.2b rendering enim ‘since in fact, inasmuch as, for, because’ B.2c rendering etenim ‘and indeed, in fact, for’ B.2d rendering autem ‘but, yet, but indeed’ B.2e glossing dumtaxat ‘to this extent; at least, at any rate’

Eornostlice’s propositional meaning as a manner adverb ‘strictly, solemnly’ (cf. its etymology ‘in an earnest way’) obviously presents little problem (cf. “A”). Much more important for the present issue are the meanings collected under “B”: after a somewhat general introduction, the lexicographer seeks help in the Latin words which eornostlice commonly renders, and their Modern English translations. The Latin and Modern English lexemes used are, however, so polysemous and multifunctional that eornostlice likewise appears to be strangely polysemous and multifunctional.

SOÞLICE AND WITODLICE

231

This lexicographical procedure is not unique to the DOE but is applied in most other dictionaries as well; as examples cf. the entries for soþlice and witodlice in one of the standard dictionaries of Old English, Bosworth-Toller’s Anglo-Saxon Dictionary (BT; 1882–98):3 (3)

soþlice:

witodlice:

I. II.

as adv. Truly, really, certainly, verily as conj. Now, then, for; representing Latin autem, ecce, enim, ergo, nam, vero adv. I. Certainly II. with a somewhat indefinite sense, translating many Latin words, indeed, surely, truly

The lexicographers of the DOE are actually the less to be ‘envied’ in the cases of soþlice and witodlice when the 325 occurrences of eornostlice are set against 1633 of witodlice and 4801 of soþlice.4 The Latin words these particles translate are even more numerous, polysemous and also more varied in their semanticsyntactic properties. As an example, cf. the evidence from the Gospels according to Matthew and John in the West-Saxon Gospels (= WSG):5 (4)

Matthew:

eornostlice soþlice witodlice

John:

eornostlice soþlice witodlice

autem (1 occ.), ergo autem, ecce, enim, ergo (1 occ.), nam, vero autem, ecce, enim, ergo, etiam, igitur, itaque, nam, quidem, siquidem, utique, vero, et factum est – autem, enim, ergo, vero autem, enim, ergo, igitur, itaque, quidem, utique

More importantly, the Modern English translations given for the Latin adverbs in the Old English dictionaries are contestable themselves as they merely reflect the general sense or adverbial categories to which the words are commonly allocated: ‘causal’ for enim and nam, ‘consecutive’ for ergo and ‘adversative’ for autem, at and vero.6 That this is a far too simplistic treatment of these particles can be illustrated by the entry for autem in the Mittellateinisches Wörterbuch, which shows that the lexicographers of Latin are confronted with exactly the same problems and that they react to them in a similar way, namely by referring to the Greek lemmata translated or to the supposed Latin synonyms:7

232

URSULA LENKER

(5)

I) II) III) IV) V) VI)

c. sensu continuationis, i.q. [idem quod] de, tunc — nun, aber, dann c. sensu contrarietatis vel diversitatis, i.q. at, vero, sed — jedoch, dagegen, aber c. sensu anaphorae, i.q. ergo — also c. sensu causae, i.q. enim — nämlich c. sensu adiunctionis, i.q. etiam — auch c. sensu concessionis — attamen, nihilominus — dennoch, trotzdem; quidem — zwar

The evidence from the Latin and Old English dictionaries shows that a full entry for soþlice or witodlice is likely to go on for pages. It would provide a great many Modern English translations, but it would still — or therefore — not be very helpful.

3.

The discourse-level approach

This paper argues that soþlice and witodlice should be investigated on a different level of language analysis since their purely semantic analysis is not only difficult but in some cases even misleading.8 Soþlice and witodlice will be shown to function as text-structuring discourse markers in Old English narrative discourse, where they are employed as highlighting devices and, more importantly, markers of episode boundaries or shifts in the narrative. This will be demonstrated by an account of the discourse functions of their Latin counterparts, in particular autem, and by an analysis of different Old English texts from the end of the tenth and the beginning of the eleventh century.9 The restriction to prose texts is not coincidental, but corresponds to the textual distribution and thus one of the important properties of soþlice and witodlice. Of the about 4800 occurrences of soþlice only 25 are found in poetry (most of them manner adverbs in direct speech); witodlice is only attested once in poetry. This evidence suggests that the discourse functions of soþlice and witodlice are restricted to prose texts as well.

4.

Morpho-syntactic analysis

Both soþlice and witodlice are adverbs derived from adjectives by means of the adverbial suffix -e. The bases soþlic and witodlic are themselves derivations

SOÞLICE AND WITODLICE

233

from soþ ‘true; (truth)’ and witod ‘certain; appointed, ordained’,10 so that both adverbs have a basic semantic meaning ‘verily, truly, assuredly, knowingly’. Modifying verbs, they can e.g. be employed as manner adjuncts (Quirk et al. 1985: 8.79):11 (6)

Nacode he scrydde, and swa ic soðlice secge,12 ealle nyd-behæfnysse he wæs dælende þam þe þæs behofodon (Eustace 9) ‘The naked he clothed; and, as I truly tell, he distributed to every necessity of them that had need thereof’ (transl. Skeat 1900: 191)13

More frequently they function as ‘emphasizers’, expressing the semantic role of modality since they have a “reinforcing effect on the truth value of the clause or part of the clause to which they apply” (Quirk et al. 1985: 8.99): (7)

Ic eom soðlice romanisc. and ic on hæftnyd hider gelæd wæs (Eustace 344) ‘I am truly a Roman, and I was brought hither in captivity’ (transl. Skeat 1900: 211)

(8)

Apolloni, ic oncnawe soðlice þæt þu eart on eallum þingum wel gelæred (Apollonius 16,24) ‘Apollonius, I know truly that you are well taught in all things’ (transl. Swanton 1975: 166; cf. Latin “Apolloni, intelligo te in omnibus esse locupletem”)

In these cases soþlice serves as a truth-intensifier and fulfils a highly subjective and speaker-oriented function which adds a strong illocutionary force to the speech acts; cf. in particular performative speech acts such as Apollonius’ promise (9)

Ic sille eow soðlice hundteontig þusenda mittan hwætes to ðam wurðe þe ic hit gebohte on minum lande (Apollonius 10,7) ‘Truly, I will supply you with a hundred thousand measures of wheat for the price I bought it in my country’ (transl. Swanton 1975: 163; cf. the Latin “Dabo itaque vobis C milia frumenti modios eo precio quo in patria mea mercatus sum”)

or the (10) consistent translation of Latin Amen, (amen) dico vobis in the West Saxon Gospels by Soþlice ic eow secge (Matthew 6,16, 10,15, 11,11 etc.; Mark 3,28, 8,12 etc.; Luke 4,24, 12,37, 13,16 etc.; John 8,51, 12,24, 13,16). These formulaic expressions are used when Jesus reinterprets the Old Testament by virtue of his authority as the Son of God.

234

URSULA LENKER

In the functions described so far — manner adjunct and emphasizer — soþlice is primarily found in direct speech with a first person (singular) subject,14 which means that in these cases the subject of the sentence is identical with the speaker. This constraint is a consequence of the propositional meaning of the lexeme, in particular the denotative and connotative features of ‘truth’ which demand a human agent with high trustworthiness, most likely the speaker himself. Yet, in the majority of their occurrences the scope of the adverbs soþlice and witodlice is not restricted to the phrase level, but extends to the whole sentence.15 Soþlice and witodlice function as disjuncts, expressing either the comment that what is being said is true (‘content disjuncts’; Quirk et al. 1985: 8.127) or conveying the speaker’s assertion that his words are the unvarnished truth (‘style disjuncts’; Quirk et al. 1985: 8.124).16 So in examples (11) and (12), soþlice is used instead of the full phrases “soþ is þæt ic secge” or “soþlice ic eow secge” (cf. examples 6 and 10).17 This change in perspective involves increased syntactic freedom and scope.18 As sentence adverbials soþlice and witodlice are not part of the core syntactic structure and are thus optional from a syntactic point of view. More importantly, there is no longer any constraint on the subject of the sentence, which may even be inanimate (cf. example 18). When used as a style disjunct, soþlice introduces the voice of the speaker — in addition to the proposition of the sentence. It functions as a speaker comment which conveys the speaker’s assertion that his words (the proposition of the sentence) are true, e.g. the assertion that Eustace (subject) is a righteous man in (11) and (12): (11)

Wæs he soðlice19 on rihtwisnysse weorcum … swiðe gefrætwod (Eustace 4) ‘Truly he was greatly adorned … with works of righteousness’ (transl. Skeat 1900: 191)

(12)

Wæs he witodlice swiðe æþele on rihtwisnysse and strang on gefeohte … (Eustace 14) ‘He was indeed very noble in righteousness, and strong in fight …’ (transl. Skeat 1900: 191)

This development from manner adverb to style disjunct corresponds to the Modern English situation where “ … a manner adverb that may co-occur with the verb tell (when tell is being used performatively) can also function as a style disjunct” (Schreiber 1972: 323).20

SOÞLICE AND WITODLICE

5.

235

Discourse particles in Latin

The usefulness of a discourse-level approach in historical linguistics has recently been demonstrated by Brinton’s (1996) account of pragmatic markers in the history of English and Kroon’s (1995) investigation of the discourse functions of a number of Latin coordinating conjunctions, in particular those which above have been shown to be the Latin counterparts of soþlice and witodlice. Kroon finds that adverbs and conjunctions which in the traditional approach are regarded as carrying a causal or consecutive meaning (nam, enim, igitur and ergo) and those with a supposedly adversative sense (autem, vero and at) actually work on very different levels of discourse, which she calls the representational, presentational and interactional level.21 While vero, ergo and at function on the interactional level of communicative acts and moves, autem, nam and igitur are connectives on the level of textual organization (Kroon 1995: 371–375). Autem, the most common Latin counterpart of soþlice and witodlice should, according to this analysis (Kroon 1995: 226–280), no longer be classified as an adversative conjunction but as a boundary marker functioning on the textual level: (13)

6.

Autem is a presentational particle which marks the discrete status of a piece of information with regard to its verbal or non-verbal context. Depending on whether the particle is applied locally (on the level of the sentence) or more globally (on the level of the text) it can be characterized as a “highlighting” or “focusing” device, or as a marker of the organization of the text (viz. of thematic discontinuity) (Kroon 1995: 226).

Visual clues: Old English initials in MS Cambridge, University Library, Ii. 2. 11

My investigation of the discourse functions of soþlice and witodlice was, however, not sparked off by textual but rather by visual clues. In a main witness for the Gospel lectionary in Anglo-Saxon England (cf. Lenker 1997), a mideleventh century manuscript of the West Saxon Gospels (Cambridge, University Library, Ii. 2. 11), the Old English Gospel text is subdivided into about 200 sections by rubrics. These Old English and Latin rubrics indicate on which day of the liturgical year the following text is commonly read during the performance of the mass. While the first words of the Latin Gospel lection are cited in the rubric, its beginning in the Old English text is highlighted by an initial; cf. the rubric for the lection beginning with Matthew 4,12 (for the Friday after Epiphany):

236

URSULA LENKER

(14)

se deofol hine and englas genealæhton and him þenodon (Mt 4,11) is sceal on frigedæg ofer twelfta dæg. Cum audisset Iesus quod iohannes traditus esset. Soþlice þa se hælend gehirde þæt iohannes belæwed wæs … (Mt 4,12) ‘[left] the devil him and angels came towards him and served him. / This shall (be read) on Friday after Epiphany. / When Jesus heard that John had been betrayed. /Truly, when the saviour heard that John was betrayed …’

Discourse-analytically, it is important that lections have to be complete episodes with a coherent structure. At their beginnings the participants, time, location etc. have to be mentioned as otherwise the congregation would not be able to understand the lection. These characteristics of the beginning of Gospel lections are strikingly similar to the characteristics of episode boundaries, which are indicated by a change in time, location, participants, the action sequence etc.22 In the texts, these changes are commonly denoted by a number of concrete linguistic clues, e.g. syntactic markers such as ‘frame-shifting’ spatial and temporal adverbials, the use of full noun phrases where anaphoric pronouns are expected, certain conjunctions or explicit metacomments and discourse particles (Brinton 1996: 44).23 Examples (15) to (18) show the text division of the first chapter of the Gospel according to Luke (“Birth and childhood of John the Baptist and Jesus”): (15)

Lk 1,26 Soþlice on þam syxtan monðe wæs asend gabriel se engel fram drihtne on galilea ceastre … (Lk 1,27–38) ‘Truly, in the sixth month was sent Gabriel the angel by the Lord to a Galilean town …’

(16)

Lk 1,39 Soþlice on þam dagum aras maria and ferde on muntland mid ofste. on iudeisce ceastre. … (Lk 1,40–55) ‘Truly, in these days Mary got ready and went to the hill-country with haste to a Judaean town …’

(17)

Lk 1,56 Soþlice maria wunude mid hyre swylce þry monþas. and gewende þa to hyre huse. Lk 1,57 a wæs gefylled elizabethe cenningtid. and heo sunu cende … (Lk 1,58–80) ‘Truly, Mary lived with her such three months and went then (back) to her house. Then came the time for Elizabeth to give birth and she gave birth to a son …’

237

SOÞLICE AND WITODLICE

(18)

Lk 2,1 Soþlice on þam dagum wæs geworden gebod fram þam casere augusto. þæt eall ymbehwyrft wære tomearcod … (Lk 2,2) ‘Truly, in these days an order was given by the Emperor Augustus that all the world should be described …’

In each of the lection-initial sentences the participants, the location and the time of the action are explicitly mentioned, while the beginning of the lection itself is denoted by sentence-initial soþlice.24 It is obvious that the sentence adverbial soþlice is semantically bleached in these cases, as its main function is no longer to convey the speaker’s assertion or comment that what is being said is true. Augustus’ order (18) to have a census taken is not a matter of the speaker’s subjective belief, but a historical fact. Soþlice is no longer a style disjunct which replaces a full phrase such as ‘soþlice ic eow secge’. Its function here is to indicate the beginning of a new lection. Soþlice thus works on the (meta)textual level, as a boundary marker with demarcating force. This use as a discourse marker obviously develops from its function as a style disjunct: by explicitly stating that what is being said is true, the speaker manages to catch the listener’s or reader’s attention at the beginning of a new episode. This use of soþlice as a boundary marker25 is not unique to the first chapters of Luke, as can be shown by an inventory of the lection-initial words which are highlighted as initials:26 (19) Mt Mk Lk John

þa

soþlice

witodlice

eornostlice

and

(others)

27 13 17 19

13 02 12 02

2 — 1 5

1 — 1 —

7 5 2 1

21 03 12 38

66

29

8

2

150

(74)

With a total of 37 instances soþlice and witodlice are the second most common of the boundary markers employed and are only outnumbered by the particle þa whose function as a discourse marker is undisputed (cf. e.g. Enkvist and Wårvik 1987; Kim 1992). For the WSG, Kim’s thorough discourse-level analysis (1992) provides convincing evidence that þa-clauses in this text signal some kind of discourse discontinuity, indicating a shift of topic, ground, time-line, scene, listener or content.27 The fact that more than half of the lections begin with þa, soþlice, witodlice or — in two instances — eornostlice thus suggests that not

238

URSULA LENKER

only þa, but also soþlice and witodlice should be regarded as explicit markers of discourse discontinuity.28

7.

Soþlice and witodlice in Old English narratives

In a next step the analysis will now concentrate on the discourse functions of soþlice and witodlice in selected passages from texts which are comparatively independent of their Latin exemplars, the Old English translation of the GreekLatin romance Apollonius of Tyre (ed. Goolden 1958) and the “Life of Saint Eustace” by Ælfric (ed. Skeat 1900: 190–218). The Old English Apollonius is a narrative text whose plot of various, sometimes unrelated adventures is structured by means of short chapters. A number of chapter beginnings (Chapters 3, 4, 6, 16, 17, 49, 50 and 51) are denoted by the boundary marker soþlice which only rarely has a Latin adverb as its counterpart in the exemplar (most probably29 vero in 6,1, 16,1 and 51,1). Soþlice here signals an interruption of a thematic chain by a change in time, aspect, participants or action sequence. At the beginning of Chapter 3 Antioch continues30 to abuse his daughter (3,1 “On þisum þingum soþlice þurhwunode …”) and then asks her admirers a riddle — a definite turn in the sequence of events which brings about the misfortune of the young men who fail to solve the riddle: their heads are set up over the town gate. (20)

3,1 On þisum þingum soðlice þurhwunode se arleasesta cyngc Antiochus … he asette ða rædels þus cweðende: “ …”. … And þa heafda ealle wurdon gesette on ufeweardan þam geate. ‘3,1 In fact the infamous king Antiochus persisted in this state of affairs … he set them a riddle, saying: ‘…’. … And their heads were all set up over the gate’ (transl. Swanton 1975: 159)

Antioch persists in this cruelty (4,1 “Mid þi soðlice … þurhwunode”) until a new protagonist, Apollonius, enters the scene. In his greeting Apollonius maintains his right to marry the king’s daughter because “ic eom soðlice of cynelicum cynne cumen” (4,8), employing in his speech alliteration and an emphasizer, the truthintensifier soðlice. As a next break in the thematic chain, indicated to the reader by þa soþlice (4,19), Apollonius receives the riddle, the precondition for the next turn of events, his successful solving of the riddle:

SOÞLICE AND WITODLICE

(21)

239

4,1 Mid þi soðlice Antiochus se wælreowa cyningc on þysse wælreownesse þurhwunode, ða wæs Apollonius gehaten sum iung man se wæs swiðe welig … 4,6 Eode þa into þam cyninge and cwæð: “Wes gesund, cyningc … Ic eom soðlice of cynelicum cynne cumen and ic bidde þinre dohtor me to gemæccan” … 4,19 Apollonius þa soðlice onfangenum rædelse hine bewænde hwon fram ðam cyninge, and mid þy þe he smeade ymbe þæt ingehyd, he hit gewan mid wisdome and mid Godes fultume he þæt soð arædde. ‘4,1 While the cruel king Antiochus in fact persisted in this cruelty, there was a certain young man called Apollonius who was very wealthy …. 4,8 He went to the king and said: “All hail, King … I come in fact from a regal family, and I ask for your daughter as my wife” … 4,15 Then, truly, having received the riddle, Apollonius turned himself a little away from the king, and when he had considered the sense he solved it with wisdom, and with God’s help he guessed the truth’ (transl. Swanton 1975: 159–60)

The transition from Chapter 50 to 51 finally demonstrates that (þa) soþlice can also, though much less frequently, mark the termination of an episode, the end of a sequence of actions involving certain participants. With þa soþlice in 51,1 Apollonius is reintroduced as the protagonist. (22)

(23)

50,2 For ða soðlice þanon to Tharsum mid his wife and mid his dohtor … 50,29 Heo ræhte þa soðlice hire handa him to and het hine gesund faran, and Philothemian, þare forscildgodan dohtor, Thasia nam to hyre. ‘50,2 Then afterwards he went to Tharsus with his wife and his daughter. … 50,29 Then indeed, she extended her hand to him and bade him go in safety; and Thasia took to herself Philothemia, the daughter of the guilty woman’. 51,1 Apollonius þa soðlice forgeaf þam folce micele gifa to blisse, and heora weallas wurdon geedstaðelode. He wunode þa þar six monðas … ‘51,1 Apollonius then, indeed, gave the people great gifts to rejoice them; and their walls were restored. Then he stayed there six months …’ (transl. Swanton 1975: 172)

The information given in the soþlice-clauses is not the most salient for these episodes. The clauses do commonly not contain the core events of the episode, but provide the background information for the events to follow. With respect to grounding, Brinton (1996: 116–143) has lucidly analyzed this for the functions of the different parts of the Old English ‘gelamp-construction’ (þa hit gelamp þæt; þa hit gewearð þæt),31 a much more obvious episode boundary marker in Old

240

URSULA LENKER

English, which “grounds episodes in the narrative … and guides the reader through the episodic structure of the text” (143). She suggests that events which are temporally or causally prior to the core events of the episode are “backgrounded in this fashion if they constitute the initiating or instigating event of the episode” (133). While the main clause serves as a metacomment upon the narrative structure, the complement clause establishes the necessary conditions for the episode to occur. Soþlice is functionally synonymous32 with the main clause of the ‘gelamp-construction’ and thus serves as a metacomment upon the narrative. Ælfric’s “Life of Saint Eustace” shows that the demarcating force of witodlice and soþlice is also found in texts which were not translated but composed in Old English. The first episode after the introduction, which is marked by means of a ‘gelamp-construction’ (“Hit gelamp sume dæge”), relates the events which lead to Placidas’ conversion (24). Placidas, who goes out hunting hart, departs from his companions and has a vision of a hart between whose horns the likeness of Christ’s holy rood glitters. In this core section of not only the episode but the whole homily, witodlice is used three times to guide the reader by explicitly indicating the most important events (cf. Skeat’s more and more emphatic translations ‘verily’, ‘indeed’ and finally ‘behold’): (24)

Hit gelamp sume dæge þæt he ferde ut on huntað mid eallum his werode and his wuldre. … þa hi ealle ymb þone huntað abysgode wæron þa æteowode him-sylfum an ormæte heort … Þa æt nixtan wurdon hi ealle geteorode and he ana unwerig him æfter fyligde. Witodlice þurh godes fore-stihtunge ne hors ne he sylf gewergod wæs … and feor fram his geferum gewat. Se heort þa witodlice astah on anne heahne clud and þær gestod … Him þa god geswutelode þæt he him swilcne dom ne gedrede ne his mægnes micelnesse ne wundrode. Witodlice betwux þæs heortes hornum glitenode gelicnys þære halgan cristes rode breohtre þonne sunnan leoma (24–43) ‘It happened one day that he went out hunting with all his company and array … When they were all busied about the hunting, then there appeared to himself an immense hart, … then at last they were all tired and he alone, unweary, followed after it … Verily through God’s predestination neither his horse nor himself was wearied … and he departed far from his companions. Then indeed the hart mounted up on a high rock and there stood. … Then God revealed to him that he should not fear such power, nor wonder at the greatness of his might. Behold, between the hart’s horns glittered the likeness of Christ’s holy rood, brighter than the sun’s beam’ (transl. Skeat 1900: 193)

SOÞLICE AND WITODLICE

241

Witodlice’s use as a highlighting device here is functionally similar to its employment as an episode boundary marker, which is also a highlighting device on the more global level of textual organization (cf. Kroon’s description of the functions of autem in [13]). For (þa) soþlice, the findings agree even more closely with those for the other texts.33 In example (25), the soþlice-clause indicates a change in the action sequence and an orientation toward a new central event when, after a lengthy dialogue, Eustace and his wife leave their home to find a priest who will baptize them. (25)

Þa cwæð Placidas to hire: “þæt ylce me sæde se þe ic geseah”. Þa soþlice to middre nihte hi ferdon swa heora menn nyston to cristenra manna sacerda … and halsodon hine þæt he hi gefullode (88–91) ‘Then said Placidas to her: “He whom I saw said the same to me”. Then verily at midnight they went, so that their servants should not know it, to the Christian men’s priest … and entreated him to baptize them’ (transl. Skeat 1900: 197)

A change in time, location, participants and action sequence is denoted by þa soþlice in (26). After the baptism and the last words of the priest, Placidas, who is now called Eustace, gathers a few companions in order to return to the place of his vision. (26)

“… and gemunað me iohannis ic bidde eow”. Þa soþlice eft on ærne mergen genam eustachius feawa geferan. and ferde to ðære stowe þær he ær þa ge-syhðe geseah (103–5) ‘ “ … and remember me, John, I pray you”. Then verily again in the early morning Eustace took a few companions and went to the place where he had before seen the vision’ (transl. Skeat 1900: 197)

In this homily, soþlice’s demarcating force is, however, restricted to marking sub-episodes. Major shifts in the narrative are indicated by ‘gelamp-constructions’, such as Hit gelamp sume dæge (24) or Æfter þissum wæs geworden (27).34 (27)

Æfter þissum wæs ge-worden micel hergung on þam lande þe eustachius ær on wæs … (222) … Þa ferdon soðlice twegen cempan þa wæron genemde antiochus and achaius …. (230) ‘After this there was made a great invasion of the country wherein Eustachius had been first … Then went two soldiers who were named Antiochus and Achaius …’ (transl. Skeat 1900: 205)

242 8.

URSULA LENKER

The two versions of Wærferth’s translation of Gregory’s Dialogues

The pragmatic functions of soþlice and witodlice can be confirmed by a comparison of chapter-initial examples taken from a text which has come down to us in two Old English versions. Gregory the Great’s Dialogues on the Lives and Miracles of the Italian Fathers (ed. Hecht 1900) is an extremely clearly structured text which in short episodes (capitula) relates miracles of holy men. It was first translated by Bishop Wærferth of Worcester in 890 and was revised anonymously about a century later. The modifications found in chapter-initial sentences from the First Book which are not caused by the Latin text35 indicate substitution possibilities and can therefore serve as a kind of historical test-frame. Cf. the following selection of examples: Wærferth of Worcester (890) (28) a. Soðlice sume dæge hit gelamp, þæt an nunne of þam ylcan mynstre … eode … þa geseah heo ænne leahtric (1900: 30) ‘Truly, one day it happened, that a nun she saw a lettuce … ‘

anonymous reviser (950–1050) b. Soðlice sumon dæge hit gelamp, þæt an nunne of þam ilcan mynstre … eode … þa geseah heo ænne leahtric and … of the same monastery … went; then

(29) a. b. þa æt nextan becom þisses ylcan weres Witodlice þa æt necstan se hlisa þyses hlisa to cyþnysse Romana bisceope … ylcan weres bodunge becom to (1900: 34) cyðnysse Romanebyri. ‘Then at next the fame of the same man became known to the Roman bishop …’ (30) a. b. Eac hit gelamp on sume tid, þæt him Soðlice on oðrum timan him comon to comon twegen men to … and þa sealde twegen men … þa sealde he him … he heom … (1900: 66) ‘It also happened some time, that two men came to him … and then he gave them …’

SOÞLICE AND WITODLICE

243

(31) a. b. Eac þeos ylce his modur gewunode, Witodlice þeos ilce his modor þæt heo hæfde hire henna and fedde … gewunode to fedenne henna … c. d. þa sume dæge stod Bonefacius se cniht soðlice sume dæge, þa þa se cniht in þam ilcan ingange; þa com se fox … Bonefatius stod on þam ylcan ingange, (1900: 69) þa com se fox … ‘Also, the mother of the same (one) had hens and fed them … then one day when the boy Boniface was standing in the same entrance, the fox came …’ Already the original translator employs a number of the particles and phrases which have above been shown to serve as episode boundary markers: the text is structured by the use of þa (29a, 31c), soþlice (28a), and ‘gelamp-constructions’ (28a, 30a), either alone or in combination. The most striking result of the comparison of the two versions is, however, that the reviser explicitly marks the beginning of the chapter by witodlice (31b) or uses sentence-initial soþlice and witodlice (29b, 30b, 31d) to replace þa or a ‘gelamp-construction’. This is a clear indication that soþlice and witodlice are functionally equal to discourse markers with a more obvious demarcating force, such as the ‘gelamp-construction’, and that they indeed functioned as episode boundary markers for a ‘native speaker’ of Old English.

9.

Conclusions

A comparison of the morpho-syntactic and functional characteristics of soþlice and witodlice with the properties of other Old and Middle English discourse particles as established by Brinton (1996: 265–267) shows that these Old English words fit well into the pattern.36 They are high-frequency words which often occur in sentence-initial position. As sentence adverbials, they exist outside the core syntactic structure and are syntactically detachable from the sentence. They show an apparent lack of semantic content and are therefore able to work at both local and global levels of discourse. More importantly, the discourse-level analysis of their functions shows them to belong to the linguistic devices which denote the textual structure of Old English narratives: from original manner adjuncts, truth-intensifying emphasizers and sentence adverbs (style disjuncts) soþlice and witodlice develop into semantically bleached37 and pragmatically enriched indicators of thematic discontinuity and are consequently employed as

244

URSULA LENKER

episode boundary markers. They thus follow exactly the stages of grammaticalization Traugott proposes for the adverbial cline “manner adverb > sentence adverbial > discourse marker” (Traugott 1995b).38

Acknowledgments I would like to thank Alfred Bammesberger, Walter Hofstetter, Lucia Kornexl, Justin Larsen, Andreas Mahler and the anonymous commentators for their most helpful comments on earlier versions of this paper.

Notes 1.

For the later development of soþlice see Lenker (1999); witodlice is not attested after the Old English period.

2.

The Dictionary of Old English has, to this point, only been issued to the letter “E”, so that the entries for soþlice and witodlice are not yet available.

3.

See also the entry for sothli in the Middle English Dictionary (Kurath et al. 1954–): “… 3) as quasi-conj. a) used as a connective (often translating L autem, enim, vero): and, for; b) with adversative sense (often translating L autem, vero): but; c) with causal sense (often translating L enim): because, for since” or the entry for soothly in the OED: “ … (2) Used to render L. autem, enim, ergo etc. Obs.”. In BT, the introductory sentences to the entry for witodlice — “with a somewhat indefinite sense, translating many Latin words” — show that the editors were well aware of the problems connected with these particles.

4.

Counts according to the frequency list of the Microfiche Concordance to Old English (Healey and Venezky 1980; MCOE); soþlice is filed under the separately issued ‘high-frequency words’.

5.

The exact version of the Latin Vulgate text used as an exemplar for the translation of the WSG (ed. Liuzza 1994) is not known (Lenker 1997: 28–41). The Vulgate text used for the present purpose is Nestle-Aland’s edition (Aland and Aland 1984).

6.

For the treatment of these particles in traditional grammars and handbooks see Kroon (1995: 1, 132–143 [nam, enim], 217–225 [autem, vero, at]).

7.

Other dictionaries, such as the Mittellateinisches Glossar by Habel and Gröbel (1959), refer to autem and enim as “meaningless fillers”. The only dictionary which includes discoursepragmatic information is the Oxford Latin Dictionary (1968), s.v. autem: “5) introducing a fresh idea or consideration …; 6) expr. indignation or surprise in questions and exclamations”.

8.

One of the more obvious problems is the rather clumsy Modern English translation ‘verily’ which is commonly used for both adverbs.

9.

Soþlice and witodlice are attested in translated texts of the earlier periods of Old English (cf. MCOE and Swan 1988: 92). These texts do not lend themselves to a discourse-pragmatic investigation as most of them are too dependent on their Latin exemplars.

SOÞLICE AND WITODLICE

245

10.

Cf. Modern English soothsayer and the archaic in sooth. Witod is the past participle of the verb witian ‘to order, to decide’, which is cognate to the verb witan ‘to know’.

11.

For the four broad categories of grammatical function of adverbials — adjuncts, subjuncts, disjuncts and conjuncts – see Quirk et al. (1985: 501–647).

12.

On the level of the phrase, soðlice is a manner adjunct modifying secgan. On the level of the sentence the whole phrase swa ic soðlice secge serves as a style disjunct.

13.

The passages from Apollonius are cited by chapter and line number, the passages from Ælfric’s “Life of St Eustace” by line numbers only. The translations, which are taken from Swanton’s Anglo-Saxon Prose (1975; Apollonius) and Skeat’s edition of Eustace (1900), demonstrate the difficulties of translation.

14.

As further examples cf. Apollonius 2,14, 4,8, 8,11, 14,30, 16,9, 21,20 etc. or Eustace 128, 137, 202, 209, 369 etc.

15.

For this functional shift see Swan (1988: 91–110) and Sato (1990). Other similar Old English introductory adverbs which may even be used as conjunctions are e.g. ær, forðon, huru, nu, ono, þær, þanon, þider, þa, þeah, þonne, siþþan, swa (Mitchell 1985: §§ 1101, 1862, 2418).

16.

Likewise in Modern English, most of the common emphasizers (certainly, indeed, surely, for certain, for sure) can function as disjuncts (cf. Quirk et al. 1985: 8.100).

17.

For further attestations of the full phrases cf. Wulfstan’s homilies (WHom 13,79 “And þæt is witodlice ful soþ …”, Hom 17,65 “Eala, eala, soð is þæt ic eow secge …”, WHom 20.1,1 “Leofan men, gecnawað þæt soð is …”; cited from MCOE).

18.

Traugott finds that “the proposed adverbial cline involves increased syntactic freedom and scope” (1995b: 1). This violates “the principles of bonding and reduced scope frequently associated with grammaticalization” (1995b: 1); for an accordingly modified definition of grammaticalization see Traugott (1995b: Chapter 5).

19.

In Modern English, sentence adverbials are mostly placed sentence-initially. This is not in all cases true for Old English, though the examples from the WSG (10, 14–18) demonstrate that there is a tendency to front them. They are always found in the left periphery of the sentence. For a more detailed discussion of this syntactic slot see Traugott (1995b: 3.1).

20.

The syntactic conditions are described as “[I]NP [[+V, +performative, +communication, +linguistic, +declarative] [you]NP (Advmanner) S]VP” (Schreiber 1972: 325).

21.

Most of the research on discourse particles is, sometimes critically, based on Schiffrin (1986) (cf. Kroon 1995: 7–57 and Brinton 1996: 29–65 for the literature on discourse analysis and its terminology). Kroon’s three levels more or less correspond to the more familiar distinction between the ideational or propositional, the textual and the interpersonal level. The ideational or propositional level considers the semantic content proper. On the textual level methods of organization which create a coherent discourse are investigated. The interpersonal level refers to the social and expressive functions of communicative acts and moves.

22.

According to Brinton, episode boundaries correspond to one of the following points of change: “a change in time”, “a change in location”, “a change in participants”, “a change in the action sequence, with an orientation toward a new central event, or the activation of a new schema”, “a change in ‘possible world’”, “a change from general to specific, or the reverse” and “a change in perspective or point of view” (Brinton 1996: 43–44).

23.

The underlined words and phrases in the following examples show the correspondences to these linguistic clues.

246

URSULA LENKER

24.

Witodlice as a boundary marker is e.g. used in John 20,1: “Witodlice on anon restedæge seo magdalenisce maria com on mergen ær hit leoht wære to þære byrgenne …”; ‘Assuredly, early on the Sabbath Mary Magdalene, in the morning before it was light, went to the tomb …’.

25.

The evidence is not as clear when we consider the whole Gospel text, in particular Chapters 5–13 of the Gospel according to Matthew, where the translator, which may be a different one than in other parts of the translation, indiscriminately renders every autem and enim he or she can find in the Latin exemplar by soþlice. The textual distribution of soþlice and witodlice for Latin autem has been repeatedly used to prove that there were at least two translators involved — one for Mt and Joh and the other for Mk and Lk (cf. Lenker 1997: 50–4).

26.

The importance of a manuscript’s layout must not be underestimated from a text-semiotic point of view. As there is commonly no paragraph indentation in Old English manuscripts, the highlighting of the first letters was certainly an important and striking feature for both the copyists and readers of the text.

27.

92% of the about 1200 sentence-initial þa-clauses investigated by Kim show one or more of these discourse shifts (see the summary in Kim 1992: 152). For points of disagreement concerning the specific discourse functions of þa see Enkvist and Wårvik (1987); Kim (1992: 1–6) and Brinton (1996: 9–11).

28.

This assumption is further supported by the differences between the four Gospels. In contrast to the predominance of þa, soþlice and witodlice in the Gospels according to Matthew, Mark and Luke, the Gospel according to John uses a much broader variety of lection-initial words, such as different temporal adverbials, determiners or pronouns (listed among “others” in [19]). This can be explained by the different textual structure of this Gospel which is not organized as a collection or sequence of episodes, but as long elaborated speeches in which the narrative parts are interwoven.

29.

The exact exemplar of the translation is not known: for the relationship of the Old English translation, of which only Chapters 1–22 and 48–51 are extant, with the reconstructed Latin exemplar see Goolden (1958: xx–xxv).

30.

Antiochus, the wicked king of Antioch, rapes and abuses his daughter and as he therefore does not want her to get married asks her admirers a supposedly insoluble riddle: whoever is successful in solving the riddle gets his daughter as a wife, whoever does not is beheaded. Soþlice here marks a change in aspect.

31.

The syntactic construction “(þa) VHAPPEN (hit) (Adv) þæt Complement Clause” occurs most frequently with the verbs (ge)limpan, (ge)weorþan, beon, but also with gebyrian, getimian, agan and gesælan in poetic texts (Brinton 1996: 116).

32.

As an interesting case of an explicit metacomment which shows that the ‘gelamp-construction’ is also semantically similar to soþlice cf. Eustace 361: “Nu ic hæbbe eall þis gesæd swa hit gelamp …”; ‘Now I have said all this as it happened …’ (transl. Skeat 1900: 213).

33.

For an example in which eornostlice marks a change in the action sequence cf. Eustace 419: “Eornostlice se casere geseah þas wundorlican wæfersyne, þæt se leo heora ne oðhran, þa het he gefeccan ænne ærenne oxan and þone onælan and þa halgan ðær-on don …”; ‘Earnestly the emperor saw this wonderful spectacle, that the lioness touched them not; then bade he fetch a brazen ox and heat it and put the saints therein’ (transl. Skeat 1900: 217).

34.

In line 153, an episode begins with “on þam dagum gelamp”, the sub-episode (162) with “soðlice æfter þam þe hi ferdon twegen dagas”. Other episodes with an initial ‘gelampconstruction’ are e.g. Eustace 141, 316, 391.

35.

Cf. Yerkes (1979: xvi): “The anonymous reviser changed thousands of words and phrases,

SOÞLICE AND WITODLICE

247

sometimes no doubt to render the Latin more closely, at other times apparently only to bring the diction of the translation up to date or into conformity with that of his own dialect”. The examples selected for comparison most probably belong to the second group. 36.

Cf. also Schiffrin’s “tentative suggestions as to what specific conditions allow an expression to be used as a marker: … it has to be syntactically detachable from a sentence, it has to be commonly used in initial position of an utterance, it has to have a range of prosodic contours (e.g. tonic stress and followed by a pause, phonological reduction), it has to be able to operate at both local and global levels of discourse … this means that it either has to have no meaning, a vague meaning …” (1986: 328). — For their “range of prosodic contours” cf. Mitchell (1985: § 2423): “I believe that in Old English ‘phonological differentiation’ existed between adverbs … and conjunctions of the same spelling”; for a similar case cf. the written and spoken form of Modern English “you know”.

37.

It is evident that the different meanings coexist for a certain period. Initially there is only a redistribution or shift, not a loss, of meaning. In the case of soþlice and witodlice the originally salient truth-intensifying meaning persists over time and also constrains the later uses of the grammaticalized form (cf. Hopper and Traugott 1993: 87–93).

Texts Aland, K. and Aland, B. (eds). 1984. Nestle-Aland. Novum Testamentum Latine. Stuttgart: Deutsche Bibelgesellschaft. Goolden, P. (ed.). 1958. The Old English Apollonius of Tyre. Oxford: Oxford University Press [= Apollonius]. Hecht, H. (ed.). 1900. Bischofs Wærferth von Worcester Übersetzung der Dialoge Gregors des Grossen. Leipzig: Georg H. Wigand’s Verlag. Liuzza, R. M. (ed.). 1994. The Old English Version of the Gospels. Volume 1: Text and Introduction [Early English Text Society 304]. Oxford: Oxford University Press [= WSG]. Skeat, W. H. (ed.). 1900. Ælfric’s Lives of Saints. Volume 2 [Early English Text Society 114]. Oxford: Oxford University Press [= Eustace]. Zupitza, J. (ed.). 1966. Ælfrics Grammatik und Glossar. Text und Varianten, 2nd edition with an introduction by Helmut Gneuss. Berlin: Weidmannsche Verlagsbuchhandlung [1880].

References Bosworth, J. and Toller, T. N. (eds). 1882–98. An Anglo-Saxon Dictionary. Oxford: Clarendon [= BT]. Brinton, L. J. 1996. Pragmatic Markers in English. Grammaticalization and Discourse Functions [Topics in English Linguistics 19]. Berlin/New York: Mouton de Gruyter. Cameron, A., Amos, A. C. and Healey, A. di Paolo (eds). 1986. Dictionary of Old English. Toronto: Pontifical Institute of Mediaeval Studies [= DOE].

248

URSULA LENKER

Enkvist, N. E. and Wårvik, B. 1987. “Old English þa, temporal chains, and narrative structure”. In Papers from the Seventh International Conference on Historical Linguistics, A. G. Ramat, O. Carruba and G. Bernini (eds), 220–237. Amsterdam/Philadelphia: John Benjamins. Habel, E. and Gröbel, F. (eds). 1959. Mittellateinisches Glossar, 2nd edition. Paderborn: Schöningh. Healey, A. di Paolo and Venezky, R. (eds). 1980. A Microfiche Concordance to Old English. Toronto: Pontifical Institute of Mediaeval Studies [=MCOE]. Hopper, P. and Traugott, E. C. 1993. Grammaticalization [Cambridge Textbooks in Linguistics]. Cambridge: Cambridge University Press. Kim, T. 1992. The Particle þa in the West Saxon Gospels: A Discourse Level Analysis [European University Studies, Series XIV, Anglo-Saxon Language and Literature 249]. Bern: Lang. König, E. 1991. The Meaning of Focus Particles: A Comparative Perspective. London: Routledge. Kroon, C. 1995. Discourse Particles in Latin: A Study of nam, enim, autem, vero and at [Amsterdam Studies in Classical Philology 4]. Amsterdam: J. C. Gieben. Kurath, H., Kuhn, S. M. and Lewis, R. E. (eds). 1954–. Middle English Dictionary. Ann Arbor: University of Michigan Press. Lenker, U. 1997. Die westsächsische Evangelienversion und die Perikopenordnungen im angelsächsischen England [Texte und Untersuchungen zur Englischen Philologie 20]. München: Fink. Lenker, U. 1999. “Actually: soþlice, treuli, indeed — Grammaticalization of truthintensifying adverbs in the history of English”. Paper presented at the international symposium on ‘New Reflections on Grammaticalization’, Potsdam, 17–19 June 1999. Mitchell, B. 1985. Old English Syntax, 2 vols. Oxford: Clarendon. Prinz, O. and Schneider, J. (eds). 1967. Mittellateinisches Wörterbuch bis zum ausgehenden 13. Jahrhundert. Vol. 1: A-B, München: C. H. Beck. Quirk, R., Greenbaum, S., Leech, G. and Svartvik, J. 1985. A Comprehensive Grammar of the English Language. London: Longman. Sato, A. 1990. “Soþlice and witodlice: A case of functional shift of Old English adverbs”. English Linguistics: Journal of the English Linguistic Society of Japan 7: 39–55. Schreiber, P. 1972. “Style disjuncts and the performative analysis”. Linguistic Inquiry 3: 321–347. Schiffrin, D. 1987. Discourse Markers [Studies in Interactional Sociolinguistics 5]. Cambridge: Cambridge University Press. Swan, T. 1988. Sentence Adverbials in English: A Synchronic and Diachronic Investigation [Tromsø-Studier i Språkvitenskap 10]. Oslo: Novus. Swanton, M. 1975. Anglo-Saxon Prose. London: Dent. Traugott, E. C. 1995a. “Subjectification in grammaticalization”. In Subjectivity and Subjectivisation, D. Stein and S. Wright (eds), 31–54. Cambridge: Cambridge University Press.

SOÞLICE AND WITODLICE

249

Traugott, E. C. 1995b. “The role of the development of discourse markers in a theory of grammaticalization”. Paper presented at ICHL XII, Manchester, 1995; version of 11/97 (http://www.stanford.edu/~traugott/ect-papersonline.html). Yerkes, D. 1979. The Two Versions of Waerferth’s Translation of Gregory’s Dialogues: An Old English Thesaurus [Toronto Old English Series 4]. Toronto: University of Toronto Press.

Onginnan/beginnan with bare and to-infinitive in Ælfric Bettelou Los Free University Amsterdam

1.

Introduction

Onginnan and beginnan, both meaning ‘begin’, allow two infinitival complements: the infinitive with to and the bare infinitive (without to). There are syntactic and semantic differences between the two constructions after these inchoative verbs which seem to point to auxiliary status when these verbs are followed by the bare infinitive, and lexical status when they are followed by the to-infinitive. As an auxiliary, onginnan/beginnan with the bare infinitive is shown to have been subject to considerable semantic bleaching. When onginnan/ beginnan are followed by a to-infinitive, there is no bleaching: they are full lexical verbs with inchoative meaning. This situation is reached at least by the time of Ælfric (late Old English). The two constructions with these verbs in his texts exhibit a number of syntactic differences; this paper will argue that these follow from the interplay of the semantic, aspectual differences observed between the two infinitival constructions with these verbs, and certain discourse effects that are typical of narrative contexts. The development of Old English verbs capable of taking a bare-infinitival complement is particularly interesting in the light of the general question of the syntactic status of the forerunners of the present-day English auxiliaries. At some stage in the Middle English period lexical verbs can no longer be complemented by a bare infinitive, only by a to-infinitive; the bare infinitive becomes the prerogative of a small set of grammaticalized verbs that have evolved into functional elements with no argument structure of their own. This study examines the evidence for auxiliary versus full lexical status of two inchoative verbs, onginnan and beginnan, at a later stage of Old English.

252

BETTELOU LOS

Callaway (1913: 67–8) notes that the co-occurrence of both infinitives after onginnan/beginnan is particularly noticeable in the works of Ælfric, which is one of the reasons why this investigation into the use of bare and to-infinitives concentrates on his texts. Moreover, restricting the corpus to the works of Ælfric makes it possible to examine every instance of onginnan/beginnan in its context, which would otherwise hardly be a feasible task because of the sheer numbers in which these verbs are attested in Old English. Studying the work of a single author also minimizes the risk of interference from dialectal or diachronic factors (although, of course, the risk of interference from scribes remains — see Allen 1992). Ælfric’s works have the further advantage of consisting, for the most part, of original, untranslated prose; and although Ælfric draws heavily on various Latin and Greek sources, his views on the translation process, explicitly stated in his preface to the translation of Genesis in 〈ÆGen Pref 93〉,1 make it unlikely that he would opt for an over-literal translation. A search for all occurrences of these verbs in the works of Ælfric resulted in the following figures. Onginnan was complemented by the bare infinitive 44 times, and by the to-infinitive 22 times. Beginnan was complemented 13 times by a bare infinitive, and 46 times by a to-infinitive. There is no obvious explanation for why we find a bare infinitive after onginnan/beginnan in some cases and a to-infinitive after another. Although Ælfric’s prose has been described as “rhythmical”, the selection of a bare or a to-infinitive in his texts does not appear to be dictated by prosodic considerations; it must also be remembered that Ælfric’s rhythmical prose is by no means as strict in its metrical rules as the relatively firm patterns of Old English poetry (see also Pope 1967: 105). Callaway studies a number of possible factors, such as meaning differences, semantic bleaching, the distance between infinitive and governing verb; but in the end he concludes that the differentiation of the two infinitives remains unexplained: Be the reason what it may, the inflected infinitive caught the fancy of Ælfric, and to him we owe 25 out of the total of 37 inflected infinitives after onginnan. On the other hand, Ælfric uses the uninflected infinitive 82 times without making any rational differentiation between the two so far as I can discover. (Callaway 1913: 67)2

We will see in the remainder of this chapter that Callaway’s comments contain powerful pointers to a coherent account of the two infinitives after these aspectualizers, and that there is no need to have recourse to Callaway’s conclusion that “the fluctuation in the two forms of the infinitive may be due in part merely to the lateness of Ælfric’s works, by whose time the distinction between

ONGINNAN/BEGINNAN WITH BARE AND TO-INFINITIVE

253

the two forms had begun to break down” (Callaway 1913: 67). The issue of semantic bleaching and the possibility of auxiliary status are indeed the prime ingredients in the account of the selection of a bare or of a to-infinitive which we will present below. We will start with a more detailed look at the syntactic differences between the two complements.

2.

General syntactic differences between bare and to-infinitival complements

In general, bare infinitive-complements exhibit a less fixed word order than to-infinitival complements; the bare infinitive very frequently precedes the finite verb, as in (1): (1)

He geseah þa standan swiþe gehende þone deofol, he saw then stand very nearby the devil and he hine orsorhlice axian ongan, and he him cheerfully ask began Hwæt stendst þu her wælhreowa deor? 〈ÆLS (Martin) 1364〉 what stand you here cruel animal? ‘He then saw the devil standing quite nearby, and he began to ask him cheerfully: what are you standing here for, cruel animal?’

With to-infinitives this order is extremely rare. This was already noticed by van der Gaaf (quoted by Callaway 1913: 65), and Georg Riggert, who adds at the very end of his dissertation, almost as an afterthought: “Bezüglich der Stellung des präpos. Inf. ist zu bemerken, daß stets der Infinitiv dem regierenden Worte folgt” (‘As regards the position of the prepositional infinitive, it may be observed that it follows the word that governs it’) (Riggert 1909: 75). The same is noticed by Bock (1931: 154). A second typical feature of bare infinitival complements after onginnan/ beginnan is that the matrix verb (i.e. onginnan/beginnan) in such a construction can be ‘pervious’ to the argument structure of the infinitive. This becomes immediately evident with impersonal infinitives: onginnan/beginnan itself does not normally occur with an impersonal argument structure, i.e. with a dative experiencer; but when followed by an impersonal infinitive, this changes, and we find constructions like (2):

254

BETTELOU LOS

(2)

a ongan me langian for minre hæftnyde and ic ongan then began me grieve because of my imprisonment and I began gyrnan, þæt … 〈LS 35 (VitPatr) 330〉 yearn that … ‘Then I began to find my imprisonment tedious and I began to yearn that …’

See also Denison (1990: 148) and Warner (1992: 186, 194–195). Both note that onginnan + bare infinitive sides with the modal auxiliaries in this respect; they too show this ‘perviousness’ to the argument structure of the following infinitive. There are more such examples with onginnan/beginnan, but only with bare infinitives; it is the first piece of evidence which seems to suggest that onginnan/beginnan with the bare infinitive is an auxiliary, with no argument structure of its own. These two syntactic features appear to exemplify typical differences between bare and to-infinitival complements in general; there are, however, some syntactic patterns that point to a further difference between the two infinitival complements which is peculiar to onginnan/beginnan.

3.

Onginnan/beginnan in Ælfric: Syntactic patterns

In (3) we find another frequent pattern, with the matrix verb in second position in a main clause, after the adverb þa: (3)

þa ongunnon ða Iudei. hine eft torfian. mid heardum stanum. then began the Jews him again stone with hard stones. and heora an hine sloh mid ormætum stencge. and of-them one him struck with enormous cudgel inn oð þæt bragen 〈ÆCHom II, 18 172.100〉 inside the brain ‘then the Jews began to stone him again with hard stones and one of them struck him a blow with an enormous cudgel in the brain’

A second frequent pattern is afforded by example (4), where we find a to-infinitive with onginnan, with the matrix verb sentence-initial:

ONGINNAN/BEGINNAN WITH BARE AND TO-INFINITIVE

(4)

255

Ongann ða Augustinus mid his munecum to geefenlæcenne began then Augustine with his monks to emulate þære apostola lif mid singalum gebedum 〈ÆCHom II, 9 78.205〉 the apostles’ lives with continuous prayers ‘Augustinus with his monks then began to emulate the lives of the apostles with continuous prayers’

In (5) we find a to-infinitive with beginnan, also in initial position of a conjoined clause: (5)

þa cempan ða æt nextan oncneowan þurh ða nytena þæt the soldiers then at last realized through the animals that hi mid godcundre mihte gefæstnode wæron, and begunnon to they with divine power fastened were and began to axienne æt oþrum wegfarendum hwæt se man wære þe hi ask from other wayfarers what the man were whom they swa wælhreowlice beoton. 〈ÆLS (Martin) 990〉 so cruelly beat ‘the soldiers then finally realized because of the animals that they had been fastened to the ground by means of divine power, and began to ask other wayfarers what the man was whom they had beaten so mercilessly.’

This construction is in some ways ambiguous: because the subject of the conjoined clause is identical to that of the preceding clause it is often ellipted, and if it is ellipted it is difficult to tell whether we are dealing with a Verb-first construction or not. Similarly, in (6) we have an example of a Verb-first construction which, though not preceded by and, shows similar ellipsis of the subject: (6)

ongunnon ða to oftorfigenne mid heardum stanum þone began then to stone with hard stones the eadigan stephanum and he cleopode and cwæð: Drihten blessed Stephen and he called and said: Lord hælend. underfoh minne gast 〈ÆCHom I, 3 48.2〉 Saviour receive my spirit ‘then [they] began to stone the blessed Stephen to death with hard stones and he called out and said: Lord Saviour, receive my soul’

For the moment, I have subsumed the few examples of V1 in patterns as in (5) and (6) also under the heading Verb-first, although they are ambiguous in that

256

BETTELOU LOS

there is no overt subject (see Mitchell 1985: § 3934; Stockwell and Minkova 1990 and Koopman 1992: 67 for discussion). Comparing such V1 patterns with the þa V construction yields interesting results, which suggest that these patterns are not entirely random: Table 1. Respective numbers of constructions with þa V and V1 with onginnan and beginnan Verbs onginnan beginnan totals

þa V with bare infinitive

V1 with bare infinitive

þa V with to-infinitive

V1 with to-infinitive

17 6 23

1 1 2

3 9 12

4 5 9

It is clear from this table that the bare infinitive exhibits a clear preference for the þa V construction, whereas V1 constructions tend to occur with to-infinitives. The Fisher exact probability test, which is particularly useful when numbers are small (Siegel and Castellan 1988: 103), returns a significant level of probability (two-tailed, p < 1%). It seems, then, that the bare infinitive is dispreferred when onginnan/beginnan is clause-initial. In the remainder of this paper we will argue that the V1 construction marks a turning point in the narrative, a thematic discontinuity, an interruption; and that this function is only compatible with a to-infinitive, not with a bare infinitive, because only to-infinitives could signal ingression by Ælfric’s time, and only the event described by a to-infinitive could be interrupted. This aspectual difference between the two infinitives will be discussed in the next section.

4.

An aspectual difference between the two infinitives after onginnan/ beginnan

4.1 Semantic bleaching It has long been noticed that ginnan ‘begin’, the stem form of onginnan and beginnan, appears to show considerable semantic bleaching in Middle English, particularly in Chaucer, with the verb losing its original inchoative meaning and evolving into some sort of auxiliary; Einenkel (1891: 89) refers to the combination of ginnan + bare infinitive as “the most beloved periphrasis of Middle English”. Some work has been done to try to establish whether there was any

ONGINNAN/BEGINNAN WITH BARE AND TO-INFINITIVE

257

degree of bleaching of these verbs already in Old English. Most writers appear to rely on intuitive judgements (see Mitchell 1985: § 676 for some discussion), although Funke (1922: 1–27) and Mossé (1938: § 238) link the intuitive interpretation of some of the Old English instances as ‘bleached’ to the Aktionsart of the following infinitive. Although Funke’s main concern is Middle English, for Old English he observes that onginnan in its earlier, unbleached form is a perfective lexical verb, and its function is to focus the attention on the beginning of the action expressed by the infinitive (Funke 1922: 5); one of his examples is (7): (7)

Swa ða drihtguman dreamum lifdon, so these noble men joys lived eadiglice, oððæt an ongan blessedly until one began fyrene fremman feond on helle. 〈Beo 99–101〉 evil perform fiend in hell ‘Thus these noble men lived in joy blessedly, until a certain one began to do evil, a fiend from hell.’

He contrasts this with examples like (13) below, which he, too, interprets as purely “descriptive”, i.e. bleached (Funke 1922: 9). Mossé (1938: § 238), too, observes that onginnan in earlier Old English was used only with lexical verbs of durative meaning, i.e. only with infinitives that allowed an inchoative reading. Both Mossé and Funke conclude that there was already some degree of bleaching in Old English, which, as we will see later, is also borne out by our own findings below. Note that, in effect, Funke and Mossé offer a diagnostic test (durativity) which goes beyond intuitions; and it is this test which we find further elaborated in Brinton (1988), who investigates the issue in greater detail. Brinton relies on the work of Freed (1979: 25–40), who suggests that “aspectualizers” (e.g. inchoatives) take events as complements, not propositions or objects. An event can be segmented into different temporal stages: it has an “onset”, a “nucleus” and a “coda”. “Ingressive aspectualizers” refer to these temporal stages. For present-day American English Freed observes, for example, that begin refers to the initial segment of the nucleus of the event, while start refers to the onset, which explains why one can start to do something and then not do it, but not begin to do it and then not do it. Aspect and Aktionsart meaning interact: for ingressive aspectualizers to be compatible with their complements, these complements must refer to situations that are segmentable — that have an onset, or a nucleus. ‘Activities’ (like laugh or run) consist of a nucleus, while ‘accomplishments’ (like bake a cake) consist of onset, nucleus and

258

BETTELOU LOS

coda; ‘achievements’ (like get married or arrive at the party) and ‘states’ (like live), however, are not segmentable and should not be compatible with ingressive aspectualizers. Brinton (1988: 85) shows that ingressive aspectualizers can indeed occur with states but achievements remain awkward. Brinton makes the important observation that the incompatibility is due to the punctual nature of achievements; “because a punctual situation begins and ends at the same time, it is not possible to focus on the beginning point apart from the rest of the situation” (Brinton 1988: 85). If the argument of an achievement is an unspecified plural or mass noun, begin is possible, because it marks the beginning of a series; cf. in this respect (8) and (9) (Brinton’s 8f and 8h): (8)

*A friend began {to arrive/arriving} at the party

(9)

{Friends/People} began to arrive at the party

In other words, it must be possible to regard the situation as durative or iterative in some way. When Brinton comes to investigate the development of the aspectualizers in the history of English, she remarks that she does not regard the presence or absence of to as significant in any way (Brinton 1988: 119). She notes, however, that the ingressive meaning is weakened in many cases, especially with Middle English ginnen, so that especially “the past tense sequence of gan+infinitive appears to be no more than a periphrastic preterite” (Brinton 1988: 120). As a diagnostic of such a pleonastic reading Brinton proposes the following criterion: when Middle English ginnen (or, of course, Old English onginnan and beginnan) occurs with a non-repeatable punctual verb (i.e. an achievement verb without an unspecified plural or mass noun object) or with a durative or iterative adverbial, an ingressive reading is impossible, because there is no segmentable nucleus that onginnan or beginnan can refer to. She gives six examples from Old English which are identified by this criterion as ‘pleonastic’, i.e. non-ingressive, five of which are with a bare infinitive, given below as (10)–(13) (her 70a,b, 71a and 72a,b; Brinton 1988: 160): 〈El 302b–3a〉

(10)

ge to deaþe þone deman ongunnon you to death that-one judge began ‘you judged (*began to judge) that one to death’

(11)

ge þa sciran miht deman ongunnon 〈El 310b–11a〉 you then bright power judge began ‘you then judged (*began to judge) the bright power (of Christ)’

ONGINNAN/BEGINNAN WITH BARE AND TO-INFINITIVE

259

(12)

and ða

mid geættredum stræle ongann sceotan 〈LS 25 (MichaelMor) 45〉 and then with poisoned arrow began shoot ‘and then with a poisoned arrow [he] shot (*began to shoot)’

(13)

Elles ne ongunnon ræran on roderum, nymþe riht and soþ, else not began perform in heavens except right and truth ærþon engla weard for oferhydge … 〈GenA, B 20b-2b〉 until angels’ guardian for pride ‘[They] performed (*began to perform) in the heavens nothing else except right and truth until the guardian of angels out of sheer pride …’3

Her only example of this ‘bleached’ use of onginnan with a to-infinitive she gives as (14) (her 71b): (14)

Witodlice … ongann se hiredes ealdor to agyldenne þone pening 〈ÆCHom II, 5 46,137〉 ‘Certainly repaid (*began to repay) the elder of the house the penny’

Its ellipsis marks it as an example from Callaway (1913: 53). Here is the sentence in its entirety: (15)

Witodlice fram ðam endenextan ongann se hiredes ealdor truly from the last-ones began the household’s lord to agyldenne þone pening. to pay the penny ‘Truly, from the last ones began the lord of the household to pay the penny.’

This sentence appears in the context of a discussion of the Parable of the Vineyard (‘the last will be the first, and the first will be the last’). Even those workers who did not start work in the vineyard until the eleventh hour will get a full day’s wages at the end of the day — one penny. The workers have lined up to receive their wages, and their employer is in the process of giving each of them a penny, starting with the workers who came in last, ‘at the eleventh hour’; he is paying out a multiplicity of pennies, one to each worker. The object, though singular in form, implies an unspecified plural, so that the situation described in (15) is clearly iterative, even though the Aktionsart meaning is one of achievement. It is certainly not ‘pleonastic’. Brinton may well have missed an important generalization here: the bleached, ‘pleonastic’ instances of these ingressive aspectualizers only occur with the bare infinitive in Old English. When the to-infinitive is used, these verbs

260

BETTELOU LOS

always have their full ingressive meaning. Thus we find in Ælfric a few examples of the ‘pleonastic’ use of onginnan/beginnan + bare infinitive, e.g. with a durative adverbial: (16)

þa ongann se apostol hi ealle læran ofer twelf monað. then began the apostle them all teach for twelve months ða deopan lare be drihtnes tocyme. to ðyssere woruld 〈ÆCHom II, 18 170.27〉 the deep lore about lord’s coming to this world ‘Then the apostle taught (*began to teach) them all for twelve months the profound doctrine of the Lord’s coming to this world’

There are, however, no examples of such ‘pleonasms’ with a to-infinitive. A final point to note here is that Ælfric himself, in his Latin Grammar, systematically employs onginnan + to-infinitive to translate Latin inchoatives, and not the bare infinitive, e.g. calesco, ic onginne to wearmigenne ‘I begin to become warm’ (〈ÆGram 212.3〉; see also Mitchell 1985: § 675). This further supports the idea that only the to-infinitive expresses strong ingression. 4.2 The grammaticalization of onginnan/beginnan with the bare infinitive If strong ingression is restricted to onginnan/beginnan with the to-infinitive, what exactly is it that is expressed by the bare infinitive with these verbs? ‘Bleaching’, after all, suggests not only a loss of lexical meaning, but also a corresponding gain in the functional domain. A suggestion of what this functional meaning might be is to be found in Mustanoja (1960: 611), who claims that Old English onginnan, when followed by a (bare) infinitive, can be said to carry out two principal functions: (1) it brings out the ingressive and perfective aspects of the action represented by the infinitive, and (2) it intensifies the descriptive force of the infinitive. Mitchell (1985: § 676) notes quite rightly that it is not easy to see how the same verb can bring out both the ingressive and perfective aspects of an action; and as we have seen, ingression, at least in the works of Ælfric, is strongly associated with the to-infinitive, not with the bare infinitive. The suggestion of the combination onginnan + bare infinitive expressing perfective aspect is an interesting one, however, especially because there are occasional examples of this construction being used to translate Latin perfects. One such example is discussed in Riggert (1909: 46): persecuti sunt ‘[they] are persecuted’ shows up as ehtan ongunnon [lit. ‘began to persecute’] in Old English (〈PPs 118.161〉), which is why he notes: “Es wird nicht der Anfang, das Beginnen einer Handlung ausgedrückt, sondern vielmehr die Tätigkeit selbst oder deren Vollendung”

ONGINNAN/BEGINNAN WITH BARE AND TO-INFINITIVE

261

(‘it [onginnan with bare infinitive] does not express the beginning of the action but rather the event itself or its completion’) (Riggert 1909: 46). If onginnan/beginnan + bare infinitive tends to express a perfective, and onginnan/ beginnan + to-infinitive tends to express an ingressive aspect, it means that only the action in the latter is viewed as temporally segmentable. This ties in with the findings of Fischer (1995: 20), who notes that bare infinitives often signal an entailment relationship and that they share the same tense domain as their governing verbs, unlike to-infinitives (Fischer 1995: 19). One of Fischer’s Middle English examples of a to-infinitive is with ginnan ‘begin’: (17)

And to thise clerkes two he [Nero] gan to preye/ To sleen hym … and to these clerks two he began to pray to kill him (Chaucer, Monk’s Tale 2545–2546; Fischer 1995: 20) ‘And to these two clerks he [Nero] began to pray to kill him’

Fischer comments: “The ‘clerks’ can still refuse Nero an easy death and walk out on him” (Fischer 1995: 20), i.e. the event is viewed as interruptable. If a bare infinitive had been used here, this would not have been the case. It is this increasing grammaticalization of onginnan/beginnan + bare infinitive that ultimately explains the anomaly in the figures of Table 1. Bare infinitives after onginnan/beginnan no longer allow temporal segmentation in Ælfric, and its selection signals to the hearer that the action described in its complement should be viewed as being completed without interruption. In the next section we will see why this property of bare infinitives makes it particularly compatible with the þa V pattern, but not with the V1 pattern.

5.

Onginnan/beginnan in Ælfric: Discourse effects

5.1 Episode boundaries and action markers The role of þa as a discourse marker has been the subject of a number of studies; many of these favour the Old English chronicle as the source of their data, as þa is particularly frequent in this text. While many scholars agree that þa serves an important discourse function, the results are not always equivocal. Sometimes þa is argued to mark the boundaries between paragraphs, or more precisely, episodes, i.e. the levels of discourse intermediate between the sentence and the text. There appears to be a general consensus that the central characteristics of an episode include thematic unity, which depends on the continuity of both participant and action (see Brinton [1993: 73–74] for more discussion). The

262

BETTELOU LOS

boundaries of episodes are characterized by both semantic and formal features. Brinton (1993: 74) lists the following ‘discontinuities’ by which episode boundaries may be formally marked: a. b. c. d. e.

f. g.

a change in time; a change in location; a change in participants; a change in the action sequence, with an orientation toward a new central event, or the activation of a new schema; a change in ‘possible world’, for example, from the real world to the fictional world, from the real world to the dream world, or from the physical world to the world of thought; a change from general to specific, or the reverse; and a change in perspective or point of view.

Sometimes þa is described as a mere sequencer of events, with the role of episode boundary marker reserved for ond þa ‘and then’ (Turville-Petre 1974), although it has been argued that þa can also head episodes when not preceded by ond, e.g. by Enkvist (1986), who calls þa an “action marker”, and adds that it tends to occur in passages and statements describing action, “as opposed to background information” (Enkvist 1986: 301). It is worthwhile quoting Enkvist’s numbered ‘theses’ in which he recapitulates his comments on the function of þa: i. ii.

Old English þa is an action marker; As actions in narrative tend to be foregrounded, þa can also be regarded as a foregrounding device; iii. When þa collocates with stative verbs it may be taken to serve as a foregrounding “dramatizer” highlighting a dramatic view of stative conditions; iv. The main structural functions of þa are not only to mark distinctions between foregrounded action and background, but also to indicate the division of narrative discourse into narrative units; v. At the same time þa indicates the sequencing of events on the main story line; vi. We may expect þa to have been common in impromptu storytelling, where a clear marking of distinctions between foreground and background, a clear ordering and sequencing of events, and a conspicuous signalling of narrative units and their hierarchies are especially important to ease the processingload under real-time conditions. Therefore a high frequency of þa might be regarded as a marker of colloquial, lively, impromptu-speech-like narrative style (Enkvist 1986: 306–307).

ONGINNAN/BEGINNAN WITH BARE AND TO-INFINITIVE

263

It seems clear, then, that the þa V-pattern is pre-eminently associated with  in terms of Hopper (1979), in that it belongs to the language of the actual storyline, “the parts of the narrative which relate events belonging to the skeletal structure of the discourse” (Hopper 1979: 213) and not to , the language of supportive material which does not itself narrate the main events. Because foregrounded clauses denote the discrete, measured events of the narrative, the verbs tend to be punctual rather than durative or iterative, and there is a further tendency for these punctual verbs to have perfective aspect (Hopper 1979: 215). When we turn our attention to the other pattern relevant to our discussion of onginnan/beginnan, V1, we find that it, too, is associated with . It is often analysed in these studies as identical to þa V at a deeper level (e.g. Stockwell 1977, and also Hopper 1979), implying that it is actually the order VS that signals an episode boundary, with þa being a kind of optional extra. It has often been remarked that the V1 pattern in Germanic, or ‘narrative inversion’, has a special stylistic function, usually loosely described as typifying “lively narrative” (e.g. Kiparsky 1995: 163; Thráinsson 1985: 172) or “vividness of action” (Stockwell 1977: 291). There are hints that this “special stylistic function” can, at least for Old English, be described more accurately in discourse analysis terms. Mitchell reports, for example, an interesting suggestion by Fred C. Robinson that V1 in Old English often seems to mark a turning-point in the narrative, a transition, or a change of pace, “just as a new paragraph does in MnE prose” (Mitchell 1985: § 3933). These are of course precisely the functions of the episode boundary marker. As þa V, the first pattern, is also credited with this function, these comments do not make it easy to tease out the salient difference between the two patterns which might explain the contrast we found in infinitival complement with onginnan/beginnan in Ælfric. If þa V and V1 are both episode boundary markers, why would the former pattern favour a bare infinitive, and the latter a to-infinitive? I would like to suggest that both patterns are indeed foregrounding devices, but that as episode boundary markers they do not have the same status. The difference in the work of Ælfric is that onginnan/beginnan in V1 position invariably indicate a thematic discontinuity, i.e. mark episodes in which the narrative takes a dramatic turn. In contrast, þa V marks the smooth flow of narrative events and is primarily the “action marker” in terms of Enkvist (1986). If it marks episode boundaries at all in Ælfric, such episodes are not likely to introduce striking turns of event, but rather represent the smooth progression of the narrative.

264

BETTELOU LOS

5.2 Þa V as episode boundary marker: Thematic continuity Enkvist compares a fragment of an Ælfric narrative text with a more archetypal example of “colloquial, lively, impromptu-speech-like narrative style”, namely the Ohthere-Wulfstan interpolation in Alfred’s Orosius, and notes that the Ælfric text is “more artful and sophisticated, consisting of a structure describable on at least three levels, which we may call text, episode (marked with þa), and action (consisting of narrative statements within the þa-marked episode)” (Enkvist 1986: 306). Although Enkvist regards all four instances of þa in his discussion of the Ælfric text (〈ÆCHom II, 37, 272.1–19〉) as a device to divide the text into four narrative units, it is worth pointing out that three of these units are headed by the þa V pattern, whereas the third unit is not; and that this may not be accidental, as there is a clear functional difference between the þa in this third unit and the þa in the þa V constructions of the other three: it is this third unit in which the plot thickens and the pace of the story is stepped up (note that this change in pace is also formally marked by hwæt, a foregrounding dramatizer [Brinton 1990: 56]). When þa occurs in a þa V pattern, it is more than an “action marker” that marks the beginning of a new narrative unit. A clue to this special function can be found in Wilbur’s (1988) discussion of the functions of thô, the Old Saxon and Old High German cognate of þa (mentioned by Brinton 1990). He also observes a difference: when thô occurs in sentence-initial position, it signals the progression of the narrative; in other positions it has another function. We will argue below that the same situation obtains in Old English, although with a further refinement: it is not simply sentence-initial þa, but particularly sentence-initial þa V that signals this narrative progression; and we will further argue that þa V is systematically used to signal straightforward narrative progression without sudden turns in the plot, i.e. thematically continuous. The following passage from Ælfric illustrates this idea in more detail, and also shows how the idea of thematic continuity interacts with onginnan/beginnan + bare infinitive. For convenience, it is cut up into narrative units in (18)–(24). After a short exposition on the way God punished entire communities for their sins in the days before Moses, the narrative proper begins (the relevant þa’s in bold). The first narrative unit has God announce his plan to Moses. For a discussion of the correlative þa þa …, þa V pattern in this unit see Enkvist (1986: 305):

ONGINNAN/BEGINNAN WITH BARE AND TO-INFINITIVE

(18)

265

Eft ða þa God wolde wrecan mid fyre þa fulan after then when God wanted destroy with fire the foul forligeras þæs fracodostan mennisces, Sodomitiscra ðeoda, fornicators of-that wickedest nation Sodomitic people þa sæde he hit Abrahame. 〈ÆLS (Pr Moses) 190〉 then said he it Abraham. ‘Afterwards when God wanted to destroy with fire the foul fornicators of that most wicked nation, the people of Sodom, then he announced it to Abraham.’

The second unit contains Moses’ response. He pleads with God to save the town if as few as fifty righteous men can still be found. The pace of the narrative quickens, formally marked by a þa which is not part of a þa V pattern: (19)

Habraham þa bæd þone ælmihtigan ðus, þu Drihten, þe Abraham then asked the Almighty thus Thou Lord who demst eallum deadlicum flæsce, ne scealt ðu þone rihtwisan judges all mortal flesh not shalt Thou the righteous ofslean mid þam arleasan. Gif ðær beoð fiftig wera wunigende destroy with the wicked If there are fifty men living on þam earde, rihtwise ætforan ðe, ara him eallum. on the Earth righteous before Thee spare them all 〈ÆLS (Pr Moses) 193–196〉 ‘Abraham then asked the Almighty thus, thou Lord, who judges all mortal flesh, you should not destroy the righteous together with the wicked. If there were fifty men living in the land, righteous before you, spare them all.’

God grants his plea in the third unit. Note again that þa V again expresses a thematic continuity, as it introduces the second member of an ‘adjacency pair’, i.e. an answer to a question: (20)

a cwæð God him to eft, Ic arige him eallum gif ic ðær finde then said God him to again, I spare them all if I there find fiftig rihtwisra. fifty righteous 〈ÆLS (Pr Moses) 198〉 ‘Then God said to him, I will spare them all if I find there fifty righteous ones.’

Abraham makes his second plea in the fourth unit. Will God save the towns if he finds as few as forty righteous men? Again þa V, and no thematic discontinuity,

266

BETTELOU LOS

because (21) is a repeat of (19), only with the number of 50 reduced by ten. Note that began is here followed by the bare infinitive biddan ‘ask’. Although this verb indicates an activity, an Aktionsart meaning that in theory allows temporal segmentation, the selection of the bare infinitive, as opposed to a to-infinitive, signals to the hearer that the event of ‘asking’ will be completed without interruption: (21)

Þa began Abraham eft biddan God georne, þæt he hi ne then began Abraham again ask God eagerly that he them not fordyde gif ðær feowertig wæron rihtwisra wera wunigende destroy if there forty were righteous men living on ðære leode 〈ÆLS (Pr Moses) 200〉 in that nation ‘Then Abraham began again to ask God eagerly, that he should not destroy them if there were forty righteous men living in that nation.’

Unit five: his plea is again granted: (22)

God him ðæs tiþode, … God him that granted ‘God granted him that, …’

〈ÆLS (Pr Moses) 203〉

Unit six sets up a little ‘loop’ telescoping several rounds of pleading on Moses’ part and granting on God’s part. Note that the phrase he began git biddan refers to a durative, not a punctual situation: ‘he kept on asking until …’. The use of the bare infinitive signals that the rounds of pleading will not be interrupted by other events, and that the hearer should infer that there are three more rounds of pleading, with the respective numbers of just men first reduced to thirty, then twenty, and finally ten: (23)

and he began git biddan oðþæt he becom to tyn mannum, and him tiðode God ða, þæt he nolde hi fordon gif he funde ðær tyn rihtwisra manna, … 〈ÆLS (Pr Moses) 203–206〉 ‘and he began to ask further until he came to ten men, and God granted him then, that he would not destroy them if he found there ten righteous men, …’

Unit seven concludes the episode: (24)

and he wende ða him fram ‘and he then turned away from him.’

〈ÆLS (Pr Moses) 207〉

ONGINNAN/BEGINNAN WITH BARE AND TO-INFINITIVE

267

The compatibility of the þa V pattern and onginnan/beginnan with the bare infinitive is due to the fact that the use of the bare infinitive after these verbs signals that the event in question will proceed without interruption, as onginnan/ beginnan here does not focus on the onset, or even beginning of the nucleus, of the action, while the þa V pattern signals a straightforward progression of the narrative without sudden changes. 5.3 V1 as an episode boundary marker: Thematic discontinuity The question remains why the bare infinitive is dispreferred when onginnan/ beginnan occurs clause-initially. We saw above that the literature on V1 constructions suggests that V1, too, functions as an episode boundary marker. What makes it so different from the þa V pattern is that V1 also signals a thematic discontinuity, i.e. a sudden change in events, or a turning point. Every single occurrence of the V1-pattern with onginnan/beginnan + to-infinitive in the work of Ælfric can be argued to signal an episode boundary of this kind. Examples are for instance 〈ÆCHom II, 38 282.89〉 in which the action switches to other participants, or 〈ÆLS (Martin) 165〉 in which the victim finally manages to turn the tables on his oppressors; similar turnings in the plot occur in 〈ÆLS (Martin) 990〉 (= (5) above) and 〈ÆLS (Chrysanthus) 170〉. A change in perspective or point of view occurs in 〈ÆLS (Oswald) 45〉, where the line of the original narrative is taken up again after a diversion; and a change in the ‘possible world’, e.g. from the real to the fictional or vice versa occurs in 〈ÆCHom I, 3 46.35–48.5〉, where clause-initial onginnan/beginnan + to-infinitive signals a lengthy interruption of the narrative by an authorial exposition. The latter instance is particularly instructive because it can be compared to a passage from 〈ÆCHom II, 18〉 which describes basically the same plot. Both passages can be divided into three narrative units : (i) the crowd decide to stone a saint (St. Stephen in 〈ÆCHom I, 3〉, St. James in 〈ÆCHom II, 18〉); (ii) the stoning is interrupted by the saint asking God to forgive the sins of the people stoning him; (iii) the stoning is resumed and brought to a close without a further hitch. The second unit (example (25a–b) below) is particularly illustrative of the way in which the ingressive aspect of onginnan/beginnan + to-infinitive and V1 word order interact, as it is the one that contains a thematic discontinuity — the interruption of the stoningevent. In the James narrative (25a) this unit is introduced by þa; note that this þa is not part of a þa V construction, as such a construction would signal continuity. Because the narrative employs the open-ended torfian ‘to stone,’ ‘to throw stones at’, which by itself does not imply ‘to the death’, the interruption of the stoning event itself need not be strongly marked, only by ac ‘but’. In the Stephen narrative

268

BETTELOU LOS

in (25b), however, the ‘resultative’ or ‘perfective’ sense of oftorfian ‘stone to death’ in contrast to the open-ended torfian ‘stone’ forces the selection of a to-infinitive after ongunnon (in bold): a bare infinitive would have created the impression that Stephen was killed off at this point, and he cleopode in the next clause would then have come as a bit of a surprise to the audience. The interruption of the action is anticipated by the to-infinitive. The discontinuity is more strongly marked than in the James fragment, as it is introduced by the V1 pattern; one reason might be that the interruption of the narrative is a long one, and does not interrupt only the stoning itself, but also the narrative: the saint’s last words are followed by an authorial exposition explaining the crucial importance of these last words, as they led to Saul’s conversion (cf. ‘change in possible worlds’, Brinton’s [g] in Section 5.1 above). (25)

a.

b.

Hi ða upastigon. and hine underbæc scufon. and mid they then ascended and him from-behind pushed and with stanum torfodon þone soðfæstan Iacob; Ac he næs acweald stones stoned the righteous James but he not-was killed acweald ðurh ðam healican fylle. ac gebigde his cneowu killed by the high drop but bent his knees on gebedum sona. and bæd þone ælmihtigan for ðam prayers at-once and asked the Almighty for the arleasum cwellerum. þæt he him forgeafe þa fyrnlican synne cruel assasins that he them forgave the earlier sins 〈ÆCHom II, 18 172.95–7〉 ‘They then ascended and pushed him from behind and stoned the righteous James with stones; But he was not killed by the high drop, but bent his knees in prayer at once and prayed to the Almighty for the cruel assassins that he would forgive them the earlier sins’ ongunnon ða to oftorfigenne mid heardum stanum þone began then to stone-to-death with hard stones the eadigan stephanum and he cleopode and cwæð: (…) min blessed Stephen and he called-out and said (…) my drihten: ne sete ðu þas dæda him to synne … lord not set thou those deeds them as sin 〈ÆCHom I, 3 48.2–5〉 ‘[they] began then to stone to death with hard stones the blessed Stephen and he called out and said: (…) my Lord, do not lay those deeds to their charge as a sin …’

ONGINNAN/BEGINNAN WITH BARE AND TO-INFINITIVE

269

In the third unit the stoning is resumed. The transition to this unit is marked in the Stephen narrative (26b) by þa, which indicates a new narrative unit; the James narrative in (26a), however, uses onginnan this time. As there is no second interruption, a bare infinitive is now possible (onginnan + infinitive in bold): (26)

a.

b.

þa ongunnon ða Iudei. hine eft torfian. mid heardum then began the Jews him again stone with hard stanum. and heora an hine sloh mid ormætum stencge. stones and of-them one him struck with enormous cudgel inn oð þæt bragen; þus wearð gemartirod se mæra apostol inside the brain thus was martyred the famous apostle 〈ÆCHom II, 18 172.100〉 ‘then the Jews began to stone him again with hard stones and one of them struck him a blow with an enormous cudgel in the brain; thus was martyred the famous apostle …’ and he mid þam worde þa gewat to þam ælmihtigum and he with that word then departed to the Almighty ælende þe he on heofenum healicne standende geseah; Saviour whom he in heavens high standing saw 〈ÆCHom I, 3 48.5〉 ‘and with those words he then departed to the Almighty Saviour whom he had seen standing high in the heavens’

To summarize, then, in both cases the stoning is interrupted by the martyr’s last words. The Stephen fragment of (25b) uses onginnan+to-infinitive to describe the first abortive attempt to stone St Stephen; this event is then interrupted by Stephen’s last words. In contrast, the James fragment of (26a) uses onginnan+bare infinitive to indicate that the procedure is resumed and brought to a close without a further hitch.

6.

The fate of inchoative markers

The change in onginnan appears to be an instance of grammaticalization, first described as “the attribution of a grammatical character to a once autonomous word” (Meillet [1912] 1948: 131). Kuryłowicz’s description of the process further elaborates on the meaning of the term ‘grammatical’: “Grammaticalization consists in the increase of the range of a morpheme advancing from a lexical to a grammatical or from a less grammatical to a more grammatical status, e.g. from a derivative formant to an inflectional one” (cited in Heine, Claudi and

270

BETTELOU LOS

Hünnemeyer 1992: 3). This idea was developed further by Haspelmath, who suggests that some grammaticalization processes appear to move along a derivational/inflectional cline (Haspelmath 1996). Derivational suffixes typically impose restrictions on the type of Head they occur with. The West-Germanic action-noun suffix -ung (Old English, Modern German) or -ing (Modern Dutch), for instance, cannot attach to just any verb; as it has an inherently perfective meaning, it can only attach to verbs that denote achievements and processes (Hoekstra and van der Putten 1988: 175); the infinitival suffix -en in these languages, however, may attach to any verb. The concepts of derivation and inflection as representing opposite poles of a grammaticalization cline could also be argued to describe the development of onginnan. In its original inchoative meaning onginnan cannot be paired with just any verb, only with verbs that carry the right Aktionsart meaning (durative or iterative, or in Vendler’s classification, activities or achievements). The bleached, perfective variant of onginnan that only occurs with a bare infinitive, on the other hand, shows affinities with the inflectional process: it is a true auxiliary in that it may occur with any verb. There are abundant precedents for ingressive markers being grammaticalized to auxiliaries or inflectional morphemes. Andrew Allen (1993) observes in his discussion of the history of the PIE *-sk- suffix that the derivational suffix -sc-, which signalled inchoativity in Latin, subsequently lost its ingressive meaning and developed into an inflectional suffix in some of its daughter languages. In Italian, for example, -isc- has become part of an irregular conjugation which uses the suffix in forms like finisco ‘I finish’, which has an unstressed personal ending, but not in the first person plural finiamo ‘we finish’ or the second person plural finito ‘you finish’, which have stressed personal endings (Allen 1993: 5). An even more relevant example here is Allen’s story of the verb cepi ‘I began’, from Classical Latin coepi. Allen cites a study by Bechtel (1902), who provides evidence from a fifth-century manuscript, the Peregrinatio Egeriae, that this verb lost its lexical ingressive meaning in this particular variety of Latin and became an auxiliary of the perfect, a development absolutely parallel to what happened to Old English onginnan + bare infinitive. A similar development took place in the Old French verb commencier ‘begin’, which in some thirteenth-century Old French translations from Latin clearly serves as a durative past tense auxiliary (Allen 1993: 5, drawing on Beer 1974: 43–48), as simple past tense forms in the Latin source text emerge in the translation as a form of commencier plus the infinitive. Why aspectualizers in particular are prone to grammaticalization is a difficult question. In the case of onginnan/beginnan it might be possible to

ONGINNAN/BEGINNAN WITH BARE AND TO-INFINITIVE

271

connect the process to the fact that these verbs are particularly frequent in narrative texts, and it is conceivable that they acquired a particular discourse function of their own. Studies of the rise and fall of various discourse markers suggest that they are transient, ephemeral entities, particularly susceptible to change (e.g Stein 1985: 300). Any element used as a device to arouse the interest of an audience by creating suspense in some way is prone to rapid devaluation through overuse; and this may also have been the case with onginnan/beginnan + bare infinitive. How far had this process progressed in Ælfric’s time? It is striking that onginnan in clause-initial position, i.e. in a V1 configuration, is robustly attested with the bare infinitive in Old English poetry (48 times) but only once in Ælfric, who tends to use the to-infinitive with V1 configurations instead. This makes it likely that the to-infinitive, which originally represented a non-finite alternant of the subjunctive that-clause, had stepped in by Æfric’s time to restore the ingressive meaning of onginnan/beginnan, with the bare infinitive after these verbs reserved for signalling that the action it describes progresses to the coda of the event.4

7.

Conclusion

This paper argues that the Old English verbs onginnan and beginnan, both originally meaning ‘begin’, show auxiliary behaviour in the works of Ælfric when governing a bare infinitive: – –

like the modals, they can either follow or precede the matrix verb; like the modals, they are transparent to the argument structure of e.g. impersonal verbs, as if they had no argument structure of their own.

We further noted that they rarely occur in a V1 construction when governing a bare infinitive, and that this is due to the fact that the events referred to by a bare infinitive are not temporally segmentable and are therefore never interrupted. Onginnan and beginnan as auxiliaries show semantic bleaching in that they have lost their original inchoative meaning in the bare infinitive construction. When onginnan and beginnan are followed by a to-infinitive, they express ingression, and refer to a temporally segmentable, interruptable event; this explains their compatibility with the V1 word order, which signals a thematic discontinuity. We conclude that bare and to-infinitives after these verbs are not in free variation in Ælfric’s prose, but carefully selected.

272

BETTELOU LOS

Acknowledgments The research reported in this article was made possible by a grant from the Netherlands Organization for Scientific Research (NWO), grant number 300–74–021. The author gratefully acknowledges helpful hints and comments from Olga Fischer, Ans van Kemenade, Anthony Kroch and Evert Wattel, and would also like to thank the editors of this volume and an anonymous referee.

Notes 1.

Throughout this article, the reference to an Old English text is enclosed in 〈 〉 and follows the system of short titles as employed in Healey and Venezky (1985) (in turn based on the system of Mitchell, Ball and Cameron 1975, 1979).

2.

Note that these figures are higher than those given above. The reason is that fewer texts are attributed to Ælfric today than in Callaway’s time.

3.

This example also occurs in Funke (1922: 9). He, too, interprets this case as an instance of bleaching. He translates: “Fürwahr nichts vollführten sie im Himmel außer gute und gerechte Taten, bis …” (‘Indeed, they did not perform anything in Heaven but good and just deeds, until …’).

4.

For the idea that to-infinitives and subjunctive that-clauses are rival structures, see Fischer (1996), Los (1998), Los (1999).

References Allen, A. 1993. “Regrammaticalization and degrammaticalization of the inchoative suffix”. In Historical Linguistics 1991, H. Andersen (ed.), 1–8. Amsterdam/ Philadelphia: John Benjamins. Allen, C. L. 1992. “Old English and the syntactician: Some remarks and a syntactician’s guide to editions of the works of Ælfric”. In Evidence for Old English [Edinburgh Studies in the English language 2], F. Colman (ed.), 1–19. Edinburgh: John Donald. Bechtel, E. A. 1902. Sanctae Silviae Peregrinatio. Studies in Classical Philology, vol. VI. Chicago: University of Chicago Press. Beer, J. 1974. “Traces of a durative use of comencier + infinitive in 13th-century French”. Romance Philology 28: 43–48. Bock, H. 1931. “Studien zum präpositionalen Infinitiv und Akkusativ mit dem to-Infinitiv”. Anglia 55: 115–249. Brinton, L. 1988. The Development of English Aspectual Systems: Aspectualizers and Postverbal Particles. Cambridge: Cambridge University Press. Brinton, L. 1990. “The stylistic function of Middle English gan reconsidered”. In Papers from the 5th International Conference on English Historical Linguistics [Current Issues in Linguistic Theory 65], S. Adamson, V. Law, N. Vincent and S. Wright (eds), 31–53. Amsterdam/Philadelphia: John Benjamins.

ONGINNAN/BEGINNAN WITH BARE AND TO-INFINITIVE

273

Brinton, L. 1993. “Episode boundary markers in Old English discourse”. In Historical Linguistics 1989: Papers from the 9th Conference on Historical Linguistics [Current Issues in Linguistic Theory 106], H. Aertsen and R. J. Jeffers (eds), 73–89. Amsterdam/Philadelphia: John Benjamins. Callaway, M. 1913. The Infinitive in Anglo-Saxon. Washington DC: Carnegie Institution of Washington. Denison, D. 1990. “Auxiliary + impersonal in Old English”. Folia Linguistica Historica IX (1): 139–166. Einenkel, E. 1891. “Der Infinitiv im Mittelenglischen”. Anglia 13: 79–104. Enkvist, N. E. 1986. “More about the textual functions of the Old English adverbial Þa”. In Linguistics across Historical and Geographical Boundaries: In honour of Jacek Fisiak on the Occasion of his Fiftieth Birthday, vol. 1, D. Kastovsky and A. Szwedek (eds), 301–309. Berlin: Mouton de Gruyter. Fischer, O. C. M. 1995. “The distinction between to and bare infinitival complements in late Middle English”. Diachronica XII (1): 1–30. Fischer, O. C. M. 1996. “Verbal complementation in early Middle English: How do the infinitives fit in?” In English Historical Linguistics 1994, D. Britton (ed.), 247–270. Amsterdam/ Philadelphia: John Benjamins. Freed, A. F. 1979. The Semantics of English Aspectual Complementation. Dordrecht/ London: D. Reidel. Funke, O. 1922. “Die Fügung Ginnen mit dem Infinitiv im Mittelenglischen”. Englische Studien 56: 1–27. Haspelmath, M. 1996. “Word-class-changing inflection and morphological theory”. In Yearbook of Morphology 1995, G. Booij and J. van Marle (eds), 43–66. Dordrecht: Kluwer. Healey, A. D. and Venezky, R. L. 1985. A Microfiche Concordance to Old English. Toronto: The Pontifical Institute of Mediaeval Studies [1980]. Heine, B., Claudi, U. and Hünnemeyer, F. (eds). 1991. Grammaticalization: A Conceptual Framework. Chicago: University of Chicago Press. Hoekstra, T. and van der Putten, F. 1988. “Inheritance phenomena”. In Morphology and Modularity, M. Everaert, A. Evers, R. Huybregts and M. Trommelen (eds), 163–186. Dordrecht: Foris Hopper, P. J. 1979. “Aspect and foregrounding in discourse”. In Syntax and Semantics 12: Discourse and Syntax, T. Givón (ed.), 213–241. New York: Academic Press. Kiparsky, P. 1995. “The Indo-European origin of Germanic syntax”. In Clause Structure and Language Change, A. Battye and I. Roberts (eds), 140–169. Oxford: Oxford University Press. Koopman, W. 1990. Word Order in Old English. Ph. D. dissertation, University of Amsterdam. Koopman, W. 1992. “Old English clitic pronouns. Some remarks”. In Evidence for Old English [Edinburgh Studies in the English language 2], F. Colman (ed.), 44–87. Edinburgh: John Donald.

274

BETTELOU LOS

Los, B. 1998. “The rise of the to-infinitive as verb complement”. English Language and Linguistics 2: 1–36. Los, B. 1999. Infinitival Complementation in Old and Middle English [LOT dissertations Series 31]. The Hague: Thesus. Meillet, A. 1948. “L’évolution des formes grammaticales”. In Linguistique Historique et Linguistique Générale, A. Meillet, 130–148. Paris: Champion [1912]. Mitchell, B. 1985. Old English Syntax, 2 vols. Oxford: Clarendon Press. Mitchell, B., Ball, C. and Cameron, A. (eds). 1975. “Short titles of Old English texts”. Anglo-Saxon England 4: 207–221. Mitchell, B., Ball, C. and Cameron, A. (eds). 1979. “Addenda and corrigenda”. AngloSaxon England 8: 331–333. Mossé, F. 1938. Histoire de la Forme Périphrastique ‘Être + Participe Présent’ en Germanique, vol I, Paris: Klincksieck. Mustanoja, T. F. 1960. A Middle English Syntax, part I. Helsinki: Société Néophilologique. Pope, J. C. (ed.). 1967. Ælfric’s Homilies: A Supplementary Collection, 2 vols. EETS 259, 260. London: Oxford University Press. Riggert, G. 1909. Der syntaktische Gebrauch des Infinitivs in der altenglischen Poesie. Inaugural Dissertation der Königlichen Christian-Albrechts-Universität, Kiel. Siegel, S. and Castellan, N. J. 1988. Nonparametric Statistics for the Behavioral Sciences, second edition. New York: McGraw-Hill. Stein, D. 1985. “Discourse markers in Early Modern English”. In Papers from the 4th International Conference on English Historical Linguistics [Current Issues in Linguistic Theory 41], R. Eaton, O. Fischer, W. Koopman and F. van der Leek (eds), 283–302. Amsterdam/Philadelphia: John Benjamins. Stockwell, R. P. 1977. “Motivations for exbraciation in Old English”. In Mechanisms of Syntactic Change, C. N. Li (ed.), 291–314. London/Austin: University of Texas Press. Stockwell, R. P. and Minkova, D. 1990. “Verb phrase conjunction in Old English”. In Historical Linguistics 1987: Papers from the 8th International Conference on Historical Linguistics, H. Andersen (ed.), 499–515. Amsterdam/Philadelphia: John Benjamins. Thráinsson, H. 1985. “V1, V2 and V3 in Icelandic”. In Verb Second Phenomena in Germanic Languages, H. Haider and M. Prinzhorn (eds), 169–194. Dordrecht: Foris. Turville-Petre, J. 1974. “The Narrative Style in Old English”. In Iceland and the Mediaeval World: Studies in Honour of Ian Maxwell, G. Turville-Petre and J. S. Martin (eds), 116–125. Victoria, Australia: Wilke & Company. Warner, A. R. 1992. “Elliptical and impersonal constructions: Evidence for auxiliaries in Old English”. In Evidence for Old English [Edinburgh Studies in the English Language 2], F. Colman (ed.), 178–210. Edinburgh: John Donald. Wilbur, T. H. 1988. “Sentence connectives in ancient Germanic texts”. In Germania: Comparative Studies in the Old Germanic Languages and Literatures, D. G. Calder and T. C. Christy (eds), 85–95. Wolfeboro, New Hampshire: Brewer.

Some suggestions for explaining the origin and development of the definite article in English Robert McColl Millar University of Aberdeen

1.

Introduction

One of the most central items in the functional inventory of the English noun phrase is the definite article the. So central is it, in fact, that it is impossible for most native speakers of the language to imagine what it would be like for there not to be what is (for us) an important — indeed essential — aid to our understanding of the world. Yet there are languages spoken in Europe today which have no such functional category. Some of these — such as Finnish — are not related to English; others, such as the majority of the Slavonic languages are, if at something of a remove. When a Russian speaker is learning English, he or she is faced with a whole functional grouping which to him or her makes no sense — indeed is highly superfluous. Yet that does not mean that Russian has no semantic category of definiteness; merely that it does not feel the need to create a single functional niche to express these ideas. Underlying this dichotomy is the fact that all the Indo-European languages which have a definite article (including English) have developed this during their history. In that sense Russian is actually closer to the Indo-European archetype in having no such overt distinction. In this essay I will suggest reasons behind the development of the definite article in English, compare these to developments in other languages, and finally suggest a rather heretical conclusion: that English has an overt, largely single-function discrete form the not so much because the language felt an overwhelming need for such a form, but rather because a gap had opened in the semantic fabric of the language due to the specialisation in meaning of that. The is, in other words, an historical accident.

276 2.

ROBERT MCCOLL MILLAR

The overt expression of definiteness: Some views on its development

As was said in the above, all languages must have an inherent sense of definiteness. Without this concept, it is practically impossible to imagine a language functioning. There is, of course, some variation over what types of definiteness are necessary for a given language, but that does not disprove the general point. Why therefore should some languages develop the overt expression of this distinction, whilst others do not? In his seminal study of the English definite article, Christophersen suggests (1939: 20) that languages develop definite articles because they shift along the typological continuum from synthetic to analytic. In other words, the morphosyntactic material which supplied these distinctions without the provision of an article is rendered obsolete and redundant, thus bringing about the development of an overt expression of deixis. We do see evidence of this kind of development. If we look at the Romance languages today, all of them have a discrete definer of some sort or other (whether this be a preposed particle or a postposed enclitic particle). Latin, whilst occasionally employing its demonstrative pronouns in situations where its daughters would use such a definer, had no such discrete form. What splits the two stages in the language’s history is a move from a synthetic structure towards an analytic one in all the dialects.1 The same must surely, it would be argued, be the case for English. Despite its being attractive at first glance, there are a number of problems with accepting such a position at surface level, however. If we look at languages such as Greek, or the Scandinavian dialects, we can see that they developed dedicated definers before they shifted along the synthetic to analytic continuum. By the same token, there are languages such as Farsi which have also lost all but the smallest vestiges of grammatical gender and case, but have not developed a definite article. At the same time, there are languages such as Modern High German which appear to be stalled somewhere along the line between being typologically synthetic and being analytic and also having no dedicated definer and having one (of which more in the following). The situation cannot be as clear-cut as Christophersen appears to think. More promising, perhaps, are those views put forward by scholars interested in the various processes of grammaticalisation. In his 1997 study, Nikolaus P. Himmelmann puts forward a scheme for the cline of development of a definite article, basing his description upon the earlier work of Greenberg (1978)2 and that of Lehmann (1982):

THE ORIGIN AND DEVELOPMENT OF THE DEFINITE ARTICLE

277

  +   →   →   →     →   →   →   (Himmelmann 1997: 23). According to this view, a definite article develops as part of a greater process of re-designation of purpose of morphological components within a system. As a brief aside, we can see that whilst English has reached either the “weakly demonstrative definite determiner” or perhaps even the “definite article” stage in the pattern, the Scandinavian “article” is actually closer to either the “affixal article” or “noun marker” stage. Native speakers of Danish, for instance, (and, indeed, native grammars) seem to consider the form manden ‘the man’ to be the ‘definite form’ of the noun (although they have no problem applying this rule on definite nature to their use of the English article). German, on the other hand, is at most at the “demonstrative determiner” stage. But whilst much of this appears highly reasonable in its discussion of the process of article formation, it does not really explain why the process should have been initiated in the first place.

3.

Previous suggestions for the development of the definite article in English

In most standard historical grammars and histories of the English language the development of the definite article is glossed over as if it were an inevitable development — which, as I have already suggested, is how it would appear to a native speaker of Modern English. As we have seen, such a view is not tenable. At most, many scholars see the development as being due to the replacement of the original demonstrative system by the use of ‘uninflected forms’ such as the, that and this. As early as 1882, Koch suggested that the development could be seen as part of a process where (1882: 475), “[n]eben den flexionen kömmt das abgeschliffene þe vor”. The logical leaps that are required by such a view will be dealt with below. Some other scholars — particularly (although not exclusively) those from a German-speaking background — have seen the birth and growth of the overt expression of definiteness as being a long drawn-out process involving a number of developments.

278

ROBERT MCCOLL MILLAR

3.1 The adjective paradigms According to this view, both the Germanic dialects and the contiguous Baltic and Slavonic dialects exhibit (or historically have exhibited) evidence of a split between those adjectives which are preceded by a determiner of some form and those which stand on their own without a preceding determiner (Kramsky 1972: 179–183). In the modern Germanic dialects which have maintained this distinction the former declension is referred to as the weak adjectives, the latter as strong adjectives. Some scholars (most notably Heinrichs [1954: 52 ff.]) have contended that underlying this distinction is an early form of overt determiner. It is their suggestion that the weak adjectives originally had some elements of deictic Kraft, whilst the strong adjectives were originally lacking in such force. Over time, this overt relationship with a definer function was lessened by phonetic attrition – the ‘weakening’ of syllables at the end of the word due to the forces of, among other things, elision. Nevertheless, the old relationship is still present in the way in which the modern Germanic determiners are attracted towards the weak adjectives. It may be the case that this relationship was originally developed to support the teetering weak adjective as determiner system as described above, in order to make certain that there was deixis in the expression. If that is the case, then as a saving mechanism, an attempt to maintain the central point of the distinction, it was bound to fail. This, along with a wide variety of other attempts at partial retention, might be described as ‘conservative radicalism’. As with other examples of this type, it was bound to fail because the sacrifice of one part of a system inevitably leads to excess pressure upon another and the gradual shift of a grammatical feature from the centre of a system to its periphery (Vachek 1980: 373). 3.2 The simple demonstrative pronoun Many European languages have based their definite article on a part of their distal demonstrative. There are exceptions to this. The Scandinavian dialects, whilst having a distal ‘simple’ demonstrative pronoun, use what was originally part of their third person pronoun, thus (1)

Modern Danish huset ‘the house’, det hus ‘that house’.

As a general rule, however, we can see that there is a semantic connection between the and that. Indeed in English (as in other languages) the two forms derive from the same source. From around the sixth century on, all of the West Germanic languages

THE ORIGIN AND DEVELOPMENT OF THE DEFINITE ARTICLE

279

began to use their distal ‘simple’ demonstrative in functional and semantic slots which nowadays we would associate with the definite article. This does not mean to say that these languages developed a definite article; only that the functional/semantic zone was rendered active (or perhaps re-activated after the ambiguisation of the original weak adjective determiner). We can see the prevalence of this usage spreading in all the West Germanic dialects. For instance, in the case of English, a poetic and probably early text such as Beowulf has considerably fewer examples of the new article use of the distal demonstrative than do later prose texts such as the works of Ælfric (Christophersen 1939: 86; Micillo 1982). Many scholars have been tempted to see in these developments the birth of the definite article. They point to article function and then to demonstrative function with forms from the same paradigm and believe that in this functional variety lies the seeds of the formal split. There is a certain circularity of argument in this; it is also illogical to suggest that because there are semantic distinctions, these must cause fissures within a paradigm. Bruce Mitchell has described (1985: I, 133) the search for a definite article in ‘classical’ Old English as a “terminological will o ‘the wisp’ ”. Quirk and Wrenn (1955: 69) said: It must be emphasised that until the very close of the Old English period se (rarely þe until very late) was simply an inflectional variant of þæt, in complementary distribution with it as Mod.E. the is with that. The existence of a definite article in Old English is a vexed question, but it seems to be one which has been raised largely by our desire to impose upon OE a terminology familiar in and suitable for Mod.E.. Where today we have three contrastive and formally distinct defining words, the, that, this, each with a name, in Old English there was but two, se and þes, and we are left, as it were, with a name to spare. The problem partly disappears when we reflect that in many instances of their use today, the and that are interchangeable … .

This problem is not helped by the fact that German-speaking scholars, whose system is essentially still the one described here (with minor modifications), often see the very real semantic distinction between their ‘article’ and einfach Demonstrativpronomen as being a formal distinction, when this is not the case. Old English, and for that matter Modern High German,3 does not have an article in the way that Modern English (or French, or Danish) has. Instead they have demonstrative pronouns which exhibit near-article function. All that we can really say is that in the development of a definer function for the distal, simple demonstrative, the semantic seeds for the later formal split were laid.

280

ROBERT MCCOLL MILLAR

3.3 Purdy (1973) Where, therefore did this distinction come from? There have been a number of attempts in recent years to explain this process; as we will see, however, not all of what was claimed has been entirely correct or successfully argued. In his 1973 article, “Did Old English have a definite article?”, D.W. Purdy argued that whilst Old English did not have an article per se, it had developed a vacuum in its pronominal and determiner systems which cried out to be filled by what eventually became our definite article. He claims correctly that there is considerable correspondence between the pronoun and determiner categories in Old English (Purdy 1973: 121). He points out further that this correspondence breaks down when it comes to the relationship between se (the determiner) and he/heo/hit (the pronoun), claiming that there are a number of occasions when the use of one does not necessarily imply the use of the other. In general these exceptions can be seen as being largely the preserve of se: (a) its use as a relative pronoun; (b) its direct contrast to þes and the remainder of the compound demonstrative pronoun paradigm, and (c) as a determiner (Purdy 1973: 123). It is this third distinction upon which he concentrates, and it is on this occasion that his argument falls down. His argument is as follows. He points out (quite correctly) that there is a lacuna in Modern English in terms of correspondence between determiners and pronouns, since whilst it is entirely possible and correct to represent essentially the same meaning with either That is a man or He is a man and That is a woman and She is a woman, there is no exact correspondent to That is a tree (*It is a tree is not possible in these contexts). Here his argument becomes highly questionable, since he claims (Purdy 1973: 123) that precisely the same distinction “may have existed in OE, but we have no way of telling since it involves the whole speech act”. The rest of his argument is based upon this highly shaky foundation. Purdy (1973: 123) continues the tentative argument by saying: We can say then that it, when used as the subject of a sentence accompanying a pointing gesture, is unacceptable [in Modern English]. This could also have been the case in OE, since hit was early chosen to refer to abstract nonpointable-at-things like the time and the weather, and it was also used as the dummy subject after extraposition had taken place.

This is somewhat spurious “logic”. Nevertheless he “tentatively” extrapolates a “fourth rule” from this of a semantic position where se and he could not both occur (Purdy 1973: 123): “as neuter subject of sentences involving a pointing

THE ORIGIN AND DEVELOPMENT OF THE DEFINITE ARTICLE

281

gesture — only þæt could occur”. This “split” between the two paradigms in semantic terms, the “vacuum” that this developed, and the subsequent development of a “definite article” to fill the slot (partially, although he does not discuss this) were rendered inevitable by (Purdy 1973: 123) “the upheaval and instability which followed the Norman Conquest”. This article is centred on a circular logic also noticed by Mitchell who states that (Mitchell 1985: I, 133, note 87) “[h]is suggestion that OE ‘nevertheless had a semantic/grammatical void which was very susceptible to being filled by a definite article’ is no doubt supported by the fact that English did develop the definite article”. We can see this circular argumentation most markedly in Purdy’s back-reference of usage patterns from the present-day to that of nearly a thousand years before as if they were truly equivalent, and as if we could say that because the usage described is a grammatical necessity now, this must have been the case at an earlier stage in the language’s development. This stability is in marked contrast to the flux which he describes for other parts of the paradigms at the same time. The article also makes a number of errors in the structure of its analysis. Firstly, it appears to accept Ross’ ‘neutralisation’ hypothesis (Ross 1936) for the development of an independent that (see 5 below). Whilst Purdy is correct in assuming that there was a correspondence between the use of hit and inanimate objects (even when of a different surface grammatical gender), it is wrong to extrapolate — immediately — from this that the semantic specialisation of that has come about as a result of this development. Further, along with many other scholars who deal with English alone, Purdy has under-estimated (or, indeed, ignored) what has happened in other languages. It is very easy to point out parallels between the semantic developments discussed here and those put forward for other Germanic languages such as German. German has exactly the same ‘vacuum’ in the relationship between determiners and pronouns; nevertheless its speakers have never felt the need to develop a system like that of English. On the other hand the Scandinavian languages have developed a usage very similar in semantic import to that found in English. There is no evidence that any tension along the lines of those postulated by Purdy was ever felt there. Whilst it is true that separate languages can (and do) have different means of development, developments in cognate languages should not be ignored in this way. The many mis-statements and misinterpretations found within this article are particularly galling because Purdy may well be on the edge of understanding something fundamental to what is discussed here: there was a fundamental flaw in the structure of the demonstrative system of Old English as it entered the

282

ROBERT MCCOLL MILLAR

‘transition period’, one which was exploited by other forces (both intra- and extra-systemic) to create a new system from the old. 3.4 Spamer 1979 The most important attempt in recent years to explain the development of the definite article is to be found in Spamer’s 1979 article, “The development of the definite article in English; a case study of syntactic change”. He suggests in this that the final fissure of the definite article from the rest of the simple demonstrative paradigm derives from the breakdown in the Middle English period of the distinction between strong and weak adjectives. He bases his ideas primarily upon the contention that this distinction is not, as is usually considered, one merely between different morphological classes of adjectives, but rather represents the remnants of a distinction between two word-classes: adjuncts (as part of a compound with the head noun) and adjectives. He begins his analysis by pointing out that there is a “striking difference” between the Modern English and Old English noun phrase. In Modern English there are two potentially recursive elements: adjectives, for example (2)

the happy man

(Spamer 1979: 242)

and adjuncts, for example (3)

stone wall

(Spamer 1979: 242).

“But the Old English has only one recursive element, the adjuncts. In Old English, the adjectives were non-recursive: each head noun could have at most only one adjective preceding it” (Spamer 1979: 243), thus producing what to speakers of present-day English seem to be rather long-winded turns of phrase, for example: (4)

þæt hi næfre ær swa clæne gold, ne swa read ne gesawon ‘that they had never before seen such pure gold, nor such red [gold]’ (Spamer 1979: 244).

He further points out that Old English syntax is in conflict with that of Modern English when it comes to the manner in which a series of adjectives can be laid out. In Modern English it is perfectly possible to have adjectives following each other, for example (5)

They had never before seen such pure red gold.

THE ORIGIN AND DEVELOPMENT OF THE DEFINITE ARTICLE

283

In Old English each adjective “almost invariably” must be followed by and (1979: 244). Spamer says that (1979: 246): … the traditional classification of adjectival endings as “strong” and “weak” in Germanic philology is actually based on a misconception. There are not two inflectional systems for adjectives at all. There is simply the adjectival declension, the traditional “strong” one, and there is the adjunct declension, the traditional “weak” one. The adjuncts would originally have been combinatives, and the surface structure of an adjective with a strong ending followed by adjectives with weak endings would have derived from what was diachronically a single adjective followed by a compound.

From this Spamer extrapolates that there were really only two classes of words in Old English which could precede a head noun in a noun phrase: modifiers (on this occasion, adjectives) and adjuncts. As we have seen, he classifies what are traditionally termed ‘weak’ adjectives as adjuncts. By the same token he classifies the ‘strong’ adjectives and the demonstrative pronoun (among other modifiers) as being highly similar if not identical in function (1979: 246): “The demonstrative and the adjective function in the same way in the noun phrase: they take essentially the same endings (in contrast to the adjuncts), they occupy the same initial position, and the use of one precludes the use of the other”. From this he postulates that there is an underlying phrasal structure along the lines of (6)

[modifier] + [adjunct] + [head noun]

inherent, although not necessarily realised, in all Old English noun phrases. Bearing this in mind, it would appear logical that only one modifier can be used in each phrase, whether such a modifier is a strong adjective or a demonstrative pronoun. This is a particularly ingenious solution because it overturns the traditional view of what lies behind the use of strong or weak adjectival forms: that they are triggered by the use or non-use of a determiner with a noun phrase (Campbell 1959: § 638). Continuing this line of logic, it becomes apparent that if what he terms the ‘adjective’ declension (in other words, the strong declension) and the determiner declensions are both modifiers, then the demonstrative pronouns can also be seen as — at least in morphological and functional terms — adjectives (1979: 247). From this, according to Spamer, we can extrapolate a reason behind the split in the Old English se paradigm, the ‘simple demonstrative’. As long as the strong/weak distinction was maintained with the adjective, there was no reason for the essentially adjectival nature of the demonstrative to be emphasised and

284

ROBERT MCCOLL MILLAR

rendered critical. However, according to Spamer, as the morphological breakdown of word classes continued apace in the Middle English period, a distinction between the adjunct class (the ‘weak’ adjectival declension) and the original adjective class (the ‘strong’ adjectival declension) was no longer possible (Spamer 1979: 248: “[a]s a result, adjectives in English become recursive”). As Spamer himself admits, there is a problem with this reasoning: “if the demonstrative itself is essentially a specialised adjective, and if adjectives become recursive, there is nothing to block the surface order represented by *old the man” (Spamer 1979: 248). To get round this problem Spamer suggests that there has been a fundamental change in the deep structure interpretation of the noun phrase and that (Spamer 1979: 248), “Middle English speakers … would have realised that the had to occur initially and could not be repeated in the same noun phrase, [and] they would have naturally concluded that the belonged to a form class different from that which included adjectives. In a word, they would have concluded that the was an article”. 3.5 Some objections to Spamer’s ideas It is not possible to accept Spamer’s findings as the sole explanation for the development of a definite article in English. The first, and most important, problem with Spamer’s explanation is his final point: that speakers would have reinterpreted the as an article. It is difficult to see why the processes that he is discussing should have brought this about. Nowhere in his article is an attempt made to explain the split between the and that, the latter originally also deriving from the simple demonstrative paradigm. Why would speakers at the time have made a distinction between these two forms when they did not make one between the various forms of the compound demonstrative paradigm, equally under pressure from ‘grammatical breakdown’? There must be other reasons behind the morphological developments discussed. Moreover, the connection postulated between the complete breakdown of the strong and weak adjective systems and the development of a definite article is not as straightforward as Spamer appears to believe. Whilst it would be wrong to claim that there is no connection between the two, this connection cannot have been intrinsic. Michael Samuels has shown that the London English of Chaucer’s time had retained — for at least some, more conservative, speakers — some remnants of the old distinctions between strong and weak adjectives (Samuels 1989a; see also Elliott 1974: 55) Yet the tripartite system of the, that and this was in full use in just these dialects in a manner very close — if not entirely identical — to that found in English today (Kerkhof 1982: § 662).

THE ORIGIN AND DEVELOPMENT OF THE DEFINITE ARTICLE

285

Spamer has also not thought to compare his findings with those for other languages. When we do so, it becomes apparent that while his analysis may be applied to some, such as Dutch, there are others, such as the North Germanic dialects, which had a fully developed article system — albeit from a different source — at a time when they also had a highly complex synthetic morphology. There cannot be the one-to-one correspondence he suggests between grammatical ‘simplification’ and the growth of an article, even if we have the suspicion that just such ‘simplification’ does encourage the development of an article system in a wide range of languages (Christophersen 1939: 20). Despite Spamer’s article’s importance in terms of our understanding of the development, we must therefore say that it only adds some pieces to the puzzle. What it might suggest is an explanation for the final stabilisation of the new tripartite system due to his postulated deep structure reinterpretation of one part of the original ‘adjective’ declension. But the reason for the split between the and that is no clearer after reading his article than it was before.

4.

Developments in the demonstrative paradigms

Up to this point mention of the demonstrative pronoun paradigms and the directions in which they developed during the period under discussion has been avoided. As we might expect, both simple demonstrative pronouns (the semantic ancestors of Modern English that) and compound demonstrative pronouns went through the same morphological levelling which all the inherited paradigms went through.4 Thus, while in Old English, there was a near one-to-one correspondence between a single form and a case/gender slot, Modern English has replaced this with a relationship based largely upon a single (or highly circumscribed) form/all function relationship. It would be convenient to see this as the replacement of many forms by one ‘radical’ form. Indeed such a view is embraced by a great many scholars, albeit not always in a systematic way.5 But when we look at the developments involved in this we can see that there are certain general tendencies at play over a relatively lengthy historical period which have caused these developments to come about. They might all be grouped together as phonological attacks upon the system. 4.1 Ambiguity in ending One of the most central of these is what might be termed ‘ambiguity in ending’. A number of the forms in both inherited demonstrative paradigms carried their

286

ROBERT MCCOLL MILLAR

functional information embedded in the ending. Thus the simple demonstrative form þære (dative feminine singular) is only distinguished from the form þæm (dative masculine/neuter singular and dative plural) by their respective endings. As ambiguity grew over what certain endings represented, and these very endings were attacked by the forces of ‘phonetic attrition’ (both processes part of the developments central to the ‘transition period’ between Old and Middle English), it is inevitable that problems would be encountered in maintaining both the external formal and the underlying functional distinctiveness. This would particularly be the case where it was only one vowel that separated the two (or more) forms. For instance, the form þæm and the form þone (accusative masculine) inhabit similar functional areas — particularly in prepositional position where semantic distinctions are sometimes not great between the cases — and are differentiated largely by the -e ending (m and n are given to variation in most languages, and there is considerable variation in both root vowels throughout the Old English period). When endings are under attack as they were at the end of the Old English period, it is not surprising that the two forms should fall together at þVn (Blake 1994: § 6.3). Many scholars (Körner 1888: 58; Seidler 1901: 23–25, 32; Landwehr 1911: 10–11; Pervaz 1958: 127–128; d’Ardenne 1961: § 92, among others) have suggested that this took place first in hiatus position. Indeed we do find examples of this type of phenomena in texts from the period: (7)

a.

b.

Vices and Virtues (late twelfth century) 13/30–1 Šewiss hafð godd forworpen þan ilche man ‘Indeed God has cast out that very man’ Owl and the Nightingale (mid-thirteenth century) 741–2 C an bidde þat hi moten iseche / þan ilche sang þat euer is eche ‘and bid that he must seek out that very song which ever is eternal’.

These are, however, often outweighed by examples of this type of variation where the form in question is followed by a consonant: (8)

a.

b.

Peterborough Chronicle First Continuation 1123/41  he sæde þone kyng þet hit wæs togeanes riht ‘and he told the king that it was against right’ LaŠamon (later thirteenth century) 1359 O Þohte Gorgwind þane king ‘it seemed to the king Gorgwind’.

THE ORIGIN AND DEVELOPMENT OF THE DEFINITE ARTICLE

287

How can we reconcile these two apparently divergent tendencies? In her recent book, Donka Minkova has argued that hiatus is not in itself a sufficient reason for a development of this type (1991: 164): “Avoidance of hiatus” is a very unenlightening characterization of the causation of early schwa loss. There is no evidence in English, at any point in its history, of any phonotactic constraint on hiatus, word internally or across word boundaries. Nor is there any universal, or even a tendency, for languages to disallow hiatus; in fact languages with and without hiatus are about evenly distributed.

It is difficult to see how we can accept everything that she says here at face value. In fact, present day non-rhotic dialects of English do evince a considerable degree of hiatus avoidance in their use of /r/ before vowels across word boundaries in phrases such as (9)

law and order.6

Having said that, there can be little doubt that what she says here is grounded in the truth. There is no immediate connection in most examples between the loss of -e and hiatus. It might be argued that the hiatus phenomenon development was expressed first in speech, and that it had already spread into other non-hiatus contexts by the time it entered the considerably more conservative written form, a form which, as we know, was profoundly influenced for a considerable period of the early Middle English period, by Anglo-Saxon models (Millar and Nicholls 1997).7 Whilst there is every likelihood that this is the case, it might be argued that beyond this is a general tendency to treat final vowel endings as being less meaningful than would previously have been the case. For instance, in the late twelfth century MS BL Cotton Vespasian A. xxii text of Ælfric’s De Initio Creaturae,8 there is much evidence of ‘misuse’ or non-use of endings, such as (10)

buton elce eorðlice federe ‘without any earthly father’

where the archetype has (11)

buton ælcum eorðlicum fæder.9

This would particularly be the case with the use of pronouns, which are not always stressed within a phrase. Minkova suggests (1991: 128–129) that “their [pronouns’] position in the sentence would not allow them to occupy a prominent position in the hierarchy of prosodic salience; these pronouns will be mere prosodic clitics when they appear within the noun phrase”.

288

ROBERT MCCOLL MILLAR

Such a position would inevitably lead in these circumstances to a situation where, as Minkova notes (1991: 162), “the W terminal node [will be] dominated by a higher W node …, that in itself can be a sufficient condition for schwa deletion”. In fact what appears to have taken place is a staggered, staged development, bearing in mind the proviso that, as we have seen, there may be more than one stage present in a given text (if not necessarily in a speaker’s idiolect) at the same time. Thus a speaker might not regularly have used a given ending or form in speech, but understood (to some extent at least) what the functional implications of its use might be when using it in the written form. By the same token this understanding might only be fragmentary (particularly if the person’s knowledge of the usage is largely a passive, reading knowledge). That this ambiguity is highly prevalent can be seen in the way in which variation can be found within texts that have more than one manuscript witness, for example: (12)

Katerine (late twelfth century) B 433–4 he bichearde þene feont ant schrencte þen alde deouel R 547–8 he bi-cherde þene feont. and schrencte þen alde deouel T 801–2 he bicherde þene feond.  schrencte þen alde deouel ‘he deceived the fiend and tricked the old devil’.

The consistency, if we can talk in such terms, of the confusion should also be noted on this occasion. This is not an example of a single idiolect’s idiosyncrasy. On the contrary, it demonstrates how widespread the variation must have been. Indeed, in her edition of Seinte Iuliene, d’Ardenne went so far as to suggest (d’Ardenne 1961: § 92) that þene, a relatively rare form in texts of the AB grouping, should probably be regarded as a longer variant of þen, and not as especially accusative in nature. Pervaz echoes these sentiments, suggesting that in the South English Legendary the -ne form (Pervaz 1958: 109) “is probably the longer variant of a generalized dative form …”.10 In his Grammatical Gender in English: 950–1250, Charles Jones suggested an alternative scenario for these developments. He suggests that (in between the complete application of the ‘classical’ West Saxon case/gender system, and the ‘morphology-free’ system found in most dialects from the early Middle English period on) there was a brief period — probably only a few generations per dialect, although he does not define this — where the original case and gender sensitive materials were employed for a similar, but not identical, functional purpose. He suggests that the -ne ending associated with accusative case contexts in ‘classical’ Old English became associated in texts such as the Peterborough

THE ORIGIN AND DEVELOPMENT OF THE DEFINITE ARTICLE

289

Chronicle First Continuation and Interpolations and the Caligula text in particular of LaŠamon’s Brut with two functions not identified with the inherited surface case structure, but rather with functions as defined in John Anderson’s On Case Grammar (1977). In essence these two contexts can be defined as  ‘case’ and in  contexts where the phrase in question has “intrinsically oriented directional properties”. The former, as found in examples such as, (13)

Peterborough Chronicle First Continuation 1125/8–9 þa benam man an  an  benæm ælc þone riht hand ‘and took each one by one and removed their right hand’,

is a means by which we might explain non-historical gender-association. According to Jones (1988: § 4.5), an example such as the þVne form “is being produced to reflect the ‘neutral’,  case relationship of those arguments which they modify bear to the sentential relationship”. There is little which can be said against such a proposition save that, as we will see in the concluding portion of this section, these new usages of þVne are often in variation with þV forms which represent exactly the same contexts. In many of the later texts discussed in this essay, þV forms have almost entirely replaced þVne. The sub-system Jones postulates was too fragile to survive. The latter is particularly useful in explaining the apparent confusion between ‘accusative’ and ‘dative’ forms discussed above: ‘accusative’ forms are moving into ‘dative’ functional zones in a non-haphazard manner. When speaking of the First Continuation and Interpolations of the Peterborough Chronicle (1988: § 4.6.c), Jones (1988: § 4.6.c) discusses a number of contexts where a þVne form is realised, but where ‘classical’ Old English would regularly have demanded the use of a þVm shape. Interestingly, most of these fall within a similar semantic field, for example: (14)

a.

b.

Peterborough First Continuation 1126/3–5 Mid him com se cwen  his dohter þet he æror hæfde giuen þone kasere Heanri of Loherenge to wife ‘with him came the queen and his daughter whom he had earlier given to the Emperor Henry of Lorraine as wife’ Peterborough Chronicle First Continuation 1129/25–26 Se kyng Henri geaf þone biscoprice æfter Michelesmesse þone abbot Henri his nefe ‘The king Henry gave the bishopric after Michaelmas to the abbot Henry his nephew’

290

ROBERT MCCOLL MILLAR

In a complex argument, Jones claims that whilst the examples above demonstrate a surface indirect object function, at a deeper level their  nature is apparent. Essentially, his argument runs that when there is no “left movement” with an indirect object, such as (15)

John gave Mary the book,

when the indirect object remains in “far right position”, for instance, (16)

John gave the book to Mary,

the indirect object takes on some of the semantic (and, it would appear on this occasion, morphological) features of the direct object. He says (Jones 1988: 153) that: [t]he semantic hallmark of the absolutive case relationship is “completion”, “exhaustiveness” and that which is inherent in the semantic reference of the sentential predicate … In the types of sentence in the Peterborough Chronicle … — “The king gave the abbot the property” — we should like to claim that not only do verbs like /give/ display an [absolutive] “direct object” lexical specification (in this case “the property”) but they also manifest inherent directionality: the act of giving assumes a locational goal (as well as a source, as we shall see below). But, more importantly for our argument here, that directionality itself has an   ; a goal-only directional path. Predications like “come” and “go”, although likewise having intrinsically directional status are nevertheless unspecified as to their source/goal route. It is just that exhaustiveness of directional orientation which, we suggest, supplies the [locational. Ergative] “indirect object” with an absolutive relational property in /give/ predications. And that particular property is signalled in these early Middle English data by the preposition-less 〈þone〉 attributive in noun phrases.11

In his discussion of LaŠamon’s usage (Jones 1988: § 5.2.b), he widens his argument somewhat. He gives examples of  phrases, such as (17)

LaŠamon (later thirteenth century) C Saturnus heo Šiuen Sætterdæi þene sunne heo Šiuen Sonedæi ‘they give Saturday to Saturn; they give Sunday to the Sun’.12

He suggests (1988: 182) that, “Perhaps under this type too we might include ‘indirect objects’ arguments in construction with predicates like ‘promise’, citing examples such as, (18)

LaŠamon (later thirteenth century) O Hii bi-hehte Goffare þane king þat hii him wolde helpe ‘They promised Goffar the king that they would help him’

THE ORIGIN AND DEVELOPMENT OF THE DEFINITE ARTICLE

291

working, no doubt, on the idea that we could paraphrase the above as ‘They gave a promise to Goffar the king that they would help him’, thus implying that we have here again a ‘locational goal only’ context. There might, in fact, be something in his argument; it certainly appears to be suited to the examples that he gives. One problem that he does not address, however, is the manner in which a native speaker would have analysed the two þVne forms found in an example such as, (19)

Peterborough Chronicle First Continuation 129/25–26 Se kyng Henri geaf þone biscoprice æfter Michelesmesse þone abbot Henri his nefe ‘The king Henry gave the bishopric after Michaelmas to the abbot Henry his nephew’.

The first example of þVne would, according to Jones’ argument given above, demonstrate the spread of a no longer gender sensitive form as a marker of absolutive function. Yet it evidently serves a different function here from the second, nearly contiguous example of þVne, at least at surface level. Even if we accept that both function/form identifications could occur at the same time, that does not say much for the long-term health of the sub-system. In effect, what we are seeing is merely a further variation on the theme of formal and functional ambiguity discussed in this essay. Jones goes on to discuss the use of þVne with prepositions in both textual groupings mentioned. In the First Continuation and Interpolations of the Peterborough Chronicle, he cites examples such as, (20)

Peterborough Chronicle First Continuation 1128/17–18 Þes ilces geares com fram Ierusalem Hugo of þe Temple to þone kyng on Normandig. ‘In this same year Hugo of the Temple came from Jerusalem to the king in Normandy’

and in LaŠamon’s Brut he cites examples such as, (21)

LaŠamon (later thirteenth century) C 164 heo hine flemden out of þane londe ‘they chased him out of the land’

as the basis for an argument for the use of þVne being conditioned by the presence of a preposition expressing movement or ‘directional motivation’. In comparing the previous  phenomenon with this prepositional phenomenon, he says (Jones 1988: 155):

292

ROBERT MCCOLL MILLAR While with the former we suggested that the ungoverned 〈þone〉 attribute was selected as a signal for the exhaustive, completive nature of ergative locatives with an absolutive characteristic, in the latter it is its “neutral”, unspecified participant function as a marker of “case” only which is manifested — specific locational information being provided by the particle serially preceding it.

It is difficult to know precisely what he means with ‘case’ here, since it is precisely this feature — whether taken as the semantic function indicators he is foregrounding in the book or the grammatical surface case — which he claims is being negated. In fact, what we have here is a reformulation of the differences between what van Kemenade (1987: § 3.1.1.2.3) called ‘oblique case’ (and Hoffmann (1909: 92–93) and Pervaz (1958: 93) ‘prepositional case’) and ‘structural’ case. Nevertheless, it is significant that it should be the þVne form that is preferred in these contexts without high functional sensitivity,13 rather than þVm and its descendants. This would suggest that þVne is developing new functions which overstep former gender and case prerequisites. But whilst many of these ideas have a degree of logic to them, they cannot tell the whole story. The problem underlying the stability of the sub-system he posits is variation. At the same time we can have, as we have seen, usage which is morphologically very close to (if not identical with) the ‘classical’ Old English position, usage which appears to represent the system he posits, and usage which is much closer to the Modern English position (in the sense that þe is preferred over all historical forms). The situation is much ‘greyer’, less clear-cut, than Jones might like us to think. Indeed he himself has to admit that when dealing with the — admittedly morphologically eccentric — LaŠamon’s Brut. When dealing with material such as, (22)

LaŠamon (later thirteenth century) C 711 heo wolden … þan kinge wið-stonden ‘they wanted to withstand the king’

Jones does concede (1988: § 5.5) that þan is being used in these contexts with  ‘case’ functions. He does verge on the idea that this demonstrates the breakdown — at least to some extent — of the distinction, but claims instead that the frequency levels for the variation in usage are too low for such a distibution to be random. He (1988: 198) concludes: One very tentative solution might be to argue for the existence in the LaŠamon data of two separate attributive outputs. One of these, and the one with which we are now familiar, would generalize the 〈-ne〉 accretion shape from an original function of marking “completive” absolutive arguments to those other environments where, in the presence of alternative signals for locational

THE ORIGIN AND DEVELOPMENT OF THE DEFINITE ARTICLE

293

direction [in other words, largely in prepositional position], “case empty” attributive forms were seen to be appropriate. The other would take as its “model” that determiner shape found increasingly intruding to the right of directionally specific locational particles with nouns of all historical genders — 〈þan〉. That “case neutral” or “case empty” form could then be generalized to non-locative absolutive slots.

Jones (1988: 199) admits himself that “it is difficult to constrain such a tentative proposal”, although he does suggest some not entirely tenable distinctions between the two usages, claiming that (1988: 199) “the ‘case neutral’ 〈þan〉 type has not intruded into those arguments whose directional characteristics are inherent in the semantics of the predicates with which they are in construction”.14 In fact, I would claim that what he is describing refers to the fact that, in general, the þVn shape has retained a connection with functions associated with dative case in Old English, the þVne shape with ‘accusative case’ in Old English, but that the -e ending, suffering from ‘phonetic attrition’ has led to considerable — and growing — ambiguity and an inevitable merger (or loss) of forms. Therefore, even if we take on board all of Jones’ provisos, we must, I would claim, assume that there was an ongoing blending between the þVne and þVn shapes. Another ambiguity is also present in these circumstances: that between these two highly ambiguated forms and þV. Earlier scholars have seen this development as representing the introduction of the new undeclined the form into the existing paradigms (or perhaps even as an alternative to it). But whilst in a very broad historical sense this is no doubt correct, it does seem as likely, if not more so, that in fact the þV form is a natural, organic result of the variation we have already seen. We can probably see this in the variation to be found in the multimanuscript texts already mentioned: (23)

a.

b.

Katerine (late twelfth century) R 522  adun warp þene wiðerwinne of helle B 437 ant adun weorp þe wiðerwinne of helle T 807–8  adun weorp ðe wiðer-wine of helle ‘and throw down the hellish adversary’ LaŠamon (later thirteenth century) 6576 C  þat child Šef þan abbede on hond : twenti sulhene lond. O and þe child Šef þe abbot on hond :: twente solŠene lond ‘and the child gave into the abbot’s hands twenty plough lands’.

294

ROBERT MCCOLL MILLAR

It would seem from this that writers (and scribes) at the time did not always make a complete distinction between these forms as representatives of a given functional zone alone. By this I mean that whilst some remnants of the historical patterns are obviously represented in the day-to-day language of these writers and scribes (in the sense that they are largely able — albeit with declining accuracy — to represent something approaching the historical system), the ‘walls’ between the different functions are by no means as stable as they once were. For instance, in one of the latest copies of an Ælfric text (beginning of the thirteenth century), De Initio Creaturae from BL Cotton Vespasian A.xxii, there appears to have been some transmission of a tradition of ‘correct’ ‘classical’ Anglo-Saxon endings, but these are on occasion completely swapped around. Thus whilst an early witness (CUL Gg.3.28) reads, (24)

seo eorðe þe is awyrige on þinum weorce ‘the earth which is cursed by your work’,

Vespasian reads, (25)

se eorðe his awirigd on þine weorcum

(Millar and Nicholls 1997: 443). It might also imply that there are other than purely linguistic reasons for the use or non-use of a ‘traditional’ or a ‘modern’ form. By this I mean that in a society where the literate and their culture are very much in a minority, and where this culture has an extreme respect for auctoritas, there is no guarantee that the form evinced on the page is necessarily the majority form in the scribe’s idiolect (Stanley 1969 and 1988). That these distinctions were eventually rendered essentially meaningless can be seen in the use of what appear to be the original declined forms in the period after all case and gender marking seems to have been jettisoned, for example: (26)

Sir Orfeo (mid-fourteenth century) 53–4 Þe fairest leuedi for þe nones / Þat miŠt gan on bodi & bones ‘the fairest lady for the nonce who could walk around in body and bones’.

Similar processes can be seen at work with the compound demonstrative pronouns, bearing in mind Rennhard’s caveat cited in the above. In Old English the form for dative singular masculine/neuter (and dative plural) contexts was þissum; the form used in accusative singular masculine contexts was þisne. By the same processes as those described in the preceding, these two forms began to fall together at þisse (particularly since for a large part of the late Old English

THE ORIGIN AND DEVELOPMENT OF THE DEFINITE ARTICLE

295

and early Middle English periods the final nasal of the dative form was represented only by /n/), (27)

Vices and Virtues (early thirteenth century) 55/31 of ðese liue ‘of this life’.

As with the above, the final -e ending appears to have been particularly prone to ‘phonetic attrition’. This means that we often find forms with the basic structure þ + vowel + s in those contexts originally associated with the declined forms already described, for example, (28)

LaŠamon (later thirteenth century) C of þis ærd O of þis erþ ‘from this country’.

5693

Variation between these forms can also be seen within the same manuscript version, for instance, (29)

a.

b.

c.

d.

Iuliane (late twelfth century) R 517 of þisse reade B 656 of þis reade LaŠamon (later thirteenth century) 7939 C vnder þissen stane O vnder þis stone ‘under this stone’ LaŠamon (later thirteenth century) 3098 C bi þisse sæ-rime O bi þis see-rime ‘by this seashore’ LaŠamon (later thirteenth century) 2670 C of þisse lond O of þis lond ‘from this country’.

Many scholars have seen this usage as representing the introduction of the undeclined, radical this form; the root vowel variation militates against this, however, even if, again, over a broad enough historical perspective they could be seen as being correct. Again I would suggest that underlying this variation is an organic, system-internal development.

296

ROBERT MCCOLL MILLAR

4.2 Ambiguity in form The corollary to the above is ‘ambiguity in form’. A number of the function specific forms of both paradigms were differentiated in Old English only by alternation in the root vowel. This became the case most particularly after the sV forms of the nominative masculine and feminine singular simple demonstrative — se and seo — were replaced in a number of dialects by þV forms by the forces of analogy — þe and þeo. Thus all that differentiated nominative and accusative functional contexts with nouns of the feminine gender class with both paradigms was the root vowel (þeo vs. þa, þeos vs. þas). The same minimal variation was to be found in nominative contexts between masculine and feminine forms in the case of the simple demonstrative pronouns (þe and þeo) and all forms in the case of the compound demonstrative (þes vs. þeos vs. þis). Given that there are a number of occasions when demonstrative pronouns are not stressed in a clause (Minkova 1991: 128–129) (particularly when, as we have seen, the same forms are being used in near-article functions), it is inevitable that there would be some confusion between root vowels, often being pronounced as schwa. This was exacerbated in the early Middle English period by a number of sound changes which rendered the ambiguity even more critical. This confusion can be seen in the way in which simple demonstrative pronoun forms are employed in contexts with which they would not have been associated in ‘classical’ Old English, (30)

a.

b.

c.

d.

Peterborough Chronicle 1st Continuation 1131/10 þa scyrte ða flescmete ‘then the meat was in short supply’ Peterborough Chronicle 2nd Continuation 1137/54 þe erthe ne bar non corn ‘the earth did not bear any corn’ LaŠamon (later thirteenth century) 149 C þa brude dead iwarð ‘the bride was dead’ Katerine (late twelfth century) B 260 tu schalt sone etsterten al þe R 323 tu schalt sone etsterten al þe strencðe T 493–4 tu schalt sone atstirten al þe strengðe ‘you must immediately escape from all the strength’

THE ORIGIN AND DEVELOPMENT OF THE DEFINITE ARTICLE

e.

f.

g.

h.

297

Vices and Virtues (late twelfth century) 91/18 Þis is Šewiss þe holiŠe mihte ‘this is indeed the holy power’ Vices and Virtues (late twelfth century) 139/8 þe fule wombe is crewlinde full of weormes ‘the foul womb/belly is crawling full of worms’ South English Legendary (late thirteenth century) Bridget 34 for al þe worlde ssel ioyuol beo ‘for all the world must be joyful’ South English Legendary (late thirteenth century) Patrick 533 for as þe sonne bynymeþ þe liŠt ‘for as the sun takes the light’.

Again, the nature of the variation is made plainer if we consider the variation found between different manuscript copies of the same text, for example (31)

a.

b.

c.

LaŠamon (later thirteenth century) 139 C þeo wiman hefde on wombe O þe womman bere ‘the woman had in her womb/ the woman bore’ LaŠamon (later thirteenth century) 219 C muchel wes þa neode O mochel was þe neode ‘great was the need’ Katerine (late twelfth century) R 275 þis meiden wes bicluset þeo hwile B 222 þeos meiden wes bicluset þe hwile T 421–2 þes meiden was bicluset þe hwile ‘this maiden was enclosed for the period’.

Nevertheless we find occasions in very similar contexts where there is no variation, the sole form remaining apparently being þe: (32)

a.

b.

LaŠamon (later thirteenth century) 630 C hou þe læfdi him sæide O wat þe leafdi saide ‘how the lady said/ what the lady said to him’ Iuliene (late twelfth century) R 402–5 hire nebscheft schininde al as schene as þe sunne B 514–6 hire nebscheft schene as þe sunne ‘her face (shining all as) bright as the sun’

298

ROBERT MCCOLL MILLAR

c.

Katerine (late twelfth century) B 347 þes is al þe lare R 436 þes is al þe lare T 645 þis is al þe lare ‘this is all the knowledge’.

This ambiguity can also be seen with the compound demonstratives, where a þVs form associated with a particular case and gender context would have been expected historically in a given context, but where, instead, a þVs form, either historically associated with another case or gender context or with no such context at all, is realised, for instance: (33)

a.

b.

d.

Vices and Virtues (late twelfth century) 69/3 ies Šungemann Šiede a-wei sari ‘this young man went away sorry’ Vices and Virtues (late twelfth century) 105/9 es ilche hali mihte iusticia ‘this same holy power justicia’ South English Legendary (late thirteenth century) Juliana 185 led þis hore fram me ‘take this whore away from me’,

and indeed where two manuscript variants of the same text contain the same ‘error’: (34)

a.

b.

Iuliene (late twelfth century) R220 bidden þeos bone B271 bidde þeos bone ‘make this request’ LaŠamon (later thirteenth century) 5898 C Conaan þeos Šeue afeng O Conan þeos Šeft onderfen[g] ‘Conan received this gift’.

Earlier scholars have seen the processes described here as demonstrating the introduction of the undeclined ‘new’ forms the and this. Again I say that this might be true if viewed over a sufficiently broad historical perspective. As I hope I have shown, however, the developments involved in this movement towards heavily circumscribed formal variation associated with all functions is actually part of ongoing, organic, evolutionary processes within the paradigms themselves.

THE ORIGIN AND DEVELOPMENT OF THE DEFINITE ARTICLE

299

4.3 Lesser phonological attacks on the system Two other processes also fall within the framework of phonological attacks upon the inherited system. The first of these is ‘formal dislocation’, most noticeable in the erosion of use with forms (not just demonstrative forms) with the -re ending, associated with ‘feminine dative singular’ and eventually ‘feminine genitive singular’ contexts. We can see this process in the way in which texts of similar dates and provenance, but of strikingly different conception realise very different proportions of such forms. For instance, in the Caligula version of LaŠamon’s Brut around 90% of all feminine nouns in dative case or in dative-triggering prepositional contexts are modified by demonstratives with -re endings (the historically ‘correct’ ending). In Seinte Katerine, whose surviving manuscripts date from considerably earlier than those of the Brut, no -re forms whatsoever are found in these contexts. The same apparent lack is true for other texts from roughly the same period, albeit without such striking exclusion. This could suggest that whilst -re forms are still part of the formal inventory of the scribes in question, that does not mean that the use of a given form is random. What I am suggesting is that the use or non-use of the -re forms in the given texts is a marked stylistic decision on the part of the author (and probably scribes). As Eric Stanley (1969) and others have pointed out, the Caligula version of the Brut (and perhaps the archetype) demonstrates considerable tendencies towards archaism in its language. Katerine (indeed the AB group in general) is, despite its undoubted influence from late Anglo-Saxon models, more interested in the ‘modern world’. The use of -re forms might therefore be seen as as demonstrating an ‘historical’ perspective on the part of the author (or scribes), the non-use a ‘contemporary’ one. This specialisation in use probably masks the fact that -re forms were the only part of the compound demonstrative paradigm — and, with the exception of the þæt form, of the simple demonstrative paradigm — which was not part of the general tendencies towards the movement towards one form found with the ambiguities discussed in 4.1 and 4.2 above. In the case of þæt this distinctiveness fed into processes described in 5 below; with the -re forms it led to a gradual dislocation from the rest of the paradigm Secondly, there is ‘genitive reinterpretation’, where a long-standing (but minority) means of describing possessive relationships (except when describing Christian concepts) along the lines of (35)

[definer [noun in genitive case] noun] se cyninges þeow ‘the servant of [the] king’

300

ROBERT MCCOLL MILLAR

replaced the earlier more prevalent (36)

[[definer and noun in genitive case] noun] þæs cyninges þeow ‘[the] servant of the king’

structure, possibly to avoid confusion between this particular simple demonstrative form — þes — and a large part of the compound demonstrative paradigm rendered confused by ‘ambiguity in form’ — þis, þeos, þes, þas. The development can be seen in examples such as (37)

Seinte Katerine (late twelfth century) B 695 þet wes þes deofles budel R 882–3 þ wes þes deoules budel T þ was te deoules budel ‘who was the devil’s door-man’.

Each of these ‘ambiguities’ would by their nature have weakened the formal diversity and individual distinctiveness of both simple and compound demonstrative pronoun paradigms. Together they were disastrous to the integrity of the link between form and defined function.

5.

Paradigm fissure

From an early period in the history of the great changes of which these developments form a part, Old English þæt, originally associated only with distal semantic contexts with nouns of the neuter gender-class in nominative and accusative case functional zones, was being used in positions with which it was not historically associated, whether this was in terms of gender, for example, (38)

a.

b.

Lindisfarne Gospels Mark XV/43–45 giuede lichoma hælendes  mid ongæt from ðæm aldormenn salde þ lichoma iosep (Latin: petit corpus Iesu et cum cognouisset a centurione donauit corpus ioseph) ‘[he] asked about the body of the saviour and learned about that from the centurion he gave that body to Joseph’ Peterborough Chronicle 1095/64–5 man syððan þet Romesceot be him sende ‘afterwards Peter’s Pence was sent with him’

THE ORIGIN AND DEVELOPMENT OF THE DEFINITE ARTICLE

c.

d.

e.

301

Iuliene (late twelfth century) R 493–4 Reue us reoweð þ sið ‘Magistrate, that journey grieves us’ LaŠamon (later thirteenth century) C 4473 al þat winter heo wuneðden here ‘all that winter they lived there’15 Vices and Virtues (late twelfth century) 45/19–20 Alswo doð ðat unwise mann ‘as does that unwise man’

or number, for example, (39)

LaŠamon (later thirteenth century) C 14996–8 Æluric wes king of londe : bi norðen þere Humbre . And Cadwan wes king sele : a suð half þere Humbre . blisse wes on hireden : mid balden þat kingen ‘Æluric was king of the land to the north of the Humber and Cadwan was the good king of the land to the south of the Humber. Bliss was among the people because of those bold kings’.

A number of scholars, most notably Ross in his 1936 article, have associated this development with what has been termed ‘neutralisation’. In languages that have three genders today, the three genders — masculine, feminine and neuter — are not associated entirely with male, female and asexual/nonhuman/inanimate classes. The process under discussion here, it is suggested, attempted to redress this balance by encouraging nouns which represented asexual or inanimate meaning to ‘move’ into the neuter gender-class. By the same token, male words would become masculine and female feminine. Whilst there is little doubt that processes of this type did occasionally take place, that does not mean to say that it was by any means common. Indeed scholars who follow this school of thought occasionally find themselves having to indulge in special pleading to explain why a given form should have ‘changed gender’, gone through Genuswechsel ‘gender change’. The idea itself is, in fact, fundamentally flawed. The grammatical gender classes are largely means by which noun classes are kept separate. Indeed it is unfortunate that the sex-terminology of the Roman (and earlier Greek) grammarians should have been carried forward into modern times, influencing the thought processes in particular of those whose language does not have this kind of distinction. More promising from the point of view of what the process of semantic specialisation with that actually consisted of is the work on the breakdown of grammatical gender carried out by Charles Jones. In his work on the relationship

302

ROBERT MCCOLL MILLAR

between form and function with the demonstrative paradigms in late Old English and early Middle English, most significantly in the English gloss to the Lindisfarne Gospels, Jones (most cogently in 1988: § 2.4–2.10) suggests that, as in present-day English, radical usages of þæt represent — from a very early period — a concentration on the ‘demonstrative’ element of the simple demonstrative’s semantic field, whilst at the same time other forms from the same paradigm still represent their full range of original semantic and functional associations. This discrepancy provokes what I term an ‘ambiguity in function’. As with the other ambiguities born from variation described before, there would have come a point where the ‘weaker’ of the two competing systems, the historical one, no longer supported by the original case-gender apparatus, would have been jettisoned in an effort to ‘streamline the system’, to avoid ambiguity. Yet neither Jones’ apparently incontestable description of the process, or the more dubious neutralisation hypotheses previously prevalent, actually ‘explain’ what the motivation behind this development was. It might be argued that the semantic specialisation of that is largely due to the breakdown of grammatical gender — yes indeed, but only because it is part of the same process. If purely the breakdown of grammatical gender were involved in this process, it would be most unlikely that a single form (from one, not both paradigms) should have been ‘plucked out’ to represent a highly specified distal demonstrative meaning. I would argue that, to do this, we have to look beyond system-internal factors and examine what was prevalent system-externally. In Western Europe today, most languages have separate distal demonstrative pronouns and definers. This is with the exception of High German, which, as we have seen, has ‘near-article’ functions associated with its distal demonstrative. In the late Old English period, when English was first developing a definite article, this was the case with at least the Celtic languages, French and the Scandinavian dialects (le Bidois and le Bidois 1967: § 65–220; Hulthén 1948: 18–86; Calder 1923 [1990]: § 117; Wessén 1958: § 128; Iversen 1973: § 148–149), albeit with somewhat less general coverage than is the case today. Could it be possible that one (or more) of these languages should have influenced English? At first glance, the most obvious choice for contact-induced change would be the Celtic languages. But the Celtic languages have had such a small influence upon Standard English — even in the easily changeable lexical sphere — that it is difficult to imagine how this might have come about in the far less susceptible to change semantic/syntactic zone. The two most far-reaching influences upon English during the period are the Scandinavian and Norman French influences. Without giving the development much thought, it is the French influence

THE ORIGIN AND DEVELOPMENT OF THE DEFINITE ARTICLE

303

that appears the more promising. It is, after all, the case that while Norse has — and had — a postposed enclitic definer derived from the third person singular personal pronoun (Modern Danish mand ‘man’ manden ‘the man’) and a preposed distal demonstrative (Modern Danish den mand ‘that man’), French — like English — has (and had) a preposed demonstrative and a similarly preposed definite article. There are problems with accepting such a conclusion at face value, however. At the time at which þæt was first developing a more circumscribed semantic/functional zone in comparison with the rest of the simple demonstrative paradigm, English was not in direct substantial contact with any of the dialects of French. It was only when the development was well underway in some dialects that the growth in Norman influence preceding the Conquest of 1066 was felt. Equally, if there were a transfer of usage of a more casual nature before the Conquest, it would be most unlikely to have been instigated in the North of England where the change appears to have started. The Scandinavian presence in England was concentrated in just such an area, however.16 When people who speak one language begin to speak another, they are inclined to carry over linguistic material from their first language to their second. Speakers of Scandinavian languages would have felt the need for a separate definer in the new language they were learning. There are two ways in which they might have gone about such a development. The first is that they would have carried over their own native mode of defining into the new language, either by using their native formation or by employing the same process using the tools of the new language (mannhe ‘the man’). The second would be to use the building blocks of the new language to create essentially the same semantic distinction without using the same morphological materials. It was, of course, the latter that came about. The reasons for this will no doubt always be obscure — these were obscure times historically for the North of England, after all. I might tentatively put forward an explanation, however. If English had had only a distal demonstrative that was not also used as a definer (the near-article), it is quite possible that the Norse speakers of English would have carried over their own pattern of formation into the new language. But since English had, as was seen in 3.2 above, just such a semantic distinction (albeit without a concomitant semantic split), it would probably have been easier for the Norse English speakers to reinterpret one phonologically distinct element of the paradigm as separate from the rest, as being equivalent to their own distal demonstrative pronoun (which in terms of origin that of course is). This association would have been encouraged by the gradual formal simplification of the simple demonstrative paradigms (although

304

ROBERT MCCOLL MILLAR

this might be a ‘chicken and egg’ type explanation since this simplification might have been permitted by the semantic simplification). It could be argued that with the growth in French-speaking influence in England during the eleventh century, the prestige nature of that language’s system might have encouraged the development of this new system forged by paradigm fissure. But it could only encourage a development that had already been initiated.

6.

Conclusions

It can therefore be argued that the development of a discrete definite article in English, whilst a not unlikely event, was by no means inevitable. If English had followed the same path as High German, it would today have a simple demonstrative pronoun which also represented ‘near-article’ meanings. Even if the great morpho-syntactic changes in the language had been carried through under such a scenario it would not be impossible for English to have maintained the main distinction between the as distal demonstrative and ‘near article’, and this as proximal demonstrative. That this is not the case has, I would argue, more to do with the development of that as a ‘pure demonstrative’, bearing only distal meaning with little in the way of article or near-article function. Whilst it would be impossible to prove that this development was not due to system-internal developments, it would seem likely that there was at least some influence from the speakers of Norse dialects settled in the North of England who already had such a semantic outcome in their own languages.

Notes 1.

Although Maria Selig (1992: § 5.3) would claim that the development has more to do with the codification of a previously only optional concept. This, I would argue, is a telescoping of two separate processes into one, however.

2.

In work more recent than his 1978 article, Greenberg has further refined his views on these developments. He now sees the development in question as being part of the process he terms ‘re-grammaticalization’. In his view, Stage I of such a development would be where demonstratives become definite articles. In Stage II their use is extended to all specific nouns, whether definite or indefinite. This type of development leads, in his view, to the article becoming ‘morphologised’ as a prefix or suffix on a noun. In Stage III, the use of this affix spreads to almost all nouns (Greenberg 1991). Heinrichs’ comment (1954: 21) that, “der gleichzeitige Gebrauch von Pronomen als Demonstrativa und als Artikel [schwächt] die

THE ORIGIN AND DEVELOPMENT OF THE DEFINITE ARTICLE

305

deiktische Kraft dieser Pronomina” (‘the simultaneous use of pronouns as demonstrative and as article weakens the deictic strength of these pronouns’) demonstrates that these views have been present in a less developed form for a considerable period. It is worth noting, however, that nowhere does he explain why (or even how) the paradigm fissure already described has taken place 3.

This statement is made with the proviso that Modern High German has developed the capability of making some distinction in some case/gender contexts between ‘simple demonstrative’ pronouns which can be used in both article and demonstrative contexts and those which can be used only in demonstrative.

4.

Rennhard (1962: 235) suggests that “[d]ie Formen des einf. Dem. zeigen sich als ziemlich stabil. Sie sind noch alle in MSS belegt, die für das zus. Dem. nur noch als Restformen bestehen” (‘The forms of the simple demonstratives show themselves to be fairly stable. They are still all exemplified in the manuscripts; the forms of the compound demonstratives only exist as residual forms’). Whilst the reason for this discrepancy is difficult to explain, it certainly has a degree of truth to it. Nevertheless, many of the processes discussed for the simple demonstrative can also be found for the compound demonstrative.

5.

See, for instance, Diehn (1901: 66), Seidler (1901: 23) and, in particular, Jones (1964: 240–241), (1967a: 105), (1967b: 303), (1983: 336), (1988: 100–101, 217–218, 221, 228) and Markus (1990: § 6.4.2).

6.

Indeed, so ubiquitous has this usage become in British dialects that it can be heard spreading into rhotic accents such as those to be found in Scotland.

7.

Indeed Minkova herself appears to have been heading along these lines in earlier work. In her 1984 essay she writes (1984: 57): Elision, both ‘metrical’ and in ‘ordinary’ speech was a widespread phenomenon in EME, ie before 1200 … It is possible to consider elision in hiatus the environmentally conditioned beginning of the more general, across the board, process of schwa deletion in English.

8.

For examples (10) and (11) see transcriptions in Millar and Nicholls (1997).

9.

For a discussion of this point see Millar and Nicholls (1997: 442–443).

10.

It is more than likely Pervaz who based her idea upon that of d’Ardenne, since d’Ardenne (1961) is a reprint of a widely disseminated dissertation of 1936.

11.

The orthographical practise here is Jones’.

12.

It is worth noting, however, that this example does not suit completely the position-based arguments he gives before.

13.

In the Peterborough Chronicle. As Jones (1988: 183) shows with LaŠamon’s Brut, examples of the phenomenon under discussion “exist more or less as occasional forms alongside prepositional phrase constructions involving locative [by which I assume he means dative] case forms”, following which he cites examples with þVn and þVre. It may be that with the LaŠamon examples he is indulging in special pleading, since, as we saw in Chapter 3, the latter, ‘dative’, type much outweighs the ‘accusative’. It may be that the phenomenon is confined largely to the First Continuation of the Peterborough Chronicle, in which case we might have to question whether þVne is preferred in these contexts because it is phonologically more marked than is þVn. If this is the case, then Cecily Clark’s remarks (Clark 1970: lxi) that “[t]he unhistorical uses of þone are to be explained by an analogous substitution of þVne for the þV found in Peterborough speech” is due to “false archaism”, which Jones (1988: 150),

306

ROBERT MCCOLL MILLAR disparages saying, “False archaism is a recourse made by a linguistic model devoid of explanatory power”, may actually have some truth underlying them. This would be the case if — as I am suggesting in this essay — there was considerable ambiguity leading to variation within the demonstrative paradigms. Rather than there being a complete ‘substitution’ of one form by another, we have instead a preference on the part of scribes for a more overtly traditional form — still present, if not necessarily common — over a new form without auctoritas, as has been suggested for other texts by Stanley (1969) and Millar and Nicholls (1997). In fact, evidence from the late Old English glosses do appear to refute Jones’ assertion that it would be impossible for a scribe to maintain an ‘artificial’ system consistently. In his essay on Karl Luick’s assertion that ‘man schrieb wie man sprach’, Eric Stanley writes (1988: 321–322): the absence of Scandinavianisms in Owun’s South Northumbrian (except when copied from Aldred’s North Northumbrian) is striking, and must mean that Owun, working in a heavily Scandinavianized part of England and presumably himself speaking the English of the area, did not readily admit Scandinavian loanwords into his writings, but he sometimes copied such words from the Lindisfarne gloss to which he had access. It might be argued, however, that such a task is easier when it comes to lexis than it would be with morphology. This does not negate the whole point, however.

14.

Jones’ argument (1988: § 5.5) that this ambiguity is peculiar to the Brut is, as we have seen, entirely untenable.

15.

Note the equivalent in O: al þan winter wonede here.

16.

See Millar (1997) for a discussion of the theoretical implications of this distribution.

Texts d’Ardenne, S. R. T. O. (ed.). 1961. Þe Liflade ant te Passiun of Seinte Iuliene [EETS os 248]. Oxford: Oxford University Press. d’Ardenne, S. R. T. O. and Dobson, E. J. (eds). 1981. Seinte Katerine. [re-edited from MS Bodley 34 and the other Manuscripts, EETS ss 7]. London/Oxford: Oxford University Press. Atkins, J. W. H. (ed.). 1922. The Owl and the Nightingale. [ed. with introduction, texts, notes, translation and glossary]. Cambridge: Cambridge University Press. Bliss, A. J. (ed.). 1966. Sir Orfeo, 2nd edition. Oxford: Oxford University Press. Brook, G. L. and Leslie, R. F. (eds). 1963/1972. LaŠamon: Brut. [edited from the British Museum MS Cotton Caligula A.ix and British Museum Ms, Cotton Otho C.xiii., EETS os 250/277], London/Oxford: Oxford University Press. Clark, C. (ed.). 1970. The Peterborough Chronicle, 1070–1154, 2nd edition. Oxford: Oxford University Press. d’Evelyn, C. and Mill, A. J. (eds). 1956. The South English Legendary, [edited from Corpus Christi Cambridge Ms. 145 and British Museum Ms. Harley 2277 with

THE ORIGIN AND DEVELOPMENT OF THE DEFINITE ARTICLE

307

variants from Bodley MS. Ashmole 43 and British Museum Ms. Cotton Julius D. IX, vol. 1, EETS os 235], London/Oxford: Oxford University Press. Holthausen, F. (ed.). 1888. Vices and Virtues. [EETS os 89]. London/Oxford: Oxford University Press. Skeat, W. W. (ed.). 1887. The Holy Gospels in Anglo-Saxon Northumbrian and Old Mercian Versions, [synoptically arranged, with collations exhibiting all the readings of all the MSS., together with the early Latin version as contained in the Lindisfarne MS., collated with the Latin version in the Rushworth MS.], Cambridge: Cambridge University Press.

References Anderson, J. 1977. On Case Grammar. London: Croom Helm. d’Ardenne, S. R. T. O. (ed.). 1961. Þe Liflade ant te Passiun of Seinte Iuliene [EETS os 248]. Oxford: Oxford University Press. le Bidois, G. and le Bidois, R. 1967. Syntaxe du francais moderne: ses fondements historiques et psychologiques, 2nd edition. Paris: Picard. Blake, B. J. 1994. Case [Cambridge Textbooks in Linguistics]. Cambridge: Cambridge University Press. Blake, N. F. and Jones, Ch. (eds). English Historical Linguistics: Studies in Development [CECTAL Conference Papers Series 3]. Sheffield: The Centre for English Cultural Tradition and Language, University of Sheffield. Calder, G. 1923. A Gaelic Grammar: Containing the Parts of Speech and the General Principles of Phonology and Etymology, with a Chapter on Proper and Place Names. Glasgow: MacLaren. Campbell, A. 1959. Old English Grammar. Oxford: Oxford University Press. Christophersen, P. 1939. The Articles: A Study of their Theory and Use in English. Oxford: Oxford University Press. Clark, C. (ed.). 1970. The Peterborough Chronicle, 1070–1154, 2nd edition. Oxford: Oxford University Press. Diehn, O. 1901. Die Pronomina im Frühmittelenglischen [Kieler Studien zur englischen Philologie 1]. Tübingen: Niemeyer. Elliott, R. W. V. 1974. Chaucer’s English. London: André Deutsch. Fisiak, J. (ed.). 1980. Historical Morphology. The Hague: Mouton de Gruyter. Greenberg, J. H., Ferguson, Ch. A. and Moravcsik, E. A. (eds). 1978. Universals of Human Language, vol. 3: Word Structure. Stanford: Stanford University Press. Greenberg, J. H. 1978. “Gender markers”. In Universals of Human Language, vol. 3: Word Structure. J. H. Greenberg, Ch. A. Ferguson and E. A. Moravcsik (eds), 48–82. Stanford: Stanford University Press. Greenberg, J. H. 1991. “The last stages of grammatical elements: Contractive and expansive desemanticization”, In Approaches to Grammaticalization [Typological

308

ROBERT MCCOLL MILLAR

Studies in Language 19], E. C. Traugott and B. Heine (eds), 301–314. Amsterdam: Benjamins. Heinrichs, H. M. 1954. Studien zum bestimmten Artikel in den germanischen Sprachen. Giessen: Wilhelm Schmitz. Himmelmann, N. P. 1997. Deiktikon, Artikel, Nominalphrase: Zur Emergenz syntaktischer Struktur [Linguistische Arbeiten 362]. Tübingen: Niemeyer. Hoffmann, P. 1909. Das grammatische Genus in LaŠamon’s Brut [Studien zur englischen Philologie 36]. Tübingen: Niemeyer. Hulthén, L. 1948. Jämförande Nunordisk Syntax II [Comparative Syntax of the Contemporary Nordic Languages II]. Göteborgs Högskolas Årsskrift LIII. Gothenburg: Elanders Boktryckeri Aktiebolag. Iversen, R. 1973. Norrøn Grammatikk [Norse Grammar], 7th edition by F. Halvorsen. Oslo: Aschehoug. Jones, Ch. 1964. Grammatical gender in late Old English and early Middle English. B. Litt. dissertation, University of Glasgow. Jones, Ch. 1967a. “The functional motivation of linguistic change”. English Studies 48: 97–111. Jones, Ch. 1967b. “The grammatical category of gender in early Middle English”. English Studies 48: 289–305. Jones, Ch. 1983. “Determiners and case-marking in Middle English: A localist approach”. Lingua 59: 331–343. Jones, Ch. 1988. Grammatical Gender in English 950–1250. London: Croom Helm. Kastovsky, D. and Bauer, G. (eds). 1988. Luick Revisited. Tübingen: Narr. Kemenade, A. van. 1987. Syntactic Case and Morphological Case in the History of English. Dordrecht: Foris Publications. Kerkhof, J. 1982. Studies in the Language of Geoffrey Chaucer, 2nd edition. Leiden: Leiden University Press. Koch, C. F. 1882. Die Laut und Flexionslehre der englischen Sprache, 2nd edition. Kassel: Georg H. Wigand. Körner, K. 1888. Beiträge zur Geschichte des Genuswechsels der englischen Substantiva. Ph. D. dissertation, University of Greifswald. Kramsky, J. 1972. The Article and the Concept of Definiteness in Language [Janua Linguarum Series Minor 125]. The Hague: Mouton. Laing, M. (ed.). 1989. Middle English Dialectology. Essays on Some Principles and Problems. Aberdeen: Aberdeen University Press. Landwehr, N. 1911. Das grammatische Geschlechts in der Ancren Riwle. Ph. D. dissertation, University of Heidelberg. Lehmann, Ch. 1982. Thoughts on Grammaticalization. A Programmatic Sketch. Cologne: Institut für Sprachwissenschaft. Markus, M. 1990. Mittelenglisches Studienbuch. Tübingen: Francke. Micillo, V. 1982. “Observazioni sulla determinazione in antico inglese”, Annali instituto universitario orientale di Napoli sezione germanica [AIUON 25], 161–202.

THE ORIGIN AND DEVELOPMENT OF THE DEFINITE ARTICLE

309

Millar, R.McC. 1997. “English a koinëoid? Some suggestions for reasons behind the creoloid-like features of a language which is not a creoloid”. Vienna English Working Papers 6: 19–49. Millar, R.McC. and Nicholls, A. 1997. “Ælfric’s De Initio Creaturae and London, British Library, Cotton Vespasian A.xxii: Omission, addition, retention and innovation”. In The Preservation and Transmission of Anglo-Saxon Culture [Studies in Medieval Culture XL], P. E. Szarmach and J. T. Rosenthal (eds), 431–463. Kalamazoo: Medieval Institute Publications. Minkova, D. 1984. “Early Middle English metre elision and schwa deletion”. In English Historical Linguistics: Studies in Development [CECTAL Conference Papers Series 3], N. F. Blake and Ch. Jones (eds), 56–66. Sheffield: The Centre for English Cultural Tradition and Language, University of Sheffield. Minkova, D. 1991. The History of Final Vowels in English. The Sound of Muting. Berlin: Mouton de Gruyter. Mitchell, B. 1985. Old English Syntax, 2 vols. Oxford: Clarendon Press. Pervaz, D. 1958. The survival of grammatical gender in LaŠamon’s Brut, the Southern Legendary and Robert of Gloucester’s Chronicle. Ph. D. dissertation, University of Edinburgh. Purdy, D. W. 1973. “Did Old English have a definite article?”. York Papers in Linguistics 7: 121–124. Quirk, R. and Wrenn, L. 1955. An Old English Grammar. London: Methuen. Rennhard, S. 1962. Das Demonstrativum im Mittelenglischen. Winterthur: P. G. Keller. Ross, A. S. C. 1936. “Sex and Gender in the Lindisfarne Gospels”. Journal of English and Germanic Philology 35: 321–330. Samuels, M. L. 1989a. “Chaucerian final ‘-E’”. In The English of Chaucer and his contemporaries, J. J. Smith (ed.), 7–12. Aberdeen: Aberdeen University Press. Samuels, M. L. 1989b. “Some applications of Middle English Dialectology”. In Middle English Dialectology. Essays on Some Principles and Problems, M. Laing (ed.), 64–80. Aberdeen: Aberdeen University Press. Seidler, O. 1901. Die Flexion des englischen einfachen Demonstrativpronomens in der Übergangzeit 1000–1200. Ph. D. dissertation, University of Greifswald. Selig, M. 1992. Die Entwicklung der Nominaldeterminanten im Spätlatein: Romanischer Sprachwandel und Lateinische Schriftlichkeit [Script Oralia 26]. Tübingen: Narr Smith, J. J. (ed.). 1989. The English of Chaucer and his Contemporaries. Aberdeen: Aberdeen University Press. Spamer, J. B. 1979. “The development of the definite article in English: A case study of syntactic change”. Glossa 13: 241–250. Stanley, E. G. 1969. “LaŠamon’s antiquarian sentiments”. Medium Ævum 38: 23–37. Stanley, E. G. 1988. “Karl Luick’s ‘Man schrieb wie man sprach’ and English Historical Philology”. In Luick Revisited, D. Kastovsky and G. Bauer (eds), 311–334. Tübingen: Narr.

310

ROBERT MCCOLL MILLAR

Szarmach, P. E. and Rosenthal, J. T. (eds). 1997. The Preservation and Transmission of Anglo-Saxon Culture [Studies in Medieval Culture XL]. Kalamazoo: Medieval Institute Publications. Traugott, E. C. and Heine, B. (eds). 1991. Approaches to Grammaticalization [Typological Studies in Language 19]. Amsterdam: Benjamins. Vachek, J. 1980. “Problems of morphology seen from the structuralist and functionalist point of view”. In Historical Morphology, J. Fisiak (ed.), 373–381. The Hague: Mouton de Gruyter. Wessén, E. 1956. Svensk Språkhistoria, vol. III, Grundlinjer till en Historisk Syntax [The History of the Swedish Language, vol. III, Outlines of an historical syntax.]. Stockholm: Almqvist & Wiksell. Wessén, E. 1958. Isländsk grammatik [Icelandic Grammar]. Stockholm: Svenska Bokforlaget.

Parallelism vs. asymmetry The case of English counterfactual conditionals Rafał Molencki University of Silesia

1.

Introduction

In the languages of the world one can come across two opposing tendencies in the form of the protasis and the apodosis in counterfactual conditionals. In line with one of these tendencies both clauses have identical verb forms — the same morphosyntactic marking — and thus they often cannot be distinguished one from another, especially when there are no subordinators. This natural crosslinguistic tendency for isomorphic verbal forms in the protasis and apodosis — what Behaghel (1923–1932) called Streben nach Parallelismus — is found in many languages (among others Papuan, Australian, various African languages, Baltic, Slavonic, Romance and Hungarian). It should, however, be noted that the verbs can be accompanied by some other markers (e.g. conjunctions, particles or inversion). The ‘strive for parallelism’ has been competing with another tendency to make protasis and apodosis different, either for economic reasons (‘the same’ does not have to be repeated) or because, being less irrealis, the protasis gets demodalized. Some other examples of the variation between parallelism and asymmetry can be found in the development of English adverbial clauses, e.g. Old English þa …, þa …; þonne …, þonne …, later replaced by when …, (then) …, or coordinate structures ge …, ge …, now rendered by both …, and … . Discussing isomorphism, Bolinger (1977: x) says that “the natural condition of language is to preserve one form for one meaning, and one meaning for one form”. Haiman (1985: 30) agrees: “Recurrent identity of form must reflect similarity of meaning”, which is justified by the economy of language. The presence of synonymy (several forms — one meaning) and homonymy (one form — several meanings) in all languages might be treated as counterevidence of

312

RAFAŁ MOLENCKI

isomorphism, but, as is well known, perfect synonyms are extremely rare and language will tolerate homonymy as long as it does not bring about confusion. The central issue in this article is whether in protasis and apodosis we have the same meaning or not. From the data presented it appears that speakers of different languages and especially speakers of English at different stages of its development have not been consistent in their choice of verb forms in counterfactual conditionals. While morphologically symmetrical protasis and apodosis indicate that the two verb forms convey the same meaning, the other tendency to make the two clauses formally distinct supports the contrary. This confirms the cognitivist view that the apodosis is epistemically more remote than the protasis (cf. James 1982; Dancygier and Sweetser 1994). The data found in the present work come from large electronically readable corpora of earlier and modern English texts, to which I have applied searching and concordancing programs. The major sources are the Diachronic Part of the Helsinki corpus (Kytö 1993; wherever possible I checked the examples in the book editions), the on-line version of the Second Edition of the Oxford English Dictionary and the CD-ROM versions of two British quality daily newspapers from the year 1993. Additionally I have used some other Early Modern English texts downloaded from various websites (see Electronic corpora in the References below) and the data collected by Visser (1963–1973), the editors of the Middle English Dictionary and some other authors. The article also includes some random examples that I have come across while reading numerous old and contemporary pieces of writing.

2.

Early Germanic parallelism

Throughout the history of English there has been interesting variation in the interaction of tense, mood and modality in expressing counterfactuality. It has been one of the most unstable categories and has kept on changing its morphosyntactic exponents at different stages of the language’s development. This lifecycle of counterfactual markers in Germanic is best described in Dahl (1997). The pre-Old English stage, as the Gothic data appear to indicate, was characterized by parallel preterite optative forms in the protasis and the apodosis: (1)

frauja, iþ veseis her, ni þau gadauþnodedi broþar meins. Lord if thou-were here not then died brother mine Vulgate: domine si fuisses hic frater meus non fuisset mortuus. Authorised Version (1611): ‘Lord if thou hadst been here, my brother had not died’ John 11,21

PARALLELISM VS. ASYMMETRY IN COUNTERFACTUALS

313

though Behre (1934: 53) believes that in earlier Germanic the preterite subjunctive was originally used for past counterfactuals only and that later its usage was extended to present counterfactual situations. In Old English and Early Middle English the preterite subjunctive/indicative forms (cf. Campbell (1959: 302) on the gradual loss of morphological contrasts between the two moods) are found in both present and past counterfactual protases and apodoses. According to Mitchell (1985: § 3607), “unreality in OE is timeless”, as is exemplified by: (2)

He nære na ælmihtig, gif him ænig gefadung earfoðe he not-were() no almighty if him any order difficult wære. were() ‘He would not be almighty if any order were difficult for him to maintain’ general present time reference (ÆDT 80 early 11th c.)

(3)

gif þu wistest hwæt þe toweard is þonne weope if thou knewest() what thee imminent is then wept() þu mid me. thou with me ‘If you knew what is to come to you, you would weep with me’ specific present time reference (ÆCHom i.404.27 early 11th c.)

(4)

ac hit wære to hrædlic gif he þa on cildcradole but it were() too quick if he then in child-cradle acweald wurde. killed were() ‘but it would have been too early if he had been killed in his cradle then’ past time reference (ÆCHom i.82.28 early 11th c.)

Similarly in Early Middle English: (5)

Witti  wise wordes hit weren Šef ha neren false. witty and wise words it were if they not-were false ‘These would be prudent and wise words if they weren’t false’ (St. Katherine 317 c1200)

314

RAFAŁ MOLENCKI

(6)

Had ic an swerd, ic sluŠe ðe. had I a sword I slew thee ‘If I had had a sword, I would have killed you’ (Genesis & Exodus 3976 a1325(c1250))

Needless to say, the same parallel preterite forms were found in those counterfactuals whose protases and apodoses had different time reference, e.g.: (7)

ge witon þæt ge giet todæge wæron Somnitum þeowe, you know that you yet today were() to-Samnites slaves gif ge him ne alugen iowra wedd  eowre aþas if you them not denied() your pledges and your oaths þe ge him sealdon. that you them gave ‘you know that you would be Samnites’ slaves still today if you had not broken the pledges and oaths that you swore to them’ (Orosius 67.5 late 9th c.)

Thus, only the presence of adverbials of time, some other context or the Latin original, if extant or relevant, allow us to distinguish between the present and past conditionals.

3.

The emergence of the pluperfect in counterfactual conditionals

In the thirteenth century the pluperfect gradually begins to replace the preterite in past counterfactuals. This must have been connected with the spread of the tense in other constructions, e.g. temporal clauses or sequence of tenses in reported speech (cf. Mustanoja 1960: 507; Fischer 1992: 259). The influence of French (so strong in vocabulary loaning) cannot be excluded, either — this is the time when so many Anglo-Norman speakers (and writers!) are switching to English as the main medium of communication. First the pluperfect subjunctive occurs, as the few contrasts left after the Late Old English morphophonemic levelling indicate, but soon analogous indicative pluperfect forms prevail; compare the earlier fourteenth century versions of the religious poem Cursor Mundi with the end-of-the century manuscripts (the dating of manuscripts after Horrall 1978: 14–20):

PARALLELISM VS. ASYMMETRY IN COUNTERFACTUALS

(8)

315

“Had þou,” sco said “ben here wit vs Had noght mi broþer deied þus.” (14296 Cotton MS c1340(a1325)) had þou bene here ho saide wiþ us had noŠt my broþer deyed þus. (Fairfax MS late 14th c.) Scho said, “had þu bene here wid vs Had noght mi broþer deiei þus.” (Göttingen MS 2nd half of 14th c.) haddestou lord ben þere wiþ vs Had not my broþer deŠed þus. (Trinity MS ?c1400) Haddestou lord ben þere wiþ vs Hadde not my broþer died þus. (Arundel MS ?c1400) ‘Lord, If you had been with us, my brother would not have died thus’

Symmetrical forms of the past subjunctive and the pluperfect for present and past counterfactuals respectively become a norm between 1250 and 1350:

4.

(9)

Yif he ne were Ich were nou ded. if he not-were I were now dead ‘If it weren’t for him, I would be dead now’

(10)

War mi hare schorn, i war noght þan Stranger þan a-noþer man. were my hair shorn I were not then stronger than another man ‘If my hair were shorn, I wouldn’t be stronger than anybody else’ (Cursor Mundi 7211 1340(a1325))

(11)

If he ne had rysen fra ded to lijf had ben us all for noght. if he not had risen from dead to life had been us all for nothing ‘If he hadn’t risen from the dead, everything would have lost its significance for us’ (Cursor Mundi 17061 1340(a1325))

(12)

Bot al his praier had ben als noght If godd self his but all his prayer had been as nothing if God himself his might had wroght. might had wrought ‘but all his prayer would have been of no use if God had shown his power’ (Cursor Mundi 2797 1340(a1325))

(Havelok 54 c1300)

Modals in the apodosis

From about the mid-fourteenth century onwards, the pluperfect in the apodosis is being replaced by the combination of a bleached modal (most frequently wolde, but scholde is not infrequent, either) + have + Past Participle, as in

316

RAFAŁ MOLENCKI

(13)

For had he knowen hit biforn A childe of a mayden born for had he known it before a child of a maiden born Wolde he neuer haue Šyuen to rede þat iesu crist shulde haue would he never have given to advice that Jesus Christ should have ben dede. been dead ‘For if he had known before about a child born of a virgin, he would never have suggested that Jesus Christ should die’ (Trinity MS Cursor Mundi, 10787 ?c1400)

(14)

oure Lord Jhesu Crist wolde nevere have descended to be born of a womman, if alle wommen hadden been wikke. (Chaucer CT Melibee 220 c1395)

(15)

For he had caught kywaert by the throte and had i not that tyme comen he sholde haue taken his lyf from hym. (Caxton Reynard 8 1483)

Analogous wolde + present infinitive forms are found in present counterfactuals: (16)

If ani barn of hir war þine

I now it held it als for mine. (Göttingen MS) If any childe of hir were þine I wolde holde hit as for myne. (Trinity MS) (Cursor Mundi 2601)

This is not an entirely new construction, as the northern versions of Old English Gospels (10th-11th centuries) almost systematically have wolde + infinitive, for the southern (West-Saxon) preterite subjunctive in apodoses: (17)

West Saxon Gospels: Gif he nære yfel-dæde. ne sealde we hine ðe. if he not-were evil-doer not gave we him thee Lindisfarne Gospels: gif nere þes yfeldoend, nalde ue gesealla hine. Authorised Version: ‘If he were not a malefactor, we would not have delivered him up unto thee’(John 18,30)

For more examples of the variation see Molencki (1998: 244–245). The replacement of the preterite subjunctive by the periphrastic construction of (pre)modal + infinitive in the northern Old English dialects first can be easily explained by the fact that the subjunctive/indicative contrast was first lost in the North, where the phonetic reduction of inflectional endings to /6/ occurred at least a century earlier than in the South. Being more counterfactual than the

PARALLELISM VS. ASYMMETRY IN COUNTERFACTUALS

317

protasis (cf. Dancygier and Sweetser 1994: 9–10), the apodosis first needed to replace the no longer distinctive preterite form of the subjunctive by the new complex modal form. The problem is that none of the later northern Early Middle English texts appear to continue the analytic forms found in the Rushworth or Lindisfarne glosses. The comparison of the different manuscripts of the Cursor Mundi (Molencki forthc.) shows that in Middle English the first analytic forms are found in the later southern versions of the turn of the fifteenth century. They are very common in the classical Middle English literature of Chaucer and Gower. In Late Middle English and Early Modern English they become prevalent, almost obligatory:

5.

(18)

If I had thought you had bene so well furnished with Booke Cases, I woulde haue bene better prouided for you. (Trial of Throckmorton I,74,C1 1554)

(19)

The former would have been ruined if he had not saved it by betraying his party. (History of Charles II 1,I,170 1674)

Early Modern English symmetry

Simultaneously, however, parallel pluperfect forms keep occurring, sporadically even as late as the nineteenth century. Rissanen (1999: § 3.3.2) believes that the pluperfect in the apodosis is “particularly common” when the protasis is marked by inverted word order (“explained by the symmetry of the two verbal groups”), but in my corpus I have come across many examples in which the pluperfect main clause follows an if-clause, as well: (20)

If I had died in Guiana, I had not left 300 Marks a Year to my Wife and Son. (Trials Raleigh I, 215 C2 1606)

(21)

If he had lived a little longer, he had broke all their Schemes. (Roger North Life of the Right Honourable Francis North 326 1742)

(22)

Had I yielded to the first generous impulse … how different had been my present situation. (Scott Waverley 23 1814)

Scott, however, is known to have used archaisms deliberately (cf. Bailey 1991: 272). To a lesser extent, parallel forms are found in present counterfactuals, e.g.: (23)

Mother, I thanke you for the …. you send mee, for yf you were not, I were not able to live. (Robert Plompton’s letter to his mother c1536)

318

RAFAŁ MOLENCKI

(24)

Had I a son by thee, the grief were less. (Marlowe Faustus V.1, p. 81 1596)

All the examples available to me have were, which had become the only morphologically distinct past subjunctive verb form by Early Modern English. This is confirmed by Rissanen (1999: § 3.3.2), who observed that were “seems to resist best the replacement by auxiliary periphrasis”. Indeed, the sequence would be is extremely rare before the eighteenth century. In most of the examples from the Early Modern English part of the Helsinki corpus it is found in dependent speech conditionals, thus probably backshifting will be, e.g.: (25)

Mistresse Winchcombe gaue her great thankes for her fauour, saying, that if she needed her helpe, she would be bold to send for her. (Deloney Jack of Newbury 71 1619)

Otherwise, the modal in the apodosis was the standard form of all other verbs, whose past subjunctives and indicatives no longer differed, as in (26)

6.

For if I toke not better heede, a knaue wold haue my hennes. (Stevenson Gammer Gurton’s Needle 55 1575)

Modals in the protasis

The tendency to have a different form in the apodosis appears to have prevailed, at least in the standard language in the modern period. Nevertheless, the parallelism is by no means dead, though a new type has emerged: this time the protasis copies the verb form of the apodosis — both clauses have a three-element structure of would have + past participle. This does not happen only in nonstandard modern English dialects, as some authors have claimed (Quirk et al. 1985: 1011; Fillmore 1990: 153; Swan 1995: 261), for even in quality newspapers (both British and American) one can come across numerous examples where the protasis copies the form of the apodosis, e.g.: (27)

He thought that if he would have been able to freeze-dry one of the more attractive men, then he would not have had a desire for the other victims. (UPI Top Stories 1992)

(28)

they would not have considered murder if it would have secured them the chance to bat first. (The Daily Telegraph 12 Feb 1993)

(29)

he avoids direct denunciations of privatisation because he would make a deal with reformers if it would bring him to power. (The Times 23 March 1993)

PARALLELISM VS. ASYMMETRY IN COUNTERFACTUALS

319

This is particularly common in concessive conditionals introduced by even if: (30)

A vast majority of American women say in a poll released Saturday that they would never consider having an extramarital affair — even if no one would ever find out. (UPI Top Stories 1992)

(31)

I could have taken more of all this, even if that would have made a long book even longer. (The Times 9 Sept 1993)

The presence of the modal would in the protasis is by no means a modern development. It has been found throughout the development of the English language (cf. Denison 1993: 312–313, 1998: § 3.6.5.3, who finds the volitional sense and foreign substratum to have been the main causes). But in the earlier periods it was mostly used in its original volitional meaning, whereas in the modern examples given above it appears to be a fully grammaticalized form, analogous to and parallel with the form of the apodosis (cf. Molencki 1999: § 4.3.2). The availability of the grammaticalized would + (have) structure following if in reported questions as in: (32)

Iohn Frenchan, went to the King and craued his Passe for England, who very courteously dema~ded of vs if we would serue him in his wars. (Robert Coverte A Trve and Almost Incredible Report of an Englishman 41 1612)

may have encouraged the use of would in protases as the analogical factor. This analogy with interrogative structures extends also to the protases with inverted word order (already found in Old English), as in (33)

Had he gotten them (which was his aime) he had don his purpose. (John Pinney’s letter to his daughter Hester, 16 August 1688)

or the etymology of the if-equivalents in numerous languages (e.g. Russian or Polish jes´li < jest-li?, i.e. is + interrogative particle). In Old English protases wolde is found mostly with the volitional sense: (34)

ond gif heo Ongolcynne lifes weg bodigan ne woldon, þæt and if they to-English-race life’s way preach not would that heo þonne wæron þurh heora honda deaðes wræc þrowigende. they then were through their hands death’s misery suffering ‘and if they would not preach the way of life to the English, they would suffer the penalty of death at their hands’ (Bede 102.23 late 9th c.)

320

RAFAŁ MOLENCKI

Yet in Late Old English and Early Middle English we find more and more examples, where wolde is clearly non-volitional or at least ambiguous. As wolde was becoming increasingly common in apodoses, new parallelism appeared: (35)

þu geswore Apollonio, gif he wolde gehirsumian minum willan thou swore to-Apollonius if he would obey my will on lare, þæt þu woldest him geinnian swa hwæt swa seo sæ in doctrine that thou wouldst him restore whatever the sea him ætbræd. him took-away ‘You have sworn Apollonius that if he obeyed my will concerning doctrine, you would restore him whatever the sea took away from him’ (Apollonius of Tyre 34.26 mid-11th c.)

(36)

Forðam God sylfa behet synfullum mannum þæt he wolde for-this God self promised sinful men that he would miltsian, gyf hi woldon earnian be-merciful if they would deserve ‘For this reason God himself promised sinful people that he would show mercy if they deserved it’ (Wulfstan Homilies 210.190 mid-11th c.)

(37)

þa cydde man into þære scipfyrde. þet hi mann eaðe then made-known one to the fleet that them one easily befaran mihte. gif man ymbe beon wolde. surround might if one around be would ‘Then information was brought to the fleet that they [Wulfnoth’s ships] could easily be surrounded if the opportunity were seized’ (transl. G. N. Garmonsway) (Chronicle 138.19 year 1009)

(38)

þo cneu seint iohan. þat gif he wolde þolen þat te king then knew Saint John that if he would suffer that the king drige his unriht he mihte liuen and ben him lief and commit his injustice he might live and be him friendly and wurð. ac gif he wolde folgen rihtwisnesse he sholde þerfore worthy but if he would follow righteousness he should therefore his lif forleten and swo dide atten ende. his life abandon and so did at end ‘Then St. John knew that if he allowed the king to commit his injustice, he might stay alive and be friendly and dear to him, but if he were to follow righteousness, then he would lose his life, and did so in the end’ (Trinity Homilies 139 a1225)

PARALLELISM VS. ASYMMETRY IN COUNTERFACTUALS

321

(39)

As Šif a lond wolde bere good corn wiþowte tylyng and donghyng þerof, it were but ydel to traueyle [work] þerfore, whonne it encressuþ not þe fruyt. (Wycliffe Sermons P I,588 a1425)

(40)

& [=if] he wolde not a followed me, I wolde haue retourned ageyn … Whereby I shulde in no wyse haue fallen in this daungier. (Earl Rivers, The Cordyal 79,12 c1479)

(37) and (39) are particularly remarkable, as the bleached wolde occurs only in the protasis and additionally (39) has an inanimate subject. More and more instances of clearly grammaticalized would appear in Early Modern English, e.g.: (41)

If the world would have begunne as I would have wished. (More Richard III 235 1513 (1641) after OED)

(42)

he was soe gredy on his bocke, that yf his master wold not have beaten him, yf he could not say his lesson well, he wold have wepte and suobbed more than yf he had byne beaten. (Biography of Forman 11 1600)

Once the structure became well established in counterfactual conditionals, its usage was extended to semantically related constructions such as optative wishes or hypothetical comparatives, which also refer to counterfactual situations, e.g.: (43)

I wish my mony would have extended itself into a larger maner, for it may be beleft I have but three shillinges to keep me untill our Lady Day. (Elizabeth Oxinden’s letter to her mother, 25 Feb 1666)

(44)

“I wish,” said he, “brother, you would have confined your care to your own daughter, and never have troubled yourself with my son …” (Fielding Tom Jones xiv.viii.231 1749)

The concordancing program has provided me with quite a few recent counterparts, e.g.: (45)

I wish the home team would have won today, but there’s always tomorrow. (UPI Top Stories 1992)

(46)

Looking back, I wish … I would have known what the appropriate student/professor relationship was. (Stanford Daily 2 June 1995)

All those very early examples above, where would appears to be deprived of the volitional meaning altogether, indicate that the construction may not be so new and not necessarily “made in America” as Fillmore (1990: 153) thinks. Thus English appears to have parallel forms again, but this time if + modal + infinitive

322

RAFAŁ MOLENCKI

(protasis), modal + infinitive (apodosis) for counterfactuals with present time reference, and if + modal + perfect infinitive (protasis), modal + perfect infinitive (apodosis) for past time, and such forms are spreading to the despair of normativists.

7.

Other languages

An interesting phenomenon concerning counterfactual constructions can be observed in a close cognate of English, viz. Modern Dutch, where the parallel and asymmetrical patterns appear to be in free variation like in Early Modern English. Wekker (1991: 140–142) discusses the alternation between the structures with the auxiliary modal zou (=should) + present or perfect infinitive (analogous to English would + present or perfect infinitive) and those with the simple past or the pluperfect in Modern Dutch in present and past counterfactuals, respectively. All four combinations of +zou and -zou structures are available in protases and apodoses, though “Dutch  s [counterfactuals] favour the use of the pluperfect structure”. Wekker complains that Dutch grammarians do not have much to say about the difference, but he himself believes that the presence of zou in the protasis “leaves it open whether the proposition is really false or not”, whereas the event expressed by the past tense or pluperfect is “contrary to fact”. The Early Modern English data, however, are too scarce to draw similar conclusions. Discussing the development of Spanish conditionals, Penny (1991: 205) speaks of a “(possibly universal) tendency … for the same verb-form to appear in both the apodosis and the protasis of conditional sentences”. Also in Harris and Vincent (1988: 72) we learn that language purists in various Romance languages strongly oppose the parallel usages of verb forms in protases and apodoses in colloquial speech (“despite continual denunciation by prescriptive grammarians” one finds “morphological harmony between the two parts of modal conditional sentences”). Interestingly, all these parallel verbal forms were derived from different sources. Only Sicilian continues the classical Latin usage, while the other languages developed new conditional forms: (47)

Italian: conditional se verrebbero ci aiuterebbero ‘if they came (would come) they would help us’

(48)

Sicilian: archaic past subjunctive vivissi si ci fussi acqua ‘I would drink if there was any water’

PARALLELISM VS. ASYMMETRY IN COUNTERFACTUALS

323

(49)

Italian: imperfect indicative se mi lasciavi lassù, era tanto meglio ‘if you had left me up there, it would have been a lot better’

(50)

French: perfect conditional je l’aurais pas fait si j’aurait su ‘I wouldn’t have done it if I would have known’ (after Harris and Vincent 1988: 240, 304)

Similar trends can be observed in as different languages as e.g. German, Baltic, Slavonic, Hungarian, e.g.: (51)

Polish Panie, gdybys´ tu był, nie umarłby Lord if-.-. here was, not died-. brat mój. (John 11,21) brother mine ‘Lord, if you had been here, my brother would not have died’

(52)

Hungarian Ha jött volna láttam volna. if came-3 be() saw-1 be() If s/he had come I would have seen him/her’ (after Haiman 1985: 54)

In all these languages, however, we find only partial parallelism (only the protasis is marked with an IF word); but there are many languages (Papuan, Australian, Moroccan Arabic, Hausa, Yoruba, Swahili) where the parallelism is complete, morphologically indistinguishable: “the parallelism is taken to an extreme in those languages where protasis and apodosis are formally indistinguishable: not only are both inflected with the same irrealis particle or verbal desinence, but neither clause has any particle analogous to either if or then” (Haiman 1985: 51), e.g. in Daga used in Papua-New Guinea: (53)

Ya wada-nege-po, ya anu-po. not say-me-, not know- ‘if s/he hadn’t told me, I wouldn’t know’ (after Murane 1974: 259)

There are languages in which the irrealis marker may appear in both protasis and apodosis or in either of them. According to Haiman (1985: 84–85) hypothetical and given conditionals are not usually characterized by such thorough morphological parallelism between protasis and apodosis in most languages; on the other

324

RAFAŁ MOLENCKI

hand, counterfactual conditionals often are. To the question why counterfactual protases and apodoses are morphologically symmetrical, why we have tense agreement or what might be called discontinuous grammaticalization across clauses, the commonsense answer is simply that both are counterfactual — the epistemic stance for both protasis and apodosis is basically the same. Why then should we have two different exponents? But Haiman (1985: 85) believes such argumentation is insufficient, as we find a similar case with hypotheticals, where protases and apodoses are more often different.

8.

Present-day English variation in the protasis

On the other hand, where there are differences between the morphosyntactic marking between the two clauses one might argue, following Haiman (1985: 56), that “since protasis and apodosis agree in mood anyway, the expression of the mood in the first is frequently omitted”, which might be explained by the laws of economy in language. Hence we find a canonical Modern English pluperfect, deprived of modality, or even the simple preterite form for past counterfactuals (Quirk et al. 1985: 1012), which is becoming more and more common in colloquial English, as in (54)

If she was near to me, I’d have done to her what Arthur Jackson did to Theresa Saldana. (UPI Top Stories 1992)

I will only partially agree with Haiman (1985: 252) that the symmetry is destroyed for “essentially economic reasons”. I believe that a more salient reason for the apodosis to have a different form is the fact that its epistemic stance is more remote than that of the protasis. The protasis of a counterfactual conditional sets up the imaginary world in which the apodosis is true (cf. James 1982: 378), which consequently is less real (or more counterfactual) than the if-clause, whence the need for its higher degree of modality — cross-linguistically the counterfactual is a highly marked category. As Haiman (1985: 57) admits, “the protasis creates a set of irrealis worlds and the apodosis represents a proposition whose validity is asserted in the framework of those worlds alone. In this sense … the apodosis “agrees” in mood with the protasis”. The problem is also discussed within the theory of possible worlds in Dancygier and Sweetser (1994) — they call it mental space embedding: “the apodosis being derived from the protasis is true in the same space in which the protasis is true”. Some English speakers, apparently unhappy with the demodalized pluperfect form, have one more variant of the protasis, namely an extra have intervening

PARALLELISM VS. ASYMMETRY IN COUNTERFACTUALS

325

between the auxiliary had and the past participle (cf. Denison 1993: 356–358, 1998: § 3.3.2.5): (55)

If we hadn’t’ve met Harry, where would we be now? (example from Fillmore 1990: 139)

Various explanations of this peculiar structure have been offered, from phonetic harmony (three-element verb form in both clauses), psychological (Boyland 1995), and morphosyntactic (the /6v/ sequence being the marker of irreality) to foreign influences. The search of earlier English texts has provided me with some remarkably early instances from the times when the subjunctive was disappearing: (56)

For had nott yit that danger have been, I mygh yit have ben at home. (Paston Letters 5.328 1475)

(57)

had tybert the catte haue ben there, he shold also somwhat haue suffred. (Caxton Reynard 46 1483)

Curiously enough, all of the modifications (which in the light of the historical data are not new at all, as might be inferred from the fact that only more recent grammars of English take account of them) concern the past counterfactual protasis only. Since the disappearance of the pluperfect from the apodosis, speakers of none of the English dialects have altered its structure: the modal (would) + ’ve + past participle, which appears to have satisfied the need for having a distinct, inherently counterfactual form in the language. Being less counterfactual, the protasis has been vacillating between parallel or nearly parallel forms (if we would have done; if we had have done) and the demodalized pluperfect or even pure preterite (if we had done; if we did). We have some evidence that all these forms must have coexisted in spoken English since Late Middle English. The paucity of Early Modern English examples of the variants at the expense of the ‘standard’ if we had done can be explained by very strong authority on the part of prescriptivists and lexicographers, who were imposing ‘the only correct form’ in numerous school grammars. Obviously, due to increasing standardization, morphosyntactic change must have been much slower than in the pre-mass media past. However, owing to the recent non-prescriptive approach to the study of language and even to language teaching, one can find more and more examples of variation in expressing past counterfactual conditionals, which is evidence of the fact that the struggle between parallelism and asymmetry is by no means over.

326

RAFAŁ MOLENCKI

Texts Bately, J. (ed.). 1980. The Old English Orosius [EETS S. S.6]. Oxford: Oxford University Press. Bethurum, D. (ed.). 1957. The Homilies of Wulfstan. Oxford: Clarendon Press. Blake, N. F. (ed.). 1970. Caxton’s ‘History of Reynard the Fox’ [EETS 263]. Oxford: Oxford University Press. Einenkel, E. (ed.). 1884. The Life of Saint Katherine [EETS 80]. London: N. Trübner and Co. Garmonsway, G. N. (ed.). 1972. The Anglo-Saxon Chronicle. London and New York: Dent and Dutton. Goolden, P. (ed.). 1958. The Old English ‘Apollonius of Tyre’. Oxford: Oxford University Press. Horrall, S. M. (ed.). 1978. The Southern Version of Cursor Mundi. Ottawa: The University of Ottawa Press. Miller, T. (ed.). 1890. The Old English Version of Bede’s ‘Ecclesiastical History of the English People’ [EETS 95, 96], London: N. Trübner and Co. Morris, R. (ed.). 1874–1893. Cursor mundi. A Northumbrian Poem of the XIVth Century in four Versions [EETS 57, 59, 62, 66, 68], 3 vols. London: K. Paul, Trench, Trübner & Co. Plummer, Ch. (ed.). 1892. Two of the Saxon Chronicles Parallel with Supplementary Extracts from Others. Oxford: Clarendon Press.

Electronic corpora CD-ROMS The Daily Telegraph and The Sunday Telegraph 1 January 1993–31 January 1993 on CD-ROM. Distributed by Chadwyck — Healey, produced by FT Profile. Helsinki Corpus. Diachronic Part of the Helsinki Corpus of English Texts. 1991 prepared by Matti Rissanen and Ossi Ihalainen et al. University of Helsinki [available online], OED. 1991. Oxford English Dictionary. 2nd edition. [available on-line], The Times and The Sunday Times 1993 on CD-ROM. Distributed by Chadwyck-Healey, produced by FT Profile. Internet Stanford University, Stanford, California Academic Text Service. 〈[email protected]〉

PARALLELISM VS. ASYMMETRY IN COUNTERFACTUALS

327

University of Michigan Middle English: 〈http://www.hti.umich.edu/english/mideng〉 Early Modern English: 〈http://www.hti.umich.edu/dict/memem〉 University of Virginia Middle English: 〈http://etext.lib.virginia.edu/eng-on.html〉 Modern English: 〈http://etext.lib.virginia.edu/modeng/modeng0.browsw.html〉 Henry Fielding Tom Jones 〈gopher://gopher.vt.edu: 10010/02/82/1〉

References Bailey, R. W. 1991. Images of English. A Cultural History of the Language. Cambridge: Cambridge University Press. Behaghel, O. 1923–32. Deutsche Syntax: eine geschichtliche Darstellung, 4 vols. Heidelberg: Winter. Behre, F. 1934. The Subjunctive in Old English Poetry. Göteborg: Elanders Boktryckeri Aktiebolag. Bolinger, D. 1977. Meaning and Form. London: Longman. Boyland, J. T. 1995. “A corpus study of would + have + past participle constructions in English: Grammaticalization in progress?” Paper presented at 12 ICHL in Manchester, England in August 1995. Campbell, A. 1959. Old English Grammar. Oxford: Clarendon Press. Dahl, Ö. 1997. “The relation between past time reference and counterfactuality: A new look”. In On Conditionals Again, A. Athanasiadou and R. Dirven (eds), 97–114. Amsterdam/Philadelphia: John Benjamins. Dancygier, B. and Sweetser E. 1994. Conditionals: reassessing structure-function relationship. Unpublished Ms., University of California at Berkeley. Denison, D. 1993. English Historical Syntax. London: Longman. Denison, D. 1998. “Syntax”. In The Cambridge History of the English Language, vol. IV, S. Romaine (ed.), 92–329. Cambridge: Cambridge University Press. Fillmore, Ch. J. 1990. “Epistemic stance and grammatical form in English conditional sentences”. In Chicago Linguistic Society 26, M. Ziolkowski, M. Noske and K. Deaton (eds): 137–162. Fischer, O. 1992. “Syntax”. In The Cambridge History of the English Language, vol. II, N. Blake (ed.), 207–408. Cambridge: Cambridge University Press. Haiman, J. 1985. Natural Syntax. Cambridge: Cambridge University Press. Harris, M. and Vincent, N. (eds). 1988. The Romance Languages. London: Routledge. James, D. 1982. “Past tense and the hypothetical. A cross-linguistic study”. Studies in Language 6.3: 375–403. Kytö, M. (ed.). 1993. Manual to the Diachronic Part of the Helsinki Corpus of English Texts: Coding Conventions and Lists of Source Texts [2nd edition]. Helsinki: University of Helsinki.

328

RAFAŁ MOLENCKI

Mitchell, B. 1985. Old English Syntax. 2 vols. Oxford: Clarendon Press. Molencki, R. 1998. “Modals in past counterfactual protases”. In English Historical Linguistics 1996, J. Fisiak and M. Krygier (eds), 241–251. Berlin/New York: Mouton de Gruyter. Molencki, R. 1999. A History of English Counterfactuals. Katowice: Wydawnictwo Uniwersytetu S´la˛skiego. Molencki, R. forthcoming. “Counterfactuals in the different manuscripts of the Cursor Mundi”. Neuphilologische Mitteilungen 101. Murane, E. 1974. Daga Grammar. Norman/Oklahoma: SIL. Mustanoja, T. 1960. Middle English Syntax. Mémoires de la Société Néophilologique de Helsinki 23. Helsinki: Société Néophilologique. Penny, R. 1991. A History of the Spanish Language. Cambridge: Cambridge University Press. Quirk, R., Greenbaum, S., Leech, G. and Svartvik, J. 1985. A Comprehensive Grammar of the English Language. London: Longman. Rissanen, M. 1999. “Syntax” In The Cambridge History of the English Language, vol. III, R. Lass (ed.), 187–331. Cambridge: Cambridge University Press. Simpson, J. A., Simon, E. and Weiner, Ch. (eds). 1989. Oxford English Dictionary [2nd edition]. Oxford: Clarendon Press. Swan, M. 1995. Practical English Usage [2nd edition]. London: Longman. Visser, F. Th. 1963–1973. An Historical Syntax of the English Language [3 parts in 4 vols]. Leiden: E. J. Brill. Wekker, H. 1991. “Modality, creolization and semantic transparency” In Problems in the Modality of Natural Language, P. Kakietek (ed.), 137–145. Opole: The Pedagogical University of Opole. Wright, J. 1954. A Grammar of the Gothic Language [2nd edition revised by O. L. Sayce]. Oxford: Clarendon Press.

The grammaticalization of the present perfect in English Tracks of change and continuity in a linguistic enclave Sali A. Tagliamonte University of York

1.

Introduction

In this paper, I examine the competition of variant forms used in   contexts and the multiple constraints conditioning them. The analysis is innovative in that it draws on data from a unique socio-cultural context — Samaná English — a classic linguistic enclave, or relic area (Anttila 1989: 294; Hock 1986: 442). Such areas, because of their peripheral geographic location or isolated social and/or political circumstances, tend to preserve older features, allowing for a reconstruction of the past. The results of the analysis shed light on the longitudinal development of the present perfect in English and provide some steps toward a fuller integration of synchronic sociolinguistic research using a variationist approach, into research on grammaticalization (cf. Hopper and Traugott 1993: 30). Linguistic change which results in grammaticalization can be identified by the evolution of lexical items into grammatical forms (Meillet 1912: 131) and by the fact that it is subject to certain general processes and mechanisms of change (Traugott and Heine 1991: 3). Such an evolution may go on for centuries, as linguistic forms or structures pass through a long series of transitions forming a path or trajectory. Hopper (1991: 22–31) has put forward a number of ‘heuristic’ diagnostics which identify stages in the development in such grammaticalization pathways, i.e. from less to more grammaticalized. In the first instance, such a process involves a good deal of variation due to the fact that emerging forms may co-exist with, and interact with an already existing layer of functionally equivalent ones (Hopper and Traugott 1993: 22) — the principle of ‘layering’.

330

SALI A. TAGLIAMONTE

This variability can provide a dynamic representation of the different degrees of grammaticalization attained by the different forms. As grammaticalization proceeds, the number of formal choices gets smaller and the survivors take on more general grammatical meanings. This is identified as ‘specialization’ (Hopper and Traugott 1993: 25). At the same time the meaning and function of a grammatical form is necessarily linked to its lexical past. Thus, some traces of the original meaning of a grammaticalizing form tend to remain. This is known as ‘persistence’ (Hopper and Traugott 1993: 28). At any stage, however, linguistic peculiarities (e.g. semantic associations, collocation patterns, the effect of one linguistic context or another, etc.) can be correlated with the evolving grammatical morphemes in order to provide crucial keys to viewing the diachronic process of grammaticalization in synchronic data (Traugott and Heine 1991: 6). The present perfect presents a particularly opportune place for exploring these aspects of grammaticalization. In contemporary prescriptive English grammar the present perfect is typically equated with the morphosyntactic construction have + past participle, as exemplified with data from Samaná English in (1). (1)

Auxiliary have + past participle: a. Some of them have regretted it already. Yes, many of ‘em have regret it already. (006/171–173)1 b. That was the first they learnt me and I’m old and it have remained here. (002/115–6) c. I have sold two bale of yams for three dollars. … I’ve sold seven grain for a cent. (021/715–17) d. Many of ‘em have died. (003/015) e. It been so long I’ve forgotten. (020/87) f. No, they don’t know that they immigrants. Because the old people have died out.

However, the present perfect is widely attested as being interchangeable with the , both in contemporary (Elsness 1996: 237–335; Mencken 1962: 533; Vanneck 1955: 237) as well as in historical varieties of English (Brunner 1963: 86; Elsness 1996: 79–229; Fridén 1948: 27; Mitchell 1985: 274; Strang 1970: 149; Traugott 1992: 190), as in (2). (2)

Preterite: a. They all died out already. (013/80) b. But I don’t know what took her now. (015/245) c. They didn’t send it to me yet. (022/390) d. They killed a lot that time, yes they killed a lot. But today we calmed off and everything and got calm. (002/116)

THE GRAMMATICALIZATION OF THE PRESENT PERFECT

331

Moreover, world-wide reports of contemporary vernacular English usage attest to the presence of additional forms for the same function:2 lone past participles,3 as in (3), pre-verbal been, as in (4), pre-verbal done, as in (5), auxiliary be (instead of have) as in (6), entirely unmarked verbs (referred to here as verb stems) as in (7), ain’t, as in (8), and even a construction that Visser (1970: 2209) refers to as a “three-verb cluster”, as in (9). (3)

Lone past participles: a. She never been a person to walk (002/349) b. They called him yonder. He been there, I don’t know for what. (001/189) c. She gone to San Martin. (005/114)

(4)

Pre-verbal been: a. She been married. b. We all been raised up speaking English. c. They been fixing the road.

(5)

Pre-verbal done: a. Plenty done gone and they’s lose their life. b. I done been to Miami, Hollywood … c. So much trouble done pass.

(QFN/112) (010/599) (015/221) (005/476) (010/1032) (002/113–4)

(6)

Auxiliary be + past participle: a. I’m pass a lot of trouble. (002/374) b. This country is all descended from the old immigrants. (005/17) c. I’m forgot all them things. (015/257) d. If a child is taken measles, and you don’t know … you get that and steep it and give them that the next morning. (002/523)

(7)

Verb stems: a. I had twelve children, but only twelve of ‘em what come to be mothers and fathers of children. (006/445) b. I never like the city. (013/113) c. I never been in one of that; everything I look from far. (005/584) d. All them die away. (005/663)

(8)

Ain’t + past participle/past tense: a. He ain’t wrote yet. b. She ain’t married no-one yet. c. I ain’t got nothing to do.

(019/236–7) (005/160) (011/1143)

332

SALI A. TAGLIAMONTE

d. e. (9)

Since we move, I ain’t had to light it once …. (019/529–30) I came in last Friday, and I ain’t been nowhere (002/1339)

Three verb cluster with auxiliary be: a. They ain’t paid us yet and I’m done spent plenty money with the documents. (006/155–6) b. I’m done been over there plenty, but I don’t like it. (005/312–3)

This variability in present perfect forms has been noted extensively in contemporary dialects (Abbott 1957; Alexander 1926; Fries 1940: 59–71; Kallen 1989, 1990; Marckwardt 1958: 148–150; McDavid and McDavid 1986; Mencken 1962: 525–542; Noseworthy 1972). Explanations typically invoke the loss of the auxiliary (Menner 1926: 238; Wright 1898–1905: 298), however some also claim there is an encroachment of the  on the preterite (Menner 1926: 238), loss of the distinction between preterite and perfect (Vanneck 1955: 234), and even the use of forms for an entirely different function (e.g. a generalized past marker). Since variability seems robust, the precise stage of development of the present perfect either with respect to its meaning or its degree of grammaticalization remains an open question. A longitudinal approach to the present perfect in English is crucial in order to contextualize this variability. Historical examinations of it are by now extensive in the literature (e.g. Brinton 1988, 1994; Curme 1977; Denison 1993; Elsness 1996; Fridén 1948; Mitchell 1985; Rydén and Brorström 1987; Traugott 1992; Visser 1970). All of them report extensive variability in form and function since the Old English period. Similar variability in the perfect has been documented in many other languages where somewhat parallel trajectories of grammaticalization are attested, including modern Greek, Latin, the Germanic languages (Traugott 1972), and Romance (Harris 1982; Vincent 1982). This suggests that this area of the grammar may also be a fruitful location for further examination with respect to the putative universal pathways of grammaticalization, as outlined in Bybee, Perkins and Pagliuca (1994).4 In the present paper I approach the extensive variability of present perfect constructions in Samaná English, assuming that their patterning can be used to infer some principles underlying their uses. I then interpret the results from the perspective of grammaticalization theory, taking into account diachronic linguistic information in the literature.

THE GRAMMATICALIZATION OF THE PRESENT PERFECT

2.

333

The data

Samaná is a small village in a remote area of the Dominican Republic, where an offshoot of African American Vernacular English has been spoken since the mid-nineteenth century. The corpus comprises approximately 160,000 words from informal interviews conducted with 21 speakers from the oldest generation of the community in 1981, all of whom were native, mostly monolingual, speakers of English in the midst of a Hispanic majority (Poplack and Sankoff 1987). The sociohistorical circumstances under which these people had retained their variety of English include the well-established conditions for longitudinal continuity, including physical contact, frequency of interaction, prestige and other factors (Pousada and Poplack 1982; Thomason and Kaufman 1988). Previous linguistic research in a number of areas of the tense/aspect system (Poplack and Sankoff 1987; Poplack and Tagliamonte 1989; Tagliamonte 1991; Tagliamonte and Poplack 1988) and on three features of nominal morpohology (Poplack and Tagliamonte 1994; Tagliamonte and Poplack 1993; Tagliamonte, Poplack and Eze 1997) have led to the conclusion that Samaná English represents African American English as it was spoken in the mid-nineteenth century and that it has not been affected by Spanish as spoken on the greater island. The connection has been used to argue against the creole origins hypothesis of African American English and to show that the latter variety has changed little since the nineteenth century. Comparison of these data with other communities populated by African Americans which did not evolve in a Hispanic context, i.e. Nova Scotia, Canada, (Poplack and Tagliamonte 1991, 1994), as well as other non-standard English vernaculars, have led to the conclusion that the patterns found in Samaná can be traced to non-standard varieties of English and that they developed by the same processes which mark the history of the English language.5 Samaná English, illustrated in (1)–(9), exhibits an extensive inventory of present perfect constructions. None of these strategies are unique to Samaná English, as all of them are attested in widely-separated contemporary English vernaculars, e.g. United States (Christian, Wolfram and Dube 1988: 85–109; Feagin 1979: 81–89), Newfoundland (Noseworthy 1972: 19–24), Tristan da Cunha (Scur 1974), the Shetland Islands (Melchers 1992), and Ireland (Filppula 1996; Harris 1984; Kallen 1989, 1990). In what follows I attempt to explain this robust variability. Is it indicative of aberrant rendition of the English present perfect? Does it herald the incipient demise of the present perfect in this community? Can it be explained as continuity of earlier patterns in English? Might the variability reflect a stage in the grammaticalization of the present perfect with the

334

SALI A. TAGLIAMONTE

characteristic patterning typical of this process? First, I turn to a consideration of these data from an historical perspective.

3.

Historical development of the perfect in English

In Old English, there were only two tenses: past and non-past. Non-past served for durative and non-durative present and future reference. The past covered everything else — not only what is represented by the simple past of today, but also durative past time (e.g. past progressive), as well as the present perfect and   of the contemporary system (Strang 1970: 311). In fact, most commentators acknowledge that the simple  and present perfect were interchangeable in most contexts, including those where either one, or the other alone would be required in contemporary usage. In one recent discussion, Denison (1993: 352) provides examples in which the present perfect “appears to be commuting with a simple past” based on parallelism with preterite usage and adverbial collocation patterns. This can be seen in (10a), where the simple past tense inflection marks a function that today would be expressed with the auxiliary + past participle/past tense form, attested in the present perfect. The Old English data is comparable with (10b–c) from Samaná, where there is an inflected preterite form of the verb, and is also suggested in (10d–e).6 (10)

a.

b. c. d. e.

Fæder min, se tima com Father mine, that time came. ‘Father, the time has come.’ (Traugott 1992: 183) She had plenty time yonder and now she came to the capital to live. (002/447) God left me here for some purpose. (002/390) I come from that now. (006/1868) The crop just come in now. (004/178)

During the change from Old to Middle English this two-tense (past vs. non-past) inflectional verb system underwent substantial elaboration (Strang 1970: 98). 3.1 Elaboration of the verb phrase One of the most important changes that took place was the development of analytic morphology within the verb phrase, in addition to the suffixal inflection on the main verb, to mark tense and/or aspect distinctions. This development was in contrast with the original, and far more general, distribution of past tense

THE GRAMMATICALIZATION OF THE PRESENT PERFECT

335

forms. A number of additional periphrastic constructions came to alternate with the preterite. One of the morphosyntactic developments during this period was the emergence of the present perfect tense. While periphrastic constructions with the verb habban were present in Old English (Mitchell 1985: 723), the characteristic perfect functional distinctions did not become well established until the Middle English period and beyond (Denison 1993: 352; Mitchell 1985: 372; Visser 1970: 2191). The development of the contemporary functions, leading to fixed word order, aspectual distinctions and specific collocation patterns, etc., only gradually evolved over the ensuing centuries (see e.g. Elsness 1996). According to some commentators the rise of the modern perfect can be described as “gradual diffusion into more and more linguistic environments”, (Harris 1984: 322), though not necessarily to the same extent in all varieties of English. 3.2 The use of the verb have The original source of present perfect is traced to the stative main verb have (meaning ‘possess’), as in (11a), which was re-analyzed as an auxiliary verb combining with a transitive main verb, as in (11b) (e.g. Fridén 1948: 40–41; Jespersen 1909/1949: IV 29–30; Mitchell 1985: § 724, 726–8; Mustanoja 1960: 500–501; Traugott 1992: 192; Visser 1970: 2189–2192, § 2001–2003): (11)

a. b.

I [VP [V have] [[NP the letter] [A written] (i.e. in a written state) I [VP [have [Vwritten [NP the letter]]

Although the first attestation of have + past participle of an intransitive verb has been documented to 1096, the development of the contemporary perfect category is traced to the Middle English period, when have started spreading to other VP types. During the same time period have and be are said to have competed as auxiliaries for the new category (Denison 1993: 358; Fridén 1948: 30–117; Strang 1970: 149), as in (12). (12)

a. b.

He took his wyf to kepe whan he is gon vs. and also to han gon to solitarie exil The yonge sonne hath in the Ram his halfe cours yronne vs. as rody and bright as dooth the yonge sonne that in the Ram is foure degrees up ronne (All from Chaucer, (Brunner 1963: 87))

Note how (12a–b) compare with (12c) from Samaná:

336

SALI A. TAGLIAMONTE

c.

I’m never been in prison. Never in my life, I’ve been in prison. (021/998–9)

A number of commentators mention that the auxiliary be tended to be used with ‘mutative’ verbs (Denison 1993: 359; Mustanoja 1960: 500).7 Notice how (13a) from Samaná compares with (13b) reported in Fridén (1948: 112). (13)

a. b.

The town is changed. The town Samaná is not Samaná no more. (003/467) These days are now chaunged. (Spenser Faerie 4.4.I.3)

However, have gradually generalized to more and more verbs and eventually prevailed over be (Curme 1977: 359). Denison (1993: 344) describes the history of perfect be in English as a “continuous retreat in the face of the advancing have perfect.” 3.3 Three verb clusters During the Middle English period a “three-verb structure” (Visser 1970: 2209) also developed, as illustrated in (14). (14)

And many other false abusion. The Paip (= Pope) hes done invent. (Traugott 1972: 146)

While there are no examples of a three-verb cluster with have in Samaná, there are four examples with auxiliary be, as in (15) (see also (9)), which appear in the same discourse contexts as have. (15)

I have seen two times. The time pass and the time now. … I’m done bought meat twelve cent a pound. … And I’m done bought fish four cents a pound. … I have bought a cow for eight dollars. (021/421)

3.4 Unmarked participles Traugott (1972: 146) also notes that the main verb in the three-verb cluster may not receive an overt inflection, as in (14) above. In Samaná English the main verb may also occur without the inflection typical of the past participle, as in (16): (16)

a. b.

And he told me that he had done pass through them English books. (006/314) But the wind and the rain has wash them away. (020/262–4)

THE GRAMMATICALIZATION OF THE PRESENT PERFECT

c.

337

There are many others that have pass through normal school and they speaks their English correct. (006/340)

3.5 Lone past participle By the mid-nineteenth century attestations of a lone past participle in present perfect contexts (Visser 1970: 1298), as in (17a–b), become prevalent, although they apparently existed earlier in the nineteenth century, as in (17c–d) (Denison 1998). Visser (1970: 1298) suggests they may be forerunners of the contemporary constructions; however, they may also be part of a broader phenomenon in which a string may be ellipted, most often in clause-initial position (Denison 1998). They were plentiful in many early twentieth century English dialects, both American (Atwood 1953: 43), and British (Wright 1898–1905: 298). Compare Samaná English where the same lexical verbs appear as lone past participle,8 as in (18), (see also (3a–c)), and occasionally with a lexically unrealized subject at the beginning of a clause, as in (18c). (17)

a.

b. c. d. (18)

a. b. c.

They never got nothing but fourteen shilling … and I seen um both a-hanging in chains by Wisbeach river. [c. 1870 Ch. Kingsley, Alton Locke (Collines) 235] I’m better than the best collect he ever done business with. [c. 1872 Shaw, Widower’s Houses 2.33] Mrs. Novello seen Altam and his Wife? [c. 1818 Keats, Letters 98 p. 254 (18 Dec.)] (Denison 1998: 143) I am an old man … and many wonders seen and heard. [c. 1300 Arth. and M. 2049] I never seen him. (001/919) They done plenty things. (P/206) I can’t remember, been so long that I’ve forgotten.

3.6 Summary A general overview of the relevant diachrony of the present perfect can be summarized in five main points: 1.

The present perfect has been realized variably by preterite and perfect surface forms in English since its inception (Brunner 1963; Denison 1993; Fridén 1948; Menner 1926; Strang 1970; Traugott 1972; Vanneck 1955; Visser 1970).

338 2.

3.

4.

5.

4.

SALI A. TAGLIAMONTE

Variability in these forms was far more extensive during the Middle English period than in the contemporary standard system (Brunner 1963; Elsness 1996; Fridén 1948; Rydén and Brorström 1987; Strang 1970; Traugott 1972; Visser 1970). The development of analytic morphology in the verb phrase during the Middle English period led to competition among auxiliaries be, have, and done to signal present perfect. Much of this variability still persists in nonstandard varieties of contemporary English in North America, in the United Kingdom and undoubtedly elsewhere (Christian et al. 1988; Curme 1977; Kallen 1989; 1990; Melchers 1992; Scur 1974; Visser 1970). Lone past participles, specifically the three verbs been, done, and seen, have been attested, particularly since the early nineteenth century, but may have originated much earlier (Visser 1970). Inflection on the main verb was variable in Middle English, at least in the three verb cluster (Traugott 1972; Visser 1970).

Method

An important methodological issue for an analysis of the present perfect is to identify all the contexts in which its functions are met in discourse. This is critical in this case where form/function asymmetry is so extensively documented. In any given corpus there may be contexts which embody the functions of the present perfect, but which are not rendered by the standard morphosyntactic construction. There may also be contexts in which the standard present perfect morphology occurs, but which do not meet the semantic function of the contemporary system. Finally, it is important to keep in mind that preterite forms will appear extensively in past temporal reference contexts more generally, but only a small proportion of these will have a present perfect reading. In order to identify the specific contexts of present perfect I adopt the criteria for which there is general consensus in the current literature on this subject, namely the three basic semantic/pragmatic functions: , , and  (Bauer 1970; Brinton 1988: 10–15; Comrie 1976: 52–65; Fenn 1987: 100–131; Leech 1987: 30–49; McCawley 1971; Zandvoort 1932). Basically, the   makes reference to a time span throughout which an event or situation obtained, as exemplified earlier in (1a–b); the   refers to a situation which has occurred once or repeatedly before the present, as in (1c–d); and the   refers to a past situation that has led to some present result or state, as in (1e-f).

THE GRAMMATICALIZATION OF THE PRESENT PERFECT

339

All these distinctions have been subsumed under a unifying category labeled the present perfect in English, which identifies some state at speech time while at the same time involving the experience of past events (Fenn 1987: 136). Innovative in this research is the fact that the analysis is circumscribed to discourse contexts which meet the functional definition of the present perfect.9 There were 372 such contexts in the Samaná corpus. Each context was categorized for a number of features of the linguistic environment claimed to influence the occurrence of the present perfect from the prescriptive and descriptive literature on this subject — temporal distance, lexical aspect, temporal disambiguation, collocation patterns and subject noun. I then proceeded with statistical examination of the data in order to isolate the factors which determine the choice of forms in different contexts (see also Elsness 1996; Rydén and Brorström 1987). In this study, I employed multivariate analysis, a technique which enables the analyst to model the multiple constraints on the variability in the data. The advantage of this approach is that it considers all the relevant parameters simultaneously and calculates which ones are significant, to what degree, and perhaps most importantly, the relative patterns of variability across the categories in each factor.10 Such a method is particularly useful in the case of the present perfect where, because of the extensive overlap between different forms, isolated examples of one tendency or the other can easily be found (cf. Mitchell 1985: 298). Here, I focus on the tendencies embodied in the data as revealed by the distribution of forms and the hierarchy of constraints for each of the linguistic factors under investigation.

5.

Results

5.1 Distributional analysis Table 1 shows the overall distribution of forms used in present perfect contexts in the Samaná corpus.11 Note the robust variability amongst preterite morphology, 25%, the have + past participle construction, 22%, verb stems, 17%, lone past participles, 11% and auxiliary be, 10%. The relatively balanced frequencies of perfect and preterite forms, which may be surprising given contemporary prescriptive grammar, is entirely consistent with earlier stages in the development of the present perfect, when alternation between preterite and perfect forms was more fluid and where alternations of have and be as auxiliaries and even multiple auxiliaries are amply attested. Indeed, the localized occurrence of this particular

340

SALI A. TAGLIAMONTE

Table 1. Overall distribution of forms in present perfect contexts in Samaná English Surface morphology:

N

%

-ed/suppletion have/has/’s + past participle verb stem lone past participle be + past participle ain’t + past participle done + past participle used to, would, V-ing etc. three-verb cluster with be had + past participle was/got passive unambiguous present three-verb cluster with had

93 82 64 40 37 26 08 08 04 04 03 02 01

25 22 17 11 10 07 02 02 01 01 01 00.5 00.3

Total: N

372

cohort of forms is consistent with the five points outlined in Section 3.6 regarding the history of the perfect.12 The extent of variability within a highly circumscribed area of the grammar in Samaná English thus presents a classic case of grammatical ‘layering’ — whereby a set of forms co-exist within the same functional domain (Hopper 1991: 23).13 However, this view of the data only provides us with the relative proportions of forms used in present perfect contexts. As Hopper (1991: 23) points out, layering can represent a transition from one phase of a developmental trajectory to another. If so, what might the patterns of variability in this enclave tell us about linguistic change within the present perfect? Specifically, can the factors which determine the choice of forms shed light on their status and relative degree of grammaticalization?14 5.2 Multivariate analysis In order to assess the grammatical function and/or functions of the forms used in present perfect contexts in Samaná English, each token was coded for a number of linguistic features. These operationalize, where possible, a number of observations that have been made regarding the developmental stages of the present

THE GRAMMATICALIZATION OF THE PRESENT PERFECT

341

perfect in the literature and test hypotheses made about the distribution of forms in contemporary varieties, as will be detailed below. Table 2 shows the results of six independent multivariate analyses of the contribution of the five factors (viz.  ,  ,  ,   never, and  ₎ to the probability that each of have, been, ain’t, be, -ed/or suppletion, and a verb stem will be used in a present perfect context.15 Table 2 shows that the two most important factors which underlie the use of forms in present perfect contexts in Samaná English are temporal distance and verbal aspect. One or the other (or both) are selected as significant to all forms.16 5.2.1 Temporal distance The most fundamental characteristic of present perfect appears to be its association with present time. Originally, the simple past tense had covered all aspects of past time, including reference which brought past time into relation with the present (Curme 1977: 358). But the gradual shift in emphasis to explicitly past time that was accorded the preterite led to the need for a distinct strategy to express relations between present and past time (Curme 1977: 358). In English there is no tense which is specifically associated with a particular point in time. Indeed, in English, differential location in time is not typically associated with any tense, except the present perfect, where it can be predicted to occur under conditions of recency and current relevance (Dahl 1984: 118). In Table 2, the constraint hierarchy of the factor of temporal distance reveals that the closer, or more associated, the event is to present time, the higher the probability that have or been will occur.17 However, the predominant pattern overall is the propensity for use of have, been, ain’t, and be in present relevance contexts, all of which are favoured here at .59, .68, .64 and .56 respectively (shaded in Table 2).18 The fact that the favouring effect of ‘continuing in present’ contexts is identical across perfect forms, even where this factor group is not selected as statistically significant, is actually testimony to the similarities among them. In contrast, preterite morphology and the verb stem pattern similarly to each other, and in exactly the opposite way. Both -ed/suppletion and the verb stem are favoured in contexts that refer to specifically past time. This distributional patterning suggests that the two may be associated with the same semantic function. The pattern of the variants have/been/be contrasts markedly with -ed/suppletion and the verb stems (see shading). This suggests that the evolution of the tense system in Samaná English has reached a stage in which the newer layer of present perfect forms is taking over a present relevance function from the more

[.33] [.52]

[.48] [.55]

Collocation with never never unspecified Range

Subject noun pronoun NP Range [.51] [.48]

.89 .44 45

[.61] [.47]

[.37] [.67] [.64]

.31 .35 .68 37

.59 .28 31

[.16[ [.55[

.70 .45 25

[.48} [.28} [60}

.40 .37 .64 27

.07

ain’t

Perfect forms been/seen/ gone/done .08

Factors not selected for any analysis: Subject noun.

[.36] [.53]

Temporal disambiguation adverb unspecified Range

.35 .52 .59 24

Temporal distance more than a year ago within last year continuing in present Range

[.52] [.35] [52]

.22

Corrected mean:

Aspect dynamic mutative stative Range

have

Number of tokens: 372

[.48] [.54]

[.47] [.50]

[.47] [.51]

.41 .75 .55 34

[.50] [.49]

[.71] [.47]

[.60] [.48]

.61 .29 .41 32

.66 .53 .38 28

.23

.10 [.45] [.44] [.56]

-ed/suppletion

be

[.49] [.53]

[.44] [.51]

[.52] [.50]

[.48] [.64] [.49]

.55 .64 .39 25

.15

verb stem

Preterit forms

267 105

38 334

68 304

197 40 135

92 111 169

Ns/cell

Table 2. Multivariate analysis of the contribution of factors selected as significant to the probability of forms used in present perfect contexts in Samaná English

342 SALI A. TAGLIAMONTE

THE GRAMMATICALIZATION OF THE PRESENT PERFECT

343

general preterite. Given the time when Samaná English split from African American English in North America, we may assume that this development was already in place in the mid-nineteenth century. 5.2.2 Verbal aspect In the evolution of the present perfect, when the auxiliaries be and have were in variation, they are said to have been used in contexts with different aspectual readings. Recent research on the early grammaticalization of the present perfect (Carey 1994, 1995a, 1995b) has demonstrated varying frequencies amongst different types of stative verbs as the new tense category spread to more and more contexts.19 Mental state verbs (think, realize etc.) were the first to favour the new perfect auxiliary over perception verbs (see, hear etc.), and these two presumably before - (Carey 1994: 110–115). According to Rydén and Brorström’s (1987: 184, 186) detailed study of late Modern English (1700–1900), have, rather than be, is highly favoured in iterative contexts. Both studies reveal that at an earlier stage in the history of the present perfect the individual forms would have been highly sensitive to verbal aspect. Because the accepted functions of the perfect — in which continuative is distinct from resultative and experiental–are actually a conflation of aspect and pragmatic interpretation,20 verbs were categorized according to the kinds of situations they describe. Verbs representing states, either of emotion or attitude, sensory perception or bodily sensation, as in (19a–b) (Quirk et al. 1985: 94–96) were coded as . Verbs were coded as  when they make reference to events which occurred once (or several times) in the past, as in (19c–d). Verbs involving a transition from one place or condition to another (Fridén 1948: 57; Mustanoja 1960: 500), as in (20), were treated separately. (19)

a. b. c. d.

I have never been yonder. (004/277) You-all have heard about the Old Jericho? (003/101) I have passed a lot of little frights. (002/388) I have been told that if they know you handling money, they give another price. (010/400)

(20)

a.

The thing is increased so bad that we don’t know. At anytime somebody could come and even break open your door. (018/992) Now they have so many houses. They all is made it one thing. (003/480–2) ‘Cause them, now, since the war is got civilize. (018/747) My children, them is mixed up the language. (006/161)

b. c. d.

344

SALI A. TAGLIAMONTE

In Bybee et al.’s (1994: 68–69) account of the grammaticalization of the perfect, they project an evolution from an early stage in which have gradually replaces the be auxiliary and then expands to dynamic verbs of all kinds. Complete generalization of the grammaticalized construction is assumed when it comes to be used with stative, as well as dynamic predicates. Table 2 reveals that verbal aspect is indeed highly implicated in the array of present perfect forms and their distribution, but it is statistically significant for only one of the perfect forms — auxiliary be — which is highly favoured with mutative verbs at .75 (the underlined box in column under be). Indeed, Denison (1993: 359) notes that the majority of lexical verbs appearing with the be perfect, especially after the Old English period, are precisely these mutative verbs. Here, the Samaná English data provides a striking synchronic reflex of that earlier tendency. A similar propensity (though not statistically significant) can be seen for been at .67 (underlined), which may provide a hint at the functional characteristics of the latter construction. On the other hand, the verbs with preterite morphology exhibit a propensity to occur with dynamic (with a factor weight of .61) rather than with mutative or stative verbs (with a factor weight of .29 and .41 respectively). The verb stem is the only other form that is disfavoured with stative verbs (with a factor weight of .49) revealing another parallel with the preterite forms. However, unlike the preterite forms, it is highly favoured with mutative verbs suggesting a relatively circumscribed lexical context for its use. The perfect cohort in general — have, been, ain’t and be have a high propensity to occur in stative contexts with favouring factor weights of .52, .64, .60 and .55 respectively (shaded). However, one form of the perfect cohort — have — is also favoured with dynamic verbs (.52). As a matter of fact, its probability of use with dynamic and stative verbs is precisely the same: .52 vs. .52.21 The generalized use of have in Samaná English is consistent with the notion of bleaching which is said to occur in more fully grammaticalized forms and corroborates a progression toward a complete generalization of the constructions with have to stative as well as dynamic predicates indiscriminately (Bybee et al. 1994: 69).22 5.2.3 Temporal disambiguation In contemporary English grammar the present perfect is reportedly preferred over the preterite in contexts where there is a temporal adjunct which refers to an expanse of time, as in (21a), or those which refer to a period of time that stretches from a point in the past to the moment of speaking, as in (21b–c) (see Visser 1970: 2192). Increasing specialization of present perfect forms for

THE GRAMMATICALIZATION OF THE PRESENT PERFECT

345

situations which not only are located within a period which extends up to speech time, but which themselves extend up to that point is part of the longitudinal development of the present perfect in English (Elsness 1996: 359). (21)

a. b. c. d.

I got a long time, I ain’t been over there. (021/913) I’ve been teaching since 1924. (002/678) I have live here twenty-one years. … I came in the ‘61. (019/82) Some of them haven’t reached yet. (009/346)

In Table 2, however, temporal disambiguation is not selected as significant to the occurrence of any form in Samaná English, with the exception of ain’t. The general lack of significance of temporal adjuncts may be due to the fact that at an earlier stage in the development of the present perfect the collocation patterns typical of those reported for the modern system had not yet developed. Since adverbs may shift from use with preterite to use with perfect in the process of grammaticalization (Denison 1993: 366), these data may simply represent a stage in-between the two. The pronounced effect with ain’t, as in (22), may simply be a reflection of these earlier patterns with a form which is itself an archaic British dialect feature (Trudgill 1990: 97). (22)

a. b. c.

They ain’t paid us yet. He ain’t got long dead. I got a good while, I ain’t been over there.

(011/155) (021/1357) (021/911)

5.2.4 Collocation with never Interestingly, there is one particular collocation pattern which is attested from the Middle English period. When the present perfect was developing, constructions with never (which had apparently favoured the simple past earlier on [Denison 1993: 366; Fridén 1948: 31]) favoured the incoming present perfect morphemes, as in (23) [see also (3a), (7b–c)], (Visser 1970: 755). This was apparently due to the fact that the temporal specification of never made additional reference to time redundant (Jespersen 1990/1949: IV: 64). This disambiguation effect may explain the absence of the auxiliary verb in (23). (23)

a. b. c.

I never like the city. (014/984) (021/525) I never been to school. I never had to beat them [my children]. Only speak with them. (019/884)

346

SALI A. TAGLIAMONTE

Since use of the lexical item never was frequent in these materials, it could be treated separately. Table 2 also reveals that been is the only preterite form which is favoured in the context of never, with a factor weight of .89. For the preterite, the same constraint hierarchy is observed, but the factor is not significant. This suggests that the pattern is a latent one and corroborates the hypothesis that it is a receding pattern. For been on the other hand, collocation with never is the most significant factor influencing its selection, with a range of 45. Based on the strength of this effect in Samaná English, I surmise that the original association of never with the new perfect forms coupled with its additional temporal specification may have led to the collocation pattern with lone past participle constructions. Here, we may be observing a reflex of this older pattern. 5.2.5 Subject noun D’Eloia, (1973: 95) cites Wright (1898–1905: 298), who claims that many dialects of English use strong past participles with a ‘zero’ or deleted have in affirmative sentences in which the subject is a pronoun, viz., I done it, He been sick, etc. This claim was verified with these data; however, it turns out that ain’t is the only form for which subject noun is selected as significant. Here pronouns favour the use of ain’t at .59 while other NPs disfavour it at .28, paralleling the effect previously reported in the literature for lone past participles. There is no such effect for any other form in these data, the factor weights hover around .50, an indication of no tendency in either direction.23 5.2.6 Summary In sum, Table 2 provides a multidimensional picture of the present perfect system in Samaná English. The most salient observation that can be made about these findings is that the constraint ranking of the factors of temporal distance and aspect affect the process under consideration in basically the same way for each of the perfect markers have, been, ain’t, and be. However, this pattern differs for the preterite forms: -ed/suppletion and to a lesser extent the verb stem. This dramatic partitioning of forms reveals their differential patterning within the well-circumscribed domain of present perfect. While a number of frequent (have, -ed/suppletion) and infrequent (be, been, ain’t) forms are available for the same function, there is evidence for a significant degree of specialization amongst the old and new layers of grammatical strategies (i.e. preterite vs. perfect) in Samaná English. This is consistent with the empirical investigation of preterite and perfect reported by Elsness (1996: 355) for other varieties of English. According to him, there is extensive fluidity between present perfect and preterite forms through the history of English, but an essential opposition

THE GRAMMATICALIZATION OF THE PRESENT PERFECT

347

between the two is based on the different kinds of temporal reference they express and the different ways they combine with adverbial and other temporal specifiers. A similar conclusion is also reached by Kallen (1990: 1332), who suggests that the different forms of the modern Hiberno-English perfect are sensitive to parameters such as dynamic vs. stative, and recent vs. remote. The receding, or older construction with be (and possibly been), retains its propensity to appear with mutative verbs, exactly as hypothesized for an earlier stage in the history of the perfect. The tendency for the be auxiliary to exhibit this effect in Samaná English seems to be a retention of this earlier lexical disposition. An account of the specific status of the lone past participle and of the contemporary non-standard form ain’t is beyond the scope of this paper. However, the patterns exhibited here provide evidence that they are both bona fide participants in the English present perfect cohort, as they pattern consistently with the auxiliaries have and be on the two most significant linguistic measures. The form have, which rapidly gained ground over the other markers early on, clearly did so by spreading into more and more contexts, thus obliterating any propensity for one context over the other, hence the lack of significance between stative and dynamic verbs shown in Table 2. The extent to which this development has been arrested and is beginning to be reversed in present-day mainstream vernacular, particularly in the US (Elsness 1996: 358) provides another unique opportunity to explore the nature of the grammaticalization process.

6.

Discussion

We can now return to the question raised earlier, viz., what can these patterns of variability in the present perfect in the Samaná enclave tell us about processes of linguistic change? First, the forms in this semantic domain can undeniably be traced to corresponding forms already in place in earlier stages of English. The present analysis suggests that many of them were still in place in the midnineteenth century. More crucial, however, is the fact that the factors conditioning the choice between competing forms are not only consistent with the trajectory of development posited for the present perfect in English in the literature, but also elucidated by the principles of grammaticalization. Three are particularly relevant to the findings I have reported here: 1.

the principle of’ ‘layering’, whereby new forms are used in a functional domain and co-vary with older ones: During the emergence of the present perfect category, periphrastic perfect constructions came to alternate with preterite forms for the same perfect functions. To date, the former have not

348

2.

3.

SALI A. TAGLIAMONTE

completely ousted the latter from this domain. The Samaná data capture this aspect of the diachrony of English quite accurately. They also reveal the range of forms and constructions that have competed for the same function since the inception of the present perfect category, some of them under specific morphosyntactic constraints. the principle of ‘persistence’, whereby traces of the original lexical meanings of forms may be reflected in constraints on their grammatical distribution: This is observed in the statistically significant effect of verbal aspect, specifically with respect to the use of the moribund be auxiliary with mutative verbs, as well as in traces of this effect with the lone past participle been. However, it has been entirely bleached out of the have form. the principle of ‘specialization’, whereby as grammaticalization proceeds, a smaller number of devices (out of a wider initial range) are used more generally and specialize for the relevant grammatical function, ousting the other alternatives and reducing their systemic and/or statistical distribution: This is obvious in standard dialects of English, in which the have + past participle construction has prevailed as the expression of present perfect (even if it continues to compete with forms of the preterite). However, have + past participle constructions and preterite forms continue to stand out as dominant expressions of present perfect in Samaná English.

In conclusion, the linguistic evidence points to the fact that the synchronic slice represented by Samaná English is a striking reflection of an earlier point along the trajectory of grammaticalization of the present perfect. I suggest therefore, that it represents a retention of patterns in earlier stages of the English language. Because of its isolation, Samaná English has not participated in the ongoing development of forms and functions of the present perfect found in other mainstream varieties of English. Therefore, it provides a glimpse of what the state of that development may have been like at the time of its separation from varieties of North American English in the mid-nineteenth century. This illustrates some of the advantages of working with data from relic linguistic areas. This study also highlights the fact that the trajectory of historical change can still be observed in some contemporary dialects of the English language, not only with respect to the existence of forms, but also their relative proportion, and, interestingly, even in the variable constraints on the grammatical distribution of the forms. Moreover, the patterns of these conditioning factors can reveal the tracks of the grammaticalization process. Taken together they provide an in-depth view of a particular milestone along the pathway of development of the present perfect in English and provide another base-line for interpreting further points along the way.

THE GRAMMATICALIZATION OF THE PRESENT PERFECT

349

Acknowledgments I gratefully acknowledge the generous support of the Social Sciences and Humanities Research Council of Canada for research grants #410–90–0336 and #410–95–0778 which enabled me to conduct some of this research, and the use of the Samaná English corpus housed at the University of Ottawa Sociolinguistics Laboratory, Shana Poplack director. I would like to thank Salikoko Mufwene, three anonymous reviewers, and the editors for their valuable feedback on this work. All remaining shortcomings are my own responsibility.

Notes 1.

Codes in parentheses identify the speaker and coordinates of the example in the Samaná English corpus.

2.

In the Samaná English data there is no evidence for the have + NP + past participle order attested in earlier stages of English, see (Denison 1993: 343; Visser 1970: 2001) and in some contemporary dialects, e.g. Hiberno-English (Harris 1984; Kallen 1989, 1990).

3.

I use the term ‘lone past participle’ to refer to constructions which contain a past participle, but no auxiliary.

4.

This becomes a particularly interesting question given the recent conclusion of Elsness (1996: 362), which suggests that the development of the present perfect in English is different from developments observed in other European languages.

5.

See Poplack and Sankoff (1987) and Tagliamonte (1991) for a detailed description of the corpus and justification for its categorization as a linguistic enclave. For analyses and argumentation related to the questions of origins of the Samaná English tense/aspect system see (e.g. Poplack and Sankoff 1987; Tagliamonte and Poplack 1988; Tagliamonte and Poplack 1993). For discussion against the possible ‘creole’ origins of the present perfect forms in Samaná see Tagliamonte (1996, 1997).

6.

The form come in Samaná English, as in many non-standard varieties of English, is not uniquely a past participle, but also functions as the preterite form.

7.

Mitchell (1985: 302) points out, however, that habban is found in the earliest prose and poetry with ‘mutative’ verbs.

8.

Of course there are relatively few verbs in English which have a distinct past participle form. In most cases, absence of the present perfect auxiliary would produce a construction indistinguishable from the preterite.

9.

Had the present perfect forms been calculated as a basic frequency of all preterite and perfect forms (e.g. Elsness 1996), the frequencies of the perfect forms would have been much lower.

10.

For further discussion of the application of statistical modeling techniques in linguistics: Cedergren and Sankoff (1974); Sankoff (1978a, 1978b, 1982); Sankoff and Labov (1979); Sankoff and Rousseau (1979).

11.

Interestingly, the unambiguous present contexts, although rare, bear some of the characteristics of the ‘extended now’ perfect, (McCoard 1978), as in (a) which are comparable to those found in contemporary Hiberno-English, as in (b), (Kallen 1989: 21).

350

SALI A. TAGLIAMONTE a. b.

Since I have him, he goes to church. You’ve been doing that since I know you.

(009/126).

12.

These present perfect contexts actually represent a relatively small proportion of all past temporal reference contexts in the data, viz., 5%. It is important to point out, however, that the non-standard variants in Table 1 are largely circumscribed to contexts of the present perfect and never occur in contexts which require the preterite (see Tagliamonte 1996, 1997).

13.

The same type of layering of present perfect contexts has been reported for Hiberno-English, though with a somewhat different cohort of surface forms (Kallen 1989, 1990).

14.

In fact, as Denison (1993: 522) points out, grammaticalization “could reasonably be applied to various stages in the development of the perfect.”

15.

The paucity of some constructions, e.g. done + past participle, made quantitative analysis impossible.

16.

The multivariate analysis tables can be interpreted as follows: The numbers are ‘factor weights’, indicating the probability of the form occurring in the context indicated. The closer these numbers are to 1; the more highly favouring the effect is; the closer they are to zero the more disfavouring the effect is. The ‘corrected mean’ at the top of the table indicates the overall tendency of the marker to surface in the data. The ‘range’, indicated by the numbers in italics, represent the relative strength of the effect. The higher the numbers; the stronger the effect. All these calculations are computed by the step-wise selection procedure incorporated in the variable rule program (Rand and Sankoff 1990). Factor weights enclosed in square brackets indicate that that factor group was not selected as statistically significant to the choice of form.

17.

The ‘constraint hierarchy’ refers to the order from more to less amongst the factor weights within a given factor.

18.

Although temporal distance is not selected as significant for be this may well be due to the paucity of forms.

19.

Unfortunately, there is insufficient representation of these individual lexical verbs in these materials to conduct a quantitative analysis. In any case, such trends were undoubtedly too early in the grammaticalization of the perfect tense to be expected here.

20.

According to Tagliamonte and Poplack (1995), there is substantial interaction between the traditional semantic/pragmatic categories of perfect and lexcial stativity. Continuative perfects tend to be stative (74%); resultative perfects tend to be punctual (80%); only experiental perfects are distributed, albeit unevenly, across the two. This is why I have adopted the stative/dynamic/mutative categorization schema outlined here.

21.

Note, however, that have is still favoured in temporally proximate contexts where it has a probability of .52 (underlined in column under have). This points up its ongoing overlap with -ed/suppletion forms.

22.

The neutrality of have amongst a variety of other perfect forms is also reported in HibernoEnglish (Kallen 1989: 34).

23.

Interestingly, Elsness (1996: 97–201, 327–332) provides empirical quantitative evidence that first person subjects have favoured present perfect forms from Old English to present day English. Unfortunately, this was not coded in the present data set.

THE GRAMMATICALIZATION OF THE PRESENT PERFECT

351

References Abbott, O. L. 1957. “The preterit and past participle of strong verbs in seventeenthcentury American English”. American Speech 32 (1): 31–42. Alexander, H. 1926. “The verbs of the vulgarate in their historical relations”. American Speech 4 (4): 307–315. Anttila, R. 1989. An Introduction to Historical and Comparative Linguistics. Amsterdam/New York: John Benjamins. Atwood, E. B. 1953. A Survey of Verb Forms in the Eastern United States. Ann Arbor: University of Michigan Press. Bauer, G. 1970. “The English “perfect” reconsidered”. Journal of Linguistics 6 (2): 189–198. Brinton, L. J. 1988. The Development of English Aspectual Systems. Cambridge: Cambridge University Press. Brinton, L. J. 1994. “The differentiation of statives and perfects in early modern English: The development of the conclusive perfect”. In Towards a Standard English, 1600–1800 [Topics in English Linguistics], D. Stein and I. Tieken-Boon van Ostade (eds), 135–70. Berlin/New York: Mouton de Gruyter. Brunner, K. 1963. An Outline of Middle English Grammar, transl. by G. K. W. Johnston. Oxford: Blackwell. Bybee, J. L., Perkins, R. D. and Pagliuca, W. 1994. The Evolution of Grammar: Tense, Aspect, and Modality in the Languages of the World. Chicago: University of Chicago Press. Carey, K. 1994. “The Grammaticalization of the perfect in Old English”. In Perspectives on Grammaticalization, W. Pagliuca (ed.), 103–117. Amsterdam/Philadelphia: John Benjamins. Carey, K. 1995a. “From resultativity to current relevance: Evidence from the history of English and Modern Castilian Spanish”. In Conceptual Structure, Discourse and Language, A. E. Goldberg (ed.), 31–48. Stanford, California: CSLI. Carey, K. 1995b. “Subjectification and the development of the English perfect”. In Subjectivity and Subjectivisation in Language, D. Stein and S. Wright (eds), 83–102. Cambridge: Cambridge University Press. Cedergren, H. J. and Sankoff, D. 1974. “Variable rules: Performance as a statistical reflection of competence”. Language 50 (2): 333–355. Christian, D., Wolfram, W. and Dube, N. 1988. Variation and Change in Geographically Isolated Communities: Appalachian English and Ozark English. Tuscaloosa, Alabama: American Dialect Society. Comrie, B. 1976. Aspect: An Introduction to the Study of Verbal Aspect and Related Problems. Cambridge: Cambridge University Press. Curme, G. O. 1977. A Grammar of the English Language. Essex, Connecticut: Verbatim. D’Eloia, S. G. 1973. “Issues in the analysis of Nonstandard Negro English: A review of J. L. Dillard’s Black English: Its history and usage in the United States”. Journal of English Linguistics 7: 87–106.

352

SALI A. TAGLIAMONTE

Dahl, Ö. 1984. “Temporal distance: Remoteness distinctions in tense-aspect systems”. In Explanations for Language Universals, B. Butterworth, B. Comrie and Ö. Dahl (eds), 105–122. Berlin: Mouton de Gruyter. Denison, D. 1993. English Historical Syntax. Harlow, Essex: Longman. Denison, D. 1998. “Syntax”. In The Cambridge History of the English Language, 1776-present day, S. Romaine (ed.), 92–329. Cambridge: Cambridge University Press. Elsness, J. 1996. The Perfect and the Preterite in Contemporary and Earlier English. Berlin: Mouton de Gruyter. Feagin, C. 1979. Variation and Change in Alabama English: A Sociolinguistic Study of the White Community. Washington, D. C.: Georgetown University Press. Fenn, P. 1987. A Semantic and Pragmatic Examination of the English Perfect. Tübingen: Gunter Narr. Filppula, M. 1996. “Investigating the origins of Hiberno-English perfects: The case of PII”. In Speech Past and Present: Studies in English Dialectology in Memory of Ossi Ihalainen: University of Bamberg Studies in English Linguistics, J. Klemola, M. Kytö and M. Rissanen (eds), 33–55. Frankfurt am Main: Peter Lang. Fridén, G. 1948. Studies on the Tenses of the English Verb from Chaucer to Shakespeare with Special Reference to the Late Sixteenth Century. Uppsala: Almqvist & Wiksells Boktryckeri AB. Fries, C. C. 1940. American English Grammar. New York: Appleton, Century, Crofts. Harris, J. 1984. “Syntactic variation and dialect divergence”. Journal of Linguistics 20 (2): 303–327. Harris, M. 1982. “The ‘past simple’ and ‘present perfect’ in Romance”. In Studies in the Romance Verb, N. Vincent and M. Harris (eds), 42–70. London: Croom Helm. Hock, H. H. 1986. Principles of Historical Linguistics [Trends in Linguistics, Studies and Monographs 34]. Berlin: Mouton de Gruyter. Hopper, P. J. 1991. “On some principles of grammaticization”. In Approaches to Grammaticalization, volume 1: Focus on Theoretical and Methodological Issues [Typological Studies in Language], E. C. Traugott and B. Heine (eds), 17–35. Amsterdam and Philadelphia: John Benjamins. Hopper, P. J. and Traugott, E. C. 1993. Grammaticalization [Cambridge Textbooks in Linguistics]. Cambridge: Cambridge University Press. Jespersen, O. H. 1909/1949. A Modern English Grammar. Part VI: Morphology. London: George Allen & Unwin. Kallen, J. L. 1989. “Tense and aspect categories in Irish English”. English World-Wide 10: 1–39. Kallen, J. L. 1990. “The Hiberno-English perfect: Grammaticalization revisited”. Irish University Review 20 (1): 120–37. Leech, G. N. 1987. Meaning and the English Verb. New York: Longman. Marckwardt, A. H. 1958. American English. New York: Oxford University Press. McCawley, J. D. 1971. “Tense and time reference in English”. In Studies in Linguistic Semantics, C. J. Fillmore and D. T. Langendoen (eds), 97–113. New York: Holt, Rinehart and Winston.

THE GRAMMATICALIZATION OF THE PRESENT PERFECT

353

McCoard, R. W. 1978. The English Perfects: Tense-Choice and Pragmatic Inferences. Amsterdam: North-Holland. McDavid, R. I., Jr. and McDavid, V. 1986. “Kentucky verb forms”. In Language Variety in the South: Perspectives in Black and White, M. Montgomery and G. Bailey (eds), 264–283. University, Alabama: University of Alabama Press. Meillet, A. 1912. Linguistique Historique et Linguistique Générale. Paris: Champion. Melchers, G. 1992. “ ‘Du’s no heard da last o’ dis’ — on the use of BE as a perfective auxiliary in Shetland dialect”. In History of Englishes, M. Rissanen, O. Ihalainen, T. Nevalainen and I. Taavitsainen (eds), 602–610. Berlin: Mouton de Gruyter. Mencken, H. L. 1962. The American Language. New York: Alfred A. Knopf. Menner, R. J. 1926. “Verbs of the vulgate in their historical relations”. American Speech 1 (4): 230–240. Mitchell, B. 1985. Old English Syntax, vols I, II. Oxford: Clarendon Press. Mustanoja, T. F. 1960. A Middle English Syntax. Helsinki: Société Néophilologique. Noseworthy, R. G. 1972. “Verb usage in Grand Bank”. Regional Language Studies Newfoundland 4: 19–24. Poplack, S. and Sankoff, D. 1987. “The Philadelphia story in the Spanish Caribbean”. American Speech 62 (4): 291–314. Poplack, S. and Tagliamonte, S. 1989. “There’s no tense like the present: Verbal -s inflection in Early Black English”. Language Variation and Change 1 (1): 47–84. Poplack, S. and Tagliamonte, S. 1991. “African American English in the diaspora: The case of old-line Nova Scotians”. Language Variation and Change 3 (3): 301–339. Poplack, S. and Tagliamonte, S. 1994. “-S or nothing: Marking the plural in the African American diaspora”. American Speech 69 (3): 227–259. Pousada, A. and Poplack, S. 1982. “No case for convergence: The Puerto Rican Spanish verb system in a language-contact situation”. In Bilingual Education for Hispanic Students in the United States, J. A. Fishman and G. D. Keller (eds), 207–237. New York: Teachers College, Columbia University. Quirk, R., Greenbaum, S., Leech, G. and Svartvik, J. 1985. A Comprehensive Grammar of the English Language. New York: Longman. Rand, D. and Sankoff, D. 1990. GoldVarb: A Variable Rule Application for the Macintosh. Montreal, Canada: Centre de recherches mathématiques, Université de Montréal. Version 2. Rydén, M. and Brorström, S. 1987. The ‘be/have’ Variation with Instransitives in English: With Special Reference to the Late Modern Period. [Stockholm Studies in English 70]. Stockholm: Almqvist and Wiksell. Sankoff, D. (ed.). 1978a. Linguistic Variation: Models and Methods. New York: Academic Press. Sankoff, D. 1978b. “Probability and linguistic variation”. Synthèse 37: 217–238. Sankoff, D. 1982. “Sociolinguistic method and linguistic theory”. In Logic, Methodology, Philosophy of Science VI [Studies in Logic and the Foundations of Mathematics], L. J. Cohen, J. Los, H. Pfeiffer and K. P. Podewski (eds), 677–689. Amsterdam: North Holland & Warsaw: Polish Scientific.

354

SALI A. TAGLIAMONTE

Sankoff, D. and Labov, W. 1979. “On the uses of variable rules”. Language in Society 8 (2): 189–222. Sankoff, D. and Rousseau, P. 1979. “Categorical contexts and variable rules”. In Papers from the Scandinavian Symposium on Syntactic Variation, Stockholm, May 18–19, 1979, S. Jacobson (ed.), 7–22. Stockholm: Almqvist and Wiksell. Scur, G. S. 1974. “On the typology of some peculiarities of the perfect in the English of Tristan da Cunha”. Orbis 23 (2): 392–396. Strang, B. M. H. 1970. A History of English. London: Methuen. Tagliamonte, S. 1991. A matter of time: Past temporal reference verbal structures in Samaná English and the ex-slave recordings. Ph. D. dissertation, University of Ottawa. Tagliamonte, S. 1996. “Has it ever been PERFECT? Uncovering the grammar of early Black English”. York Papers in Linguistics 17: 351–396. Tagliamonte, S. 1997. “Obsolescence in the English perfect? Evidence from Samaná English”. American Speech 72 (1): 33–68. Tagliamonte, S. and Poplack, S. 1988. “How Black English past got to the present: Evidence from Samaná”. Language in Society 17 (4): 513–533. Tagliamonte, S. and Poplack, S. 1993. “The zero-marked verb: Testing the creole hypothesis”. Journal of Pidgin and Creole Languages 8 (2): 171–206. Tagliamonte, S. and Poplack, S. 1995. “Obsolescence in the English perfect: Evidence from African Nova Scotian English”, paper presented at American Dialect Society Meeting, special session on dialect obsolescence. Chicago, Illinois. Tagliamonte, S., Poplack, S. and Eze, E. 1997. “Pluralization patterns in Nigerian Pidgin English”. Journal of Pidgin and Creole Languages 12 (1): 103–129. Thomason, S. G. and Kaufman, T. 1988. Language Contact, Creolization and Genetic Linguistics. Berkeley /Los Angeles, California: University of California Press. Traugott, E. C. 1972. A History of English Syntax. A Transformational Approach to the History of English Sentence Structures. New York: Holt, Rinehart & Winston. Traugott, E. C. 1992. “Syntax”. In The Cambridge History of the English Language, R. M. Hogg (ed.), 168–289. Cambridge: Cambridge University Press. Traugott, E. C. and Heine, B. 1991. “Introduction”. In Approaches to Grammaticalization [Typological studies in language], E. C. Traugott and B. Heine (eds), 2–14. Amsterdam and Philadelphia: John Benjamins. Trudgill, P. 1990. The Dialects of England. Oxford: Blackwell. Vanneck, G. 1955. “The colloquial preterite in Modern American English”. Word 14: 237–42. Vincent, N. 1982. “The development of the auxiliaries HABERE and ESSE in Romance”. In Studies in the Romance Verb, N. Vincent and M. Harris (eds), 71–96. London: Croom Helm. Visser, F. T. 1970. An Historical Syntax of the English Language. Leiden: E. J. Brill. Wright, J. 1898–1905. The English Dialect Grammar. Oxford: Clarendon Press. Zandvoort, R. W. 1932. “On the perfect of experience”. English Studies 14 (1): 11–20.

Grammaticalization versus lexicalization ‘Methinks’ there is some confusion Ilse Wischer University of Potsdam

1.

Introduction

Grammaticalization and lexicalization are both processes of linguistic change codifying language material in different ways. While the former is concerned with the development of grammemes, the latter describes emerging lexemes. According to the different output there ought to be a clear distinction between the two processes. Some particular instances of linguistic change, however, seem to be difficult to classify as one or the other, especially in connection with the idiomaticization of syntagms, which take on a modal or discourse function. Methinks is such an example. In Early Modern English it developed from a free syntactic structure into a marker of evidentiality. On the basis of a detailed analysis of this process I will show that mechanisms of both grammaticalization and lexicalization are involved in the development of methinks, and that both processes in general are not at all contradictory. They exhibit many similarities, but operate on different levels of the language.

2.

Grammaticalization

Grammaticalization is generally considered to be a process that turns linguistic elements (lexical, pragmatic, sometimes even phonetic items)1 into grammatical elements, or that renders grammatical items still more grammatical. Traugott and Heine (1991: 1) define it as “the linguistic process, both through time and synchronically, of organization of categories and of coding”. Recently it turned out that there are two subtypes of grammaticalization. One of them, the more

356

ILSE WISCHER

traditional one, operates on the propositional level. It is based on Meillet’s assumption, which defines this process as “le passage d’un mot autonome au rôle d’élément grammatical” [the transition of an independent word into a grammatical element] (Meillet 1912: 131).2 The most influential approach in this respect is probably Lehmann’s (1985) article on processes and parameters of grammaticalization.3 Based on Givón (1979: 209) he characterizes the process of grammaticalization as one of cyclic waves passing through the following levels: (1)

discourse → syntax → morphology → morphophonemics → zero.4

Within this cycle a linguistic element can be more or less grammaticalized. Lehmann (1985: 306) argues that “the more freedom with which a sign is used, the more autonomous”, i.e., the less grammaticalized, it is. According to Lehmann (1985: 307–308), the degree of grammaticalization can be determined in terms of the following paradigmatic and syntagmatic parameters: 1. 2. 3. 4. 5. 6.

“attrition”: the gradual loss of semantic and phonological substance; “paradigmaticization”: the increasing integration of a syntactic element into a morphological paradigm; “obligatorification”: the choice of the linguistic item becomes rule-governed; “condensation”: the shrinking of scope; “coalescence”: once a syntactic unit has become morphological, it gains in bondedness and may even fuse with the constituent it governs; “fixation”: the item loses its syntagmatic variability and becomes a slot filler.

Whereas this first subtype of grammaticalization (subtype I) refers to the transformation of free syntactic units into highly constrained grammatical morphemes, which operate on the level of the proposition, the second subtype (subtype II) operates on the textual or discourse level and concerns the development of textual or discourse markers. The most influential work in this respect is probably Traugott (1982). According to her theory linguistic elements move in the following direction: (2)

proposition → text → discourse.

Again, an element can be more or less grammaticalized. But compared to subtype I, the processes involved are only partly the same. ‘Attrition’ occurs as phonetic reduction and semantic bleaching, the second generally being specified as subjectification, i.e., the foregrounded generalized meaning becomes more subjective, more expressive and less referential than the previous lexical meaning.

GRAMMATICALIZATION VERSUS LEXICALIZATION

357

In opposition to subtype I no ‘paradigmaticization’ seems to be necessary or even possible. Although the grammaticalized item increases in frequency, there is no strict ‘obligatorification’, i.e., the use of the textual or discourse marker has not become rule-governed in the sense that its application is the only choice in this particular context. In terms of syntagmatic relationships an extension instead of a shrinking of scope is most likely to be noticed, so that ‘coalescence’, or a fusion with the constituents situated within the scope of the textual or discourse marker, is not possible. In adopting a textual or discourse function, it prefers certain positions in the sentence, yet is not fixed to fill a particular slot (for this, see also Lenker, this volume). In a recent study of I think as an English modal particle, Aijmer (1997: 2) distinguishes between grammaticalization and pragmaticalization: “Grammaticalization is concerned with the derivation of grammatical forms and constructions (mood, aspect, tense, etc.) from words and lexicalized structures … Discourse markers such as you know, you see, etc., are typically ‘pragmaticalized’, since they involve the speaker’s attitude to the hearer.” However, she admits that “there are many similarities between the processes involved in pragmaticalization and grammaticalization” (Aijmer 1997: 6). I will consider them subtypes of grammaticalization. Both have in common that language material undergoes a process of recategorization, changing over from a more open categorial system to a closer one. For subtype I (grammaticalization on the propositional level) this implies that lexical categories, such as nouns, verbs, adjectives become grammatical categories: prepositions, conjunctions, auxiliaries, etc. These, again, can be further grammaticalized in the direction of affixes, which are integrated in closer morphological paradigms and their scope is reduced to the stem that they are attached to. They finally function as intra-propositional operators.5 In subtype II (grammaticalization on the text or discourse level), lexical material enters a relatively close categorial system of textual or discourse markers via further subjectification and extension of scope, resulting in extra-propositional or discourse operators. In both cases a lexical meaning turns into an operational meaning. This categorial change is accompanied by a process of phonetic reduction. Thus, grammaticalization is not merely a syntactic change, but a global one, including — apart from syntax — morphology, phonology, semantics and discourse as well.

358 3.

ILSE WISCHER

Lexicalization

Lexicalization, on the other hand, is a process less carefully and less systematically studied than grammaticalization. Thus, there is no general agreement about the use of the term itself. Some authors describe lexicalization as the process that turns free syntactic groups or ad-hoc-formations into lexical units by adding a specific semantic component, thus identifying it with idiomaticization,6 others consider idiomaticization a consequence of lexicalization.7 Some authors regard lexicalization as the transfer of any linguistic material into the lexicon of a language.8 As soon as something is stored as a unit in the lexicon, it is treated like a formula. According to this approach the development of any new lexeme is an instance of lexicalization. Keller (1995: 219) points out: “‘Lexikalisierung’ nennt man den Prozess, der darin besteht, daß sprachliche Ausdrücke Teil des Lexikons werden” [‘When linguistic elements become part of the lexicon, this process is called ‘lexicalization’’]. The metonymical or metaphorical use of a collocation becomes conventionalized. An ‘icon’ turns into a ‘symbol’, so that the semantic interpretation becomes rule-based.9 This symbolification of metonymies or metaphors is accompanied by different degrees of desyntacticization and demotivation. The whole process can be considered to be a kind of codification of language material, which is to be distinguished from a regular word formation process. As several mechanisms are involved in the process of lexicalization, not necessarily proceeding simultaneously, Bauer (1983: 50–61) distinguishes between different “types of lexicalization”: 1. 2. 3. 4. 5.

Changes of stress patterns and/or phonetic reductions are features of “phonological lexicalization” (cf. famous [ÁfeIm6s] — infamous [ÁInf6m6s]). Linking elements and/or non-productive roots or affixes are features of “morphological lexicalization” (cf. eat — edible). Lack of semantic compositionality is a feature of “semantic lexicalization” (cf. understand). Non-productive syntactic patterns and/or unusual functions of syntactic patterns are features of “syntactic lexicalization” (cf. pickpocket). Many examples are “mixed lexicalizations”, which can lead to a complete demotivation, so that the results have to be treated as simplex lexemes (cf. husband).

I will follow Keller’s view of lexicalization, but like Bauer, I consider it a gradual process. The types that he lists are obviously accompanying mechanisms, which can be more or less powerful. In the same sense as grammaticalization is

GRAMMATICALIZATION VERSUS LEXICALIZATION

359

defined as a process which turns linguistic elements into grammatical items, or renders grammatical items still more grammatical, I want to define lexicalization as the process that turns linguistic material into lexical items, i.e., into lexemes, and renders them still more lexical. Both processes, grammaticalization and lexicalization, are related to each other in different ways. Some scholars regard lexicalization as some kind of degrammaticalization, in a way that grammatical constructions lose their grammatical function and degenerate into idioms or lexical items (cf. e.g. Chen 1998). Others point out that grammaticalization can be a further development of lexicalization, in the sense of a stricter codification of the lexicalized item (cf. Keller 1995: 227).10 Obviously there is a close connection between the two processes, which might be due to the fact that in both cases similar mechanisms are at work: Der Übergang vom Ikon zum Symbol, [spielt] bei dem, was man “Lexikalisierung” nennt, eine entscheidende Rolle … [‘The transition of an icon into a symbol plays an important part in what is called “lexicalization” ’] (Keller 1995: 167–168) Grammaticalisation … is the gradual fusion of icons into symbols … (McMahon 1994: 172) Die Prozesse der Symbolwerdung von Metonymien und Metaphern faßt man gemeinhin zusammen unter der Bezeichnung “Lexikalisierung”. [‘The processes of symbolification of metonymies and metaphors are usually subsumed under the term “lexicalization”’] (Keller 1995: 183) … semantic change in the early stages of grammaticalization … usually involves specification achieved through inferencing …the inferencing is of two kinds: Metaphor and metonymy, … (Traugott and König 1991: 212)

Both processes involve symbolifications of icons, and these icons seem to be metaphorically or metonymically conditioned. There has always been a great deal of confusion between the two processes. Even Meillet provides examples of grammaticalization that clearly meet all the requirements of lexicalization. In example (3), phonological, morphological and semantic lexicalization have led to a complete demotivation and thus turned the former free syntactic unit into a lexical item. (3)

heute < *hiu tagu today < this- day- ‘this day’

(Meillet 1925: 25)

However, it must be admitted that the new ‘lexical’ unit belongs to a rather closed class of adverbials, which opens up the question whether it has really

360

ILSE WISCHER

become an element of the lexicon or indeed a grammatical item. This leads to the following two conclusions: a.

b.

4.

It is necessary to base the concept of grammaticalization on a clear definition of the grammatical system in the same way as the concept of lexicalization must be based on a clear definition of the lexicon of a language. There seem to be similar processes involved in both the development of new grammatical material as well as of new lexemes.

The development of methinks

In the following I will try to find out what happened to the construction me ðynceð > methinks in the history of English, which mechanisms led to its change and whether to characterize it as a case of grammaticalization or lexicalization. From this I hope to gain some insight into similarities and differences between the two processes of linguistic development. In Old English me ðynceð was a free syntactic unit constituting an impersonal construction among quite a number of other similar constructions. It occurred in all person and number categories as well as in all possible tense and mood forms with the exception of the imperative, since it expressed a nonintentional state, viz. (it) seems/seemed to me/you …. Syntactically it formed a clause with the  always marked by the dative case, the  — if surfaced — by the nominative (cf. examples (4) and (5)). The verb itself was usually complementized by an adjective(phrase) (examples (4) and (5)), a noun phrase (example (6)) or a ðæt-clause (example (7)). The word order was flexible: (4)

Se me ðincð gesæligra þe hwæthwugu hæfð. he- me- seems happier who something has ‘He who possesses something seems happier to me.’ (O2 XX Philo Boethal 119)11

(5)

ða steorran ðe us lytle ðincað, sind swiðe brade the stars- which us- small seem, are very big ‘The stars, which seem to us to be small, are very big.’ (O3 Ex Scia Temp 12)

(6)

þeah hit wisra gehwæm wundor ðince … although it of-the-wise-(men) everyone- (a)-miracle seem … ‘Although it may seem a miracle to each of the wise men …’ (O2/3 XX XX MBO 199)

GRAMMATICALIZATION VERSUS LEXICALIZATION

(7)

361

þeah monnum ðynce þæt hit long sie. although men- seem-3.. that it long be ‘Although to men it may appear to be long.’ (O2XX Philo Boethal 117)

Apart from ðyncan, there was a personal verb, ðencan in Old English, which was formally and semantically very close to ðyncan. Since it expressed a mental activity (think, remember, intend, learn, …), it was always accompanied by an  in subject function and an  of the mental activity: (8)

he ne mæg witan hwæt he þencð; he not can know what() he() thinks ‘He cannot know what he thinks.’ (O2 XX Philo Boethal 132)

(9)

þa þohte he þæt he sceolde worulde wiðsacan, then thought he() [that he should world- forsake]() ‘And he thought that he should forsake the world.’ (O2 NN Hist Bedehe 264)

In Middle English (1150–1500), the situation is to a considerable extent similar to that of Old English, as me þincð is still used as an impersonal construction among other impersonals and in all person and number categories, tenses and moods (cf. example (10)). The  is still marked by the dative case, as long as a personal pronoun is used. It is mostly preposed, but can be found in postposed position, too (cf. example (11)). The meaning is still the same. But the syntax has changed. An  participant has become very rare, as well as a complementation by an adjective (phrase).  and verb are normally in close contact position without being interrupted by other elements. Thus the verb is more and more losing its typical valence features and degenerating into a formula-like expression.12 Furthermore, the impersonal construction is frequently complemented by that-clauses with the subordinator in many examples deleted. Since in Middle English the word order has already changed to SVO in main and subordinated clauses, the former complementing clause can easily be analysed as the main clause,13 and the impersonal construction, formerly the dominating clause, is now consequently interpreted as a subordinated disjunct, which allows for its parenthetical use (cf. examples (12)–(13)). The denotation of evidentiality, formerly implied in the denotation of an act of cognition, is foregrounded (cf. Brinton 1996: 275–276).

362

ILSE WISCHER

(10)

þe þincheð þat he ne mihte his sinne forlete. you- thinks that he not might his sin- relinquish ‘It seems to you that he might not relinquish his sin.’ (MX/1 Ir Hom Trin 12, 73)

(11)

Thy wombe is waxen grete, thynke me, þou arte with barne, your womb is grown big think me- you are with child ‘Your womb has grown big. You seem to be pregnant.’ (M4 XX Myst York 119)

(12)

My lord me thynketh / my lady here hath saide to you trouthe and gyuen yow good counseyl […] (M4 Ni Fict Reynard 54)

(13)

I se on the firmament, Me thynk, the seven starnes. (M4 XX Myst Town 25)

For reasons of sound change the impersonal stative verb þyncan ‘seem’ and the personal dynamic verb þencan ‘believe, cogitate’ become formally more and more alike, and, as they are semantically very close to each other too, it is not surprising that at the end of the Middle English period both verbs can no longer formally be distinguished and personal constructions become more and more prominent, even with a non-intentional stative meaning14 (cf. examples (14) and (15), but notice (16), which is still impersonal): (14)

O ye lordes thynke ye that this is good ‘you’-

(M4 Ni Fict Reynard 9)

(15)

Sche thowt it was heuy to hir to takyn sweche labowr vp-on hir ‘she’- (M4 Ir Relt Kempe I,226)

(16)

hir thowt sche had neuyr so mech be-forn. ‘her’- (M4 Ir Relt Kempe I,229)

In Early Modern English (1500–1750) it is only the first person singular methinks/methought that survives in its impersonal form. A comparison of its use in the three sections of the Early Modern English part of the Helsinki Corpus shows that this construction is increasingly analysed as one item (cf. Table 1). Postposition of the pronoun occurs only in one single instance in the Early Modern English part of the Helsinki Corpus, and this is in section E1, the earliest period: (17)

But yet it thinketh me, loe, that if I may not declare the causes without perill, than to leaue them vndeclared is no obstinacy. (E1 XX Corp Morelet 505)

363

GRAMMATICALIZATION VERSUS LEXICALIZATION

Table 1. Total number of occurrences of methinks/methought in the three sections of the Early Modern English part of the Helsinki Corpus spelled separately or as one word

spelled separately spelled as one word

E1 1500–1570

E2 1570–1640

E3 1640–1710

19 01

8 8

03 16

Methinks/methought is no longer complemented by other elements and as a rule it functions syntactically as a disjunct. Formally it resembles an adverb, except for its present and past tense forms, which are traits of its verbal origin, and as such, viz. as a verb, it is usually treated in dictionaries. If we find methinks listed in dictionaries, should we not assume that it represents a lexeme, that it has undergone a process of lexicalization? On the other hand, function words are also listed in dictionaries, although they belong to the sphere of grammar and have undergone a process of grammaticalization. What has happened to methinks? It has definitely passed through a syntactic change, and semantically it has lost its original compositional meaning denoting an individual process of cognition and has acquired an interpersonal function, that of marking evidentiality. Methinks has been restricted to the first person singular, i.e. the speaker/writer, and is therefore highly subjective in meaning. Constructions with other referents than the speaker (thee, him, …) fall out of use, so that the only surviving one is that with the highest degree of subjectivity. With respect to I think, Aijmer (1997: 6) calls this process “specialization”. So it can be argued that towards Middle and Early Modern English the impersonal construction methinks had undergone a process of grammaticalization to develop into an adverbial marker of evidentiality (cf. Palander-Collin 1996; Brinton 1996). This would fit into subtype II of grammaticalization mentioned above. Processes like syntactic reanalysis, phonetic attrition and subjectification, extension of scope, the preference of certain positions in the sentence,15 point to this explanation. On the other hand, a once productively formed impersonal construction has been fossilized, ‘conventionalized’, partly demotivated (since impersonal constructions have become unproductive) and therefore changed into a symbol, a formula, and as such it has to be stored as a whole entity in the lexicon. This would be perfectly in line with the definition of lexicalization given above. The newly developed ‘lexeme’, however, is classified as an adverb. Syntactically it operates as a disjunct to mark evidentiality. As such it belongs to a relatively closed class. As Larreya (1986) points out, there is only a limited

364

ILSE WISCHER

number of verbs that express a mental position. Such items tend to become categorized, hence grammaticalized. Thus mechanisms of both processes are involved in the development of methinks: Lexicalization occurs as ‘syntactic lexicalization’, i. e. the symbolification of a former free collocation, the syntactic pattern of which has become unproductive. The new sign, however, immediately takes over a grammatical function on the discourse level, and is thus grammaticalized according to subtype II. The lexicalization of methinks becomes obvious when its historical development is compared to that of I think in Modern English. Aijmer’s (1997) description of the pragmaticalization of I think shows many parallels with the grammaticalization (type II) of methinks, with the only difference that I think is not lexicalized. None of Bauer’s (1983) types of lexicalization can be applied to I think. It has not become a symbol to be listed as such in the lexicon of English.

5.

Summary

Summarizing these facts, it becomes obvious that lexicalization and grammaticalization (both, subtypes I and II) are not at all contradictory processes. They show many similarities, but operate on different levels of the language (see Figure 1). Such processes as phonetic attrition and semantic bleaching seem to be irreversible. Lüdtke (1996: 537) convincingly points to the fact that the phonic substance of linguistic signs is constantly eroding in the course of time. This erosion cannot be reversed. It is compensated by morpho-syntactic means. Likewise he provides evidence of irreversible semantic processes that turn specific meanings into more general and finally into very abstract (operational) meanings.17 It is therefore highly unlikely that a grammatical item (a function word or an inflectional morpheme) turns back into a lexeme,18 i.e., becomes lexicalized again. So lexicalization cannot be considered to be the reverse of grammaticalization or ‘degrammaticalization’. However, it can be related to desyntacticization, in the sense of a syntagmatic structure losing its syntactic transparency and merging into one single lexical item, (which also happened with methinks). Both processes, grammaticalization and lexicalization, are accompanied by very similar syntactic and phonetic mechanisms: gradual phonetic reduction, syntactic reanalysis, demotivation, fossilization, conventionalization. The semantic changes, however, differ: When a free collocation or an ordinary word formation is lexicalized, a specific semantic component is added, so that the new lexical meaning differs from the former compositional meaning. Both are related to each other in a metaphorical or metonymical sense.19 When a linguistic term

GRAMMATICALIZATION VERSUS LEXICALIZATION LEXICON

365

GRAMMAR

Language use

LEXICON 1.1. syntagm becomes new lexical item (mother-in-law; ...)

GRAMMAR 2.1. syntagm becomes new grammatical item on the propositional level (I) (be going to; in front of ... ) or on the textual or interpersonal level (II) (methinks; indeed; ... ) 2.2. lexeme becomes grammatical item (I) (will; have; ... ) or (II) ( well; ... )

1.2. lexeme becomes more lexical16 (OE: hlaf-weard ‘bread-protector’' > ModE: lord,... )

2.3. grammatical item becomes more grammatical (I) (function words > clitics; ...) (II) (marker of dynamic modality > marker of epistemic modality; ...)

Figure 1. Similar processes on the lexical and the grammatical level of the language

is grammaticalized, specific semantic components get lost and an implied categorial or operational meaning is foregrounded. Methinks passed through a syntactic lexicalization process which was not accompanied by an addition of a specific semantic component, but rather by an opposite semantic change: Impersonal think has no longer a meaning of its own. It cannot combine with other persons any more (*himthinks, *us thought …). So, methinks has lost its original propositional meaning denoting an act of cognition and has acquired an exclusively speaker-oriented, or interpersonal, meaning. This change can be regarded as an increase in subjectivity. Thus it operates as a marker of evidentiality, and as such it has become a member of a relatively closed class. Although we can notice an increase in frequency in Early Modern English, methinks never becomes obligatory,20 but always stands in free variation with expressions like I think, it seems to me, obviously etc. Grammatical elements of this kind are less constrained and therefore situated at the periphery of grammar.21 And this must also be the reason for the relatively shortlasting existence of methinks. Linguistic elements, once they are grammaticalized, normally have a rather stable position in the language, whereas lexical units are less constrained, more flexible and pass out of existence more easily.

366

ILSE WISCHER

Acknowledgments I would like to thank the many people who commented on earlier versions of this paper, especially Olga Fischer and Dieter Stein, Elizabeth Traugott, Laurel Brinton, as well as the anonymous reviewers of this paper.

Notes 1.

E.g. in Irish, palatalisation and initial mutations have been functionalized/grammaticalized to distinguish different cases, number and gender, etc.

2.

This approach was further developed by Kuryłowicz (1965).

3.

Cf. also Heine and Reh (1984: 15) “[Grammaticalization is] an evolution whereby linguistic units lose in semantic complexity, pragmatic significance, syntactic freedom, and phonetic substance …”.

4.

Lehmann (1985: 311); Cf. also Lehmann (1995: 13); Bybee (1985); Hopper (1988); Li and Thompson (1974).

5.

The term ‘operator’ is used with several different meanings. The usage here refers to a specifying or modifiying element with no denotational, but only an “operational”, i.e. grammatical meaning (cf. Hansen et al. 1990: 155–156).

6.

Cf. Hansen et al. (1990); Talmy (1985); Brekle and Kastovsky (1977); Kastovsky (1974).

7.

Lipka (1977: 155) considers lexicalization a process by which a “complex lexeme” loses its syntagmatic character and its compositional meaning due to or accompanied by phonological changes, loss of motivation and semantic changes, which finally lead to its idiomaticization.

8.

Cf. Coulmas (1985); Schwarze and Wunderlich (1985); Keller (1995).

9.

According to Keller (1995: 114–118) ‘icons’ and ‘symbols’ are defined in terms of the procedure that leads to their interpretation. ‘Icons’ are signs that are interpreted by means of associative conclusions. Between signifier and signified there is a relation of similarity. New ‘icons’ can principally be created ad hoc. They are necessarily motivated and transparent. ‘Symbols’, on the other hand, are arbitrary signs that are interpreted by means of rule-based conclusions. For a detailed description of the process of symbolification of icons cf. Keller (1995: 167–173).

10.

Aijmer (1997: 2) implies the same idea in her definition of grammaticalization.

11.

The examples are taken from the Diachronic Part of the Helsinki Corpus of English Texts. More detailed bibliographical information about the references is given in the Appendix.

12.

However, such a reanalysis seems only to be possible, if the meaning of the former main clause allows a certain categorization, for instance as a marker of evidentiality, of epistemic modality, or others (it seems, es scheint, mne kažetsa [Russ.]; I assume, the point is, I expect, … ).

13.

Further investigations might provide evidence for a general relationship between the deletion of the subordinator that and a probable reinterpretation of main and subordinated clauses in English.

14.

Non-intentional stative meanings (mental states) were formerly expressed by impersonal constructions with þyncan ‘seem’. Þencan ‘believe, cogitate’, on the other hand, implied a mental activity, often combined with an intention, cf. Þeah hwa mæge ongitan hwæt oðer do, he

GRAMMATICALIZATION VERSUS LEXICALIZATION

367

ne mæg witan hwæt he þencð; þeah he mæge sume his willan ongitan, þonne ne mæg he eallne, ‘though someone may understand what somebody else is doing, he cannot know what he is thinking; though he may understand part of his desire, yet he cannot (understand) all (of it).’ (O2 XX Philo Boethal 132). 15.

It can be preposed, postposed or parenthetically embedded into the proposition.

16.

According to Bauer’s (1983) types of lexicalization an item can only be phonologically or morphologically, etc. lexicalized. The more levels (phonology, morphology, syntax, semantics) are involved in the lexicalization process, the stronger the item seems to be lexicalized, with its total demotivation as its final stage.

17.

Cf. also Brinton (1996: 52 and 273); Heine, Claudi and Hünnemeyer (1991: 150).

18.

There has, however, developed a dispute on the unidirectionality hypothesis of grammaticalization, based on probable counterexamples like case markers turning into clitics (Kahr 1976), or words as elements of a rather closed class changing over into an open, lexical class (Givón 1975: 96). Whether especially the latter are merely instances of divergence, which is normal in any kind of grammaticalization or whether they are due to some reanalysis, that may indeed be described as ‘degrammaticalization’, must be the subject of further investigation.

19.

According to Hansen et al. (1990: 39) reader in The reader of that article will forgive us if … is an ad-hoc formation, whereas in The newspaper lost many readers it is lexicalized. It has acquired the additional component 〈+〉. Both meanings are metonymically related.

20.

In contrast to a case marker, an article, or an auxiliary. Aijmer (1997: 3) states: “pragmatic elements tend to be optional in the sentence while grammaticalization results in forms which are an obligatory part of the grammatical ‘core’ such as tense and mood.” Cf. also Hopper and Traugott (1993: 113).

21.

Based on the assumption that there is no strict borderline between grammar and lexicon or grammar and pragmatics (for grammaticalization is a gradual process), I consider those grammatical morphemes that are less categorized and that exhibit a certain degree of variability as peripheral. Rules or morphemes that are obligatory (e.g. word order, case markers, or articles) belong to the core of grammar. Cf. also “grammatical ‘core’” in note 20.

Appendix (4)

Alfred’s Boethius. King Alfred’s Old English Version of Boethius De Consolatione Philosophiae, W. J. Sedgefield (ed.). Oxford: The Clarendon Press, 1899. (5) Aelfric’s De Temporibus Anni. Early English Text Society, 21, Henel (ed.). London, 1942. (6) The Meters of Boethius. The Paris Psalter and the Meters of Boethius. The Anglo-Saxon Poetic Records, V, G. Ph. Krapp (ed.). London. George Routledge & Sons, Limited and New York: Columbia University Press, 1933. (7) Cf. (4). (8) Cf. (4). (9) Bede’s Ecclesiastical History. The Old English Version of ‘Bede’s Ecclesiastical History of the English People’, Parts I,1; I,2. Early English Text Society, O.S. 95, 96, Th. Miller (ed.). London, 1959 (1890; 1891). (10) Trinity Homilies. [1225]. Old English Homilies of the Twelfth Century. Second Series. Early English Text Society, O.S. 53, R. Morris (ed.). London, 1873. (11) The York Plays. [1450], Richard Beadle (ed.). London: Edward Arnold Ltd, 1982.

368

ILSE WISCHER

(12) Caxton, William. [1481]. The History of Reynard the Fox. Translated from the Dutch Original by William Caxton. Early English Text Society, 263, N. F. Blake (ed.). London, 1970. (13) The Wakefield Pageants in the Towneley Cycle. [1500] Old and Middle English Texts, A. C. Cawley (ed.). Manchester: The Manchester University Press, 1958. (14) Cf. (12). (15) Kempe, Margery. [1438]. The Book of Margery Kempe, vol. I. Early English Text Society. S. B. Meech and H. E. Allen (eds). London, 1940. (16) Cf. (15). (17) More, Thomas. [1529]. Letter(s). The Correspondence of Sir Thomas More, E. F. Rogers (ed.). Princeton: Princeton University Press, 1947.

References Aijmer, K. 1997. “I think — an English modal particle”. In Modality in Germanic Languages. Historical and Comparative Perspective, T. Swan and O. J. Westvik (eds), 1–47. Berlin/New York: Mouton de Gruyter. Bauer, L. 1983. English Word-formation. Cambridge: Cambridge University Press. Brekle, H. E. and Kastovsky, D. “Wortbildungsforschung: Entwicklung und Positionen”. In Perspektiven der Wortbildungsforschung, H. E. Brekle and D. Kastovsky (eds), 7–19. Bonn: Bouvier. Brinton, L. J. 1996. Pragmatic Markers in English. Grammaticalization and Discourse Function. Berlin/New York: Mouton de Gruyter. Bybee, J. 1985. Morphology. A Study of the Relation between Meaning and Form. Amsterdam: John Benjamins. Chen, G. 1998. “The degrammaticalization of addressee-satisfaction conditionals in Early Modern English”. In Advances in English Historical Linguistics, J. Fisiak and M. Krygier (eds), 23–32. Berlin/New York: Mouton de Gruyter. Coulmas, F. 1985. “Lexikalisierung von Syntagmen”. In Handbuch der Lexikologie, Chr. Schwarze and D. Wunderlich (eds), 250–268. Königstein /Ts: Athenäum. Givón, Talmy. 1975. “Serial verbs and syntactic change: Niger-Congo”. In Word Order and Word Order Change, C. N. Li (ed.), 47–112. Austin/London: University of Texas Press. Givón, T. 1979. On Understanding Grammar. New York: Academic Press. Hansen, B., Hansen, K., Neubert, A. and Schentke, M. 1990. Englische Lexikologie. Leipzig: Verlag Enzyklopädie. Heine, B. and Reh, M. 1984. Grammaticalization and Reanalysis in African Languages. Hamburg: Helmut Buske. Heine, B., Claudi, U. and Hünnemeyer, F. 1991. “From cognition to grammar: Evidence from African languages”. In Approaches to Grammaticalization, vol. 1, E. C. Traugott and B. Heine (eds), 149–187. Amsterdam/Philadelphia: John Benjamins. Hopper, P. J. 1988. “Emergent Grammar and the a priori grammar postulate”. In Linguistics in Context: Connecting, Observations, and Understanding, D. Tannen (ed.), 117–134. Norwood: Ablex.

GRAMMATICALIZATION VERSUS LEXICALIZATION

369

Hopper, P. and Traugott, E. C. 1993. Grammaticalization. Cambridge: Cambridge University Press. Kahr, J. C. 1976. “The renewal of case morphology: Sources and constraints”. Working Papers on Language Universals 20: 107–151. Kastovsky, D. 1974. “Word-formation, case grammar, and denominal adjectives”. Anglia 92: 1–54. Keller, R. 1995. Zeichentheorie. Tübingen/Basel: A. Francke Verlag. Kuryłowicz, J. 1965. “The evolution of grammatical categories”. Esquisses Linguistiques 2: 38–54. Larreya, P. 1986. “Amalgames Lexicaux et Verbs de Jugement en Anglais”. LexiqueGrammaire-Domaine Anglais, 79–94. Université de Saint-Étienne, Travaux XXXXVII. Lehmann, Chr. 1985. “Grammaticalisation: Synchronic variation and diachronic change”. Lingua e stile 20: 303–318. Lehmann, Chr. 1995. Thoughts on Grammaticalization. München/Newcastle: LINCOM EUROPA. Li, C. N. and Thompson, S. A. 1974. “Historical change and word order: A case study in Chinese and its implications”. In Historical Linguistics I: Syntax, Morphology, Internal and Comparative Reconstruction, J. M. Anderson and C. Jones (eds), 199–218. Amsterdam: North-Holland. Lipka, L. 1977. “Lexikalisierung, Idiomatisierung und Hypostasierung als Probleme einer synchronischen Wortbildungslehre”. In Perspektiven der Wortbildungsforschung, H. E. Brekle and D. Kastovsky (eds), 155–164. Bonn: Bouvier. Lüdtke, H. 1996. “Changement linguistique”. In Kontaktlinguistik — Contact Linguistics — Linguistique de contact, vol. 1, H. Goebl, P. H. Nelde, Z. Starý and W. Wölck (eds), 526–540. Berlin/New York: Walter de Gruyter. McMahon, A. M. S. 1994. Understanding Language Change. Cambridge: Cambridge University Press. Meillet, A. 1912. “L’évolution des formes grammaticales”. In Linguistique Historique et Linguistique Générale, A. Meillet (ed.), 131–148. Paris: Champion. Meillet, A. 1925. La méthode comparative en linguistique historique. Paris: Champion. Palander-Collin, M. 1996. “The rise and fall of METHINKS”. In Socio-linguistics and Language History — Studies based on the Corpus of Early English Correspondence, Part III: Individual changes in social focus [Language and Computers: Studies in practical linguistics. 15], T. Nevalainen and H. Raumolin-Brunberg (eds), 131–149. Amsterdam — Atlanta, GA: Rodopi. Schwarze, Chr. and Wunderlich, D. 1985. Handbuch der Lexikologie. Königstein /Ts: Athenäum. Talmy, L. 1985. “Lexicalization patterns: Semantic structure in lexical forms”. In Language Typology and Syntactic Description, vol. 3: Grammatical Categories and the Lexicon, T. Shopen (ed.), 57–149. Cambridge: Cambridge University Press.

370

ILSE WISCHER

Traugott, E. C. 1982. “From propositional to textual and expressive meanings: Some semantic-pragmatic aspects of grammaticalization” In Perspectives on Historical Linguistics, W. P. Lehmann and Y. Malkiel (eds), 245–271. Amsterdam: Benjamins. Traugott, E. C. and Heine, B. (eds). 1991. Approaches to Grammaticalization. 2 vols. Amsterdam/Philadelphia: John Benjamins. Traugott, E. C. and König, E. 1991. “The semantics-pragmatics of grammaticalization revisited”. In Approaches to Grammaticalization, vol. 1, E. C. Traugott and B. Heine (eds), 189–218. Amsterdam/Philadelphia: John Benjamins. Wischer, I. 1997. “Lexikalisierung versus Grammatikalisierung — Gemeinsamkeiten und Unterschiede”. Papiere zur Linguistik der Universität Klagenfurt 2/97: 121–134.

Name index

A Abbott 332 Abraham 9–11, 189 Adamson 1, 8, 19, 23, 27, 175 Aijmer 357, 363–364, 366–367 Akimoto 4–5 Aland 244 Alexander 332 Allen, A. 252, 270 Allen, C. 7, 29 Andersen 225 Anderson, J 289 Anderson, N. 220, 224 Anttila 329 d’Ardenne 286, 288, 305 Atwood 337 Auwera, van der 92, 164 B Bailey 317 Ball 272 Bammesberger 244 Barber 68 Bauer, G. 338 Bauer, L. 358, 364, 367 Bechtel 270 Beekes 225 Beer 270 Behaghel 311 Behre 313 Benson 162 Berndt 202–203 Besten 189

Biber 61, 178–179, 182, 185 Biber et al. 61 le Bidois 302 Blake 286 Bobaljik 188–189, 198 Bock 253 Bolinger 57–59, 63, 311 Bopp 151 Bosworth 231 Boyland 119, 325 Brekle 366 Brinton 69, 78–79, 81–82, 141, 149, 173, 229, 235–236, 239, 243, 245–246, 257–259, 261–262, 264, 268, 332, 338, 361, 363, 366–367 Brorström 120, 332, 338, 339, 343 Brown, K. 60 Brown, P. 82 Brown, R. 68–69 Brunner 194–195, 198, 330, 335, 337–338 Bruyn 26, 164 Burnham 104 Bybee 13, 16–17, 19, 28, 116, 149, 151–152, 164, 332, 344, 366 C Calder 302 Callaway 252–253, 259, 272 Cameron 230, 272 Campbell, A. 104, 188–189, 196, 283, 313

372

NAME INDEX

Campbell, L. 17–18, 32, 149, 153, 213 Carey 343 Carlson 59 Castellan 256 Cedergren 349 Chafe 6 Chen 5, 7, 92–94, 359 Chomsky 187–188, 199 Christian et al. 333, 338 Christophersen 276, 279, 285 Clark 305 Claudi 6, 13, 269, 367 Coard 349 Collinder 219, 226 Comrie 338 Conrad 178, 182, 185 Cosmides 213 Coulmas 366 Coulomb 55, 58 Cowie 224 de la Cruz 115 Crystal 82, 184 Curme 332, 336, 338, 341 D Dabrowska 32 Dahl 312, 341 Dancygier 312, 317 David 332 Davies 224 D’Eloia 346 Demske 27–28 Denison 1, 6, 60, 113–114, 118, 121, 125, 128, 130–132, 134, 140–142, 144, 182, 254, 319, 325, 332, 334–337, 344–345, 349–350 Deumert 224 Deutscher 22 Diehn 305 Diewald 14 Dixon 42–45, 47, 48, 50, 61, 64 Domingue 7

Doyle 164 Dressler 150 Dube 333 DuBois 29 E Einenkel 256 Elhanon 47 Ellegård 114, 126 Elliott 284 Elsness 330, 332, 335, 338–339, 345–347, 349–350 Enkvist 237, 246, 262–264 Eze 333 F Farquhar 79 Feagin 333 Fenn 338–339 Fillmore 318, 321, 325 Filppula 333 Finegan 178, 182, 185 Finell 40 Fischer 1, 6–7, 11, 17, 22–23, 27, 30, 33, 63–64, 90, 108, 149–150, 153–156, 158–159, 161–165, 172–173, 224, 261, 272, 314, 366 Fitzmaurice 1, 7, 11, 14, 22–23, 31, 33, 165, 179 see also Wright, S. Fónagy 30 Fraser 60 Freed 257 Fridén 330, 332, 335–338, 343, 345 Fries 332 Funk 182 Funke 257, 272 G Gaaf 253 Gabelentz 9 Geerts et al. 118, 143 Gelderen, van 8, 11, 14, 187, 198 Gerritsen 156

NAME INDEX Gilman 61, 68–69 Givón 12, 13, 150, 155, 216, 356, 367 Goolden 238, 246 Görlach 7 Greenberg 276, 304 Gronemeyer 135 H Haegeman 137 Haiman 28–29, 47, 85, 107, 109, 150, 154, 311, 323–324 Halliday 67, 81 Hansen et al. 366–367 Harris, A.C. 17–18, 32, 149, 153, 213 Harris, J. 333, 335, 349 Harris, M. 85, 91, 94, 96, 104, 322–323, 332 Haspelmath 9–10, 12, 21, 32, 153, 161, 164, 225, 270 Healey 244, 272 Hecht 242 Heine 6–7, 13–16, 20, 26, 32, 149, 151–152, 269, 329–330, 355, 366–367 Heinrichs 278, 304 Herold 191 Hetzron 47 Himmelmann 276–277 Hock 329 Hoekstra 270 Hoffmann 292 Hofstetter 244 Holmberg 189, 198 Holmqvist 192 Holt 157 Hopper 1–4, 10, 12–17, 23, 78, 81–82, 85–86, 134, 149–152, 159, 165, 173, 183, 207, 209, 218, 222, 229, 247, 263, 329–330, 340, 366–367 Horn 175 Horrall 314

373

Huang 199 Huddleston 61, 174, 177 Hulthén 302 Humboldt 151, 154 Hünnemeyer 13, 270, 367 Hurford 224 I Iversen 302 J Jaeggli 187, 199 James 312, 324 Janda 165 Jespersen 30, 133–135, 137, 154, 188, 196, 198, 335, 345 Johansson 178, 182, 185 Jonas 188–189, 198 Jones 288–293, 301–302, 305–306 Joseph 153 K Kahr 367 Kallen 332–333, 338, 347, 349–350 Kastovsky 366 Kaufman 7, 333 Keller 32, 358–359, 366 Kemenade, van 165, 188, 272, 292 Kemmer 13 Kerkhof 284 Kim 237, 246 Kiparsky 41, 263 Klein-Andreu 61 Koch 277 Kolln 182 König 7, 18, 85–88, 90–91, 94, 96–98, 100, 103–104, 109, 359 Koopman 256 Kopytko 69 Körner 286 Kornexl 244 Kossuth 140, 144 Krahe 225

374

NAME INDEX

Kramsky 278 Kroch 272 Kroon 235, 241, 244, 245 Kuhn 195 Kurath et al. 244 Kuryłowicz 269, 366 Kytö 61, 108–109, 312 L Laanest 226 Labov 349 Lakoff 129 Landwehr 286 Langacker 150 Larreya 363 Larsen 244 Lass 1, 7–8, 20–22, 27, 164, 209, 212–213, 220, 225 Leech 82, 178, 182, 185, 338 Leek 63, 162–163 Lehmann 5, 12, 23–25, 28, 149, 158, 159, 165, 276, 356, 366 Lenker 4–5, 19, 235, 244, 246, 357 Leuschner 85, 88 Levinson 69, 82 Li 366 Lightfoot 10, 22, 32, 59, 63, 153, 164 Lipka 366 Liuzza 244 Los 5–6, 8, 162–163, 272 Lüdtke 25, 32, 364 Luick 306 Lyell 216 Lyons 55, 61, 63 M Macauley 156, 162 Mahler 244 Manabe 162 Marckwardt 332 Markus 305 Matthews 60–61, 175 Mätzner 105 Mayerthaler 150

McCawley 338 McColl Millar 6, 26, 287, 294, 305–306 McMahon 28, 359 Meillet 2, 9, 18, 32, 151, 225–226, 269, 329, 356, 359 Melchers 333, 338 Mencken 330, 332 Menner 332, 337 Mey 82 Micillo 279 Milner 55 Minkova 256, 287–288, 296, 305 Mitchell 90, 104, 126, 144, 188, 230, 245, 247, 256–257, 260, 263, 272, 279, 281, 313, 330, 332, 335, 339, 349 Mithun 6, 27 Molencki 6, 31, 316–317, 319 Montgomery 115 Morris 157–158 Mossé 129, 132–133, 143, 257 Mufwene 349 Mulac 71, 77–78 Murane 323 Mustanoja 154, 156, 158, 260, 314, 335–336, 343 N Nagle 114–115 Nakamura 131, 143 Nänny 150 Nehls 131 Nevalainen 69 Newmeyer 9–10, 32, 224 Nicholls 287, 294, 305–306 Nichols 183 Norde 164 Nordlinger 177 Noseworthy 332–333 O Ogura 114, 143 Overdiep 157–159

NAME INDEX P Pagliuca 116, 149, 332 Palander-Collin 363 Palmer 176 Penny 322 Perkins 116, 149, 332 Pervaz 286, 288, 292, 305 Peters 54 Phillipps 143 Pinker 225 Pintzuk 10 Plank 12, 29, 59, 115, 154, 161, 173 Platzack 189, 198 Pope 252 Poplack 333, 349–350 Popper 20 Pousada 333 Poussa 7 Pratt 130, 144 Purdy 280–281 Putten, van der 270 Q Quirk 104, 188, 196, 279 Quirk et al. 43, 45, 55, 60, 67, 81, 82, 86–88, 108–109, 233, 245, 318, 343 R Radford 57 Ramat 5, 6, 61 Rand 350 Raumolin-Brunberg 69 Reh 6, 151, 366 Rennhard 305 Reppen 179 Ricca 61 Riggert 253, 260–261 Ringe 224–225 Rissanen 317–318 Roberts 11 Robinson 81, 263 Rohdenburg 164 Ropers 201

375

Rosenbach 29, 32, 184 Ross, A.S.C. 281, 301 Ross, J.R. 68 Rousseau 349 Rubba 17, 151 Rydén 120, 332, 338–339, 343 S Safir 187, 199 Samuels 284 Sankoff 333, 349–350 Sato 245 Saussure 9 Schiffrin 82, 245, 247 Schrader 200 Schreiber 234, 245 Schwarze 366 Scur 333, 338 Seidler 286, 305 Selig 304 Siegel 256 Skeat 201, 203, 233, 238, 240, 245–246 Smith, C. 59 Smith, JC 224 Solá 203 Spamer 59, 63, 282–285 Stampe 150 Stanley 294, 299, 306 Stein 19, 32, 61, 108, 163, 271, 366 Stockwell 256, 263 Stoett 156–157 Strang 130, 134–136, 330, 334–335, 337–338 Swan 39–40, 244–245, 318 Swanton 239, 245 Sweet 197, 220 Sweetser 4, 18, 82, 173, 182, 312, 317 Szemerényi 225 T Tabakowska 30 Tabor 25

376

NAME INDEX

Tagliamonte 4, 6, 8, 14, 26, 333, 349–350 Talmy 366 Taraldsen 199 Teyssier 55, 63 Thomas 82 Thomason 7, 207, 216, 333 Thompson 13, 71, 77–78, 150–151, 366 Thráinsson 263 Toller 231 Tooby 213 Traugott 2–5, 10, 12, 14–19, 25–26, 40–41, 60–61, 67, 79, 81–82, 85–86, 94, 103, 149, 151–152, 159–161, 163, 165, 173, 177, 183, 207, 209, 218, 222, 224, 229, 244–245, 247, 329–330, 332, 335–338, 355–356, 359, 366–367 Trudgill 345 U Ukaji 67–68 Ura 199 V Vachek 278 Vanneck 330, 332, 337 Vendler 270 Venezky 244, 272 Vezzosi 29, 32

Vincent 20, 60–61, 151, 322–323, 332 Visser 95, 104, 114, 117–120, 125–126, 128–135, 137, 142–143, 157–158, 199–200, 202, 312, 331–332, 335–338, 344–345, 349 W Wal 204 Wales 82 Warner 134, 144, 254 Wårvik 237, 246 Wattel 272 Wekker 118, 322 Wessén 302 Whewell 224 Whorf 43 Wilbur 264 Wischer 6, 25 Wolfram 333 Wrenn 188, 196, 279 Wright, J. 332, 337, 346 Wright, S. 19, 61 see also Fitzmaurice Wunderlich 366 Wurzel 150 Y Yerkes 246 Z Zandvoort 338

Subject index

A ablative see case accomplishment verbs see verb(s) accusative see case and infinitive constructions (aci) 163 achievement verbs see verb(s) action marker 261–264 activity verbs see verb(s) adjectives 232 see also grammaticalization, cline(s) affective 19, 23–24, 39–66 see also lovely characterising 56–59, 64 classifying 56–58, 60 see also adjective, denominal denominal 63 descriptive 19, 23–24, 39–66 see also lovely double 52 identifying 57, 59 postnominal 61 predicative 57, 62, 63 prenominal 61 referent-oriented see adjectives, descriptive semantic sub-classes of 42–43 speaker-oriented see adjectives, affective strong/weak declension 63, 278, 282–284 word order see word order

adjunct(s) 3, 60, 245 see also adjectives, (strong)/weak declension; see also adverb, manner; see also adverbial, adjuncts manner 233–234, 243, 245 nominal 56, 63 see also compound style 234 temporal 344–345 admit see concessive, markers adverb 78, 363 adversative 235 causal 235 conjunctive 230 consecutive 235 epistemic modal see epistemic formation 2 manner 5, 39–40, 229–230, 234 scope of see scope sentence 39–40, 61 see also disjunct sentential 230–249 word order see word order adverbial adjuncts 347 clauses 311 cline see grammaticalization, cline(s) durative 258, 260 iterative 258 scope of see scope spatial 236 temporal 24–25, 236, 246, 314

378

SUBJECT INDEX

adversative 87–88, 92–93, 103, 106 conjunction 235 meaning 105 particle 105 adversity 91, 96–100 notion of 87  360–361 African American English 333, 343 creole origins hypothesis 333 languages 6–7 Afro-Asiatic 220 agreement checking 198 endings 187–206 participle 203 Aktionsart 257, 259, 266, 270 albeit see concessive, markers alliteration 238 (al)though see concessive, markers ambiguity 41, 57–58, 299–300, 302, 306 avoidance of 302 in ending 285–295 in form 296–298, 300 in function 302 scope ambiguity see scope American English 23, 31, 171–186 dialects 142 languages 6 anacolutha 123 analogy 14–16, 32, 149–150, 152, 154, 296, 319 analytic morphology 334 negation see negation synthetic to analytic see synthetic anaphor 199 anaphoric pronouns 236 animacy 29 hierarchy 159

Arabic, Moroccan 323 ARCHER corpus 49, 51, 61 article 56, 367 affixal 277 definite article 6, 26, 28, 55 see also grammaticalization, cline(s) formation 275–310 indefinite 4 aspect 246, 346 ingressive 251–275 perfective 251–275 verbal 341, 343–344, 348 attrition see semantic, bleaching; see phonetic, reduction Australian 323 autonomous see grammar, autonomous auxiliaries 6, 18, 251–275, 367 see also NICE criteria combination of 111–147 contracted 119 doubling of 114–122 double be 127, 132–133 double do 114–116 double modals 114–116 double passive 121–122 double perfect 116–120 double have 117–120 perfect have + perfect be 120 double progressive 120–121 fixed order of 111–147 multiple 339 quasi- see auxiliaries, semisemi- 3, 17, 23, 31, 171–186 be going to 3, 5–6, 16–17, 19, 23–24, 27, 152, 159, 161, 165, 171, 173, 175, 182–184 be supposed to 173, 175, 182, 184 be to 175, 175, 184

SUBJECT INDEX auxiliaries, semi- (Continued) have (got) to 18, 23, 153, 161, 171, 173–174, 176, 182, 184 need to 176, 184 negation of see negation used to 182 want to/wanna 23, 161–162, 173, 175, 182–183 auxiliary future 159 have see have periphrasis 318 status 253 B background(ing) 239–240, 262–263 Baltic 278, 323 Balto-Finnic 211–212, 218–220, 226 be double see auxiliaries, doubling of lexical 124 main verb 127 mutative 113, 329–354 passival 130–131 passive 121–123, 126–132, 134, 136, 138–141 perfect 120, 127, 136, 138 see also be, mutative progressive 129–132, 134, 137–141, 144 be going to see auxiliaries, semibe supposed to see auxiliaries, semibe to see auxiliaries, semibleaching see semantic, bleaching bondedness 22–24, 31, 158, 172, 356 boundary marker 229, 235, 237–238, 246 C case 11, 187, 199, 276, 285–286, 288–290, 292–294, 305 ablative 219 accusative 219, 286, 288–289, 293–294, 296, 305

379

dative 289, 293, 299, 305, 360–361 essive 226 genitive 299–300 inessive 211, 219 lative 219 locative 226 marker 211, 367 nominative 296, 360 oblique 292 prepositional 292 structural 292 translative 211 categoriality 59 formal 212 category shift 52–55 see also reanalysis Celtic 302 change in progress 13–14, 179 characteriser 23, 56–57, 59, 60, 63 see also adjectives, characterising; see also referent modification classifier 23, 57, 59, 63 see also adjectives, classifying; see also reference modification cline(s) see grammaticalization, cline(s) clitic 86, 165, 198, 210, 223–224, 287, 367 closed-class 112, 359, 367 coalescence 171, 173, 183, 356 codification 304, 359 Cohesion 24 collocability 62 collocative restrictions loosening of 101 comparative method 217, 221 comparatives, hypothetical 321 C(omplementizer) position 11, 187–189, 199 compound 56, 282–283 see also adjuncts, nominal compounding 28

380

SUBJECT INDEX

conceptual distance see distance principle concession see also concessive counterfactual 91 factual 7 hypothetical 7 notion of 87 concessive clauses 85–110 clause order 107–108 use of indicative 105 use of subjunctive 90, 97, 105 conditionals 7, 85–110, 319 conjunct 108 correlative 87, 106 factual 7, 85–110 markers 7 admit 92–94 albeit 86, 91, 94–96, 108 (al)though 7, 86–87. 89–91, 95, 97–99, 103–105, 108 inspite of/(in) despite of 86, 100–102 even if 89, 91, 107, 319 for all 91, 99, 108 grant 92–94 never so 99, 108 nevertheless 105–106 notwithstanding 100, 102–103, 108 typological pathway of 7 while 103 yet 86–87, 105–106, 108 reinforcement 91, 98 subordinator 86, 89, 92, 96 factual- 104 verbs see verb(s) concessives factual see concessive, clauses hypothetical see concessive, conditionals

concomitance notion of 103–104 condensation 356 conditional clauses 85–110 counterfactual 311–328 parenthetical 107 subordinator 89, 93 conjunctions 236 adversative 235 causal 235 consecutive 235 co-ordinating 235 temporal 103 conjuncts 245 correlative 86–87, 105 contact 26, 164, Celtic 302 Norman French 302–304 Norse 303–304 Scandinavian 302–303, 306 context 32, 164 contrast 103 notion of 87–88 contrastive 87 conventionalization 23, 32, 56, 62, 173–174, 363–364 correlative 264 see also conjuncts, correlative counterfactual 105 conditional clauses see conditional, clauses counterfactuality 98 expression of 31 courtesy marker 67, 78–81 see also pray subjuncts 82 creole 6–7, 14, 26, 164, 333 origins see also African American English; of Present Perfect in Samaná 349

SUBJECT INDEX D Danish 277–278 dative see case experiencer 253 decategorialization 78 definite article see article definiteness 56 functional category 275–310 semantic category 275–310 degemination 211 degrammaticalization 5–7, 21–23, 29, 31, 33, 149–169, 171–186, 207, 213, 359, 364 deictic 218–219 particle 26, 277 terms 45 deixis 275–310 demonstrative 56 distal 278–279, 302–304 pronouns 275–310 proximal 304 denotation 55–56 see also scope modification 63 see also reference modification deontic 19 derivation concept of 270 derivational see also grammaticalization, cline(s) morphology 216–217, 219 morphs 219 suffixes 270 desubjectification 23, 60 desyntacticization 364 determiner 56–57, 59, 63, 246, 275–310 as a category 59 distal 26 functions of 56 dictionaries 230–232, 244, 363 discourse

381

marker 5, 63, 67, 82, 229–249, 261, 271, 356- 357 see also soþlice; see also witodlice operators 357 particle 40–41 see also discourse, marker; see also indeed disjunct 60, 234, 245, 363 content 234 marking evidentiality 363 style 234, 237, 243, 245 distance principle 28–29, 31 divergence 4, 23, 78, 164, 367 backward 33 do causative 114 double see auxiliaries, doubling of dummy auxiliary 123 see also do, periphrastic periphrastic 114, 123–126, 141 substitute 123 double-base hypothesis 10, 14 doubling of auxiliaries see auxiliaries Dravidian 217, 221 drift see synthetic to analytic durative 258, 266, 270 adverbial see adverbial verbs see verb(s) durativity diagnostic test 257 Dutch 118, 154–157, 159–162, 165, 172, 189, 285, 322 Middle 156–157 Old 204 dynamic see verb(s) E economy 29, 311, 324 economic motivation 154 elision 278, 305 ellipsis 62–63 of the subject 255

382

SUBJECT INDEX

emphatic particle 94, 96–98 universal quantifier 96 emphasizer 233–234, 238, 243 empirical 7–8 enrichment lexical 25 of meaning 19 pragmatic 25 entailment 261 episode boundaries 236, 245, 261–269 marker 232, 239, 241, 243–244, 264, 267 see also V1, as episode boundary marker epistemic meaning 18–19, 160 modal adverb 41 modality see modality modals see modals parentheticals see parentheticals stance 324 essive see case Estonian 220, 226 even if see concessive, markers event coda 257, 271 nucleus 257, 267 onset 257, 267 temporal stages of 257 evidentiality marker of 355, 363, 365–366 existentials 129 explanation 9, 12–13, 22, 164, 207, 223 by universals 212 expressivity 29, 152 loss of 154 F factual see concession; see concessive falsification 7, 10, 12, 213

Farsi 276 feature checking 188–189 licensing 199 strength of 201 strong 188 Finnish 212, 218, 220, 226 focus marker 107 folk-etymology 6 for all see concessive, markers foregrounding 5, 262–264, 361, 365 forms of address 69 French 219, 302, 323 Old 270 frequency 8, 26, 53, 213–214, 219, 243–244, 271, 357, 365 functional categories 10–11, 13, 187 specialization 104 fusion 24–25 future 3, 16 auxiliary see auxiliary infinitive see infinitive G gelamp-construction 239–241, 243, 246 gender 285, 288, 292, 294, 296, 300–301, 305 change 301 grammatical 276, 281 breakdown of 301–302 generalisation 164 genitive 210, 224 see also case group 165, 210, 224–225 his-genitive 225 reinterpreation 299 s-genitive 29, 32, 165, 210, 224–225 see also possessives German 59, 161, 164–165, 276- 277, 281, 302, 304–305, 323 dialects 59 Old High 196, 264

SUBJECT INDEX get passive 121–122, 135–139, 141, 143 progressive 144 Gothic 193, 195, 312 gradience 10, 59 synchronic 59 graduality 32, 151 see also grammaticalization, gradualness of grammar as system 5, 367 autonomous 4, 9, 13 change 11, 164 see also language, change depth 164 diachronic 153 theory of 9, 164 grammatical categories 210 gender see gender items origin of 209–212, 218–221 see also unidirectionality, strong form of grammaticality 20, 212–221 see also lexicality hierarchy 210 see also grammaticalization, cline(s) grammaticalization see also regrammaticalization actuation of 12–13 as a theory 20–21, 32, 213, 224 as re-analysis 13 cline(s) 2, 5, 18, 22, 28, 59, 86, 207, 221–224, 270 adjective/intensifier cline 62 adverbial cline 40, 229, 244–245 characterizer/classifier cline 23 definite article 276–277 derivational/inflectional cline 270

383

lenition cline 222–223 of nominal premodifiers 59 subjectification see subjectification, cline counterexamples 207 see also degrammaticalization; see also unidirectionality, counterexamples cycle(s) 11, 18, 25, 28–30, 152, 356 definition of 2–9 see also grammaticalization, subtypes of diagnostics of 23–25, 27, 165, 329, 356 discontinuous 324 formal changes 2, 25 formal/generative approaches to 2, 9, 13–14 formal vs. semantic factors 2, 25 functional approaches to 2, 4, 9, 12–14 generative approaches to 9 gradualness of 10, 12–13, 20 implementation of 12–13 mechanisms/causes of 11, 14–31, 151–152 parameters see grammaticalization, diagnostics role in English 6–8 role of context 32 semantic-pragmatic factors 9–10 semantic changes 2, 25 status of 20, 152–153 types of 2, 25, 355–357 typological approaches to 4, 12–14 universality of 153, 213 grammemes 355 grant see concessive, markers Greek 231, 276 grounding 239

384

SUBJECT INDEX

H Hausa 323 have auxiliary 329–354 double see auxiliaries, doubling of: double perfect German and Dutch cognates 18 modal 138 perfect 112, 126–129, 136, 138–139, 141 possessive 184 have (got) to see auxiliaries, semihedge 184 Helsinki Corpus 40, 51, 85, 92–93, 95–97, 99–101, 103, 106–109, 156, 165, 191, 195–197, 312, 318, 366 hiatus 286–287, 305 avoidance of 287 homonomy 60, 311–312 Hungarian 323 I I(nflection) position 188 I think 357 icon(s) 359, 366 iconic 154–155, 163 motivation 29–30, 154 pole 29, 31, 154 iconicity 28–31, 43, 149–151 see also isomorphism; see also onomatopoeia; see also distance principle diagrammatic 150 imagic 150 ideational level see language, functions/levels identifier 57, 59–60, 63 see also adjectives, identifying idiomaticization 355, 358, 366 if-clause 107 imperative see mood impersonal 6, 253 verbs see verb(s)

implicatures 17, 26 inchoative marker 5 verbs see verb(s) indeed as a discourse particle 41, 63 as an intensifier 63 indicative see mood indirect object 290 requests 68 Indo-European 209, 217–221, 225 Indo-Uralic 220 inductive generalizations 214–217 inessive see case inferences 17–18, 26, 152, 359 infinitival see also word order, infinitival constructions complements 162, 251–275 marker to 7, 22–23, 27, 30–31, 153–169, 171–186 see also infinitive, to-infinitive infinitive 3 see also infinitival; see also negation, and infinitive VPs bare infinitive 154, 157, 163, 251–275 coordinated 159 Finnish 225 future 159 infinitival complements 251 inflected see infinitive, to-infinitive negative (split) infinitives 23, 31, 171–186 passive 160 plain see infinitive, bare infinitive purposive 17 split infinitives 158, 165, 171–186 to-infinitive 5, 18, 70, 72, 149–169, 251–275 see also infinitival, marker to complements 70, 72 intransitive 18 uninflected see infinitive, bare infinitive

SUBJECT INDEX inflection 219, 223, 270 see also grammaticalization, cline(s): derivational/inflectional cline overt 188–189 reduced 187–206 rich 188–189 inflectional endings see also past tense, endings loss of 63 morphology 216–217, 219 passive 112 ingressive aspectualizers see verb(s), inchoative inspite of/(in) despite of see concessive, markers intensifier 19, 23–24, 27, 52–57, 62–64 see also very truth- 233, 238, 243, 247 indeed see indeed lovely see lovely very 53–54 interpersonal/-actional level see language functions/levels interjection 67, 78, 80 interjectional function 81 marker 78 inversion 99–100, 108, 196, 317, 319 Irish 366 isolating 216–217, 221 isomorphic 155, 163 isomorphism 31, 149–150, 154–155, 163 between protasis and apodosis 311–328 lack of 155 Italian 199, 270, 322 dialects 203 iterative 258, 263, 270 adverbial see adverbial contexts 343

385

J juxtaposition 63 K Karelian 218, 220 Kartvelian 220 L language change 10–11, 164 see also grammar, change contact see contact functions/levels 235, 245, 271 ideational 81, 245 interactional/-personal 81, 235, 245 textual 79, 81, 235, 237, 241, 245, 357 locus of 10–13, 22, 164 see also language, change; see also grammar, change processing 29 Latin 231, 235, 244, 270, 276 aci 163 classical 322 copula 218 dictionaries 232 discourse particles 235 futures 218 imperfects 218 translations 201, 230–231, 233, 238, 244, 246–247, 260 lative see case layering 4, 8, 10, 13, 23, 78, 229, 329–354 lenition cline see grammaticalization, cline(s) letter salutations 69 levelling morphological 285 morphophonemic 314 lexical diffusion 223 lexicality 20, 210–212 see also grammaticality

386

SUBJECT INDEX

lexicalization 5–6, 21, 56, 60, 86, 122, 355–371 in generative grammar 189 types of 358–360, 367 lexicography 230–231 lexicon 5, 358, 360, 363–364, 367 literary language 15–16 Longman Spoken and Written English Corpus 178, 182, 185 lovely 39–66 M manner adjuncts see adjuncts adverbs see adverbs markedness 212, 214, 225 Me-Thee 220–221 men’s speech mental space embedding 324 metaphor 14–16, 149, 152, 164, 359, 364 methinks 6, 355–371 metonym(y) 16–17, 152, 164, 359, 364, 367 Minimalist framework 187 Mitian see Me-Thee modal auxiliaries see modals bleached (in counterfactual conditionals) 315 particle 357 see also I think periphrasis 31 modality 233 see also mood epistemic 366 optative 92, 94, 321 use in counterfactual conditionals 312 use of subjunctive 94 modals 10, 141, 153, 161, 165, 254, 271 double see auxiliaries, doubling of epistemic 78, 115 quasi- see auxiliaries, semiroot 115

subjectification of 19 use in counterfactual conditionals (apodosis) 315–317 use in counterfactual conditionals (protasis) 318–322 modifier 59, 283 see also premodifiers non-restrictive 57 post-adjectival 42, 56 pre-adjectival 42, 44, 56 restrictive 57 monosyllables 218–219 mood 324 see also modals imperative 68, 70–73, 75–77, 80, 188–189, 192, 196, 360 indicative 192, 203 use in counterfactual conditionals 313, 318 subjunctive 31, 90–100, 188–189, 192–193, 196 see also concessive, clauses; see also modality, optative /indicative contrast 313, 316 past subjunctive use in counterfactual conditionals 315, 318 preterite subjunctive use in counterfactual conditionals 316 that-clauses 271–272 use in counterfactual conditionals 313–314, 317–318, 325 motivation 9, 28–30, 102, 108 see also iconic; see also iconicity; see also economic competing 29 pragmatic 180 movement 10, 188–189, 198 overt 189–190, 198 Multivariate analysis 339, 342, 350 mutative 127 be see be verbs see verb(s)

SUBJECT INDEX N naturalness 150, 212 need to see auxiliaries, seminegation 31 see also operator analytic 175 and infinitive VPs 177–178 Finnish 212 of semi-auxiliaries 175–176 synthetic 175 negative see also polarity double 99, 184 particle 99 raising 31, 175, 179 scope see scope split infinitive see infinitive neutralisation 281, 301–302 never so see concessive, markers nevertheless see concessive, markers New Zealand English 215 NICE criteria 182 Norse, Old 26 Nostratic 217, 220, 225 notwithstanding see concessive, markers noun phrase German 27 subjectivity in see subjectivity, in NP syntactic-semantic change 39–66 word order in see word order number 11, 16, 187–206, 301 O obligatorification 356–357 OED quotation database 51, 61 onginnan/beginnan 5, 141, 144, 251–275 onomatopoeia 150 operator 357, 366 discourse see discourse grammatical 212 negating 212

387

optative see modality Oxford Text Archive 204 P palatalisation 366 Papua-New Guinea 323 paradigmaticization 356–357 parenthetical(s) 63, 71, 77–78, 80, 82 see also pray conditionals see conditional, clauses epistemic 77–78 suppositional 78 participles lone past participles 331, 337–338, 346–349 unmarked 336–337 passival 121, 128, 130 passive 112 be see be double see auxiliaries, doubling of get see get infinitive see infinitive progressive 131–132, 134, 140 weorðan 144 wurthe 128, 138 past tense see also preterite backshifted 98, 105 endings 188 marker 24 pathway(s) see grammaticaliztion, cline(s) perfect see also present perfect be see be continuative 338, 343, 350 double see auxiliaries, doubling of experiental 338, 343, 350 have see have resultative 338, 343, 350 perfective markers 5 verb 257 performative 68, 234, 245 speech acts 233

388

SUBJECT INDEX

periphrasis 256 auxiliary see auxiliary periphrastic constructions 111–147, 316, 335 do see do preterite see preterite persistence 23, 78, 134, 149, 154–155, 163, 330, 348 person 11 features 187, 198–199, 204 first/second/third 187–206 split 201 phi-features 199 phonetic attrition see phonetic, reduction harmony 325 reduction 2–3, 22, 25, 32, 74, 155–156, 161–162, 172, 247, 278, 286, 293, 316, 356–357, 363–364 phonogenesis 219 pleonastic 258 pluperfect 31 use in counterfactual conditionals 314–315, 317, 322, 324- 325 plural 16, 210 polarity in negative VPs 174–178 politeness 67–69, 79, 81–82 marker of 5 polysemous 23, 45, 230–231 possessives 28, 56, 59 see also genitive possible worlds 245, 262, 267–268 pragmatic enrichment see enrichment inferences see inferences motivation see motivation particle 62 pragmaticalization 63, 357, 364 pray 4–5, 67–84

premodifiers functional classification of 55–56 prescriptive grammar/grammarians 172, 322, 325, 339 tradition 117 present perfect 4, 26, 329–354 specialization of see specialization present tense endings 187–206 present-preterite verbs see verb(s) prestige 304, 333 preterite 329–354 see also past tense periphrastic 258 preterite subjunctive see mood, subjunctive use in counterfactual conditionals 324 pro-drop 11, 187, 199–204 progressive see also passive be see be double see auxiliaries, doubling of get see get grammaticalization of 141, 144 pronouns 11, 246 anaphoric see anaphoric demonstrative see demonstrative personal 11, 280 relative 280 prosodic contours 247 salience 287 protolanguage 217 prototypicality 212 Proto(Balto)-Finnic 219–220 IE 220 Uralic 218–220 punctual situation 266 verb see verb(s) purposive 3, 23, 31 see also infinitive

SUBJECT INDEX Q quantifier 56, 59, 63 see also emphatic, universal quantifier quantity principle 155 question complements 75–76 main clause 188 R  360–361 reconstruction 7, 21, 209, 216–219, 221 reanalysis 10, 13–15, 17–19, 27, 32, 52–55, 59, 63, 112, 122, 134, 141, 149, 152, 183, 363–364, 366–367 in generative grammar 10 recursive qualification 43–44 recursiveness 63 reference 55–56 modification 63 referent modification 63 regrammaticalization 6, 304 renouvellement 226 resultative perfect see perfect retrogression 102 see also degrammaticalization routinization 29, 152 Russian 275, 319 S Samaná English 4, 14, 26, 329–354 Scandinavian 276, 281, 302 schwa deletion 288, 305 loss 287 scope 356–357 ambiguity 57 denotative 56–57, 60 increase in 25, 245, 357, 363 of adverbials 41 of adverbs 39–40, 234 of negative 39, 174–178

389

reduction of 22, 24–25, 43–44, 158–159, 172, 245, 357 Scots 115, 125–126, 215, 305 semantic attrition see semantic, bleaching bleaching 17–19, 25, 54, 62, 80, 82, 112, 141, 151, 173, 229, 237, 243, 251–253, 256–260, 270–271, 272, 315, 321, 344, 348, 356, 364 change 3, 79, 151, 164, 359 see also unidirectionality, of semantic change gap 26, 275 integrity 155–157, 159–161, 172 loading 210 shift 52–53 specialization see specialization semanticization 152, 173 semi-auxiliaries see auxiliaries, semiSemitic 217, 221 sentence adverbial see adverb, sentential Sicilian 322 Slavonic 278, 323 soþlice 5, 229–249 source concepts 15–16, 26 South African English 215 Spanish 199, 322 specialization 26, 78, 299, 301–302, 330, 348, 363 of present perfect 344, 346, 348 standardization 325 statistical modeling techniques 349 structuralism 4, 9, 28 style adjunct see adjunct, style disjunct see disjunct, style stylistic function 263 subject agentive 160 expletive 159 inanimate 160, 234, 321

390

SUBJECT INDEX

subject (Continued) non-agentive 159 null 187–188, 199 see also pro-drop subjectification 5, 19, 27, 32, 39–66, 79, 229–249, 356–357, 363 see also desubjectification cline 4–5, 61, 67, 79, 82, 356 subjective 19, 176, 363 -objective polarity 55 subjectivity 19, 23, 363, 365 and word order 19, 23, 32, 39–66 in NP 19, 23, 32, 39–66 subjectivisation see subjectification subjunctive see mood subjuncts 245 see also courtesy, subjuncts substratum 26, 164 superlatives 56 suppletive forms 188 Swahili 323 syllabification 211 symbol 359, 363, 366 symbolic pole 29, 154 symbolification 359, 364, 366 synchrony vs. diachrony 4, 9, 12–13 synonymy 311–312 syntactic change 3, 9, 16, 207 see also grammaticalization, formal changes synthetic to analytic 6, 187, 276 T temporal adjunct see adjunct conjunction see conjunction disambiguation 341, 344–345 distance 341–343, 346, 350 see also perfect, continuative tense marker 24 shift 162–163 textual function/level see language, functions/levels

that clauses 162–163 deletion 71, 77–78, 366 determiner 26 thematic 218 discontinuity 229, 235, 237–238, 243, 256, 262–263, 265, 267–269, 271 marker of 5 though see concessive, markers topicality 29 topicalisation 41 Topic-drop 199–200 translative see case truth-intensifier see intensifier typological 7, 26, 183, 217, 276 diversity 216–217 U unidirectionality 7, 19–23, 27, 79, 149, 150–153, 164, 184, 207–227, 364, 367 counterexamples 20–22, 165, 208, 213–214, 219, 225, 367 see also grammaticalization, counterexamples empirical basis of 213–215 of semantic change 60 strong form 209, 216–218 weak form of 216 uniformitarianism 21–22, 216–218 unreality 313 expression of 116–120 Uralic 209, 217, 220, 226 used to see auxiliaries, semiV V1 254–256, 261, 263, 271 as episode boundary marker 267–269 variation 3, 8, 10, 14 see also layering

SUBJECT INDEX Variability 24 paradigmatic 24 syntagmatic 24 variationist approach 329–354 Vepsian 226 Verb first see V1 morphology 11 movement 204 second 188–189, 254 Subject (VS) 190–198, 263 see also inversion three-verb cluster 142 verbal agreement loss of 187–206 aspect see aspect endings 11 see also verbal, inflections fusion of 28 preservation of 11 reduction of 11 inflections 187–206 ordering of 28 verb(s) accomplishment 257 achievement 258 activity 257 concessive 92–94 directional 3, 16–17 durative 257 dynamic 343–344, 347, 350 impersonal 355–371 inchoative 251–275 see also onginnan/beginnan irregular 188 mental state 343 mutative 344, 347–348, 350 see also be, mutative non-stative 343 perception 162–163, 343 present-preterite 204

391

punctual 258, 263 stative 132, 258, 343–344, 347, 350 strong 188, 195 weak 188, 195, 218 volitionality 31 vowel harmony 226 W want to/wanna see auxiliaries, semiwhclause 72–73 complements 76 question 72–77 while see concessive, markers witodlice 5, 229–249 women’s speech 79, 82 word order 11, 18, 102, 187, 190, 192, 194, 197, 360–361, 367 see also inversion adjectives 42–45 adverbs 39–40, 61 and subjectivity see subjectivity fixed 335 infinitival constructions 253–256 noun phrase 23, 32, 45, 282–284 of nominal premodifiers 39–66 possessive constructions 29 Y yet see concessive, markers Yiddish 215 Yoruba 323 þa 5, 237, 243 as a discourse marker 237, 261–263 clauses 237 sentence-initial 246 þa V constructions 254, 256, 261, 263 þeah 89, 104

In the STUDIES IN LANGUAGE COMPANION SERIES (SLCS) the following volumes have been published thus far or are scheduled for publication: 1. ABRAHAM, Werner (ed.): Valence, Semantic Case, and Grammatical Relations. Workshop studies prepared for the 12th Conference of Linguistics, Vienna, August 29th to September 3rd, 1977. Amsterdam, 1978. 2. ANWAR, Mohamed Sami: BE and Equational Sentences in Egyptian Colloquial Arabic. Amsterdam, 1979. 3. MALKIEL, Yakov: From Particular to General Linguistics. Selected Essays 1965-1978. With an introd. by the author + indices. Amsterdam, 1983. 4. LLOYD, Albert L.: Anatomy of the Verb: The Gothic Verb as a Model for a Unified Theory of Aspect, Actional Types, and Verbal Velocity. Amsterdam, 1979. 5. HAIMAN, John: Hua: A Papuan Language of the Eastern Highlands of New Guinea. Amsterdam, 1980. 6. VAGO, Robert (ed.): Issues in Vowel Harmony. Proceedings of the CUNY Linguistics Conference on Vowel Harmony (May 14, 1977). Amsterdam, 1980. 7. PARRET, H., J. VERSCHUEREN, M. SBISÀ (eds): Possibilities and Limitations of Pragmatics. Proceedings of the Conference on Pragmatics, Urbino, July 8-14, 1979. Amsterdam, 1981. 8. BARTH, E.M. & J.L. MARTENS (eds): Argumentation: Approaches to Theory Formation. Containing the Contributions to the Groningen Conference on the Theory of Argumentation, Groningen, October 1978. Amsterdam, 1982. 9. LANG, Ewald: The Semantics of Coordination. Amsterdam, 1984.(English transl. by John Pheby from the German orig. edition “Semantik der koordinativen Verknüpfung”, Berlin, 1977.) 10. DRESSLER, Wolfgang U., Willi MAYERTHALER, Oswald PANAGL & Wolfgang U. WURZEL: Leitmotifs in Natural Morphology. Amsterdam, 1987. 11. PANHUIS, Dirk G.J.: The Communicative Perspective in the Sentence: A Study of Latin Word Order. Amsterdam, 1982. 12. PINKSTER, Harm (ed.): Latin Linguistics and Linguistic Theory. Proceedings of the 1st Intern. Coll. on Latin Linguistics, Amsterdam, April 1981. Amsterdam, 1983. 13. REESINK, G.: Structures and their Functions in Usan. Amsterdam, 1987. 14. BENSON, Morton, Evelyn BENSON & Robert ILSON: Lexicographic Description of English. Amsterdam, 1986. 15. JUSTICE, David: The Semantics of Form in Arabic, in the mirror of European languages. Amsterdam, 1987. 16. CONTE, M.E., J.S. PETÖFI, and E. SÖZER (eds): Text and Discourse Connectedness. Amsterdam/Philadelphia, 1989. 17. CALBOLI, Gualtiero (ed.): Subordination and other Topics in Latin. Proceedings of the Third Colloquium on Latin Linguistics, Bologna, 1-5 April 1985. Amsterdam/Philadelphia, 1989. 18. WIERZBICKA, Anna: The Semantics of Grammar. Amsterdam/Philadelphia, 1988. 19. BLUST, Robert A.: Austronesian Root Theory. An Essay on the Limits of Morphology. Amsterdam/Philadelphia, 1988. 20. VERHAAR, John W.M. (ed.): Melanesian Pidgin and Tok Pisin. Proceedings of the First International Conference on Pidgins and Creoles on Melanesia. Amsterdam/Philadelphia, 1990.

21. COLEMAN, Robert (ed.): New Studies in Latin Linguistics. Proceedings of the 4th International Colloquium on Latin Linguistics, Cambridge, April 1987. Amsterdam/ Philadelphia, 1991. 22. McGREGOR, William: A Functional Grammar of Gooniyandi. Amsterdam/Philadelphia, 1990. 23. COMRIE, Bernard and Maria POLINSKY (eds): Causatives and Transitivity. Amsterdam/Philadelphia, 1993. 24. BHAT, D.N.S. The Adjectival Category. Criteria for differentiation and identification. Amsterdam/Philadelphia, 1994. 25. GODDARD, Cliff and Anna WIERZBICKA (eds): Semantics and Lexical Universals. Theory and empirical findings. Amsterdam/Philadelphia, 1994. 26. LIMA, Susan D., Roberta L. CORRIGAN and Gregory K. IVERSON (eds): The Reality of Linguistic Rules. Amsterdam/Philadelphia, 1994. 27. ABRAHAM, Werner, T. GIVÓN and Sandra A. THOMPSON (eds): Discourse Grammar and Typology. Amsterdam/Philadelphia, 1995. 28. HERMAN, József: Linguistic Studies on Latin: Selected papers from the 6th international colloquium on Latin linguistics, Budapest, 2-27 March, 1991. Amsterdam/Philadelphia, 1994. 29. ENGBERG-PEDERSEN, Elisabeth et al. (eds): Content, Expression and Structure. Studies in Danish functional grammar. Amsterdam/Philadelphia, 1996. 30. HUFFMAN, Alan: The Categories of Grammar. French lui and le. Amsterdam/Philadelphia, 1997. 31. WANNER, Leo (ed.): Lexical Functions in Lexicography and Natural Language Processing. Amsterdam/Philadelphia, 1996. 32. FRAJZYNGIER, Zygmunt: Grammaticalization of the Complex Sentence. A case study in Chadic. Amsterdam/Philadelphia, 1996. 33. VELAZQUEZ-CASTILLO, Maura: The Grammar of Possession. Inalienability, incorporation and possessor ascension in Guaraní. Amsterdam/Philadelphia, 1996. 34. HATAV, Galia: The Semantics of Aspect and Modality. Evidence from English and Biblical Hebrew. Amsterdam/Philadelphia, 1997. 35. MATSUMOTO, Yoshiko: Noun-Modifying Constructions in Japanese. A frame semantic approach. Amsterdam/Philadelphia, 1997. 36. KAMIO, Akio (ed.): Directions in Functional Linguistics. Amsterdam/Philadelphia, 1997. 37. HARVEY, Mark and Nicholas REID (eds): Nominal Classification in Aboriginal Australia. Amsterdam/Philadelphia, 1997. 38. HACKING, Jane F.: Coding the Hypothetical. A Comparative Typology of Conditionals in Russian and Macedonian. Amsterdam/Philadelphia, 1998. 39. WANNER, Leo (ed.): Recent Trends in Meaning-Text Theory. Amsterdam/Philadelphia, 1997. 40. BIRNER, Betty and Gregory WARD: Information Status and Noncanonical Word Order in English. Amsterdam/Philadelphia, 1998. 41. DARNELL, Michael, Edith MORAVSCIK, Michael NOONAN, Frederick NEWMEYER and Kathleen WHEATLY (eds): Functionalism and Formalism in Linguistics. Volume I: General papers. Amsterdam/Philadelphia, 1999.

42. DARNELL, Michael, Edith MORAVSCIK, Michael NOONAN, Frederick NEWMEYER and Kathleen WHEATLY (eds): Functionalism and Formalism in Linguistics. Volume II: Case studies. Amsterdam/Philadelphia, 1999. 43. OLBERTZ, Hella, Kees HENGEVELD and Jesús Sánchez GARCÍA (eds): The Structure of the Lexicon in Functional Grammar. Amsterdam/Philadelphia, 1998. 44. HANNAY, Mike and A. Machtelt BOLKESTEIN (eds): Functional Grammar and Verbal Interaction. 1998. 45. COLLINS, Peter and David LEE (eds): The Clause in English. In honour of Rodney Huddleston. 1999. 46. YAMAMOTO, Mutsumi: Animacy and Reference. A cognitive approach to corpus linguistics. 1999. 47. BRINTON, Laurel J. and Minoji AKIMOTO (eds): ollocational and Idiomatic Aspects of Composite Predicates in the History of English. 1999. 48. MANNEY, Linda Joyce: Middle Voice in Modern Greek. Meaning and function of an inflectional category. 2000. 49. BHAT, D.N.S.: The Prominence of Tense, Aspect and Mood. 1999. 50. ABRAHAM, Werner and Leonid KULIKOV (eds): Transitivity, Causativity, and TAM. In honour of Vladimir Nedjalkov. 1999. 51. ZIEGELER, Debra: Hypothetical Modality. Grammaticalisation in an L2 dialect. 2000. 52. TORRES CACOULLOS, Rena: Grammaticization, Synchronic Variation, and Language Contact.A study of Spanish progressive -ndo constructions. 2000. 53. FISCHER, Olga, Anette ROSENBACH and Dieter STEIN (eds.): Pathways of Change. Grammaticalization in English. 2000. 54. DAHL, Östen and Maria KOPTJEVSKAJA TAMM (eds.): Circum-Baltic Languages. Volume 1: Past and Present. n.y.p. 55. DAHL, Östen and Maria KOPTJEVSKAJA TAMM (eds.): Circum-Baltic Languages. Volume 2: Grammar and Typology. n.y.p. 56. FAARLUND, Jan Terje (ed.): Grammatical Relations in Change. 2001. 57. MEL’C
UK, Igor: Communicative Organization in Natural Language. The semanticcommunicative structure of sentences. n.y.p. 58. MAYLOR, Brian Roger: Lexical Template Morphology. Change of state and the verbal prefixes in German. n.y.p.