Subjectification, Intersubjectification and Grammaticalization 9783110226102, 9783110205886

Citation preview

Subjectification, Intersubjectification and Grammaticalization

Topics in English Linguistics 66

Editors

Elizabeth Closs Traugott Bernd Kortmann

De Gruyter Mouton

Subjectification, Intersubjectification and Grammaticalization Edited by

Kristin Davidse Lieven Vandelanotte Hubert Cuyckens

De Gruyter Mouton

ISBN 978-3-11-020588-6 e-ISBN 978-3-11-022610-2 ISSN 1434-3452 Library of Congress Cataloging-in-Publication Data Subjectification, intersubjectification and grammaticalization / edited by Kristin Davidse, Lieven Vandelanotte, Hubert Cuyckens. p. cm. ⫺ (Topics in English linguistics; 66) Includes bibliographical references and index. ISBN 978-3-11-020588-6 (alk. paper) 1. Subjectivity (Linguistics) 2. Grammar, Comparative and general. I. Davidse, Kristin. II. Vandelanotte, Lieven, 1978⫺ III. Cuyckens, H. P299.S89.S85 2010 4011.43⫺dc22 2010009228

Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available in the Internet at http://dnb.d-nb.de. ” 2010 Walter de Gruyter GmbH & Co. KG, 10785 Berlin/New York Cover image: Brian Stablyk/Photographer’s Choice RF/Getty Images Printing: Hubert & Co. GmbH & Co. KG, Göttingen ⬁ Printed on acid-free paper Printed in Germany www.degruyter.com

Acknowledgements

The present volume finds its origin in the conference “From ideational to interpersonal: Perspectives from grammaticalization” (FITIGRA), held at the University of Leuven from 10 to 12 February 2005. This conference was financed through a research community on interpersonal and ideational grammar sponsored by the Fund for Scientific Research – Flanders (Scientific Research Network WO.018.00N) from 2000 to 2005. The editors would also like to acknowledge the generous support of the Interuniversity Attraction Pole (IAP) – Phase VI, project P6/44 of the Belgian Science Policy Office on ‘Grammaticalization and (Inter)Subjectification’ and that of the Spanish Ministry for Science and Innovation and the European Regional Development Fund (grant HUM2007-60706/FILO). Bringing this project to fruition would not have been possible without the support and input of the series editor, Elizabeth Traugott. Over and above her ever incisive comments on all the papers included in this volume, we would like to thank her for keeping things moving and spurring us on to completion. To Mouton de Gruyter’s editor Birgit Sievert and its production editor Wolfgang Konwitschny we say thanks for the pleasant exchanges and helpful feedback at different stages of the publication process. We owe a deep debt of gratitude to Lobke Ghesquière and Sigi Vandewinkel, both appointed to the ‘Grammaticalization and (Inter) Subjectification’ IAP-project at different intervals, for their invaluable help in formatting the manuscript. While naturally we accept the final responsibility for the choices made, we were very lucky in being guided by the advice of external referees (listed at the back of the volume) who did a tremendously conscientious and thorough job, benefitting both the papers selected for this volume and those eventually recommended to other publication outlets. For this we are very grateful to them. Finally, kudos are due to the contributors who all agreed to develop, and in some cases re-orient, their original studies with the themes and overall coherence of this volume in mind. For the good spirit in which everything was conducted and, of course, for the stimulating contributions, many thanks. K.D. – L.V. – H.C.

Contents

Acknowledgements ...................................................................................... v Introduction Hubert Cuyckens, Kristin Davidse and Lieven Vandelanotte ...................... 1 Part 1. Theoretical issues (Inter)subjectivity and (inter)subjectification: A reassessment Elizabeth Closs Traugott ............................................................................ 29 Part 2. (Inter)subjectification and grammaticalization involving adverbials Presupposition accommodation and language change Scott A. Schwenter and Richard Waltereit ................................................. 75 How prosody reflects semantic change: A synchronic case study of of course Anne Wichmann, Anne-Marie Simon-Vandenbergen and Karin Aijmer...................................................................................... 103 The semantic-pragmatic development of well from the viewpoint of (inter)subjectification Tine Defour .............................................................................................. 155 Paths in the development of elaborative discourse markers: Evidence from Spanish Teresa Fanego .......................................................................................... 197

viii

Contents

Part 3. (Inter)subjectification and grammaticalization in the noun phrase The emergence of the definite article: ille in competition with ipse in Late Latin Anne Carlier and Walter De Mulder........................................................ 241 On the subjectification and intersubjectification paths followed by the adjectives of completeness Lobke Ghesquière..................................................................................... 277 On the rise of (inter)subjective meaning in the grammaticalization of kind of/kinda Hélène Margerie ...................................................................................... 315 Part 4. (Inter)subjectification and grammaticalization involving modals Better as a verb David Denison and Alison Cort ............................................................... 349 (Inter)subjectification in the domain of modality and mood – Concepts and cross-linguistic realities Heiko Narrog ........................................................................................... 385 Contributors.............................................................................................. 431 Referees .................................................................................................... 433 Author index............................................................................................. 435 Subject index ............................................................................................ 441

Introduction Hubert Cuyckens, Kristin Davidse and Lieven Vandelanotte This volume sets out to investigate how the concepts of subjectification and intersubjectification can be insightfully employed in grammaticalization research. In exploring this relationship between (inter)subjectification and grammaticalization, it focuses on three distinct areas of the grammar: adverbials, modals, and the noun phrase. Together, these studies represent recent thinking about subjectification and push further the reflection about the less researched notion of intersubjectification: what is involved in indexing speaker attitude and belief, and expressing attention to the hearer’s self in the meaning of adverbials (e.g. discourse markers), NPs (e.g. determiners) and VPs (modality)? As the structure of the volume is in terms of these three domains, it will naturally invite the reader into this systematic exploration. In this introduction we will focus on the key concepts of language change addressed in this book, viz. subjectification and intersubjectification in relation to grammaticalization, and how the main theoretical claims and descriptive findings of the contributions relate to them. 1. Subjectivity in linguistics As subjectification and intersubjectification denote processes whereby a linguistic element acquires increased (inter)subjective meaning, we will set the scene with a general consideration of the key concept of subjectivity. Defining the notion ‘subjectivity’ is not a straightforward matter, as it has several uses and interpretations, some of them connoted negatively. As a non-technical term used in everyday language, it refers to the situation of being influenced by personal opinion, often unfairly so (as in His judgements are far too subjective), or of existing in the mind or imagination. In scientific investigation, subjectivity suggests something not based on empirical fact, “something ‘soft’, unverifiable, even suspicious” (Finegan 1995: 1), and therefore not worthy of scholarly investigation. In linguistic theorizing, it refers, broadly speaking, to the centrality of the speaker in language. While this is in itself a neutral characterization, linguistics – like

2

Hubert Cuyckens, Kristin Davidse and Lieven Vandelanotte

other social sciences – has for a good part of the twentieth century been influenced by objectivist epistemology and metaphysics (cf. Lakoff 1987), so that subjectivity and speaker-reference have until recently not been seen as paramount to linguistic research. For a good part of the twentieth century, subjectivity and speakerreference in language was not a research topic of major concern. While the importance of subjectivity in language was recognized as early as in Bréal’s ([1900] 1964) writings, it is only the last two or three decades that have seen a resurgence of interest, mainly in the cognitive-functional tradition. Of particular importance in this respect is Benveniste, who introduced the notion of ‘sujet d’énonciation’ (speaking subject) (vs. ‘sujet d’énoncé’ or syntactic subject) and who pointed out that “language is so strongly marked by subjectivity that one might wonder whether it could still function as, or be called, language if it was organized differently” (Benveniste 1966: 261; our translation).1 The importance of subjectivity was further underscored by Lyons (1977, 1982), and was subsequently given increasing attention in the cognitive-functional literature, with hallmark publications such as Traugott (1989), Stein and Wright (1995) and Langacker (1990). Recent publications testifying to the importance of subjectivity in linguistics include Nuyts (2001), Traugott and Dasher (2002), Langacker (1999, 2002, 2006), Verhagen (2005), Athanasiadou, Canakis and Cornillie (2006), De Smet and Verstraete (2006). The two main research strains to have developed on subjectivity and subjectification are those referring to the thought of Traugott and Langacker. In Traugott’s approach, (inter)subjectification is a semantic process whereby a linguistic element (morpheme, word, phrase or construction) develops new senses that involve speaker-reference or speaker-perspective. For instance, Traugott (1989) characterizes the development of English sculan from main verb (‘owe’) to its various modal uses as a shift from objective (‘propositional’) to subjective meanings. From its original concrete sense (‘owe financially’) it first acquired internal evaluative, but still propositional meanings (‘owe morally’). These shifted to deontic speechact-like meanings, which in turn led to epistemic meanings primarily focused on speaker’s belief and knowledge states. For Langacker, an expression or its meaning, taken as a whole, is neither subjective nor objective; as such, “a semantic shift does not … result in a global meaning becoming 1. Benveniste’s original (1966: 261) French wording was as follows: Il est marqué si profondément par l’expression de la subjectivité qu’on se demande si, autrement construit, il pourrait encore fonctionner et s’appeler langage.

Introduction

3

more subjective” (Langacker 2006: 18). Instead, in his view, an expression’s meaning always comprises both subjectively and objectively construed elements, and it is individual conceptual elements within an expression’s meaning that, over time, may come to be construed with a greater degree of subjectivity or objectivity (cf. Langacker 2006: 18). As such, the semantic shift in English modals such as may, will and must do not constitute a development from objective to subjective meaning; rather, the development from main verb to modal involves particular elements in the verb’s conceptualization coming to be construed with greater subjectivity. For instance, in its main verb use in Old English, magan involves an objectively construed force (“onstage” as a focused object of conception) emanating from the trajector for carrying out the situation described in the infinitival complement. In its modal verb use in Present Day English, this modal force is now subjectively construed and emanates from the speaker (subject of conception). Narrog (this volume) compares Traugott’s and Langacker’s theoretical approach to subjectification – and descriptive trade-offs of the differences – in the domain of the development of English modals. Thus, in Langacker’s approach, the term ‘subjective’ (as well as ‘objective’) does not pertain to the semantic content of a linguistic expression; rather, it should be understood in terms of the viewing arrangement or perspective assumed to construe a particular conceptual content. The notions subjective and objective are “characterized relative to the asymmetry between the observer in a perceptual situation and the entity that is observed” (1985: 120), whereby maximal subjectivity is characteristic of “the viewer’s role as such” and maximal objectivity characterizes “the onstage focus of attention” (2008: 77) on the object of perception. “For linguistic purposes,” Langacker (2008: 77) holds, “we are interested in the general conceptual analog of this perceptual asymmetry.” Speaker and hearer are construed with maximal subjectivity when they function as a tacit conceptualizing presence that is not itself made explicit. “At the opposite extreme, construed with maximal objectivity, is the focused object of attention: the entity an expression puts onstage and profiles” (2008: 77). Athanasiadou, Canakis and Cornillie’s (2006) volume Subjectification: Various Paths to Subjectivity is devoted mainly to Langacker’s approach, which is shown at work in predominantly synchronic studies of modals, adjectives and syntax. This volume is complementary to it in that the take on subjectification is Traugottian, and that a good number of the studies, distributed over the domains of modals, the NP and discourse markers, are

4

Hubert Cuyckens, Kristin Davidse and Lieven Vandelanotte

diachronic. As well, a concern specific to this volume is the investigation of intersubjectivity and intersubjectification. 2. A form-meaning approach to language change Traugott’s chapter “(Inter)subjectivity and (inter)subjectification: a reassessment” forms the theoretical cornerstone of the volume. To her, subjectivity “involves the expression of self and the representation of a speaker’s ... perspective or point of view in discourse” (Finegan 1995: 1, quoted in Traugott and Dasher 2002: 20). Subjectification is the semasiological process whereby linguistic expressions acquire subjective meaning. In particular, it refers to the tendency whereby lexical material “tend[s] to become increasingly based in the SP[eaker]/W[riter]’s subjective belief state or attitude to what is being said and how it is being said” (Traugott 2003: 125; see also Traugott 1989: 35, 1995: 47, 1997, 1999). While expressions of subjectivity index speaker attitude or viewpoint, markers of intersubjectivity index the speaker’s “attention to the addressee’s self-image” (Traugott this volume: 29). Intersubjectification is the semasiological process “whereby meanings come over time to encode … SP/W’s attention to the ‘self’ of AD[dressee]/R[eader] in both an epistemic and social sense” (Traugott 2003: 130). Importantly, in Traugott’s view, intersubjectification follows, and arises out of, subjectification. Both these changes crucially involve “the reanalysis as coded meanings of pragmatic meanings arising in the context of speaker-hearer negotiation of meaning” (Traugott this volume: 60). As subjectification and intersubjectification are sometimes invoked as purely semantic notions, elucidation is warranted of how Traugott grounds them in a functional theory of the linguistic sign and its use. It has been a constant of Traugott’s thinking to define subjectification and intersubjectification as semanticization, which requires the new (inter)subjective meanings to be conventionally coded by the forms, with new form-meaning pairs as result.2 At the origin of the change, there is the old form-meaning pair, a specific correlation of a – lexical and/or grammatical – form and the meaning coded by it, for instance, in the case of a bit, a noun with its full lexical weight and the potential for syntactic modification going with it (Traugott this volume: 46–49). 2. As Traugott (this volume: 35) remarks, her views of (inter)subjectification have sometimes been misconstrued as primarily pragmatic, as in Athanasiadou, Canakis and Cornillie (2006).

Introduction

5

This fixed correlation is encountered in specific contexts, which add both pragmatic specifications and formal (e.g. prosodic) differentiations to it. This can be represented as in Figure 1, which is indebted to McGregor (1997: 44). contextual pragmatic specifications

coded meaning ‘biting/bite’ coding form fully lexical noun bit

formal (e.g. phonological) differentiations

Figure 1. Source form-meaning pair: lexical noun bit

Change begins when the original coded meaning is enriched with pragmatic values, the stage Traugott (1989) refers to as “pragmatic strengthening”, as happened with the invited inference of ‘small quantity’ associated with a bit in specific contexts. However, it is only when this meaning comes to be conventionally associated with the form that Traugott speaks of subjectification as a language change. The new form-meaning pairing presupposes reanalysis of the original lexical/grammatical form, which may be accompanied by phonological change. In the case of a bit, the form coding the new quantificational, scalar meaning is the string a bit of (phonologically rendered as abitta), whose quantifier status – as opposed to its original head function in the source construction – can be demonstrated by the fact that it can only be substituted by established quantifiers such as little/few/some, etc. This new form-meaning pairing can be represented as in Figure 2. Intersubjectification likewise starts with shifting pragmatic values in specific contexts, but can only be claimed as full-fledged change with the establishment of a new form-meaning pair. Demonstrating this for languages such as English which do not have elaborate systems of regular intersubjectivity markers providing clear models may not be so straightforward.

6

Hubert Cuyckens, Kristin Davidse and Lieven Vandelanotte

coded meaning ‘little/few/some’ coding form quantifier a bit of

Figure 2. Form-meaning pair resulting from reanalysis: quantifier a bit of

Full-fledged intersubjectification is found more in languages coding honorific meaning, as does e.g. the Japanese pronoun system.3 For English, Traugott (this volume: 37) mentions, among others, hedges arising from subjectified discourse markers such as some uses of well, perhaps and sort of. In any case, consistent with the principled distinction between lexicogrammatically coded and pragmatic meanings, she insists that intersubjectification too is about the semanticization of what were pragmatic implicatures and pragmatic values of the earlier form-meaning pair. This does not render the study of pragmatic intersubjectivity uninteresting. Meanings involving attention to the addressee’s ‘self’ have been relatively understudied and have to be systematically explored and put on the map – both those that are coded and those that are due to inference and contextualization. (Inter)subjectification often involves grammaticalization, but they are different types of changes which may occur independently of each other. Thus, cases of lexical subjectification are not accompanied by grammaticalization, as illustrated by the use of a speaker-evaluation such as pig for an impolite, slobbering eater. The non-subjective and subjective uses of pig are equally lexical. There can also be grammaticalization without subjectification: we can think here of the development of certain prepositions such as to into an infinitive-marker and by into a marker of the passive. However, the general form-meaning approach to language change advocated by Traugott also applies to grammaticalization, requiring clear semantic and formal argumentation showing that reanalysis has taken place and has led to a new form-meaning pair with (more) grammatical semantics.

3. Another example is particles marking the intimacy/distance between speaker and hearer. Korean, for instance, has particles marking six such different levels (Sohn forthcoming).

Introduction

7

In this volume the chapters by Fanego and by Wichmann, SimonVandenbergen and Aijmer focus strongly on the question of the formal coding of new (inter)subjective meanings. Fanego investigates the question of the formal evidence correlating with the semantic-pragmatic changes undergone by the Spanish manner adverbial de hecho (‘in practice, de facto, illegally’). It developed analogously to English adverbials such as in fact and particularly indeed: it first acquired sentence adverbial uses, in which it could be employed either adversatively to reject a prior position or confirmatively to signal agreement. The confirmatory reading of de hecho is the more likely source of its elaborative discourse marker use. When sentence adverbial uses developed from the manner adverbial, this was accompanied by phonological and morphosyntactic reflexes: a special intonation contour was used for the sentence adverbial which appeared in positions not normally occupied by VP-adverbs (e.g. clause initially) and it lost the syntactic behaviour associated with manner adverbs such as the possibility to be the focus of clefts and to be coordinated with another manner adverbial. By contrast, Fanego does not find clear formal reflexes of the shift from sentence adverbial to discourse marker: they are not distinguished by intonation and share the same syntactic positions (initial, medial and final) in the same discourse contexts. She therefore assumes the same position for de hecho as Aijmer and Simon-Vandenbergen (2004) did for in fact. They had questioned Schwenter and Traugott’s (2000: 21) analysis of the adversative adverb and discourse marker uses of in fact as “distinct polysemies” and proposed instead to view them “as pragmatic implicatures which are conventionalised to a greater or lesser extent, as some contextual meanings are more frequent and more conventionalised than others” (Aijmer and SimonVandenbergen 2004: 1788). Fanego likewise concludes that there is no evidence to regard the shift from the sentence adverb to the discourse marker use of de hecho as a further step in the grammaticalization process. Wichmann, Simon-Vandenbergen and Aijmer in their paper begin by noting that the empirical study of phonological form has been neglected in grammaticalization studies. Mechanisms such as reduction have been pointed to as formal reflexes, but intonation patterns are rarely invoked by way of argument, let alone systematically investigated. They want to start remedying this gap by studying the prosody of of course in relation to meaning and structural position. As to the latter, of course occurs in their – synchronic spoken – data in the three positions typically associated with discourse markers: initially, medially and finally in the utterance. Half of them occur in initial position, which is consistent with the hypothesis that

8

Hubert Cuyckens, Kristin Davidse and Lieven Vandelanotte

subjectifying elements tend to migrate to the left periphery in a language such as English (Traugott this volume: 60; see also below). To get a handle on the relation between meaning and prosody, they turn to theories of intonational meaning. Intonation theory claims that prosodic prominence is directly related to semantic weight. Therefore, they predict that grammaticalization will involve a lower frequency of stress than lexical uses. Of the tokens they examined – which are all grammaticalized uses – only a third was clearly stressed, which confirmed their prediction. As another possible corollary of discourse marker status, they consider realization by a separate intonation unit, one of the few hypotheses touching on prosody to have been formulated in the literature on discourse markers (e.g. Altenberg 1990). This hypothesis was not confirmed by the of course data as the majority were intonationally integrated. They explain this by the fact that discourse markers are often unstressed, which renders it unlikely for them to have separate tone units. They also investigate the correlation between prosody and the basic semantic classes of of course, viz. subjective epistemic-evidential versus intersubjective uses invoking shared knowledge. Here their prediction is that prominence will be more common with the epistemic uses, which have more residual lexical meaning, than with the intersubjective uses, which have bleached further. These expectations were confirmed by statistically significant results, showing again the mileage that can be got out of prosody for the study of grammaticalization and (inter)subjectification. 3. Subjectivity and subjectification With the recent surge of interest in subjectivity, there is, as stressed by De Smet and Verstraete (2006), a danger of the notion being invoked indiscriminately, without an eye to basic distinctions such as pragmatic and coded meaning. The term ‘subjectivity’ subsumes various ways in which an expression may involve speaker-reference: for instance, a word may contain a speaker-related, appreciative or pejorative component as part of its semantic content (e.g. snug, idiot); or it may encode nothing but speaker-attitude with respect to a particular entity or state of affairs (e.g. bloody expressing negative stance towards an entity; may expressing epistemic stance with respect to a state of affairs); or different still, it may function as a deictic element with the speaker as the (implicit) reference point (e.g. the determiner the signalling the speaker’s assessment of an entity’s identifiability;

Introduction

9

tense markers anchoring a state of affairs to the speaker’s time of utterance). Importantly, in a general sense, subjectivity pervades all linguistic expressions, and all language is thus subjective by definition. Langacker (1991, 2008: 55–89) argues that a linguistic encoding of a scene, and thus an expression’s meaning, is not solely a matter of conceptual content but also evokes how that content is construed (or how that scene is viewed). The various facets of construal include: (i) choice of degree of specificity of linguistic expressions, e.g. rat versus rodent; (ii) focus on particular portions of a scene by foregrounding them and backgrounding others (in information structural terms: given vs. new, or topic vs. focus); (iii) degree of prominence of participants in a scene, resulting in different alignments (e.g. A is above B vs. B is below A); (iv) perspective or viewing arrangement, an important component of which is the vantage point (by default, the actual location of speaker or hearer in the spatial domain, and the time of speaking in the temporal domain). It can readily be seen that in these various construal phenomena, the speaker or conceptualizing subject plays an important role: an expression’s specificity of lexicalization, its information structure, and the relative prominence of its participants are speakerdependent (or even speaker-imposed), and can thus be said to be subjective. A similar view is expressed by Traugott and Dasher (2002: 20): “Synchronically, SP/W [speaker/writer] selects not only the content, but also the expression of that content – which entity is chosen as syntactic subject, whether topicalization is used, present or past tense, etc.” In a general sense, then, language can be said to be strongly marked by subjectivity in that any selection from the lexical and/or grammatical repertoire passes through the speaker. De Smet and Verstraete (2006: 384) also distinguish this general notion of subjectivity “which is inherent in language use and is independent of the semantics of a particular expression”. Verstraete (2007: 10) further explains it as follows: “the choice of specific lexical items or construction types to refer to a particular entity or situation always imposes a specific perspective on that entity or situation, and it is part of the speaker’s role in the utterance that he/she is the one who bears responsibility for choosing that specific perspective.” As this general type of subjectivity is inherent “in the use of any expression” (De Smet and Verstraete 2006: 384), they refer to it as ‘pragmatic subjectivity’. It is, in Traugott’s terms, non-coded subjectivity, or subjectivity not conventionalized as a form-meaning pair.

10

Hubert Cuyckens, Kristin Davidse and Lieven Vandelanotte

In contrast to this general, or ‘pragmatic’, subjectivity, De Smet and Verstraete (2006) distinguish two types of what they call ‘semantic subjectivity’ (cf. Traugott’s ‘semanticized’ subjectivity), in which speakerrelatedness is part “of the expression’s inherent meaning” (2006: 385), viz. ‘ideational’ and ‘interpersonal’ subjectivity. This terminology refers to Halliday’s (1994) distinction between the ideational functional component of language, which is concerned with the representation of our experience of the world, and the interpersonal component, which expresses speakerpositioning and speaker-interlocutor interaction. Ideational subjectivity “involves the description of a content internal to the speaker” (De Smet and Verstraete 2006: 385). Typical examples are evaluative adjectives and nouns such as stupid and pig. Ideational subjectivity in fact corresponds roughly to Traugott’s (1989: 34) ‘internal propositional’ meaning, defined as “meanings based in the internal (evaluative/perceptual/cognitive) described situation”. Interpersonal subjectivity refers to expressions that enact speaker position with respect to a particular content, such as deontic and epistemic auxiliaries, discourse markers, illocutionary speech acts, modal adverbs, and intensifying uses of adjectives such as bloody, complete. Importantly, De Smet and Verstraete’s division is not just semantic, but the ideational and interpersonal expressions show divergent syntactic behavior. For her definition of subjectivity, Traugott’s (this volume: 33; 2003: 125) starting point is – and always has been – Lyons’s (1982: 102) characterization: “The term subjectivity refers 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 his own attitudes and beliefs” (Lyons 1982: 102). This standard definition contains a more general component, “the locutionary agent’s expression of himself”, which is coordinated with a more specific component, “and his own attitudes and beliefs”. This creates a certain vagueness, allowing for a more narrow or a broader definition. The narrow interpretation finds support in the information Lyons provides regarding his characterization of subjectivity: “selfexpression is nothing other than the expression, or externalization, of one’s beliefs, attitudes, etc.” (1982: 103), and it is also reflected in Traugott’s definition of the process of subjectification as a process whereby “meanings become increasingly based in the speaker’s subjective belief state/attitude towards what the speaker is talking about” (1989: 35; repeated in Traugott 1995: 31; our emphasis).

Introduction

11

In Traugott’s various case studies of subjectivity and subjectification (e.g., 1989, 1995, 2003), as summarized in Traugott and Dasher 2002, the main grammatical expressions of speaker-stance discussed are the following: – auxiliaries and quasi-auxiliaries expressing deontic and/or epistemic modality, e.g. may, can, will, must, ought to, going to (Traugott 1989; Traugott and Dasher 2002: 105–151, and references therein; Hopper and Traugott 2003; see also Krug 2000); – various discourse particles expressing the speaker’s epistemic attitude; e.g. the epistemic phrases I think and I guess (Thompson and Mulac 1991; Traugott 1995; Vandelanotte 2004), stance adverbs such as actually, strictly, loosely, really, frankly (see also Powell 1992); – discourse markers as explicit markers of speaker attitude towards discourse structure, i.e. towards the relationship between what precedes and what follows, or of the connectivity between propositions (cf. Traugott and Dasher 2002: 23, 153–189); examples are in fact, indeed, besides, actually, well; – hortative let’s, as in Let’s go, which has come to express the speaker’s (condescending) support; – rather than expressing speaker preference, and scalar adverbs such as even reflecting the speaker’s relative ranking of alternatives (Traugott 1995); – subordinating conjunctions such as while expressing the speaker’s attitude of surprise (Traugott and König 1991; Traugott 1995). Examples of lexical coding of speaker attitude and belief (Traugott 2003: 125) are illocutionary speech-act verbs (including performatives), such as the “assertives” observe, insist, state, claim, and hypothesize, whose meaning is similar to that of epistemic modals, and the “directives” (e.g. request, command, insist), functioning similarly to deontic modals (see also Traugott 1989: 43–45; Traugott and Dasher 2002: 190–225). Other lexical expressions with subjective meaning are modal adverbs such as obviously, possibly, probably, evidently, apparently (Traugott 1989: 46–47), which express the speaker’s epistemic attitude with respect to a state of affairs, and nouns and adjectives expressing the speaker’s evaluation; e.g. bloody in This bloody car and complete in He is a complete idiot.

12

Hubert Cuyckens, Kristin Davidse and Lieven Vandelanotte

To this set, Traugott has more recently added linguistic expressions marking spatial, temporal, and person deixis4 (Traugott and Dasher 2002: 23, Traugott 2003: 125). This widening of the class can be accommodated by the broader definition invoking the speaker’s expression of self. Given that the basic function of deixis is to relate entities and states of affairs to a spatio-temporal zero-point or deictic centre – typically egocentric or involving the speaker – and that deictic expressions are thus (typically) speaker-based, they can also be viewed as expressions of the speaker’s self, and thus as subjective. Note that Traugott’s inclusion of deictic expressions in the category of subjective expression ties in with Lyons’s view of the interdependence of deixis and subjectivity, and ultimately goes back to Bühler (1934), who “insists explicitly on the expressive and subjective nature of deixis” (Lyons 1982: 106). A cover term that nicely captures the existence of deictic as well as non-deictic subjective expressions, and which is also used by Traugott, is ‘perspective’ or ‘point of view’. As such, subjectivity can be said to explicitly encode “SP[eaker]/W[riter]’s point of view, for example in deixis, modality, and marking of discourse strategies” (Traugott and Dasher 2002: 21–22).5 This volume presents new analyses of the development of subjective meanings in almost all of the areas just discussed: – discourse markers de hecho (‘indeed, in fact’) (Fanego), of course (Wichmann, Simon-Vandenbergen and Aijmer), well (Defour) and kind of/kinda (Margerie) – additive versus counterargumentative too (Schwenter and Waltereit) – volitive and non-volitive modals (Narrog), (‘d) better (Denison and Cort) – scalar intensifying uses of kind of/kinda (Margerie) and complete, total, whole (Ghesquière) – ille and the French definite article (Carlier and De Mulder) As they insert themselves in a tradition established over a few decades, it is intriguing to see how the same new accents and nuances of the main 4. Person deixis is not mentioned explicitly, but its subjective status can be inferred from the following quote: “I and you are crucially grounded in the point of view of the speaker, and so they exhibit subjectivity, as do all deictics” (Traugott and Dasher 2002: 22). 5. The term is also used, for instance, in Finegan (1995) and De Smet and Verstraete (2006).

Introduction

13

hypotheses guiding that research tradition crop up in these studies. One recurring observation is that the source expression of subjectified uses is rarely purely propositional, but may already have subjective elements, at least on the pragmatic level, associated with it (e.g. Defour; Carlier and De Mulder; Schwenter and Waltereit). Another is that the source may be formed not by one but by several expressions, and that it is the complex interaction between them that leads to the new-form meaning pairing with subjective meaning (Denison and Cort; Defour; Margerie). Some studies integrate gradient and indeterminate uses explicitly in their analysis rather than representing change in terms of discrete classifications (e.g. Wichmann et al.; Denison and Cort; Ghesquière). It is also striking that many studies identify the contexts leading to items coding new subjective meanings in terms of lexical collocations and their positive or negative semantic prosody, e.g. Traugott, Ghesquière, Defour and Margerie (see also Lorenz 2000; Brems 2003, 2010). Margerie pays particular attention to the role played by function words, viz. determiners, in kind of’s differentiation into distinct uses.6 All these tendencies converge under the heading of close study of the micro-processes of subjectification and grammaticalization. While they may, at first sight, appear as mere nuances of existing hypotheses, they may in time lead to a refashioning of them. For instance, processes of change may come to be seen as taking place along multiple interlocking paths, rather than along unitary paths (see also Vandewinkel and Davidse 2008). 4. Intersubjectivity and intersubjectification Just as all language use can be said to be subjective in a general sense, it is intersubjective in a general sense, reflecting the impact of the speech situation which not just involves a speaker but a communicative relationship between speaker and hearer. Benveniste (1971 [1958]) saw this speaker– hearer dyad – and in particular the speaking subject’s awareness and attention to another participant as speaking subject – as a fundamental condition 6. Firth (1957), whose pathfinding insights into collocation have been coming to fruition in the last decades, would have spoken of “colligation” in this case, i.e. “groups of words considered as members of word classes relating to each other in syntactic structures” (Robins 1971: 225). However, it may well be that the linguistic community will choose to use the term “collocation” for patterns of co-occurrence with both lexical and grammatical items.

14

Hubert Cuyckens, Kristin Davidse and Lieven Vandelanotte

for linguistic communication (see also Traugott 2003: 128, this volume: 32). In a similar vein, Schiffrin (1990: 142) has characterized intersubjectivity as mutual information management (the “interaction between what an actor does … and an audience’s interpretation of all available information”, Schiffrin 1990: 142). Traugott (2003: 128, this volume: 32) also stresses that intersubjectivity is a general characteristic of all language use: “Of course, in a general sense the very fact of communicating with another person entails general intersubjectivity.”7 As well, it is intersubjective speech situations that provide the crucial context for ‘invited inferences’ (cf. Traugott and König 1991). In this volume the main focus, however, is on the types of linguistic markers that encode intersubjective meaning: “Intersubjectivity is the explicit expression of the SP/W’s attention to the ‘self’ of addressee/reader in both an epistemic sense (paying attention to their presumed attitudes to the content of what is said), and in a more social sense (paying attention to their ‘face’ or ‘image needs’ associated with social stance and identity)” (Traugott 2003: 128). As examples of coded intersubjectification in English, Traugott (this volume: 37) refers to hedging uses of well, perhaps and sort of. Three studies in this volume include discussions of these or related cases, viz. well (Defour), of course (Wichmann et al.), and kind of/kinda (Margerie). Wichmann et al. make a basic distinction between subjective uses of of course that index speaker-judgement and intersubjective ones which draw in the hearer’s expectations and knowledge or take the hearer’s face into account. The subjective uses cluster around epistemic-evidential senses such as ‘as a natural result’ or ‘predictably’, in which the original lexical meaning of course (‘path taken by e.g. a river’) persists. The intersubjective uses include hearer-involving extensions of the evidential senses (‘as you may expect’, ‘as you and I know’) as well as more bleached markers of speech-as-interaction (cf. you know, Holmes 1988: 69) which co-occur with fillers and hedges. Defour shows that the hearer-oriented uses of well had several sources, the main ones being the now obsolete adjectival use expressing acceptance of the previous speaker’s turn and adverbial well in expressions such as 7. On other, general interpretations, intersubjectivity is seen as referring to evaluations/evidence “known to or accessible to a larger group of people who share the same conclusion as the SP[eaker]/W[riter]” (Traugott 2003: 128; see also Nuyts 2005).

Introduction

15

well said, which could serve a similar pragmatic function. To the hearerturn accepting meaning were added disagreeing or concessive comments, e.g. Well … but …, in contexts in which speaker and hearer have diverging views. In a later stage well came to express on its own this combination of acceptance and concession. Continuous pragmatic diversification thus lay at the basis of the semanticization of intersubjective meanings such as facethreat mitigator, qualifier and pause filler (as well as of subjective uses not recapitulated here). In her synchronic study of kind of/kinda, Margerie stresses that the hedging uses are situated on a cline in which the ‘source’ uses, nominal and intensifier, already show a mix of propositional and pragmatic (inter)subjective meanings. Its coded intersubjective uses, whose analogy to (just) like is noted, are further subdivided into approximator, hedge, focus marker and filler. In the light of Traugott’s (this volume: 59) point that imperative utterances are “in themselves initially intersubjective” but “may be subjectified”, it is interesting to consider Denison and Cort’s discussion of the deontic uses of modal auxiliary ‘(d) better. These have invited inferencing of co-opting the hearer into accepting the advice, or of being held responsible and suffering adverse consequences for non-compliance. This “suggest[s] a great degree of intersubjectivity” (379), which is, however, situated on the pragmatic contextual level. They also point out that deontic modality pragmatically requires, besides the speaker, a person on whom the obligation falls. On this count too, their conclusion is that the deontic meanings are only pragmatically intersubjective. Traugott (2003: 129) has further noted that “in so far as subjectification involves the recruitment of meanings not only to encode but also to regulate attitudes and beliefs, it inevitably involves intersubjectivity to a certain degree”. In accordance with this, Denison and Cort (this volume: 379) note that epistemic/deontic uses of ‘(d) better can be seen to have intersubjective inferences, as “speaker/writer projects the hope [of a favourable outcome] onto the addressee as well”. Whereas all the analyses of intersubjectivity/intersubjectification just discussed adhere to Traugott’s definitions, two contributions argue for a revision of these definitions. Both do this with reference to the development of determiner meanings, Carlier and De Mulder in dealing with distal Latin ille as source of the definite article in French and Ghesquière in characterizing the secondary determiner uses of completeness adjectives. They invoke, in this context, Diessel’s (1999, 2006) cognitively-oriented work on

16

Hubert Cuyckens, Kristin Davidse and Lieven Vandelanotte

deictics, pointing out that demonstratives serve to coordinate the speaker and hearer’s joint attention on a particular reference object. Similarly, the hearer’s focus of attention is an integral part of Langacker’s (1991: 91) definition of definite reference: “the speaker (S) and hearer (H) … face the task of coordinating their mental reference to some instance … .” One could add to this the emphasis on the great amount of interactional work done by speaker and hearer to ensure that the hearer focuses on the intended referent in studies of talk-in-interaction (e.g. Enfield and Stivers 2007). The attention given by the speaker to the hearer’s awareness and tracking of referents is, in other words, well-established in various linguistic approaches. Carlier and De Mulder and Ghesquière propose that extending intersubjectivity to the meaning of determiners is coherent with the general approach espoused in this volume. Carlier and De Mulder (this volume: 268) recall Traugott’s (1995: 47) definition of subjectification as “the tendency to recruit lexical material for purposes of creating text and indicating attitudes in discourse situations”, and they subscribe to Breban’s (2010: 111) position that “attitudinal and textual subjectivity are … two different manifestations of the same phenomenon neither of which can be excluded at the expense of the other”. In another wording of Breban (2010: 111), “subjectivity involves text showing the signs of the presence of the speaker. This general definition of subjective meaning covers both attitudinal and textual subjectivity.” Like Breban (2010), Ghesquière treats the meaning of all determining elements as generally intersubjective. According to Carlier and De Mulder, by contrast, some determiners may encode a predominantly speakeroriented meaning and others a more hearer-oriented one. In their account of the transition of Latin ille to the French definite article, they distinguish two stages. In the first stage, ille, which already in Classical Latin could index positive or negative speaker-attitude, came to be used as a marker of discourse prominence, which they treat as subjectification since it involves mainly speaker attitude. In the second stage, ille, in virtue of its deictic force which evokes ‘new’ or ‘renewed’ identification by the hearer, came to be used with first mentions of proper names and with recognitional uses of first mentions followed by relative clauses as well as with associative anaphoric uses. Carlier and De Mulder characterize this stage as intersubjectification, marked by a shift to the perspective of the hearer who shares the knowledge of the whole speech community. Not only is intersubjectivity a general characteristic of all language use, it is also “the ambient context in which linguistic change takes place and to

Introduction

17

which linguistic change contributes” (Traugott this volume: 32). As Traugott (ibid.) explains, “it is precisely to emphasize the intersubjectivity of the speech situation that I have referred to ‘invited inferences’ rather than ‘implicatures’ in theorizing semantic change as the semanticization of pragmatics”. In their contribution, Schwenter and Waltereit take this point further and point out that hearers can play an active role by assigning novel interpretations to forms and by using these in their own subsequent turn as speakers. They illustrate this with a type of pragmatic meaning that has received much less attention than conversational implicatures in the study of language change, viz. presupposition: “Hearers unable or unwilling to accommodate presuppositions assume a novel interpretation of the erstwhile presupposition-trigger and eventually pass this new interpretation on to other people, thereby changing the language” (Schwenter and Waltereit this volume: 80). Their study focuses on additive particles, mainly English too. The presuppositional structure of additive too invokes parallelism between the sentence it is used with and a context sentence, which are both answers to an (explicit or implicit) question given in the discourse, i.e. they have shared topicality. In bridging contexts including a third proposition, too acquired, besides its additive value vis-à-vis a parallel sentence, a refutational value vis-à-vis the third proposition. Due to backgrounding of the original additive sense, too was reanalysed as a novel form-meaning pairing, which denies a previous utterance, as in You didn’t do your homework. – I did too. 5. The (uni)directionality of change Entrenched in the diachronic perspective on (inter)subjectification is the issue of the directionality of these changes. Again, theory formation by Traugott serves as a reference point and its developments are revisited by Fanego and Traugott herself in this volume. Traugott’s early theorizing of subjectification drew on Halliday and Hasan’s (1976) proposal that, synchronically, there are three functional domains of language, the ideational, the textual, and the interpersonal. It was Traugott’s (1982) insight that, diachronically, lexical items which originate in the ideational domain tend to acquire textual and interpersonal meanings. Instead of the terms ‘ideational’ and ‘interpersonal’, however, she preferred to use ‘propositional’ and ‘expressive’ respectively. While this cline did not yet include the term ‘subjective’, there is a good deal of overlap between ‘expressiveness’ and

18

Hubert Cuyckens, Kristin Davidse and Lieven Vandelanotte

subjectivity (cf. Traugott and Dasher 2002: 94).8 The semasiological development she proposed, then, was represented by means of the following cline: propositional > (textual >) expressive (Traugott 1982: 257)

Later, she proposed a more complex model, which included several correlated clines and in which intersubjectification appears as a distinct stage presupposing subjectification (see Fanego this volume: 202): Table 1. Correlated paths of directionality in semantic change (Traugott and Dasher 2002: 40) truth-conditional

>

non-truth-conditional

content

> content/procedural

> procedural

scope within proposition

> scope over proposition

> scope over discourse

nonsubjective

> subjective

> intersubjective

The replacement of the propositional > textual > expressive path by the non-subjective > subjective > intersubjective cline has raised questions for some authors, including Carlier and De Mulder and Ghesquière in their studies of determiners. As noted above, both studies insist on keeping the textual as an essential distinction on the cline, and argue that determiners encode textual intersubjective meaning. Ghesquière’s study, which deals with all the various uses developed by completeness adjectives, focuses on the question of how these relate to Traugott’s earlier and more recent hypotheses of unidirectionality, and how textual intersubjectification can be fitted into them. She relates the unidirectionality hypothesis to the leftward hypothesis, which predicts that increasing subjectification goes together with structural movement to increasingly peripheral positions, viz. towards the left periphery in a language such as English (Adamson 2000). Starting from objective descriptive modifier uses, the adjectives complete, total and whole developed subjective descriptive modifier uses, subjective intensifying uses and secondary determiner uses (which she views as textually intersubjective; see above). Complete and total developed their noun-intensifying uses from subjective descriptive uses. Both these adjectives had earlier on derived secondary determiner uses from objective descriptive modifying meanings but the secondary 8. For a detailed comparison of Halliday and Hasan’s and Traugott’s terminology, see Traugott and Dasher (2002: 94).

Introduction

19

determiners played no role in the shift to noun-intensifying uses. With whole, by contrast, objective descriptive modifier uses led to secondary determiner uses, which in turn led to noun-intensifying uses. This chronology, Ghesquière argues, is nicely captured by Traugott’s (1989) subjectification cline, which also accommodates all the relevant meaning shifts, viz. from externally propositional (objective descriptive modifier) to internally propositional (subjective descriptive modifier), textual (secondary determiner) and expressive (noun-intensifier). She proposes one modification in that both the textual and expressive are argued to comprise subjective as well as intersubjective meaning. This gives the following cline (Ghesquière this volume: 309): externally propositional objective

> internally propositional subjective

non-bleached

> textual > expressive subjective and intersubjective bleached

Figure 3. Model of semantic change

As for Adamson’s leftward hypothesis, it is confirmed in a general sense in that all (inter)subjectified uses involve leftward movement in the structure of the NP. However, there is not always consistent progressive leftward movement for the fine-grained subjective functions developed by these adjectives. Whole, for instance, first acquired the more leftward secondary determiner use, which then led to the noun-intensifying use, which occupies a structural position to the right of the former. Another contributor who deals with the question of unidirectionality is Narrog. He too proposes modifications of the Traugottian pathways, in his case on the basis of crosslinguistic study of semantic change within the area of modality and from modality to mood. He starts from the observation that prominent studies of subjectification (e.g. Langacker 1998, 2003; Traugott 1989) have focused on the increased subjectification in changes from deontic to epistemic modality manifested by modal verbs such as English must and ought to, but that this change is not very common crosslinguistically. If this change is viewed as a general part of unidirectionality, the typologically more common changes, e.g. from epistemic possibility to deontic permission (Bybee et al. 1994), are made to appear counter-directional. Narrog therefore advocates an alternative general tendency, viz. to increased speaker-orientation. He relates this to a semantic map of modality which distinguishes between volitive and non-volitive modality and links up mod-

20

Hubert Cuyckens, Kristin Davidse and Lieven Vandelanotte

ality with mood (see also Narrog 2005). In the modal part of this map obligation, permission and volition are volitive, while epistemic modality, ability and evidentiality are non-volitive. Irrespective of the volitive/nonvolitive distinction, modality ranges from event-oriented to speakeroriented. The map also integrates modality with mood, the latter understood functionally as illocutionary force modification, subsuming moods such as imperative and admonitive. The dimension of increased speaker-orientation which applies to this map “refers to the speaker her- or himself and the speech situation, including the hearer, and thus subsumes both subjectivity and intersubjectivity in Traugott’s sense” (this volume: 394). As Narrog has argued elsewhere (2005: 692), “as speaker-orientation means orientation towards the speaker and the speech situation, it potentially also includes an increasing orientation towards the addressee”. According to Narrog, modal meanings always shift in the direction of increased speaker-orientation, independently of the dimensions of volitive/non-volitive, but incorporating mood. This model does capture the shifts that are crosslinguistically most common according to Bybee et al. (1994), viz. (i)

from non-volitive to volitive, including epistemic to deontic, where the shift often incorporates a change from no control to control over the action (ii) within volitive, e.g. from obligation emanating from external circumstances or higher (religious, social) forces to obligation imposed by a person’s will (cf. Traugott and Dasher 2002: 125, 141). To this category also belongs the change from obligation to imperative, in which speaker-assessment of obligation shifts to direct imposition on the hearer, interpreted as a higher degree of orientation on the speech situation and its participants. (iii) within non-volitive, e.g. from ability/root possibility to epistemic possibility and from future prediction to probability – changes in which the elements of speaker inference and expectation increase.

Narrog further reflects about the cultural embeddedness of notions such as (personal) imposition of obligation. Strong obligation markers are crosslinguistically not very common and even if available in a language, often not commonly used. Narrog ascribes this to the fact that obligations are impositions on the hearer; they are potentially face-threatening and their direct expression might therefore be avoided. Moreover, the very notion of requiring, on personal authority, another person to carry out an action is not congenital to many cultures, and was not either to older speech communities of languages such as English and German. An important cultural historical

Introduction

21

alternative is that of necessities/possibilities being determined by natural and religious laws, morals, shared values, etc. Narrog therefore ends on the salutory plea to include sociocultural conditions into the pragmatic contexts in which to study semantic change. In this introduction we have surveyed the main questions relating to grammaticalization and (inter)subjectification tackled by the contributions. As we have seen, they raise issues (such as the question of textual intersubjectivity) which form a higher-level ‘intersubjective’ dialogue inspiring further theoretical probing. By the same token, it has been impossible to do justice to the descriptive thorougness and insightful analyses of the changes and/or layering of uses of the linguistic expressions studied. It is our hope that both in terms of theoretical reflection and descriptive technique, this volume will prove an important stepping stone towards greater insight into processes and tendencies of language change and in their relevance to understanding language generally. References Adamson, Sylvia 2000 A lovely little example: Word order options and category shift in the premodifying string. In Pathways of Change: Grammaticalization in English, Olga Fischer, Anette Rosenbach, and Dieter Stein (eds.), 39–66. Amsterdam/Philadelphia: Benjamins. Aijmer, Karin and Anne-Marie Simon-Vandenbergen 2004 A model and a methodology for the study of pragmatic markers: The semantic field of expectation. Journal of Pragmatics 36: 1781–1805. Altenberg, Bengt 1990 Automatic text segmentation into tone units. In The London–Lund Corpus of Spoken English: Description and Research, Jan Svartvik (ed.), 287–323. Lund: Lund University Press. Athanasiadou, Angeliki, Costas Canakis, and Bert Cornillie (eds.) 2006 Subjectification: Various Paths to Subjectivity. Berlin/New York: Mouton de Gruyter. Athanasiadou, Angeliki, Costas Canakis, and Bert Cornillie 2006 Introduction. In Athanasiadou, Canakis, and Cornillie (eds.), 1–13. Benveniste, Emile 1966 De la subjectivité dans le langage. In Problèmes de Linguistique Générale, Emile Benveniste, 258–266. Paris: Gallimard.

22

Hubert Cuyckens, Kristin Davidse and Lieven Vandelanotte

Benveniste, Emile 1971 Subjectivity in language. In Problems in General Linguistics, 223– 230. Trans. by Mary Elizabeth Meek. Coral Gables: FL: University of Miami Press. Original French publication, 1958. Bréal, Michel 1964 Semantics: Studies in the Science of Meaning. Trans. by Mrs. Henry Cust. New York: Dover. Original publication, 1900. Breban, Tine 2010 English Adjectives of Comparison: Lexical and Grammaticalized Uses. Berlin/New York: Mouton de Gruyter. Brems, Lieselotte 2003 Measure noun constructions: An instance of semantically-driven grammaticalization. International Journal of Corpus Linguistics 8: 283–312. Brems, Lieselotte 2010 Size noun constructions as collocationally constrained constructions: Lexical and grammaticalized uses. English Language and Linguistics 14: 83–109. Bühler, Karl 1934 Sprachtheorie. Jena: Fischer. Bybee, Joan, Revere Perkins and William Pagliuca 1994 The Evolution of Grammar: Tense, Aspect, and Modality in the Languages of the World. Chicago: Chicago University Press. De Smet, Hendrik and Jean-Christophe Verstraete 2006 Coming to terms with subjectivity. Cognitive Linguistics 17: 365– 392. Diessel, Holger 1999 Demonstratives. Form, Function, and Grammaticalization. Amsterdam/Philadelphia: Benjamins. Diessel, Holger 2006 Demonstratives, joint attention and the emergence of grammar. Cognitive Linguistics 17: 463–489. Enfield, Nicholas J. and Tanya Stivers (eds.) 2007 Person Reference in Interaction: Linguistic, Cultural and Social Perspectives. Cambridge: Cambridge University Press. Finegan, Edward 1995 Subjectivity and subjectivisation: An introduction. In Subjectivity and Subjectivisation, Dieter Stein and Susan Wright (eds.), 1–15. Cambridge: Cambridge University Press. Firth, John Rupert 1957 A synopsis of linguistic theory, 1930–1955. In Studies in Linguistic Analysis (Special volume of the Philological Society), John R. Firth et al. (eds.), 1–32. Oxford: Blackwell.

Introduction

23

Halliday, M. A. K. 1994 An Introduction to Functional Grammar. 2nd ed. London: Arnold. Halliday, M. A. K. and Ruqaiya Hasan 1976 Cohesion in English. London: Longman. Holmes, Janet 1988 Of course: a pragmatic particle in New Zealand women’s and men’s speech. Australian Journal of Linguistics 2: 49–74. Hopper, Paul J. and Elizabeth Closs Traugott 2003 Grammaticalization. 2nd rev. ed. [1st ed., 1993.] Cambridge: Cambridge University Press. Krug, Manfred 2000 Emerging English Modals: A Corpus-Based Study of Grammaticalization. Berlin/New York: Mouton de Gruyter. Lakoff, George 1987 Woman, Fire, and Dangerous Things: What Categories Reveal about the Mind. Chicago: University of Chicago Press. Langacker, Ronald W. 1985 Observations and speculations on subjectivity. In Iconicity in Syntax, John Haiman (ed.), 109–150. Amsterdam/Philadelphia: Benjamins. Langacker, Ronald W. 1990 Subjectification. Cognitive Linguistics 1: 5–38. Langacker, Ronald W. 1991 Foundations of Cognitive Grammar. Volume 2: Descriptive Application. Stanford: Stanford University Press. Langacker, Ronald W. 1998 On subjectification and grammaticalization. In Discourse and Cognition: Bridging the Gap, Jean-Pierre Koenig (ed.), 71–89. Stanford: CSLI. Langacker, Ronald W. 1999 Losing control: grammaticization, subjectification, and transparency. In Historical Semantics and Cognition, Andreas Blank and Peter Koch (eds.), 147–175. Berlin/New York: Mouton de Gruyter. Langacker, Ronald W. 2002 Deixis and subjectivity. In Grounding: The Epistemic Footing of Deixis and Reference, Frank Brisard (ed.), 1–28. Berlin/New York: Mouton de Gruyter. Langacker, Ronald W. 2003 Extreme subjectification: English tense and modals. In Motivation in Language: Studies in Honor of Günter Radden, Hubert Cuyckens, Thomas Berg, René Dirven and Klaus-Uwe Panther (eds.), 3–26. Amsterdam/Philadelphia: Benjamins.

24

Hubert Cuyckens, Kristin Davidse and Lieven Vandelanotte

Langacker, Ronald W. 2006 Subjectification, grammaticization, and conceptual archetypes. In Athanasiadou, Canakis, and Cornillie (eds.), 17–40. Langacker, Ronald W. 2008 Cognitive Grammar: A Basic Introduction. Oxford: Oxford University Press. Lorenz, Gunter 2002 Really worthwhile or not really significant? A corpus–based approach to delexicalization and grammaticalization of intensifiers in Modern English. In New Reflections on Grammaticalization, Ilse Wischer and Gabriele Diewald (eds.), 143–161. Amsterdam/Philadelphia: Benjamins. Lyons, John 1977 Semantics. Cambridge: Cambridge University Press. Lyons, John 1982 Deixis and subjectivity: Loquor, ergo sum? In Speech, Place, and Action: Studies in Deixis and Related Topics, Robert J. Jarvella and Wolfgang Klein (eds.), 101–124. New York: Wiley. McGregor, William B. 1997 Semiotic Grammar. Oxford: Clarendon Press. Narrog, Heiko 2005 Modality, mood, and change of modal meanings – a new perspective. Cognitive Linguistics 16: 677–731. Nuyts, Jan 2001 Epistemic Modality, Language and Conceptualization. Amsterdam/Philadelphia: Benjamins. Nuyts, Jan 2005 Modality: Overview and linguistic issues. In The Expression of Modality, William Frawley (ed.), 1–26. Berlin/New York: Mouton de Gruyter. Powell, Mava J. 1992 The systematic development of correlated interpersonal and metalinguistic uses in stance adverbs. Cognitive Linguistics 3: 75–110. Robins, Robert H. 1971 General Linguistics: An Introductory Survey. 2nd ed. London: Longman. Schiffrin, Deborah 1990 The principle of intersubjectivity in communication and conversation. Semiotica 80: 121–151. Schwenter, Scott A. and Elizabeth Closs Traugott 2000 Invoking scalarity: The development of in fact. Journal of Historical Pragmatics 1: 7–25.

Introduction

25

Sohn, Sung-Ock Forthc. The role of frequency and prosody in the grammaticalization of Korean -canh-. In Formal Evidence in Grammaticalization Research, An Van linden, Jean-Christophe Verstraete and Kristin Davidse (eds.). Amsterdam/Philadelphia: Benjamins. Stein, Dieter and Susan Wright 1995 Subjectivity and Subjectivisation. Cambridge: Cambridge University Press. Thompson, Sandra A. and Anthony Mulac 1991 A quantitative perspective on the grammaticization of epistemic parentheticals in English. In Approaches to Grammaticalization. Volume II: Focus on Types of Grammatical Markers, Elizabeth Closs Traugott and Bernd Heine (eds.), 313–329. Amsterdam/Philadelphia: Benjamins. Traugott, Elizabeth Closs 1982 From propositional to textual and expressive meanings: Some semantic-pragmatic aspects of grammaticalization. In Perspectives on Historical Linguistics, Winfred P. Lehmann and Yakov Malkiel (eds.), 245–271. Amsterdam/Philadelphia: Benjamins. Traugott, Elizabeth Closs 1989 On the rise of epistemic meanings in English: An example of subjectification in semantic change. Language 57: 33–65. Traugott, Elizabeth Closs 1995 Subjectification in grammaticalisation. In Subjectivity and Subjectivisation, Dieter Stein and Susan Wright (eds.), 31–54. Cambridge: Cambridge University Press. Traugott, Elizabeth Closs 1997 Subjectification and the development of epistemic meaning: The case of promise and threaten. In Modality in Germanic languages, Toril Swan and Olaf Jansen Westvik (eds.), 185–210. Berlin/New York: Mouton de Gruyter. Traugott, Elizabeth Closs 2003 From subjectification to intersubjectification. In Motives for Language Change, Raymond Hickey (ed.), 124–139. Cambridge: Cambridge University Press. Traugott, Elizabeth Closs and Richard B. Dasher 2002 Regularity in Semantic Change. Cambridge: Cambridge University Press. Traugott, Elizabeth Closs and Ekkehard König 1991 The semantics-pragmatics of grammaticalization revisited. In Approaches to Grammaticalization. Volume I: Focus on Theoretical and Methodological Issues, Elizabeth Closs Traugott and Bernd Heine (eds), 189–218. Amsterdam/Philadelphia: Benjamins.

26

Hubert Cuyckens, Kristin Davidse and Lieven Vandelanotte

Vandelanotte, Lieven 2004 From representational to scopal ‘distancing indirect speech or thought’: A cline of subjectification. Text 24: 547–585. Vandewinkel, Sigi and Kristin Davidse 2008 The interlocking paths of development to emphasizer adjective pure. Journal of Historical Pragmatics 9: 255–287. Verhagen, Arie 2005 Constructions of Intersubjectivity: Discourse, Syntax, and Cognition. Oxford: Oxford University Press. Verstraete, Jean-Christophe 2007 Rethinking the Coordinate-Subordinate Dichotomy: Interpersonal Grammar and the Analysis of Adverbial Clauses in English. Berlin/New York: Mouton de Gruyter.

Part 1 Theoretical issues

(Inter)subjectivity and (inter)subjectification: A reassessment Elizabeth Closs Traugott Abstract This paper reviews and updates my work on subjectification and intersubjectification over twenty-five years. Distinctions are made between subjectivity and intersubjectivity as the ambient context in which change takes place, most especially the development of expressions the prime semantic or pragmatic meaning of which is to index the speaker’s viewpoint on what she or he is saying or speaking about (subjectivity). In some cases expressions also develop marking attention to the addressee’s self-image (intersubjectivity). The hypothesis is that subjectification and intersubjectification involve the reanalysis of pragmatic meanings as coded semantic meanings in the context of speaker-hearer negotiation of meaning. Subjectification and intersubjectification are shown to be independent of processes of grammaticalization, but linked to them for reasons relating to the various functions of grammar. The paper ends with discussion of some studies suggesting the types of linguistic context in which one might expect to find evidence for subjectification.

1. Introduction1 The topic of subjectivity has been discussed in semantics at least since Bréal ([1900] 1964). Benveniste’s ([1958] 1971) landmark paper distinguished subjectivity and intersubjectivity. These are synchronic notions, and can be theorized in many ways, from cognitive construal (Langacker, e.g. 1990, 2003 and references therein) to the basis of human interaction and the procedures for producing and understanding talk (Schiffrin 1990). One branch of my own work over the twenty-five years since the publication of Traugott (1982) has been to study the semanticization over time of 1. Many thanks to Brian D. Joseph and to an anonymous reviewer for valuable comments and suggestions. Any remaining errors or failures to solve the intersubjective problem of being clear and informative are of course my own responsibility.

30

Elizabeth Closs Traugott

subjectivity, understood as relationship to the speaker and the speaker’s beliefs and attitudes,2 and of intersubjectivity, understood as relationship to the addressee and addressee’s face. I have called the diachronic process of semanticization “(inter)subjectification”, assuming that an important (though not rigid) distinction is to be made between -ity (synchronic state) and -ation (diachronic process) (see also De Smet and Verstraete 2006). I have also attempted to understand what the relationship between subjectification and intersubjectification is. In this research endeavor I have drawn extensively on neoGricean pragmatics (see e.g. Horn 1984; Levinson 2000) and on discourse analysis (see e.g. Schiffrin 1987; Prince 1988). Another branch of my work has been grammaticalization (e.g. Traugott and Heine (eds.) 1991; Hopper and Traugott [1993] 2003). In the (1982) paper, and in others since then, I have sought to see where and why (inter)subjectification and grammaticalization intersect. The work has been based in historical texts and the evidence we can draw from them, assuming that language change is change in use (see Croft 2000), and that there is a distinction between semantics and pragmatics.3 My purpose in the present paper is to outline my current thinking on the relationship between (inter)subjectivity, (inter)subjectification, and grammaticalization (see also Traugott 2007 on the first two topics). I start with discussion of (inter)subjectivity and (inter)subjectification (Section 2), and then of grammaticality and grammaticalization (Section 3). After three brief case studies (a piece of, a bit of, a shred of) (Section 4),4 I provide some thoughts about the mechanisms and motivations for grammaticalization and (inter)subjectification (Section 5). The paper ends with some consideration of suggestions about how one might go about operationalizing subjectification in historical corpus studies (Section 6), and Section 7 sums up.

2. De Smet and Verstraete (2006: 369) usefully contrast Langacker’s perspective on subjectification: “In Langacker’s terms, the difference is not whether something is speaker-related or not, but how explicitly reference to the speaker figures in the form of the utterance”. 3. Because I make this distinction between pragmatics and semantics, my purposes do not and have not coincided with those of cognitive grammar, although there are obvious points of contact (see discussion in Athanasiadou, Canakis, and Cornillie 2006, and Brisard 2006). 4. For studies of these and other binominal NP of NP phrases in the context of grammaticalization theory and construction grammar, see also Traugott 2007, 2008b, c).

(Inter)subjectivity and (inter)subjectification: A reassessment

31

2. (Inter)subjectivity, (inter)subjectification In Traugott (1982) I drew in part on Halliday and Hasan’s (1976) distinction between “ideational”, “textual” and “interpersonal” components of the linguistic system, but used the terms “propositional”, “textual”, and “expressive”. Theirs was a synchronic analysis, but the insight I had was that historically in many cases a lexical item that originated in the ideational component later developed polysemies in what Halliday and Hasan called the textual and interpersonal domains. Subsequent work has shown that both of these terms encompass two different types of structure. “Textual”, as understood then, included various connectives such as and and therefore, as well as anaphoric and cataphoric pronouns, topicalizers, relativizers, complementizers, etc. In the nineteen-eighties it became clear that while all are essential ingredients of grammar, some of these serve more contentful (and sometimes truth-conditional) purposes of local connectivity (e.g. relativizers, complementizers), whereas others serve the procedural purposes of expressing speaker’s attitude to the text under production (topicalizers, discourse markers). Indeed, many connectives have dual functions, e.g. and, then, in fact. Halliday and Hasan’s term “interpersonal” likewise covered a broad spectrum of phenomena, such as expressions of speech function, exchange structures, and attitude. In a more recent discussion, Halliday has proposed that, most simply put, “interpersonal” concerns “clause as exchange” (Halliday 1994: 179), and includes both subjective and intersubjective elements, e.g. modal, and mood-marking elements, vocative, interactive acts of speaking including illocutionary acts, deictic person pronouns, attitudinal lexical items like splendid, and prosodic voice features. In Traugott (1982) I preferred the term “expressive” to “interpersonal” since it was unclear to me where the “inter-” fit in chronologically. Later, following Benveniste (1971) I came to distinguish “subjective” and “intersubjective” (Traugott and Dasher 2002; Traugott 2003a), and I will make this distinction here. Historically, and, I would argue, synchronically, there is a difference with respect to what has to be learned, and therefore specified in the inventory,5 between subjective possibly, even (markers of speaker assessment) and intersubjective please (a marker of speaker’s acknowledgment of and attention to the addressee). 5. I use the term “inventory” in preference to “lexicon”, since the latter invokes only lexical items, while “inventory” is a neutral term covering lexical and grammatical items (Brinton and Traugott 2005: 90).

32

Elizabeth Closs Traugott

Of course, in a general sense the very fact of communicating with another person entails general intersubjectivity. The “I” is constituted in part by conceptualizing the other member of the communicative dyad “you” (Benveniste 1971; Lyons 1994) and discourse is communicatively successful only if speakers pay attention to audience needs, and if “mutual manifestness” or “mutual management” is worked on (Schiffrin 1990; Nuyts 2001; Verhagen 2005). Indeed, it is precisely to emphasize the intersubjectivity of the speech situation that I have referred to “invited inferences” rather than “implicatures” in theorizing semantic change as the semanticization of pragmatics. The term “invited inferences” was chosen “to elide the complexities of communication in which the speaker/writer evokes implicatures and invites the addressee/reader to infer them”6 (Traugott and Dasher 2002: 5). Intersubjectivity is the ambient context in which linguistic change takes place and to which linguistic change contributes. My main concern is not with this context, but with linguistic MARKERS and EXPRESSIONS that index subjectivity and intersubjectivity and how they arise. These expressions of subjectivity and intersubjectivity are expressions the prime semantic or pragmatic meaning of which is to index speaker attitude or viewpoint (subjectivity) and speaker’s attention to addressee selfimage (intersubjectivity). At issue is the development of semantic (coded) polysemies that have to be learned with subjective or intersubjective meanings, and how these come into being. These polysemies may later be reinterpreted as homonymies (e.g. fairly ‘in a fair manner’ and ‘somewhat’), or one or more of the polysemies may cease to be used (e.g. villain ‘peasant’ and ‘evil person’), but by hypothesis most new semantic developments emerge as polysemies, pragmatic to begin with, then semantic. Subjectified polysemies may index evaluation of others (silly ‘blessed, innocent’ > ‘stupid’), of relative position on a scale (adverbs like pretty ‘cleverly’ > ‘attractively’ > ‘rather’), of attitude toward the truth of a proposition (epistemics like probably ‘provably’ > ‘in all likelihood’); they may index information structure (e.g. the topicalizer as far as), connectivity of clauses to each other (anyway), the speech act being undertaken (promise in its illocutionary uses), or the relationship of chunks/episodes of speech to each other (then in its discourse marker use). Intersubjectified polysemies may index euphemisms (the Lord ‘god’, pass ‘die’, etc., see Allen and Burridge 1991), politeness (please < formulae like If you please, where the surrounding 6. Speakers/writers may of course evoke unintentionally (Keller 1994), and addressees/readers may fail to make the inference (or the appropriate inference).

(Inter)subjectivity and (inter)subjectification: A reassessment

33

context has been absorbed into the meaning of please). While I have not been concerned with culturally driven shifts in stylistic preferences, genre preferences (Fitzmaurice 2000), honorification preferences (cf. Japanese vs. Dutch or English), rhetorical play whereby speakers attempt to position interlocutors (Fitzmaurice 2004), or shifts in “habits of mind” (Wierzbicka 2006), these are all important areas of research that, among other things, require detailed knowledge of attested linguistic changes, and how to analyse them. I take my work as potentially providing the linguistic underpinnings for such larger-scale studies of the relationship between language, culture and cognition. My starting point for thinking about subjectivity is and has been Lyons’s characterization of subjectivity: The term subjectivity refers 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 his own attitudes and beliefs. (Lyons 1982: 102)

Examples include such subjective expressions as: (1)

–

– –

raising constructions, in which the “speaking subject” differs from the syntactic subject (Benveniste’s “sujet d’énonciation” vs. “sujet d’enoncé”) (She’s going to give a lecture vs. There’s going to be an earthquake) illocutionary uses of speech act and mental verbs (I recognize the Senator from California) epistemic modals (That must be wrong), concessives (while), focus particles (even), discourse markers (besides).

To adapt Lyons’s words about subjectivity, intersubjectivity in my view refers to the way in which natural languages, in their structure and their normal manner of operation, provide for the locutionary agent’s expression of his or her awareness of the addressee’s attitudes and beliefs, most especially their “face” or “self-image” (Traugott 2003a). Such intersubjective expressions include expletives (“in your face” expressions such as insults; within groups some can signal solidarity, but to or from outsiders can signal aggression, e.g. “N-words”). It should be noted that this is a somewhat different view of intersubjectivity from that of Verhagen (2005), which is based in argumentation theory, and focuses on coordination between speaker’s and addressee’s cognitive systems, rather than on differences in linguistic coding. It is an entirely different view of subjectivity and inter-

34

Elizabeth Closs Traugott

subjectivity from that of Nuyts (2001, 2005) and Cornillie (2004), and indeed orthogonal to theirs. Nuyts defines subjectivity and intersubjectivity as follows (with reference to modality, but in ways generalizable beyond this domain): [A]n evaluation is subjective if the issuer presents it as being strictly his/her own responsibility; it is intersubjective if (s)he indicates that (s)he shares it with a wider group of people, possibly including the hearer. (Nuyts 2005: 14)

This is a view of intersubjectivity that intersects with objectivity, although it is not identified with it: [I]t is often essential to be able to make it clear whether one is alone in one’s views, has backing for them (subjectivity vs. intersubjectivity), is neutral, or is subjectively biased in one’s assessment (subjectivity vs. objectivity). (Nuyts 2005: 18)

If this view of intersubjectivity were dynamicized and taken as the basis of intersubjectification one would have to argue that intersubjectification precedes subjectification historically, since generalized epistemic meanings of the type All men must die (= ‘It is necessarily true that all men must die’) precede inferential ones of the type The fruit must be delicious (= ‘I infer that the fruit is delicious’). However, on the view that intersubjectification involves coding of greater attention to the addressee, the earlier modal meaning is a minimally subjective epistemic generalization, and does not enter into discussion of intersubjectification in my sense. On my view, one may organize expressions along a cline of (inter)subjectivity as in (2). On the second line I suggest approximate matches between my terminology and that of Halliday and Hasan. (2)

non-/less subjective ideational

– –

subjective – interpersonal

intersubjective

Like all synchronic clines, this is simply a way of organizing data on a continuum. It is based on a historical cline that has emerged because diachronic work has shown repeatedly that for some lexical item or construction X,7 subjectified polysemies of that item or construction arise later than ideational ones (subjectification), and for some lexical item or construction X,

7. Here “construction” is to be understood as “syntactic string”.

(Inter)subjectivity and (inter)subjectification: A reassessment

35

intersubjectified polysemies of that item or construction arise later than subjectified ones (intersubjectification). In my view, subjectification and intersubjectification are the mechanisms by which: (3)

a. b.

meanings are recruited by the speaker to encode and regulate attitudes and beliefs (subjectification), and, once subjectified, may be recruited to encode meanings centred on the addressee (intersubjectification).

This is schematized in (4) (see e.g. Traugott and Dasher 2002: 225): (4)

non-/less subjective > subjective > intersubjective

One issue that needs to be flagged here is that the formulation in (4) obscures the fact that “(inter)subjective” does not mean ‘has pragmatic (inter)subjective meanings in relevant contexts’ but rather ‘has a newly coded (inter)subjective meaning’.8 We need to distinguish between the intersubjectivity that may pragmatically accompany the use of a form from its development into a coded meaning (see also De Smet and Verstraete 2006). For example, in her discussion of the development in the later fourteenth century of the parenthetical reformulation marker I mean Brinton (2007) shows it was a subjectified form of the literal by X I mean (= ‘intend’). By virtue of the fact that it is a marker used by speakers and writers to negotiate meaning “by presenting more explicit phrasing of the preceding NP, phrasing they believe will make their meaning clearer to readers” (Brinton 2007: 48), parenthetical I mean has always been pragmatically intersubjective. Over time it has been used more intersubjectively to express emphasis and assertion of the veracity of an utterance (Brinton 2007) but it has not been intersubjectified, except in fixed phrases like You know what I mean? Heuristically, therefore, it may be better to think of (4) as (4ƍ), although from a theoretical point of view this is not a viable formulation, since language-users do not have direct access to history and may therefore be un-

8. Athanasiadou, Canakis, and Cornillie (2006) construe my view of subjectification as primarily pragmatic; while pragmatic strengthening of subjective meanings is without question a pre-condition for subjectification, subjectification itself is not pragmaticalization, but semanticization (on the assumption that there is a difference between pragmatics and semantics).

36

Elizabeth Closs Traugott

aware of what is subjectivized (though they often have indirect evidence of history from awareness of age-graded practices): (4ƍ)

(Heuristic for linguistic analysis) non-/less subjectivized > subjectivized > intersubjectivized

Another issue is that Fitzmaurice (2004) suggests that there is a further step from intersubjective to “interactive”, especially in the case of discourse markers like you know, you see, you say, from the eighteenth century on. Her hypothesis is that this is a shift away from attention to the addressee to simply “keep[ing] things going in a conversation” (Fitzmaurice 2004: 438) and grabbing the interlocutor’s attention (Fitzmaurice 2004: 439). The examples cited suggest that there is probably not a newly coded meaning, but rather a pragmatic one arising in less clearly intersubjective contexts. Indeed, Fitzmaurice ends the article with “A question for further investigation is whether the interactive discourse marker function of an expression is strongest only where no implication [of speaker meanings, ECT] may be drawn from the use of the expression” (Fitzmaurice 2004: 446). Examples of subjectification include the following, the first two of which are dynamicized versions of (1): (5)

–

raising constructions arise from non-raising ones (Langacker 1990, 1995); in the case of be going to, we find expressions of motion with intent to act in the sixteenth century (5a), intentional non-motion expressions in the seventeenth century (5b), and finally raising ones in the nineteenth (5c), which express speaker assessment of the future:9 a. I am going to visit the prisoner. Fare you well. (Exit) (1604 Shakespeare, Measure for Measure III.iii.273 [LION; Oxford Shakespeare])10 b. I ha’ forgot what I was going to say to you. (1663 Cowley, Cutter of Coleman Street V.ii [LION; English Prose Drama]) c. I am afraid there is going to be such a calm among us, that we must be forced to invent some mock Quarrels (1725

9. For a modified hypothesis about this change, see Garrett (Forthcoming). Thanks to Brian D. Joseph for this reference. 10. This example, like many others of the time, is potentially ambiguous, but the “exit” context suggests it is intended as a motion verb.

(Inter)subjectivity and (inter)subjectification: A reassessment

–

–

37

Odingsells, The Bath Unmask’d V.iii. [LION; English Prose Drama]) epistemic modals may arise from verbs of desire or volition (will), concessives from temporals (Early Middle English while ‘during’ > Early Modern English ‘although’), focus markers from manner adverbials (Old English anlice ‘simply, especially’ > only), illocutionary from non-illocutionary uses of speech act verbs (promise originates in the Latin past participle of promittere ‘send forth’) (Traugott and Dasher 2002) referent (“T/V”) honorifics may arise from non-honorifics (e.g., cooption in the Late Middle English period of plural second person ye for polite address to a singular second person) (Traugott and Dasher 2002).

Examples of intersubjectification that I have cited in the past include the co-option of subjectified meanings specifically to signal addresseeorientation and interpersonal meanings. For example, I have used intersubjectified hedges arising from subjectified discourse markers as examples of intersubjectified meanings (e.g., some uses of well (Jucker 1997), perhaps, and hedged uses of sort of) (see Traugott and Dasher 2002). But, as indicated above, what may look like it is a case of intersubjectification actually may not be. If it is derivable from the context, it is only a case of increased pragmatic intersubjectivity. In other words, there may be more addresseeoriented uses, but unless a form–meaning pair has come to code intersubjectivity, we are not seeing intersubjectification (-ation being the important item here). This makes the formulation in (4) difficult to work with, hence the heuristic usefulness of (4ƍ). Where a “dedicated” (coded) intersubjective meaning arises, however, it does by hypothesis arise from a previously subjectified meaning. An example is provided by the rise of addressee honorifics in Japanese from referent honorifics. To simplify, a referent honorific in Japanese points either respectfully or humiliatively to the subject referent.11 The construal of relationships is subjective, although since the referent is the addressee, intersubjectivity is inevitably involved. An addressee honorific, by contrast, indexes politeness or intimacy with respect to the addressee, and is part of a more general speech style or register, in which lexical items for eating or food may also be indexes; here attention is

11. Note interlocutors have an argument structure role in referent honorific (T/V) systems, including European ones.

38

Elizabeth Closs Traugott

explicitly paid to the addressee, whether or not that individual is referred to. (6) summarizes the changes undergone by one example, saburahu:

(6)

Old Japanese saburahu ‘wait (for an occasion or order) in a specific location’ (non-honorific) > Late Old Japanese ‘Humble Subject be in the vicinity of Respected Referent’ (referent honorific; subjectified) > Early Middle Japanese -saburau/-soorau ‘be-Polite’ (addressee-honorific style; intersubjectified). (Traugott and Dasher 2002: 263–276)

As is true of all diachronic clines, the cline of (inter)subjectification in (4) is a testable schema summarizing tendencies for change observed over long periods of time and over several languages (Andersen 2001: 241–245). Once started down it does not have to be gone down all the way. It is “layered” (Hopper 1991) in the sense that earlier and later forms coexist and are in variation. Neither subjectification nor intersubjectification entails grammaticalization. Evidence in the domain of subjectification comes from the development of speech act verbs and especially of their illocutionary uses. In the domain of intersubjectification it comes from the development in Japanese of polite uses of lexical items from earlier humiliative ones (e.g. moosu ‘say’, Traugott and Dasher 2002: 261). Nevertheless, there is a strong correlation between grammaticalization and subjectification, and a weaker one between grammaticalization and intersubjectification. In order to understand this we need to consider the distinction between grammaticality and grammaticalization. 3. Grammaticality and grammaticalization Just as we need to distinguish the synchronic cline of (inter)subjectivity from the diachronic cline of (inter)subjectification, so we need to distinguish the synchronic cline of grammaticality (a way of organizing data) from the diachronic cline of grammaticalization (a schema of attested tendencies over time). Synchronic clines of grammaticality can be established with varying degrees of granularity. Typically this is done with reference to degrees of fusion. A highly schematic example appears in (7) (based on Bybee, Perkins, and Pagliuca 1994: 40):

(Inter)subjectivity and (inter)subjectification: A reassessment

(7)

39

phrase or word – non-bound gram – bound gram

A more articulated cline, also based on degrees of fusion, is proposed in Brinton and Traugott (2005: 93). Here a distinction is made between periphrases that are relatively free, but have considerable internal fusion, e.g. be going to, as far as; semi-bound function words and clitics like of, ‘ll, possessive clitic -’s; semi-productive affixes like grammatical derivational -er; and productive affixes like inflectional plural -s. While degree of fusion is synchronically an independently valid measure of form–meaning pairs, interest in it derives from the fact that it is seen at the outcome of diachronic changes known as grammaticalization: [T]he change whereby lexical items and constructions come in certain linguistic contexts to serve grammatical functions, and once grammaticalized, continue to develop new grammatical functions. (Hopper and Traugott 2003: 18)12

The synchronic cline in (7) therefore derives its theoretical interest from diachronic ones such as those in (8): (8)

a. b.

phrase or word > non-bound gram > bound gram lexical/constructional item > grammatical item > more grammatical item

Again for heuristic purposes, it may be useful to think of (8b) in terms of (8bƍ), although, like (4ƍ), this has no theoretical validity: (8bƍ)

(Heuristic of linguistic analysis) lexical/constructional item > grammaticalized item > more grammaticalized item

While not restricted to grammaticalization, subjectification is more likely to occur in grammaticalization than in lexicalization or in semantic change in general,13 presumably because grammaticalization by definition 12. For discussion of how to assess “less” and “more” grammatical, see Brinton and Traugott (2005: 93, 147–150). 13. For the distinctions see Brinton and Traugott (2005: 144–145). Briefly, grammaticalization involves the development of items expressing grammatical function, whereas lexicalization involves the development of items with contentful meaning.

40

Elizabeth Closs Traugott

involves recruitment of items to mark the speaker’s perspective on factors such as: (9)

– – – – – – –

who does what to whom (argument structure) how the proposition (ideational expression) is related to speech time or to the temporality of another proposition (tense) whether the situation is perspectivized as continuing or not (aspect) whether the situation is relativized to the speaker’s beliefs (modality, mood) whether entities referred to are construed as same or different (pronouns, indexicals) which part of a clause is viewed as topic or focus how utterances are connected to each other (connectives, discourse markers)

However, not all grammaticalization is equally likely to involve equal degrees of subjectification,14 and some may involve little or no subjectification.15 For example, case markers are often derived from terms for relational space or body parts (Heine and Kuteva 2002), e.g. Old Hungarian vilag ‘world’ + béle ‘world + guts:into’ > Modern Hungarian világba ‘world:into’ (Anttila [1972] 1989: 149). Subjectification is minimally involved here, since the basic function of argument structure is ideational: to express events or situations and the participants in them. However, ideational case may be used pragmatically for non-ideational purposes, see e.g. the choice of accusative or dative case in German as “a function of the degree to which the person is affected, as well as the speaker’s subjective assessment of this effect” (Zubin 1975: 186). Subjectification is more likely to occur in primary grammaticalization (the shift from lexical/constructional to grammatical) than in secondary 14. De Smet and Verstraete (2006) seeks to distinguish degrees of subjectivity using syntactic criteria; to what extent these criteria can be identified with grammaticalization remains to be investigated. 15. Cornillie’s statement that “The combined process of grammaticalization and pragmatic strengthening crucially involves a shift toward linguistic expressions that carry a more speaker-based view. More particularly it entails subjectification” (Cornillie 2004: 52) is too strong. There is a correlation, not an entailment relationship.

(Inter)subjectivity and (inter)subjectification: A reassessment

41

grammaticalization (the development of already grammatical material into more grammatical material). This is because primary grammaticalization often requires prior strengthening of pragmatic inferences that arise in very specific linguistic contexts prior to their semanticization and reanalysis as grammatical elements. Further grammaticalization, however, often involves development into automatized structures (especially in the case of inflections). The fewer the options become, the less likely subjectification will be. This can be regarded as part of the larger constraint, noted in Dahl (2004: 84), that none of “the usual Gricean principles” are operative if a morpheme is truly obligatory. As will be discussed in more detail at the end of Section 6, subjectified elements tend to be positioned at the periphery of a constituent or clause (see Bybee 1985 on the shifting of epistemic modals to the periphery of a complex verbal construction; Suzuki 1998 on the development of the Japanese noun wake ‘reason’ into an utterance-final ‘explanatory discoursemarker’ translatable as ‘no wonder’ or ‘you see’). The migration of subjectified elements to the periphery of the phrase, clause, or sentence would not be considered a case of grammaticalization by those who take structural scope-reduction and condensation to be criterial to grammaticalization (Lehmann [1982] 1995: 144).16 Limiting grammaticalization to reduction and condensation appears to be too restrictive, however. They pertain to certain domains of grammaticalization such as the development of case and tense, but not to other domains such as epistemic modality, connectives, discourse markers, etc., where scope increase is typical of grammaticalization (Tabor and Traugott 1998; Traugott 2003b). Intersubjectification intersects less extensively with grammaticalization. In most languages it is grammaticalized only into some discourse markers and interjections. It is strongly grammaticalized, in the sense of being expressed morphologically, in only a few languages, e.g., Japanese, where verbal endings may index politeness (see 6 above).17

16. For example, Aijmer (1997) argues for the term “pragmaticalization” rather than “grammaticalization” for items that expand their scope such as discourse markers. 17. Given the example in (6), the anonymous reviewer raised the question whether intersubjectification may not correlate with secondary grammaticalization. This is a research question that needs further study.

42

Elizabeth Closs Traugott

4. Three brief case studies In this section I provide brief sketches of the history of three NP of NP patterns: a piece of, a bit of and a shred of. They reveal sufficient similarities and yet differences that they illustrate well what has become a truism in work on grammaticalization: each construction has its own history, but conforms to general schematic change-types in ways that are partly constrained by the particularities of the original meaning–form relationship. In Present Day English at least a bit of and a shred of are polysemous, with Partitive (‘unit of’), Quantifier (‘some, many’) and Degree Modifier (‘rather’) functions, but in earlier English only the Partitive use is found. The development of the Quantifier and later Degree Modifier uses illustrates grammaticalization, and, in two of the cases discussed here, further partial restriction to a semantically negative polarity context. It also illustrates subjectification, and, in the case of a bit of, pragmatic intersubjectification. The three constructions are part of a larger set of Partitive > Quantifier > Degree Modifier changes that occurred mainly in the eighteenth and nineteenth centuries (for historical studies of kind of, sort of see e.g., Denison 2002 and references therein; of a bunch of, lots of Brems 2003; and of a kind of, a sort of, a bit of, a lot of, a shred of Traugott 2008a, b, c). Partitive constructions in Present Day English have rather idiosyncratic characteristics with respect to presence or absence of the indefinite article a in either NP and the extent to which they allow an unbounded (non-mass) complement in NP2 (see a piece/?shred of an apple vs. a piece/shred of apple), but an idealized, prototypical schema for Partitives is: (9)

[NP1 [of NP2]] Head + Modifier/Complement Unit + Unbounded N (not) a shred of apple

The Partitive construction came into being in the Middle English period when the -’s genitive inflection as in an bite brædess ‘a bit bread:GEN’, was replaced by the prepositional phrase with of (which meant ‘out of’ in Old English). The NP1 of Partitives can usually be substituted by a piece/portion/unit of. By contrast, binominal Quantifier constructions typically have the schema in (10), and NP1 can usually be substituted by a quantifier like some/much (negative no/not any/not much):

(Inter)subjectivity and (inter)subjectification: A reassessment

(10)

43

[[NP1 of] NP2] Modifier + Head Quant + Unbounded N (not) a shred of beauty

There are several criteria for distinguishing Quantifier constructions with NP heads from Partitive constructions (see Denison 2002).18 These include: (11)

a.

b. c. d.

e.

agreement patterns: in the Partitive the initial determiner agrees in number with N1 (these kinds of mouse), but in the Quantifier construction it can agree with N2 (these kind of mice), at least in colloquial use in the Partitive NP2 may be preposed (of an apple a bit), but not in the Quantifier construction (*of a fraud a bit) in the Quantifier but not the Partitive construction, a N1 of can be replaced by one word (a bit of a beauty = rather/quite a beauty) only the Quantifier may develop into a Degree Modifier Adverb form that collocates with Adjectives (a bit/*piece squeamish) or Verbs (I sort of/*unit of regretted it) (although not all do so) some Partitives do not have or have only marginal Quantifier construction polysemies (a piece/unit/portion of)

4.1. A sketch of the history of a piece of Piece was borrowed from French in Early Middle English in the meaning ‘fragment, scrap, slice, portion’. Initially it is found in a binominal Partitive construction with a concrete complement (food, garment, building, precious metal, etc). In later uses, it can be found with negative evaluations implying ‘too small, inadequate’ (12c).

18. Denison (2002) calls Degree Modifiers with the first three properties PreDeterminers; Aarts (2001) calls them Complex Specifiers.

44 (12)

Elizabeth Closs Traugott

Stage I: Partitive a. In þe assaut some … breke a pece of þe wal ‘In the assault some … broke a piece of the wall’ (c1325 Glo. Chron.A 11590 [MED pece 2a]) b. Gave them a piece of a honeycomb to eat (1713 Henry, Catech. Youth in Wks. (1853) II. 169/1 [OED piece 2a]) c. The spirit which animated her father when he went to housekeeping in a piece of a house without any front window (1884 Harper’s New Monthly Mag. 69 303 [OED piece 6d, “US regional”])

By the end of the fourteenth century the complement had been generalized to animate and also abstract NP2 contexts. A piece of is still Partitive, but had been enriched with the quantifier meaning. It could now imply ‘a small amount’. In (13b) we find peece used either meaning ‘exemplar’ or (an early usage) ‘person’.19 (13)

Stage II: Extended Partitive a. Dorus, whilom king of Grece ... hadde of infortune a piece ‘Dorus, once king of Greece … had of misfortunate a piece’ (a1393 Gower, Confessio Amantis (Fairf.) V. 1338 [OED piece, 5a]; note pre-posed of infortune) b. O Pretious peece of villany! are you vnchang’d? (1615 Tomkis, Albumazar V. ix. sig. I2v [OED piece, 5c])

Toward the end of the sixteenth century we find examples in which a human subject is characterized as a ‘piece’ of a profession (logician/doctor/poet). There is some indeterminacy between the Partitive reading (a small part/exemplar of a poet/logician) and a Quantifier reading (a poet/logician to a certain degree, ‘somewhat of a poet’). This appears to be the only context in which a Quantifier use developed robustly, and here it borders on a Degree Modifier in meaning. (14)

Stage III: Quantifier a. If I had not beene a peece of a Logician before I came to him. (1586 Sidney, Apol. Poetrie (1595) sig. B1v [OED piece, 6d])

19. This was used of men and women, presumably somewhat derogatorily.

(Inter)subjectivity and (inter)subjectification: A reassessment

b.

45

methinkes I am a piece of a Poet already, there’s such a whistling in my pate. (1640 Anon, The Knave in Graine I.i (LION; English Prose Drama])

A diary entry shows that at least one writer analogized the Quantifier with Adverbs like greatly that do not require a following NP: (15)

A mayd ... dyd cutt her thrott a-pesse, and after she lepyd in-to a welle and drownyd yr seyllff ‘A maid … did cut her throat a bit, and afterwards she lept into a well and drowned her self’ (1559 Machyn, Diary July (1848) 205 [OED piece, 10b])

(15) is the only example of its kind in my data base. Although indicative of an innovation by a speaker, it does not represent a change, as far as the data base allows us to identify one. From the perspective taken here, change requires both innovation and shared (replicated) adoption by a community of speakers (Weinreich, Labov, and Herzog 1968). A piece of underwent grammaticalization to the extent that it was generalized to non-breakable (house) and mass noun (villany) complements (“host-class” expansion; see Himmelmann 2004) at Stage II; in this context there was pragmatic expansion in so far as ‘unit of mass noun’ came to imply scalar quantification. At Stage III there was syntactic expansion in that the NP of NP string now had two potential syntactic analyses, and also semantic/pragmatic expansion in that in the new polysemy NP1 (a piece) was reanalysed as ‘to some extent’ and bleached of partitive meaning (but only in restricted contexts). Himmelmann (2004) considers host-class, syntactic, and semantic/pragmatic expansion as criterial for grammaticalization.20 Another way to think of this is that, as a new polysemy arises that is an instance of primary grammaticalization, all three aspects of form– meaning pairings (lexical collocation, syntactic structure, and semantics– pragmatics) change (but not necessarily at the same time). 20. While the extensions Himmelmann identifies may be true of primary grammaticalization, there is often later restriction syntactically: a grammaticalized element may become fixed to just one context; or lose most meaning (typical of inflections); or, as a grammaticalized item comes to be obsolescent, it may become more and more frozen, as in the now long non-productive Demonstrative + Instrumental case þy which is restricted to the X-er the Y-er constructions such as The more the merrier.

46

Elizabeth Closs Traugott

4.2. A sketch of the history of a bit of It seems that a piece of is primarily a Partitive in Modern English, despite some marginal Quantifier developments. In this respect it contrasts sharply with a bit of. The history of a bit of appears to start with a nominalized expression meaning ‘biting’ (OED). We may call this Stage 0 (PrePartitive): (16)

Stage 0: Pre-Partitive a. In to the pyne of helle .. for the bytt of an Appel ‘Into the suffering of Hell…for the biting of an apple’ (c.1400 Ancr. Recl. 22/25 [MED bite 3.b.]) b. this appyl…a bete therof thou take ‘This apple, a bite of it take’ (c.1475 Ludus C. 23/220 [Ibid.]; note the preposing of a bete)

Recall that Old English of meant ‘out of’ before it came to be the default Preposition. Metonymic transfer from biting out of something to the result of the biting was semanticized, and bite came to mean ‘piece bitten out, morsel’ (i.e. a unit of a size that could be bitten out). Indeed, bete in (16b) could be interpreted not as ‘a bite’ but as ‘a mouthful’. This is a semantic/pragmatic “bridging” context (see Heine 2002: 84; Enfield 2003: 28). We may call this ‘mouthful’ polysemy Stage I: Partitive. An early example is (17), still with the genitive case: (17)

Stage I: Partitive ‘morsel, unit bitten out’ He badd tatt gho shollde himm ec / An bite brædess brinngenn ‘He commanded that she should him also a bite/bit of bread bring’ (c.1200 Orm 8640 [MED bite 3.a.])

By the earlier part of the seventeenth century the complement had been generalized to non-food, often with a contextual implicature that the unit was a small, insufficient, or inadequate part of the pragmatic Focus (NP2). This is still Partitive, but importantly for later reinterpretation as a Quantifier, the unit (NP1) has been bleached of the literal meaning ‘mouthful’, and NP2 (the complement) can be abstract (18a).

(Inter)subjectivity and (inter)subjectification: A reassessment

(18)

47

Stage II: Extended Partitive a. The fragments, scraps, the bits, and greazie reliques of her ore-eaten faith (1606 Shakespeare, Troilus and Cressida, V.ii.159 [OED bit 2, 3.a.]; note the metaphorical context of ore-eaten faith here) b. If so be as ow you ’ont blab, I’ll tell you a bit of a secret ‘If it be as how you won’t blab, ….’ (1833 Clifford, The Highwayman of 1770, I.ii. [LION; English Prose Drama])

By the eighteenth century we begin to find a bit of meaning ‘somewhat of/rather/quite’, typically before a head with an indefinite article, and an evaluative behavioral term (jilt, bastard, hypocrite, fool). In other words, it is a Quantifier, scaling NP2 (the head) down (19a, b), or approximating it (19c). This is Stage III. Some examples are (from our modern perspective at least) ambiguous bridging examples: does (19b) refer to a small amount of bastard business, or to somewhat bastard(ly) business, does (19d) refer to a small secret or (more probably in the context) something somewhat secret? It is instructive to compare (18b) with (19d). In (18b) a bit of a secret appears to mean ‘a small confidence, a small piece of information that is meant to be kept secret’, given the context in which the speaker allegedly knows something and is willing to impart it if it won’t be blabbed. But in (19d) the secret is not known, indeed, the speaker does not even know if there is a secret; what remains to be found out is whether there is a secret and how big it is. (19)

Stage III: Quantifier a. Your beauty is a little bit of a jilt (1771 S. Foote, Maid of Bath [OED bit 2, 4.b.]) b. If you be a lord, it must be a bit of bastard business (1810 W. Hickey, Mem [OED bit 2, 4.h.]) c. Wal. A sword–have I– (confused) –Why yes, it is a bit of a kind of a sword, as you say, to be sure– (1794 Morton, The Children in the Wood I.iii. [LION; English Prose Drama]; not a piece of a sword, but like one) d. (Aside) Oh, ho! There’s a bit of a secret, and I must be master of it (1769 Morton, Way to Get Married I.ii. [LION; English Prose Drama])

48

Elizabeth Closs Traugott

e.

“I’ve got something to tell you, my dear,” said Caleb in his hesitating way… “You see, I’ve been a bit of a fool again, and put my name to a bill” (1871 Eliott, Middlemarch [UVa]).

A bit of was reduced and extended to syntactic contexts prototypical for members of the Degree Modifier class, specifically pre-Adjectival position, where a bit functions as an Adverb. This may well have been by direct analogy with other Degree Modifiers like quite and very. (20)

Stage IV: Adverb Degree Modifier I would not be a bit wiser, a bit richer, a bit taller, a bit shorter, than I am at this Instant (1723 Steele, The Conscious Lovers III.i [LION; English Prose Drama])

This Adverb Degree Modifier use has been available since the seventeenth century but is considered “more or less vulgar” by Stoffel (1901) and “slang” by the OED. In the eighteenth century it developed further into an Adverb that can be used without a complement (21a), and even as a free adjunct in responses (21b). The latter usually occurs in the context of a negative, or of negative-conveying intonation. (21)

Stage V: Adjunct a. I tell ye, I zee’d un gi’ Susan a letter, an’ I dan’t like it a bit. (1800 Morton, Speed the Plough II.iii. [LION; English Prose Drama]) b. A. Hear me. B. Not a bit (1739 Baker, The Cit Turn’d Gentleman [LION; English Prose Drama])

A bit of has undergone grammaticalization to the extent that it was generalized to non-edible and abstract noun complements (host-class expansion) at Stage II; in this context there was pragmatic expansion in so far as ‘unit of mass noun’ came to imply negatively evaluated scalar quantification (the bits of her ore-eaten faith are small parts of faith, implying ‘too small’, in this context, even ‘left-over’). At Stage III there was syntactic expansion in that the NP of NP string now had two syntactic analyses, and also semantic/pragmatic expansion in that in the new polysemy NP1 (a bit) was enriched as a quantifier and bleached of partitive meaning. At Stages IV and V there was further syntactic expansion to Adverb and free adjunct use.

(Inter)subjectivity and (inter)subjectification: A reassessment

49

With respect to subjectification, a bit of was subjectified at Stage III (endowed with quantificational scalar meaning ‘somewhat’, a “downtoning” or understating meaning like that of a little). In contexts where the head is lexically negative or can be expected to be negatively interpreted, but not in negative syntax, a bit (of) is used intersubjectively as a hedge (see 19a, b). We may say it has been pragmatically intersubjectified, but not semantically so because it does not code semantic intersubjectivity. Note that in (19e) the speaker is attempting to save his own face, but clearly in the context of paying attention to the addressee: I’ve got something to tell you … You see). Being a downtoner, a bit of is favored with negatively evaluated heads as in (19a, b, e), or with neutral ones (19c, d), but it is not likely to be used, except in special circumstances, such as irony, with positive heads, especially of people: (22)

Your friend is a bit of a beauty.

4.3. A sketch of the history of a shred of In Old English a shred of meant ‘a fragment cut or broken off from fruit, vegetable, textile, coin, vessel’. In Middle English it was generalized to bodies (physical bodies and the symbolic Host) but it still meant ‘unit of X’. (23)

Stage I: Partitive With strengthe of his blast / The white [dragon] brent than rede, / That of him nas founden a schrede / Bot dust ‘With the strength of his blast, the white dragon burned the red, so that of him (the red) not a shred was found, only dust’ (c1300 Arth & M 1540 [MED shrede a)]; note the preposing)

In the sixteenth century it was generalized to further contexts including language, mankind and nature. The effect of a shred of was that the pragmatic focus (NP2) was often evaluated as small, insufficient, inadequate, but the construction is still Partitive; NP1 may still be modified by an Adjective.

50 (24)

Elizabeth Closs Traugott

Stage II: Extended Partitive a. Suche shredis of sentence strowed in the shop of ancient Aritippus ‘Such scraps of wisdom strewn in ancient Aritippus’ shop’ (1529 Skelton Sp. Parrot 94 [OED shred 6]) b. A despis’d Shred of mankind (1645 G. Daniel, Poems [OED shred 6])

This extension appears to have been an essential step in the development of the Quantifier use. In the nineteenth century a polysemy arose from reanalysis of NP1 as a Quantifier (not ‘(smallest) part of’ but ‘some/any’) with NP2 as head only when NP2 was a mass noun (e.g. mankind). As a Quantifier, a shred of requires an abstract mass noun (typically evidence, character, hope, reputation, credibility) that is normally neutrally or positively evaluated. By the twentieth century this Quantifier was largely, but not obligatorily, restricted to negative polarity syntax (negation, interrogative, conditional, comparative, etc., see Israel 1996, 2004). (25)

Stage III: Quantifier a. Loto has not a shred of beauty. She is a big, angular, rawboned Normande, with a rough voice, and a villainous patois (1867 Ouida, Under Two Flags [LION: 19thC Fiction]) b. You’re so worthless, you can’t even recognize the shred of petty virtues in others, some of which I still have (1965 Osborne, A Patriot for Me, III. v. [LION; 20thC drama]; note singular shred with plural head, not shreds of petty virtue)

A shred of has undergone grammaticalization to the extent that it was generalized to abstract Ns (host-class expansion) at Stage II; in this context there was pragmatic expansion in so far as ‘unit of mass noun’ came to imply negatively evaluated scalar quantification (shredis of sentence are small parts of wisdom, implying ‘too small’). At Stage III there was syntactic expansion in that the NP of NP string now had two syntactic analyses, and also semantic/pragmatic expansion in that in the new polysemy NP1 (a shred) was enriched as a quantifier and bleached of partitive meaning. To date a shred of has not undergone further, secondary grammaticalization, for example, expansion to pre-Adjectival position (*a shred nice) or to

(Inter)subjectivity and (inter)subjectification: A reassessment

51

adjunct status (*I opened it a shred; Q. Did you like the movie? A. *A shred), in contrast with e.g. a bit/sort of (see 21).21 With respect to subjectification, the semanticization of quantifier meaning is a type of subjectification (to scalar meaning). But there has been no pragmatic, let alone, coded intersubjectification: a shred of is not used as a hedge (*Your beauty is a shred of a jilt). This appears to be related to the fact that a shred of as a Quantifier is restricted to positively evaluated heads (honesty, evidence, decency, reputation). 4.4. Summary and some comments By hypothesis, a shift to meanings that are used evaluatively to assess not just more or less quantity, but also more or less quality, as in the case of Quantifiers or Focus Particles (see König 1991), always involves subjectification: the evaluative invited inferences are semanticized as part of the meaning of the item that comes to index scalarity. Those lexical items that initially refer to small quantities, and, by extension, low quality, tend to become negative polarity items. This involves both further grammaticalization (restriction in terms of syntactic contexts, see ft. 20) and further subjectification in that the speaker evaluates not only elements on a scale, but the scale itself. One feature of negative polarity is that meanings that would normally be understating in positive contexts (a bit (of), a shred of) are reversed and in this context are emphatic (Israel 1996). While a bit means ‘to some extent’, not a bit means ‘to no extent, not at all’. In discussing the development of the complex negative phrase in French ne … pas ‘not’, Eckardt (2006: 169, drawing on Krifka 1995) hypothesizes that negative polarity sensitivity is “the result of emphatic focusing and a specific kind of focus alternatives”. In other words, speakers assessing alternatives on a scale of probability or saliency, may use a quantifier such as bit, shred, Fr. pas (‘step’), once it has lost its original literal meaning, as an emphatic focus marker to signal that the following NP is the least probable or salient among a set of alternatives. The discourse reasons for such assessments 21. Further evidence for the low degree of grammaticalization of a shred of is its low frequency: 4 hits in LION, English Prose Drama, 32 in UVa, as opposed to 623 and 1702 respectively for a bit of (in all its uses). The numbers no doubt in part reflect the fact that the Degree Modifier use of a shred developed a hundred years later than that of a bit, but this alone hardly accounts for the dramatic differences in numbers between a shred of and a bit of.

52

Elizabeth Closs Traugott

include the intention “to contravene scalar expectations among interlocutors, whether explicit or implicit” (Gutiérrez-Rexach and Schwenter 2002). Most emphatics are speaker-oriented claims about extreme ends of scales. But some may be used intersubjectively in certain contexts for aggression (being “in” rather than “saving” Addressee’s face), e.g., Du. kanker ‘cancer’, pest ‘plague’, pokken ‘smallpox’ (Hoeksema 2001: 178) (all “positive” polarity items, i.e. they do not require negative polarity contexts). Some negative polarity items may eventually come to be negation markers in a process usually referred to as the Jespersen Cycle (Jespersen 1917), whereby a negation marker such as French ne comes to be associated in a complex negative phrase with a term referring to a small unit, in this case pas ‘step’, and eventually pas comes to be used alone as the marker of negation (see Eckardt 2006 on the history of French; Schwenter 2006 on cognates in contemporary Catalan, Italian, and Brazilian Portuguese). It appears that at least subjectification and perhaps pragmatic intersubjectification are involved in the early development of the Jespersen Cycle. However, they cease to operate once the subjectivized element comes to be the stand-alone grammatical marker (at which point it has been bleached not only of its original ‘unit’ meaning, but also of its subjective emphatic meaning). 5. Mechanisms of and motivations for change in grammaticalization and (inter)subjectification Mechanisms of change have to do with the ‘how’ of change in the mind of speaker and hearer. Harris and Campbell (1995) have claimed that there are only two basic internal mechanisms for syntactic change: reanalysis and extension.22 Recently, other possible mechanisms have been explored, such as priming, routinization, and other production/parsing processes (see Garrod n.d.). As Campbell and Harris see it, while reanalysis involves changes in the underlying structure, extension (which, in their view, overlaps with analogy) involves changes in the surface manifestation of a pattern (Harris and Campbell 1995: 50–53). This distinction is problematic in some of its details, since any extension that becomes institutionalized and acquires the status of a change rather than merely an innovation, entails at least some 22. They also cite an external mechanism, borrowing, but that does not concern us here.

(Inter)subjectivity and (inter)subjectification: A reassessment

53

minimal reanalysis of the underlying specification. For example, analogical extension of Partitives to non-countable hosts (e.g. a bit of to faith) entails a reanalysis of the restriction that the complement (NP2) is countable. Nevertheless, the distinction between changes that concern constituency, hierarchical structure, category, grammatical relations, or boundary types, and similar primarily “covert” aspects of syntagmatic linguistic structure on the one hand (reanalysis), and changes based on overt patterns and templates that serve as exemplars on the other (analogy), is a useful one (Hopper and Traugott 2003). Famously, Meillet initially conceived of grammaticalization as reanalysis (Meillet ([1912] 1958), and much has since been written about the extent to which grammaticalization and reanalysis are independent of each other (see Dahl 2004: 170–178 for a summary of arguments). It seems safe to say that grammaticalization requires structural reanalysis, but that not all reanalysis, e.g. word order change, is necessarily grammaticalization. The more work has been done on the contexts for grammaticalization, the more it has become clear that it is a reanalysis that occurs very locally and involves fine-grained changes that are not necessarily totally new in the language in the way Meillet appears to have envisioned when he said that grammaticalization “introduit des catégories qui n’avaient pas d’expression linguistique, transforme l’ensemble du système”23 (Meillet 1958: 133). For example, addressing the issue from radically different perspectives, Bybee (2006: 727) has recently suggested that “new constructions are created out of specific instances of old general constructions”, while Kiparsky (Forthcoming) has argued that much of grammaticalization arises out of optimization (generalization) of structures in a restrictive theory of analogy.24 The development of the Partitives into Quanitifiers, and, in the case of a bit of, into Degree Modifier Adverb and Adjunct illustrate well the interdependence of reanalysis and extension. The head-modifier shift is clearly reanalysis, the other changes are largely analogies of various kinds. For the constructions in question, each change was new for the string. Quantifiers and Degree Modifiers had existed since earliest times in English (e.g. manig ‘many’, swiðe ‘very’). The rise of a whole class of binominal Quantifiers in the eighteenth century (a bit of, lot of, (a) sort/kind of, etc.) was new. So was the rise of binominal Degree Modifiers, since until then they had been monomorphemic or derived Adverbs. 23. “Introduces categories which previously had no linguistic expression, transforms the totality of the system”. 24. See also e.g. Himmelmann (2004), Fischer (2007).

54

Elizabeth Closs Traugott

Just as two internal mechanisms have been proposed for syntactic change, so two have been proposed for semantic change: conceptual metonymy and metaphor (Traugott and Dasher 2002: 27–34). The first mechanism, conceptual metonymy, is analogous to reanalysis, and the second, metaphor, is analogous to analogy or generalization based on pattern match (Anttila 1992). Conceptual metonymic shift in the sense intended here involves a covert shift, e.g. the semanticization of an earlier pragmatic implicature or invited inference; in the case of Quantifiers, the semanticization of the implicature of quantity from units and portions, in certain contexts. Metaphorical change is, by contrast, based on analogical similarities. Just as reanalysis and analogy are interdependent, so are conceptual metonymy and metaphor. Much of what looks like metaphorical change can actually be seen to be the result of metonymic change when contexts for the change are taken into account. Presumably, in terms of usage, Shakespeare’s ‘The fragments, scraps, the bits, and greazie reliques of her ore-eaten faith’ cited in (18a) is an instance of metaphor. Shattered by Cressida’s unfaithfulness to him when she leaves him for Diomedes, Troilus disgustedly conceptualizes her new promises to Diomedes in terms of nasty morsels of food left over from others. Because such metaphors have mass and abstract heads, they enabled the reanalysis as Quantifier, but the semantic change Partitive > Quantifier is not a case of metaphorization: it is subjectification, the reanalysis of the pragmatically inferrable scalar evaluations arising by conceptual metonymy in the context of utterances like (17) and (18) as semantically coded scalar, quantitative meanings. Subjectification is therefore a subtype of semantic reanalysis, a mechanism. Likewise, when intersubjectification occurs, pragmatic intersubjective meanings that are pragmatically inferrable from the context (or, as in the case of If you please, syntactic contexts) come to be coded as part of the semantics of an item. Genuine cases of intersubjectification as opposed to intersubjective uses of items are hard to identify outside of languages like Japanese. But where it occurs, like subjectification it is a semantic reanalysis, and a mechanism. What might motivate grammaticalization, in other words, what might be the reason for the change? As we have seen, in many cases, primary grammaticalization follows pragmatic changes and changes in distribution. In the case of a bit of and a shred of, it followed the pragmatic strengthening of the literal Partitives as quantificational (pragmatic subjectification), and host-class expansion to more nominal complements, especially mass com-

(Inter)subjectivity and (inter)subjectification: A reassessment

55

plements. But we cannot say it was caused by these changes.25 Although Partitive a piece of underwent several of the same types of changes as a bit of and a shred of, it did not become as productive as they in the Quantifier function, nor did it participate in negative polarity (perhaps because piece was conceptualized as larger than bit or shred?).26 Haspelmath (1999: 1054–1055) has provided a fairly detailed account of how grammaticalization might be motivated. He invokes a number of “ecological” factors, among them unconscious processing, routinization, maxims of action and invisible hand processes whereby unintended results arise from action (see Keller 1994). Haspelmath suggests that grammaticalization arises out of speakers’ desire to “be extravagant”, i.e. “Talk in such as way that you are noticed” (1999: 1055). The term “extravagant” is, as he admits, unfortunate, since most grammaticalization involves fairly general meanings to start with. Furthermore, “Talk in such a way that you are noticed” implies a degree of consciousness that is inconsistent with the concept of unconscious acts leading to change. However, the examples discussed here do suggest that some combination of maxims must be at work, such as “Talk like the others talk” (with pragmatic invited inferences of quantification) and “Try something very similar that will not be incomprehensible” (and therefore does not violate the maxim “Talk in such a way that you are understood”). What might motivate subjectification? By hypothesis it is online production in the flow of speech (in certain cases, writing may be involved too, as in the case of the development of discourse connectives such as in fact, concerning). In other words, it is the subjectivity of the speech event. This hypothesis is in direct opposition to the many models of language change that are based on the assumption that language change is triggered by child language acquisition, and are therefore hearer models (e.g. Lightfoot 1999). A hearer/perception model explains little or nothing about why subjectification occurs at all. As speakers, we tend to understand in terms of our own schemas, so why would we constantly try to process from the perspective of the interlocutor, enrich the interlocutor’s subjective perspective, and semanticize it? Furthermore, it explains nothing about why subjectification would precede intersubjectification. 25. Similarly, although frequency is often construed as a factor contributing to “the mechanisms of change associated with grammaticalization” (Bybee 2003: 602), it is not a cause. 26. How the relative sizes of bit, piece, iota, shred, bunch, lot, etc. can be established is hard to determine in historical data.

56

Elizabeth Closs Traugott

Grammaticalization and especially (inter)subjectification suggest that a passive view of change is not sufficient. The fine-grainedness of the collocations and the social implications of the changes suggest not only children but also teens and adults play a role in change (see Bergs 2005 on evidence for change as communally as well as generationally motivated; also Croft 2000; Traugott and Dasher 2002; Milroy 2007). Recent work on frequency as a contributing factor in grammaticalization (e.g. Bybee and Hopper 2001; Bybee 2003, 2006) further shows that a production model is needed, and recent processing studies have suggested that production factors may override consideration of hearers’ likely ease of perception (Wasow 1997). Because of the deictic shifts of Speaker and Addressee in the speaking dyad, we need a model of change that accounts for how members of the dyad interact, and above all a production model of how speakers construct “arguments about propositions, and [assign] degrees or statements of confidence to those propositions” (Moxey and Sanford 1997: 229). 6. Operationalizing subjectification Recently there have been some interesting attempts to operationalize subjectification. The aim is to move from the admittedly rather vague notion of ambient subjectivity and interlocutor interaction as a motivating force in subjectification, toward identifying the types of linguistic context in which one might expect to find evidence for subjectification. This work draws on quantitative analysis of variation involving subjectivity in contemporary corpora (cf. Scheibman 2002). Investigating “the increasingly speaker-based construal of (counter-) expectation” as represented by middle-marked salirse ‘leave (despite obstacles/surreptitiously)’ from Old Spanish to present-day Mexico City Spanish, Aaron and Torres Cacoullous (2005: 609) suggest the following as structural correlates of subjectivity which can be considered likely contexts in which the subjectification of salirse occurred: (26)

a. b. c.

1st or 3rd person singular referents close to the speaker positive polarity past tenses

The first factor, person, can be understood as reflecting the hypothesis that 1st persons and 3rd persons with a close relationship to the speaker (e.g.

(Inter)subjectivity and (inter)subjectification: A reassessment

57

family membership), are emotionally salient, in contrast to casual acquaintances (Aaron and Torres Cacoullous 2005: 616). Likewise the temporal factor reflects emotional engagement in past narrative. The positive polarity factor (understood as affirmative contexts) reflects the fact that expression of counter-expectation would be incongruous with negation (Aaron and Torres Cacoullous 2005: 618). The study shows that, even though salirse has counter-expectation meanings from the time of the earliest documents (twelfth century), there has been a dramatic increase in the use of 1st person subjects, and there is robust evidence of the kind of change that is often mentioned in grammaticalization studies: the gradual loss of the preposition de ‘from’ and its absorption of its contextual meaning “into the salirse form itself” (Aaron and Torres Cacoullous 2005: 621). Overall there has been “increasingly speaker-based construal of (counter-) expectation, as the use of the form extends to situations involving non-physical abstract, even social, force dynamics” (Aaron and Torres Cacoullos 2005: 629). In a complementary paper Torres Cacoullos and Schwenter (2006) explore ways to operationalize degrees of subjectification as exemplified by the development of Old Spanish pesar de ‘regret of’ into the concessive connectives a pesar de (que) ‘in spite of (that)’. They construe subjectification in this case as “the evolution of the opposition by an outside force into the superimposition of the speaker’s viewpoint onto the viewpoint of another” (Torres Cacoullos and Schwenter 2006: 352). Like Aaron and Torres Cacoullos they seek to identify subjectification in the expansion of certain types of functional range. The key factors in their study are: (27)

a. b. c.

shift from non-coreferentiality to co-referentiality of the subject of the main verb and the adnominal genitive extension to subjunctive verb forms preposing of a pesar de, i.e. increase in syntactic scope

Without seeking to operationalize subjectification, but nevertheless to identify likely linguistic contexts for subjectification, Horie and Kondo (2004) show that, in the Kansai Dialect of Japanese, what they consider to be the subjectivized negative appears predominantly in: (28)

a. b.

first person contexts with verbs of evaluative attitude and cognition.

58

Elizabeth Closs Traugott

These are important studies that seek to find structural groundings for the admittedly rather imprecise notion of subjectification. However, because the relevant factors are so different, these variation- and multivariate analysis-based studies raise the question whether it is possible to identify factors of subjectification that are replicable across languages and construction-types, independently of those that might be particular to a construction. By hypothesis, it seems likely that shifts in the referent of the subject would generally be relevant to the development of subjective meanings. Shifts toward first person subjects are not necessary correlates of or indicators of subjectivity since subjectification may be most apparent precisely where there is no overt subject, first person or otherwise (Scheibman 2002: 167 herself cautions that “the presence of I does not necessarily subjectify on its own”). Obvious examples include the case of raising constructions (There’s going to be an earthquake), and discourse markers like y’know, God wot, etc. (Brinton 1996). It seems likely that when the subjectification is orientation toward a negative evaluation, the locus of change will be comments about non-first persons, especially third persons (“They” are negatively evaluated in some way), see Kranich (2007) on the development of progressives with always, as in Paul is always writing me letters (ĺ ‘and I find this aggravating’) as opposed to non-evaluative Paul always writes me letters. Another likely locus for subjectification is transitivity. In an early study, Zubin (1975: 19–20) suggested that in German when there is case alternation between dative and accusative, the speaker subjectively perspectivizes the situation, and assesses a participant’s contribution (including his or her own) as less vivid, relatively static (accusative) or more vivid, affected (dative). Thompson and Hopper (2001) suggested that a correlation between low transitivity and subjectivity in conversation is a function of the fact that “our talk is mostly about ‘how things are from our perspective’ … these are reflections of subjectivity in our everyday use of language” (Thompson and Hopper 2001: 53; italics original). The extent to which changes in transitivity relates to subjectification remains to be determined, except in the domain of the development of raising constructions (Langacker 1995). Yet another likely locus is semantic polarity sensitivity. However, the particular change in question may show different specific correlations. While the counter-expectation of salirse may be unlikely to arise in negative environments, expressions that evaluate small degree as below the norm or less than adequate may favor negative polarity (a bit, a shred of).

(Inter)subjectivity and (inter)subjectification: A reassessment

59

Relevant polarity sensitivity is not only clause- but also phrase-internal. As we have seen, in the case of (not) a shred of the head is almost exclusively positively evaluated (honor, dignity, evidence) (phrase-internal polarity). Hoeksema (1996–1997) has shown that during the twentieth century a division of labor has developed between two Dutch Degree Adverbs, bar and bijster, both meaning ‘very, all that’. Bar has come to be favored in positive clauses but with negative members of an antonym pair (e.g. ‘bad’), while bijster has come to be favored in negative clauses with positive members of an antonym pair (e.g. ‘good, uninspired’): (29)

a. b.

Real Madrid was niet bijster geïnspireerd tegen Albacete. ‘Real Madrid was not terribly inspired against Albacete’ Werpers zijn hard nodig bij de bar slecht spelende Giants. ‘Pitchers are badly needed with the very badly playing Giants’

The evaluations emerging from such distributional shifts strongly point to the importance of investigating polarity as a context for subjectification, but do not seem to be generalizable beyond the particular examples. To my knowledge, there have not been attempts to operationalize intersubjectification. Here again there is a need to distinguish the context of 2nd person subjects (addressees) from intersubjectivity that is relevant to intersubjectification. While you see and y’know, with their routinized 2nd person subjects, might appear to be markers of intersubjectivity (example 19e is a case in point), they are actually often used for subjective purposes, to negotiate speaker meaning, as in the case of subjectified a bit of. Even imperative utterances, which are clearly in themselves initially intersubjective, may be subjectified. Examples include the development of hortative let’s (as in Let’s go, shall we?) from (you) let us X, (as in Let us go, will you?) (Traugott and Dasher 2002: 176–178), and of discourse marker say in many of its meanings (‘assume, about, for example, tell me’). Say derives from an intersubjective, imperative use, but is subjectified over time, e.g. the ‘about’ use is a type of topicalizer, and the ‘tell me’ use expresses speaker’s impatience (Brinton 2005). One context that is useful to track in the search for structural correlates of (inter)subjectification is changing position in the phrase or clause. In her study of tense, aspect and mood, Bybee (1985) showed that mood (defined as epistemic modality) typically occurs on the periphery of the verbal complex (Bybee 1985: 34–35, 196–200). She correlated this position with degree of relevance to the verb (mood has sentential scope, and therefore is

60

Elizabeth Closs Traugott

less relevant to the verb than tense and aspect). Mood is typically grammaticalized from a variety of sources (see Bybee, Perkins, and Pagliuca 1994: 240–241 for one proposal), and subjectified. A growing number of studies have suggested that as they are subjectified linguistic elements are used in increasingly peripheral positions. Typically the shift is leftward in VO languages, and rightward in OV languages. In English many discourse markers are associated with left (sometimes right) periphery, and their use in this position can be correlated with subjectification of their meaning (see e.g. Traugott and Dasher 2002 on indeed, in fact, actually; Brinton 2007 on I mean). It has further been suggested that subjectified meanings of adjectives are to be found in the left periphery of the NP, see e.g. Adamson (2000) on the development from descriptive to affective meanings of lovely as in a lovely little example, and Breban (2008) on the word order correlations of subjectification and grammaticalization in the development of adjectives like different, distinct.27 Likewise, in Japanese many items that are subjectified or intersubjectified come be used on the periphery of the clause (see e.g. Onodera 2005; Onodera and Suzuki 2007). To what extent there are cross-constructional and cross-linguistic constraints on the leftward (or rightward) shift over time of expressions that undergo (inter)subjectification remains to be determined (Traugott 2007). 7. Conclusion I have hypothesized that subjectification and intersubjectification involve the reanalysis as coded meanings of pragmatic meanings arising in the context of speaker-hearer negotiation of meaning. Subjectification is the development of meanings that express speaker attitude or viewpoint, while intersubjectification is the development of the speaker’s attention to addressee self-image. This requires rethinking Halliday and Hasan’s “interpersonal” meanings as having dual functions: subjective and oriented toward the speaker, and intersubjective, oriented toward the addressee. It also requires more careful distinction than I have previously made between increases in

27. Breban (2008) relates the leftward shift of evaluating adjectives to the intersection of three things: i) the constructional properties of NPs in English, in which the rightward elements have descriptively specific semantics, while the leftward ones denote quantity and grounding (Langacker 1991), ii) grammaticalization, iii) subjectification.

(Inter)subjectivity and (inter)subjectification: A reassessment

61

pragmatic (inter)subjective invited inferences in specific contexts and the semanticization of these invited inferences. I have argued that subjectification and intersubjectification are independent of grammaticalization. However, since grammaticalization involves the development of markers of speaker attitude toward the ideational component and toward textual connectivity (among many other things), there is inevitably a close interaction between grammaticalization and subjectification. The intersection of grammaticalization and intersubjectification is less common, since the latter largely involves expressions of politeness, and cross-linguistically these tend to be associated with lexical choices rather than with grammatical ones. Attempts to identify evidence for the structural contexts of subjectification have confirmed how extensive its reach can be. However, it appears that each construction needs to be studied in its own terms before any cross-constructional let alone cross-linguistic predictions can be made about which contexts are the most likely to enable subjectification to occur. A fuller understanding of contexts for subjectification will presumably arise eventually from studies of the collocations not only within local constructions (“collostructions”, see Gries and Stefanowitsch 2004), but also across a large number of distributional contexts in a large number of languages over time (see Hilpert 2008).28 Sources of data LION MED OED UVa

Chadwyck Healey website. http://lion.chadwyck.com The Middle English Dictionary. 1956–2001. Ann Arbor: University of Michigan Press. (see also http://www.hti.umich.edu/dict/med/) Oxford English Dictionary, 3rd ed. (in progress). http://dictionary.oed.com/ University of Virginia, Electronic Text Center, Modern English Collection. http://etext.lib.virginia.edu/modeng/modeng0.browse.html

28. Such studies will presumably also test to what extent the Invited Inferencing Theory of Semantic Change (Traugott and Dasher 2002) is correct that utterance-token meanings (particularized implicatures or IIN’s) precede the general acceptance of utterance-type meanings that develop out of them (generalized implicatures or GIIN’s). In their critique of this hypothesis, Hansen and Waltereit (2006) focus not so much on the token (IIN) vs. type (GIIN) distinction, but on a distinction between foregrounding (PCI) vs. backgrounding (GCI).

62

Elizabeth Closs Traugott

References Aaron, Jessi Elana and Rena Torres Cacoullos 2005 Quantitative measures of subjectification: A variationist study of Spanish salir(se). Cognitive Linguistics 16: 607–633. Aarts, Bas 2001 English Syntax and Argumentation. 2nd ed. New York: Palgrave. Achard, Michel and Suzanne Kemmer (eds.) 2004 Language, Culture, and Mind. Stanford, CA: CSLI. Adamson, Sylvia 2000 A lovely little example: Word order options and category shift in the premodifying string. In Pathways of Change: Grammaticalization in English, Olga Fischer, Anette Rosenbach, and Dieter Stein (eds.), 39–66. Amsterdam/Philadelphia: Benjamins. Aijmer, Karin 1997 I think – an English modal particle. In Modality in Germanic languages, Toril Swan and Olaf Jansen Westvik (eds.), 1–47. Berlin/New York: Mouton de Gruyter. Allen, Keith and Kate Burridge 1991 Euphemism and Dysphemism: Language Used as Shield and Weapon. New York: Oxford University Press Andersen, Henning 2001 Actualization and the (uni)directionality of change. In Actualization: Linguistic Change in Progress, Henning Andersen (ed.), 225–248. Amsterdam/Philadelphia: Benjamins. Anttila, Raimo 1989 Historical and Comparative Linguistics. 2nd ed. Amsterdam/Philadelphia: Benjamins. Original edition, New York: Macmillan, 1972. Anttila, Raimo 1992 Historical explanation and historical linguistics. In Explanation in Historical Linguistics, Garry W. Davis and Gregory K. Iverson (eds.), 17–39. Amsterdam/Philadelphia: Benjamins. Athanasiadou, Angeliki, Costas Canakis, and Bert Cornillie 2006 Introduction. In Athanasiadou, Canakis, and Cornillie (eds.), 1–13. Athanasiadou, Angeliki, Costas Canakis, and Bert Cornillie (eds.) 2006 Subjectification: Various Paths to Subjectivity. Berlin/New York: Mouton de Gruyter. Benveniste, Emile 1971 Subjectivity in language. In Problems in General Linguistics, 223– 230. Trans. by Mary Elizabeth Meek. Coral Gables: FL: University of Miami Press. Original French publication, 1958.

(Inter)subjectivity and (inter)subjectification: A reassessment

63

Bergs, Alexander 2005 Social Networks and Historical Linguistics. Studies in Morphosyntactic Variation in the Paston Letters (1421–1503). Berlin/New York: Mouton de Gruyter. Bréal, Michel 1964 Semantics: Studies in the Science of Meaning. Trans. by Mrs. Henry Cust. New York: Dover. Original publication, 1900. Breban, Tine 2008 The grammaticalization and subjectification of English adjectives expressing difference into plurality/distributivity markers and quantifiers. Folia Linguistica 42: 259–306. Brems, Lieselotte 2003 Measure noun constructions: An instance of semantically-driven grammaticalization. International Journal of Corpus Linguistics 8: 283–312. Brinton, Laurel J. 1996 Pragmatic Markers in English: Grammaticalization and Discourse Function. Berlin/New York: Mouton de Gruyter. Brinton, Laurel J. 2005 Processes underlying the development of pragmatic markers: The case of (I) say. In Opening Windows on Texts and Discourses of the Past, Janne Skaffari, Matti Peikola, Ruth Carroll, Risto Hiltunen, and Brita Wårvik (eds.), 279–299. Amsterdam /Philadelphia: Benjamins. Brinton, Laurel J. 2007 The development of I mean: Implications for the study of historical pragmatics. In Methods in Historical Pragmatics, Susan Fitzmaurice and Irma Taavitsainen (eds.), 37–80. Berlin/New York: Mouton de Gruyter. Brinton, Laurel J. and Elizabeth Closs Traugott 2005 Lexicalization and Language Change. Cambridge: Cambridge University Press. Brisard, Frank 2006 Logic, subjectivity, and the semantics/pragmatics distinction. In Athanasiadou et al. (eds.), 41–74. Bybee, Joan L. 1985 Morphology: A Study of the Relation between Meaning and Form. Amsterdam/Philadelphia: Benjamins. Bybee, Joan L. 2003 Mechanisms of change in grammaticalization: The role of frequency. In Joseph and Janda (eds.), 602–623. Bybee, Joan L. 2006 From usage to grammar: The mind’s response to repetition. Language 82: 711–733.

64

Elizabeth Closs Traugott

Bybee, Joan L. and Paul Hopper (eds.) 2001 Frequency and the Emergence of Linguistic Structure. Amsterdam/Philadelphia: Benjamins. Bybee, Joan L., Revere Perkins, and William Pagliuca 1994 The Evolution of Grammar: Tense, Aspect, and Modality in the Languages of the World. Chicago: University of Chicago Press. Cornillie, Bert 2004 Evidentiality and Epistemic Modality in Spanish (Semi-) Auxiliaries: A Functional-pragmatic and Cognitive-linguistic Account. PhD. dissertation, Katholieke Universiteit Leuven. Croft, William 2000 Explaining Language Change: An Evolutionary Approach. Harlow, Essex: Pearson Education. Dahl, Östen 2004 The Growth and Maintenance of Linguistic Complexity. Amsterdam/Philadelphia: Benjamins. De Smet, Hendrik and Jean-Christophe Verstraete 2006 Coming to terms with subjectivity. Cognitive Linguistics 17: 365– 392. Denison, David 2002 Semantic pathways with sort of, kind of, type of, and their relation to grammatical gradience. Paper presented at Stanford University, May 2002 (see also History of the sort of construction family, ICCG2, Helsinki (http://lings.ln.man.ac.uk/info/staff/dd/default.html). Eckardt, Regine 2006 Meaning Change in Grammaticalization: An Enquiry into Semantic Reanalysis. Oxford: Oxford University Press. Enfield, Nicholas J. 2003 Linguistic Epidemiology: Semantics and Grammar of Language Contact in Mainland Southeast Asia. London/New York: Routledge. Fischer, Olga 2007 Morphosyntactic Change: Functional and Formal Perspectives. Oxford/ New York: Oxford University Press. Fitzmaurice, Susan 2000 Coalitions and the investigation of social influence in linguistic history. European Journal of English Studies 4: 265–276. Fitzmaurice, Susan 2004 Subjectivity, intersubjectivity and the historical construction of interlocutor stance: From stance markers to discourse markers. Discourse Studies 6: 427–448. Garrett, Andrew Forthc. The historical syntax problem: Reanalysis and directionality. In Grammatical Change: Origins, Nature, Outcomes, Dianne Jonas,

(Inter)subjectivity and (inter)subjectification: A reassessment

65

John Whitman and Andrew Garrett (eds.). Oxford: Oxford University Press. http://linguistics.berkeley.edu/~garrett/publications.html Garrod, Simon n.d. Interactive alignment and routinization as mechanisms for language change. www.psy.gla.ac.uk/~simon/Erice.pdf Gries, Stefan Th. and Anatol Stefanowitsch 2004 Co-varying collexemes in the into-causative. In Achard and Kemmer (eds.), 225–236. Gutiérrez-Rexach, Javier and Scott A. Schwenter 2002 Propositional NPIs and the scalar nature of polarity. In From Words to Discourse: Trends in Spanish Semantics and Pragmatics, Javier Gutiérrez-Rexach (ed.), 237–262. Amsterdam: Elsevier. Halliday, M. A. K. 1994 An Introduction to Functional Grammar. 2nd ed. London: Arnold. Halliday, M. A. K. and Ruqaiya Hasan 1976 Cohesion in English. London: Longman. Hansen, Maj-Britt Mosegaard and Richard Waltereit 2006 GCI theory and language change. Acta Linguistica Hafniensia 38: 235–268. Harris, Alice C. and Lyle Campbell 1995 Historical Syntax in Cross-Linguistic Perspective. Cambridge: Cambridge University Press. Haspelmath, Martin 1999 Why is grammaticalization irreversible? Linguistics 37: 1043–1068. Heine, Bernd 2002 On the role of context in grammaticalization. In New Reflections on Grammaticalization, Ilse Wischer and Gabriele Diewald (eds.), 83– 101. Amsterdam/Philadelphia: Benjamins. Heine, Bernd and Tania Kuteva 2002 World Lexicon of Grammaticalization. Cambridge: Cambridge University Press. Hilpert, Martin 2008 Germanic Future Constructions—A Usage-based Approach to Language Change. Amsterdam/Philadelphia: Benjamins. Himmelmann, Nikolaus P. 2004 Lexicalization and grammaticalization: Opposite or orthogonal? In What Makes Grammaticalization—a Look from its Fringes and its Components, Walter Bisang, Nikolaus Himmelmann, and Björn Wiemer (eds.), 19–40. Berlin/New York: Mouton de Gruyter.

66

Elizabeth Closs Traugott

Hoeksema, Jack 1996–97 Corpus study of negative polarity items. IV–V Jornades de corpus linguistics 1996–1997. Barcelona: Universitat Pompeu Fabre. http://odur.let.rug.nl/~hoeksema/docs/barcelona.html Hoeksema, Jack 2001 Change among expletive polarity items. In Historical Linguistics 1999: Selected Papers from the 14th International Conference on Historical Linguistics, Vancouver, 9–13 August 1999, Laurel J. Brinton (ed.), 175–186. Amsterdam/ Philadelphia: Benjamins. Hopper, Paul J. 1991 On some principles of grammaticization. In Traugott and Heine (eds.), Vol. I, 17–35. Hopper, Paul J. and Elizabeth Closs Traugott 2003 Grammaticalization. 2nd rev. ed. [1st ed., 1993.] Cambridge: Cambridge University Press. Horie, Karuo and Emi Kondo 2004 Subjectification and synchronic variation: Two negation forms in Kansai Dialect of Japanese. In Achard and Kemmer (eds.), 445–459. Horn, Laurence R. 1984 Toward a new taxonomy for pragmatic inference: Q-based and Rbased implicature. In Meaning, Form, and Use in Context: Linguistic Applications; Georgetown University Round Table ’84, Deborah Schiffrin (ed.), 11–42. Georgetown University Press, Washington D. C. Israel, Michael 1996 Polarity sensitivity as lexical semantics. Linguistics and Philosophy 19: 619–666. Israel, Michael 2004 The pragmatics of polarity. In The Handbook of Pragmatics, Laurence R. Horn and Gregory Ward (eds.), 701–723. Oxford/Malden, MA: Blackwell. Jespersen, Otto 1917 Negation in English and other Languages. Copenhagen: A. Høst. Joseph, Brian D. and Richard D. Janda (eds.) 2003 The Handbook of Historical Linguistics. Malden, MA: Blackwell. Jucker, Andreas H. 1997 The discourse marker well in the history of English. English Language and Linguistics 1: 91–110. Keller, Rudi 1994 On Language Change: The Invisible Hand in Language. Trans. by Brigitte Nerlich. London: Routledge.

(Inter)subjectivity and (inter)subjectification: A reassessment

67

Kiparsky, Paul Forthc. Grammaticalization as optimization. In Grammatical Change: Origins, Nature, Outcomes, Dianne Jonas, John Whitman and Andrew Garrett (eds.). Oxford: Oxford University Press. http://www.stanford.edu/~kiparsky/ König, Ekkehard 1991 The Meaning of Focus Particles: A Comparative Perspective. London/New York: Routledge. Kranich, Svenja 2007 Subjectification and the English progressive: The history of ALWAYS + progressive constructions. York Papers in Linguistics 2: 120–137. Krifka, Manfred 1995 The semantics and pragmatics of polarity expressions. Linguistic Analysis 25: 209–257. Langacker, Ronald W. 1990 Subjectification. Cognitive Linguistics 1: 5–38. Langacker, Ronald W. 1991 Foundations of Cognitive Grammar. Volume 2: Descriptive Application. Stanford: Stanford University Press. Langacker, Ronald W. 1995 Raising and transparency. Language 71: 1–62. Langacker, Ronald W. 2003 Extreme subjectification: English tense and modals. In Motivation in Language: Studies in Honor of Günther Radden, Hubert Cuyckens, Thomas Berg, René Dirven, and Klaus-Uwe Panther (eds.), 3–26. Amsterdam/Philadelphia: Benjamins. Lehmann, Christian 1995 Thoughts on Grammaticalization. Munich/Newcastle: LINCOM EUROPA. Original publication in Arbeiten der Kölner UniversalienProjektes 48, University of Cologne, 1982. Levinson, Stephen C. 2000 Presumptive Meanings: The Theory of Generalized Conversational Implicature. Cambridge, MA: The MIT Press, A Bradford Book. Lightfoot, David 1999 The Development of Language: Acquisition, Change, and Evolution. Malden, MA: Blackwell. Lyons, John 1982 Deixis and subjectivity: Loquor, ergo sum? In Speech, Place, and Action: Studies in Deixis and Related Topics, Robert J. Jarvella and Wolfgang Klein (eds.), 101–124. New York: Wiley.

68

Elizabeth Closs Traugott

Lyons, John 1994 Subjecthood and subjectivity. In Subjecthood and Subjectivity: The Status of the Subject in Linguistic Theory, Marina Yaguello (ed.), 9– 17. Paris: Ophrys/London: Institut Français du Royaume-Uni. Meillet, Antoine 1958 Reprint. L’évolution des formes grammaticales. In Linguistique historique et linguistique générale, Antoine Meillet, 130–148. Paris: Champion. Original publication in Scientia (Rivista di Scienza) 12, 1912. Milroy, Lesley 2007 Off the shelf or under the counter? On the social dynamics of sound change. In Studies in the History of the English Language III: Managing Chaos: Strategies for Identifying Change in English, Christopher M. Cain and Geoffrey Russom (eds.), 149–172. Berlin/New York: Mouton de Gruyter. Moxey, Linda M. and Anthony J. Sanford 1997 Choosing the right quantifier: Usage in the context of communication. In Conversation: Cognitive, Communicative and Social Perspectives, Talmy Givón (ed.), 207–231. Amsterdam/Philadelphia: Benjamins. Nuyts, Jan 2001 Epistemic Modality, Language and Conceptualization. Amsterdam/Philadelphia: Benjamins. Nuyts, Jan 2005 Modality: Overview and linguistic issues. In The Expression of Modality, William Frawley (ed.), 1–26. Berlin/New York: Mouton de Gruyter. Onodera, Noriko O. 2004 Japanese Discourse Markers: Synchronic and Diachronic Discourse Analysis. Amsterdam/Philadelphia: Benjamins. Onodera, Noriko O. and Ryoko Suzuki (eds.) 2007 Journal of Historical Pragmatics 8; special issue on subjectivity, intersubjectivity and historical changes in Japanese. Prince, Ellen 1988 Discourse analysis: A part of the study of linguistic competence. In Linguistics: The Cambridge Survey, Vol. II Linguistic Theory: Extensions and Applications, Frederick J. Newmeyer (ed.), 164–182. Cambridge: Cambridge University Press. Scheibman, Joanne 2002 Point of View and Grammar: Structural Patterns of Subjectivity in American English Conversation. Amsterdam/Philadelphia: Benjamins.

(Inter)subjectivity and (inter)subjectification: A reassessment

69

Schiffrin, Deborah 1987 Discourse Markers. Cambridge: Cambridge University Press. Schiffrin, Deborah 1990 The management of a co-operative self during argument: The role of opinions and stories. In Conflict Talk: Sociolinguistic Investigations of Arguments in Conversations, Allen D. Grimshaw (ed.), 241–259. Cambridge: Cambridge University Press. Schwenter, Scott A. 2006 Fine-tuning Jespersen’s Cycle. In Drawing the Boundaries of Meaning: Neo-Gricean Studies in Pragmatics and Semantics in Honor of Laurence R. Horn, Betty J. Birner and Gregory Ward (eds.), 327– 344. Amsterdam/Philadelphia: Benjamins. Stoffel, Cornelis 1901 Intensives and Down-Toners: A Study in English Adverbs. Heidelberg: Carl Winter. Suzuki, Ryoko 1998 From a lexical noun to an utterance-final pragmatic particle: wake. In Studies in Japanese Grammaticalization, Toshio Ohori (ed.), 67–92. Tokyo: Kurosio Publishers. Tabor, Whitney and Elizabeth Closs Traugott 1998 Structural scope expansion and grammaticalization. In The Limits of Grammaticalization, Anna Giacalone Ramat and Paul J. Hopper (eds.), 229–272. Amsterdam/Philadelphia: Benjamins. Thompson, Sandra A. and Paul J. Hopper 2001 Transitivity, clause structure, and argument structure: Evidence from conversation. In Bybee and Hopper (eds.), 27–60. Torres Cacoullos, Rena and Scott A. Schwenter 2006 Towards an operational notion of subjectification. In Proceedings of the 31st Annual Meeting of the Berkeley Linguistics Society: General Session and Parasession on Prosodic Variation and Change, Rebecca T. Cover and Yuni Kim (eds.), 347–358. Berkeley, CA: Berkeley Linguistics Society. Traugott, Elizabeth Closs 1982 From propositional to textual and expressive meanings: Some semantic-pragmatic aspects of grammaticalization. In Perspectives on Historical Linguistics, Winfred P. Lehmann and Yakov Malkiel (eds.), 245–271. Amsterdam/Philadelphia: Benjamins. Traugott, Elizabeth Closs 2003a From subjectification to intersubjectification. In Motives for Language Change, Raymond Hickey (ed.), 124–139. Cambridge: Cambridge University Press.

70

Elizabeth Closs Traugott

Traugott, Elizabeth Closs 2003b Constructions in grammaticalization. In Joseph and Janda (eds.), 624–647. Traugott, Elizabeth Closs 2007 (Inter)subjectification and unidirectionality. Journal of Historical Pragmatics 8: 295–309. Traugott, Elizabeth Closs 2008a Grammatikalisierung und emergente Konstruktionen. In Konstruktionsgrammatik und grammatische Konstruktionen, Anatol Stefanowitsch and Kerstin Fischer (eds.), 5–32. Trans. by Arne Zechel. Tübingen: Stauffenburg. Traugott, Elizabeth Closs 2008b The grammaticalization of NP of NP constructions. In Constructions and Language Change, Alexander Bergs and Gabriele Diewald (eds.), 21–43. Berlin/New York: Mouton de Gruyter. Traugott, Elizabeth Closs 2008c Grammaticalization, constructions and the incremental development of language: Suggestions from the development of degree modifiers in English. In Language Evolution: Cognitive and Cultural Factors, Regine Eckardt, Gerhard Jaeger, and Tonjes Veenstra (eds.), 219– 240. Berlin/New York: Mouton de Gruyter. Traugott, Elizabeth Closs and Richard B. Dasher 2002 Regularity in Semantic Change. Cambridge: Cambridge University Press. Traugott, Elizabeth Closs and Bernd Heine (eds.) 1991 Approaches to Grammaticalization. Amsterdam/Philadelphia: Benjamins, 2 Vols. Verhagen, Arie 2005 Constructions of Intersubjectivity: Discourse, Syntax, and Cognition. Oxford: Oxford University Press. Wasow, Thomas 1997 Remarks on grammatical weight. Language Variation and Change 9: 81–105. Weinreich, Uriel, William Labov, and Marvin I. Herzog 1968 Empirical foundations for a theory of language change. In Directions for Historical Linguistics, W.P. Lehmann and Yakov Malkiel (eds.), 97–195. Austin and London: University of Texas Press.

(Inter)subjectivity and (inter)subjectification: A reassessment

71

Wierzbicka, Anna 2006 English: Meaning and Culture. Oxford/New York: Oxford University Press. Zubin, David 1975 On the distributional properties of surface morphology and their consequences for semantic analysis. Columbia Working Papers in Linguistics 2: 189–218.

Part 2 (Inter)subjectification and grammaticalization involving adverbials

Presupposition accommodation and language change Scott A. Schwenter and Richard Waltereit Abstract Innovations leading to language change are often routinely taken as speaker-based. In this article, we look at how hearers can create new meaning in the case of presupposition triggers. Hearers unable or unwilling to accommodate presuppositions assume a novel interpretation of the erstwhile presupposition-trigger and eventually pass this new interpretation on to other people, thereby changing the language. We discuss one particular lexeme, namely the additive particle too and some of its counterparts in Spanish and German (and to a lesser extent in a range of other languages). We find that this particle provides particularly fertile ground for what we consider to be hearer-initiated changes motivated by too-costly presupposition accommodation. We argue that adversative properties of the dialogual discourse context – prototypically in the form of an argument between two interlocutors – appear to have led hearers to reanalyse too as expressing a new, rhetoricallystrategic meaning with strong counter-argumentative force. The trajectory of change thereby produces a clear path from the ideational/textual meaning of additive too to the more clearly interactionally-bound interpersonal meanings associated with non-additive too.

1. Introduction: Presupposition accommodation as a possible motivation for semantic change 1.1. Hearer-induced semantic change Recent research on language change, and especially on semantic change, has focused on the ways speakers and hearers use language innovatively and create new uses of existing words or constructions in the process (Traugott and Dasher 2002). By “innovative” we mean that the relation between form and meaning is novel, rather than given by the conventions

76

Scott A. Schwenter and Richard Waltereit

of the language or other conventions.1 If the innovative language use by speakers “catches on” and hearers in turn make use of it in their own speech, then gradually that particular form–meaning pairing may enter the inventory of the language and a change will have occurred. One kind of language use which is creative by definition are conversational implicatures. Consider, for example, the following dialogue: (1)

A: Do you want to go to a pub after the film? B: I have to get up early tomorrow morning.

Assuming that B wants to decline A’s offer with this answer, B will have conversationally implicated that content. The very fact of it being conversationally implicated means that the speaker has expressed that content by a form or expression that does not carry it by convention; rather, they rely on the hearer associating that specific interpretational content to the form used solely by contextual inferences. In other words, the speaker “invites” the inference that I have to get up early tomorrow morning means (among other things) ‘no’ in that particular context of utterance. We will refer to nonconventionalized ways of associating form and content such as the preceding as pragmatic meaning. Such a novel use may set off semantic change, and the conversational implicature may eventually conventionalize, i.e. become the coded semantic content of the forms in question. Traugott and Dasher (2002), and Traugott in many other places, have most prominently and fruitfully articulated this view in their Invited Inferencing Theory of Semantic Change. The assumption that at least a salient subset of instances of semantic change consists of the conventionalization of conversational implicatures has become commonplace in the historical linguistics literature. Many researchers in historical linguistics tend to place the burden of the labour necessary to set off any change in the language on the speaker’s creativity. For example, Haspelmath (1999) and Keller (1994) invoke speaker’s maxims such as “speak like others” or “speak extravagantly” as responsible for language change, where the former maxim disfavours language change and the latter favours it. If people just speak like other speakers do, then their behaviour will not lead to language change. If some speakers however succeed in speaking extravagantly, that may lead to others imitating them (in line with “speak like others”) and the language 1. See Morgan’s (1978) distinction of conventions of language and of conventions of language use.

Presupposition accommodation and language change

77

changing in the process. Rarely has the function of hearers been discussed. However it is easy to see that they have an equally important role, since “speakers may innovate as much as they please, but if hearers do not perceive those innovations, they obviously will not take hold” (Hansen 2008: 100). It is not straightforward to predict the relation between the extent of the difference between what is said and what is meant, inherent in any innovation, on the one hand and the likelihood of that innovation catching on in subsequent speakers’ discourse on the other hand. Clearly, the difference between what is said and what is meant must be big enough to be noted in the first place; but at the same time that difference must not be so big as to make it too difficult or even impossible for the hearer to figure out what is really meant. Certainly it may be argued that the greater the mismatch between what is said and what is meant the less easily perceptible will that mismatch be for hearers, and consequently less likely for them to employ it in their role as speakers. However very light and subtle mismatches may not take hold either, because the very complexity of verbal communication implies that any participant in the communication will necessarily have to allow for some mismatch between what is said and what they want to convey (Croft 2000). Hence hearers have a clear “regulatory” role in innovations, as their ability, or willingness, to follow speakers’ innovations places a cap on an innovation’s likelihood to be propagated in the linguistic community. However, the contribution that hearers make to semantic change is not limited to constraining speakers’ creativity. Hearers can indeed have a very active role in that process, namely by assigning novel interpretations to forms, constructions, or utterances they hear and by using these interpretations in their own subsequent use as speakers. A type of pragmatic meaning that we believe illustrates the potentially active role of hearers and that has received much less, if any, attention than conversational implicatures in the overall realm of historical linguistics, are presuppositions. To take a simple example of the phenomenon, the utterance in (2), (2)

I have to take my sister to the station.

presupposes that the proposition ‘the speaker has a sister’ forms part of the speaker and hearer’s Common Ground. However, as is well known, utterances like (2) are very common and actually perfectly felicitous even if the hearer did not know previously that the speaker had a sister. The hearer in such a case accommodates this presupposition (cf. Lewis 1979; von Fintel

78

Scott A. Schwenter and Richard Waltereit

2000; Lambrecht 1994; Beaver and Zeevat 2004; among many others): under the assumption that the speaker is communicating cooperatively, the hearer will accept this accommodated presupposition into the Common Ground. Presupposition accommodation is a pragmatic operation that falls genuinely within the remit of the hearer. Only if the Common Ground, as the set of mutually projected assumptions about each discourse participant’s knowledge, does not contain that proposition will the hearer have to accommodate it in the first place. In the case of an utterance like (2) the hearer may perform this operation noticeably, but smoothly, and will maybe not assume that any deliberate action by the speaker was involved. The speaker could certainly have avoided this operation and have introduced the proposition into the Common Ground as an independent assertion: (2ƍ)

I have a sister. I have to take her to the station.

In the case of (2ƍ), no presupposition accommodation would be involved. However this might be considered a rather cumbersome way of speaking, and intuitively (2) seems to be the more natural and indeed more common variant. Presupposition accommodation, in examples like (2), is economybased: if a speaker, for whatever reason, does not anticipate objections, comment etc. for a proposition, she may as well bring it into the common ground as a presupposition (cf. Givón 1995). Crucial here to the process of accommodation is the “non-controversiality” of the accommodated material (whether a discourse referent or a proposition), oft-cited in the literature (Grice 1981; Atlas and Levinson 1981; Heim 1992). People are typically interested in the course of the conversation rather than in the Common Ground as such, and they will normally be happy to make a reasonable effort to update the Common Ground, as required by an utterance like (2), if this enables them to follow and to successfully contribute to the conversation. The Common Ground is normally in the background of the speakers’ attention, and by the same token, it has a secondary role for the conversation or indeed for the hearer. The accommodated material should therefore be not only non-controversial but also deemed unworthy of being the current focus of attention. Presupposition accommodation may however be used in a “strategic” way, namely in order to avoid any subsequent discussion of the proposition, i.e. to keep it in the background. Imagine a speaker announcing to her father,

Presupposition accommodation and language change

(3)

79

O Dad, I forgot to tell you that my fiancée and I are moving to Seattle next week. (von Fintel 2000)

without telling him beforehand that she got engaged. The proposition that the hearer will have to accommodate is so remarkable (at least in our cultures) that it would certainly have merited introduction into the Common Ground as a full assertion, and thereby to have become a topic of conversation. The ultimately extra-linguistic (cultural) difference between (2) and (3) is that a casual acquaintance having a sister is for many a rather anodyne and insignificant piece of information whereas one’s daughter having a fiancée is normally considered to be highly relevant and significant. Packing this information into a proposition that the hearer has to accommodate, rather than expressing it in a full assertion that the hearer can elaborate on or even challenge, is a way of communicating something without talking about it. It is of course up to the hearer whether he accepts such a move and simply accommodates that probably very important piece of information or whether he challenges it. A hearer who is confronted with an utterance and assigns an interpretation to it that deviates from that utterance’s literal meaning has, in principle, two options: (i)

They can assume that regular pragmatic operations, such as conversational inference and the accommodation of presuppositions, mediate between the literal meaning and the chosen non-literal interpretation. (ii) They can assume a novel conventional meaning for some element of that utterance. In that case, fewer pragmatic computations, and/or less costly ones, may be required, since there is no need anymore for them to mediate between the chosen interpretation and the traditional (previous) conventional meaning.

If a hearer chooses (ii) and uses the novel form–meaning pairing in their own discourse, the language will have changed. But when will a hearer prefer (ii) over (i)? Probably in a situation where the regular pragmatic operations required for (i) are considered too costly, or if s/he is unable to make them in the first place. Eckardt (forthcoming) has suggested the principle “Avoid pragmatic overload” in semantic change: Assume that u [=an utterance] in the old sense Iold requires unbacked presuppositions. The speaker makes his utterance under the assumption that the interpreter will accommodate them. The interpreter may see this possibility but finds the required accommodations implausible. As an interpretive al-

80

Scott A. Schwenter and Richard Waltereit ternative, H [=the hearer] hypothesizes a new message Inew, leading to reanalysis.

This principle predicts that if the pragmatic operations required for option (i) are implausible the hearer will choose option (ii). The principle is adumbrated in Detges and Waltereit’s (2002) “Principle of reference” in reanalysis, which assumes that reanalyses may be guided by an assumption on the part of the hearer that the relation between what is said and what is meant should be maximally simple. In this article, we want to investigate how the strategy (ii) may lead to language change in the case of presupposition triggers.2 Hearers unable or unwilling to accommodate presuppositions assume a novel interpretation of the erstwhile presupposition-trigger and eventually pass this new interpretation on to other people, thereby changing the language. We choose to discuss one particular lexeme, namely the additive particle too and some of its counterparts in Spanish and German (and to a lesser extent in a range of other languages). We found that this particle proved particularly fertile and versatile, as it were, for what we consider hearer-initiated changes motivated by too-costly presupposition accommodation in the sense of “Avoid pragmatic overload”. We will use the upper-case word TOO as a shorthand for the English particle too and its congeners in Spanish (también), German (auch) and in other languages, as well as its respective negative polarity counterparts English either and Spanish tampoco. 1.2. The additive particle too: a presupposition trigger (usually) unable to be accommodated The additive particle too and its cross-linguistic congeners are oft-cited examples of a presupposition trigger that will not permit accommodation of 2. Our study may therefore be considered a contribution to Lambrecht’s (1994: 65–73) work on conventionalization of accommodation. Lambrecht looks at the use of English it-clefts without the clefted proposition being presupposed and of the emphatic do-construction (as in “I did pay the money back!”) without explicit activation of the proposition in question in discourse. Whereas Lambrecht focuses on cases where “the presuppositional structure of a frequently used construction is exploited so regularly that it loses some of its force, sometimes resulting in a new meaning for the construction” (1994: 70), we are studying strictly hearer-led changes with no implication regarding the frequency of the construction.

Presupposition accommodation and language change

81

the presupposed first conjunct, i.e. the corresponding “additive” proposition (Kripke 1990; Beaver and Zeevat 2004): (4)

(#)John had dinner in New York, too.

Simply adding to the Common Ground that someone else (e.g. some unspecified diner) has had dinner in New York does not make too felicitous in this example, even though it would seem to make sense to do so, given the trivial assumption that at least one person besides John had dinner in New York. An utterance like (4) is likely to provoke the question “Who else had dinner in New York?”, reflecting the improbability of the hearer simply assuming (and thereby adding to the Common Ground) that someone else had dinner in New York, without asking who that may be. However, additive too (or TOO) without a first conjunct does seem to be possible in certain contexts: (5) (6) (7)

You too can haiku [http://artsedge.kennedy-center.org/content/2246/] (title of an introduction to haiku poem writing) Prova anche te il nuovo gusto del cafè XXX [brand name] ‘Try you, too, the new flavor of XXX coffee’ (advertisement read in a bar in Loano, Italy) Hey, {men, fish, we} have feelings, too! (Beaver and Zeevat 2004: 32)

It seems that the TOOs in (5–7) appear in a type of context such that the hearer can readily construe a suitable parallelism between the first and the second conjunct, necessary for the felicitous use of TOO (cf. Schwenter and Zulaica-Hernández 2003), to which we will return in Section 3.1 below. In example (7), this parallelism is afforded by semantic contrast evoked by a salient set of entities: men – women, fish – other animals / humans, we – other people etc. (Beaver and Zeevat 2004: 32). The first conjunct that is required to be accommodated is simply the second conjunct’s contrasting complement, so that the hearer can easily assume what the first conjunct would be. Furthermore, the semantic inference from ‘men’ to ‘women’ or from ‘we’ to ‘the others [of a given set]’ or from ‘fish’ to ‘other animals’ provides information of the same degree of specificity about the two conjuncts.

82

Scott A. Schwenter and Richard Waltereit

Examples (5) and (6) are not from conversation, but from written communication targeted at an anonymous readership. Anyone reading them is aware that many other people whom neither the current reader nor the author know personally will have read them as well, so the current reader cannot reasonably assume that any information about who else can write haiku poems or who else has tried or may try a particular coffee specialty is deliberately being withheld from them. The underspecified pragmatic parallelism proposed in Schwenter and Zulaica-Hernández (2003) as the principal meaning component of TOO sheds more light on the difficulty to accommodate the first conjunct in utterances such as (4). What is really disruptive in (4) is not the fact that the first conjunct has been omitted as such. The hearer can in fact easily accommodate that “someone else” besides John had dinner in New York. However the underspecified pragmatic parallelism of TOO implies that if the speaker gives a specific name in the second conjunct then s/he could do the same in the first conjunct, so that saying (4) means that the speaker is withholding the identity of the first diner. In other words, uses of TOO like in (4) are unacceptable not because the hearer is unable to accommodate that someone else has had dinner, but because the pragmatic parallelism of the two conjuncts suggests that the speaker is not giving information they could reasonably be expected to give, hence violating Grice’s Quantity maxim. 2. Non-presupposition-triggering TOO / EITHER Despite the generalization about TOO and accommodation just detailed, in many languages there exist non-presupposition-triggering uses of additive markers all of which, albeit in rather different fashion, have specific, highly constrained discourse-structure- and speech-act-related functions. It appears, in every case, that these uses appear diachronically later than the respective canonical additive use. Besides the very ancient additive meaning of English too, there is a more recent use especially common in American English, where too is not additive and does not imply the topical parallelism between propositions noted above. Rather, the function of too in this use is to deny the truth validity of the propositional content of a previous speaker’s utterance:

Presupposition accommodation and language change

(8)

83

A: You didn’t/won’t do your homework! B: I did/will too!

Spanish tampoco ‘either’ has a “canonical” additive reading, like in the following example: (9)

Pablo no vino a la fiesta, y Maria tampoco. ‘P. did not come to the party, and Maria [did not come] either.’

However, tampoco also has an independent reading (Schwenter 2003), wherein its main discourse function is to attenuate or “de-realize” (Ducrot 1995) an inference issuing from a previous speaker’s utterance. These independent uses appear to exist since the 17th century (we will return to the issue of dating the examples later): (10)

A: Pablo está borracho. ‘P. is drunk.’ (ĺ He cannot drive) B: Sí, pero tampoco ha bebido tanto. ‘Yes, but he hasn’t drunk that much’ (ĺ He’s OK to drive)

Similar non-additive uses of additive markers are found in other languages as well. The German particle auch ‘too’ also displays independent readings attested since the 15th century: (11)

A: Das IST auch ein Problem! ‘Hey, that’s a real problem!’

The French additive particle encore ‘more’ has a non-additive reading as in (12): (12)

Elle m’a engueulé – et devant mon copain encore! (Hansen 2002: 163) ‘She told me off – and in front of my friend, too!’

Likewise, in Danish, the additive particle også ‘too’ and its negative polarity counterpart heller ikke ‘either’ each have non-additive readings when in combination with the modal particle (MP) da (this observation and the following examples are from Maj-Britt Mosegaard Hansen, p.c.):

84 (13)

Scott A. Schwenter and Richard Waltereit

a. b.

A: Du skulle ikke have skældt Jette ud. B: Det har jeg (da) [heller ikke] gjort. ‘A. You shouldn’t have told Jette off. B. I didn’t (MP), either.’ A: Hvorfor har du ikke ringet til Ole? B: Det har jeg (da) [også] gjort. ‘A. Why haven’t you called Ole? B. I have (MP), too.’

Similar non-additive uses of additive particles exist in European and Brazilian Portuguese (Patrícia Matos Amaral and Gláucia Silva, p.c.). Crucially, in all these examples the erstwhile presupposition trigger (TOO) does not force the hearer to accommodate an “additive” presupposition. Indeed, no such proposition is even potentially available to the hearer in the discourse record. Summarizing so far, the examples adduced up till now suggest that there is a cross-linguistic pathway recruiting additive particles to non-additive uses. This pathway seems to confirm Traugott’s subjectification hypothesis: the additive function is clearly an ideational one, as its most fundamental conventional element of meaning is the truth-functional conjunction of two propositions p and q. The requirement of shared topicality that we assumed constrains the conjunction of p and q belongs, again with respect to Traugott’s hypothesis, to the textual realm as it makes reference to textual organization and discourse structure. In other words, the additive particle already contains a non-ideational element of meaning as part of its conventional semantics. Finally, the non-additive uses of the particles, as the diachronic outcome of the process of change, may be characterized as interpersonal (intersubjective), again with respect to Traugott’s subjectification tendencies, to the extent that they imply a dialogual situation where a speaker directly refers to an addressee’s utterance (most typically one that is maximally salient in the ongoing discourse context). 3. Pathways from additive (ideational) to independent (interpersonal) uses Having mentioned a number of independent uses of TOO all of which arose later than the canonical additive meaning of TOO, we will now tackle the question as to how these independent uses may have arisen. First however we will have to provide a clear idea of the structure of additive TOO in

Presupposition accommodation and language change

85

order to be able hypothesize how the independent uses may have arisen from it. 3.1. The structure of additive TOO Additive TOO links two propositions p and q. Many researchers have struggled with the question as to what exactly the terminology “additive” means. Under what conditions can two sentences or utterances p and q be linked by TOO? The simplest instances of additive constructions seem to be those with q involving a syntactic ellipsis of p, such as (14): (14)

John had ice cream, and Mary had ice cream too.

Here there is an obvious syntactic parallelism between p and q. However, as (15) from Spanish shows, there need not be any syntactic or presuppositional parallelism between the two propositions: (15)

La biblioteca de OSU no tiene muchas computadoras y tampoco tiene buen ambiente para estudiar. ‘The OSU library doesn’t have many computers and it doesn’t have a good studying environment either.’

This suggests that constraints on TOO are pragmatic, rather than syntactic, in nature. Rejecting previous analyses that claim that q is either implicated or even entailed by p, Schwenter and Zulaica-Hernández (2003) suggest that “the relationship between the text sentences (marked by tampoco) and the context sentence is that they must display the property of shared topicality. ... we claim that the parallelism between the text and context sentences results from their shared status as parallel, but at the same time independent, answers to the same explicit or implicit question induced in the discourse.” In other words, the two propositions p and q contribute to the same point in discourse, and TOO is signalling precisely that. In (14), the two propositions jointly answer the question “who had ice cream?”, and in (15) they jointly answer the question “what is the OSU library like?” The condition of shared topicality is a conventional one. Only if the two propositions actually make the same point, i.e. if the hearer can construct the conjunction of two propositions as answering the same underlying ques-

86

Scott A. Schwenter and Richard Waltereit

tion, is the use of TOO acceptable (cf. Schwenter and Zulaica-Hernández 2003). Now, shared topicality makes additive TOO suitable for argumentative contexts. Argumentation theory (cf. Anscombre and Ducrot 1980 for an overview) assumes that speakers provide arguments in conversation to support or reject (explicit or implicit) contextual conclusions. These arguments may be of varying strength. Arguments that support the identical conclusion are called co-oriented arguments. The fact that TOO marks shared topicality means that two propositions used as arguments p and q conjoined by TOO are typically co-oriented arguments. Furthermore, two co-oriented arguments p and q implicate greater argumentative strength for the speaker’s desired conclusion than p alone, or even than the simple conjunction of p^q, whose topical parallelism must be inferred. Consider the following utterances: (16)

a. That car is affordable. (ĺ that’s why I want to buy it) b. That car has a nice color. (ĺ that’s why I want to buy it) c. That car is affordable, and it’s a nice color. (ĺ that’s why I want to buy it) d. That car is affordable, and it’s a nice color too. (ĺ that’s why I want to buy it)

(16a) and (16b) present That car is affordable and That car has a nice color, respectively, as an argument for the conclusion I want to buy this car. The two conjoined arguments That car is affordable and it has a nice color in (16c) support the same conclusion. They are co-oriented arguments. However this is not explicitly marked, their co-orientation and their topical parallelism must be inferred. (16d) contains the same two conjoined arguments as (16c), and they are linked by TOO. Their topical parallelism is explicitly coded by the additive particle. While two co-oriented arguments do not necessarily provide greater argumentative strength for a given conclusion than any other single argument (one could perhaps imagine a single stronger argument for the purchase of the car than the two conjoined arguments in 16c), they do provide at least the same (and normally a greater) strength than any one of their two conjuncts alone. So, (16c) provides greater strength for the same conclusion than each of (16a) or (16b) alone would. Moreover, the very fact of two arguments being co-oriented implicates that the speaker requires addi-

Presupposition accommodation and language change

87

tonal argumentative strength for the conclusion than offered by one of the arguments alone. Still, the co-orientation of the two arguments in (16c) is not overtly marked. The hearer will have to infer that the two propositions share topicality, and that they are being used to make the same point. In (16d) however, shared topicality is overtly marked by TOO. The hearer/reader can easily infer that the two conjoined propositions p and q are being used to make the same point, and that they are co-oriented arguments, which together provide greater argumentative strength than one of the two arguments on its own. While argumentative co-orientation is a typical property of TOO, it is not a conventional one. In fact, additive TOO is compatible with counterargumentation conveyed by but: (17)

Our new job candidate knows a lot, but he’s pretty young, too.

The q proposition (he’s pretty young) is argumentatively countering the p proposition (our new job candidate knows a lot), with the argumentative force of q superseding that of p. The utterance (17) would be interpreted as an argument against hiring the candidate in question. It seems that even though TOO implicates ‘high argumentative strength’, it does not necessarily indicate high argumentative strength of the proposition q it is attached to. It may indeed very well be that q as the second argument in the pair of arguments is the weaker one. This in turn suggests that the use of TOO conversationally implicates “high argumentative strength”. The three languages we are considering make use of this particular structure in different ways, as we will show in the following sections. Note further that the two propositions p and q need not necessarily be adjacent. There may be material separating them, as in the following example: (18)

Admits Juliette Binoche, “I wanted the English bed badly – and I asked for it first. Ralph and I had a big chat about it, and he wanted it too.” [http://www.petey.com/kk/docs/binousa.txt]

Hence, TOO has an anaphoric structure – the particle, attached to q (he wanted it too), refers back to a previous proposition p (I wanted the English bed badly) which, in this case, does not immediately precede q. It is this

88

Scott A. Schwenter and Richard Waltereit

structure that has led scholars working on presupposition to classify TOO as one of the most typical members of the class of “anaphoric presupposition triggers” (Roberts to appear). Summarizing so far, the pragmatic condition of shared topicality of propositions p and q makes TOO particularly suitable for argumentative contexts. In fact, if TOO is used in argumentative contexts, it typically indicates argumentative co-orientation and implicates high argumentative strength. However it seems that only shared topicality, not co-orientation or strength, is part of TOO’s conventional meaning. 3.2. English: refutational too While the English additive too is very ancient and, according to the OED, attested since the 9th century, the independent use mentioned in (8) above is a much more recent use especially in American English, where too is not additive and does not imply the topical parallelism discussed earlier. Rather, its function is to deny the veracity of the propositional content of a previous speaker’s utterance: (19)

(=8) A: You didn’t do your homework! B: I did too!

B denies the propositional content of A’s foregoing utterance (that the addressee, i.e. B, did not do their homework) and asserts, by implication, that B in fact did their homework. The earliest mention of this use of too in the OED is from 1914. Just like the additive too, this use makes reference in discourse to a contextually-salient proposition, most often one derivable from a previous utterance, namely the syntactically negative proposition it denies. However it is obvious that this way of referring back to a previous utterance is completely different from the way the additive too operates. In order to distinguish the two “foregoing utterances” implied by the additive and the refutational too, respectively, we will refer to them with the letters p (for the additive interpretation) and r (for the refutational interpretation). The refutational use of too is at first sight quite puzzling, as one struggles to see the semantic/pragmatic connection with the additive too. Now, in order to find out how the independent use may have arisen from the additive one, it is useful to look for bridging contexts (Heine 2002). Bridging

Presupposition accommodation and language change

89

contexts are contexts that, from today’s perspective, are compatible with both the old and the new meaning of the item or the construction in question. In other words, there is at least some ambiguity as to whether the speaker actually wanted to convey the old or the new interpretation. Such contexts are methodologically important in at least two respects. From a genuinely semantic point of view, they may enable us to see the connection between the old and the new – i.e. what led from the old to the new meaning. Secondly, in a diachronic corpus they may serve as a guidepost directing us to when in history the new meaning arose. The first potential bridging context we found where both the additive and refutational interpretations of too appear to be accessible comes from the novel The American Baron, by James de Mille (1871): (20)

“Oh yes; I did get acquainted with some in–in Canada.” “Oh; and is this man a Canadian?” “No, Dowdy darling; only an American.” “Well, if he’s a friend of yours, I suppose you know something about him. But how singular it is that you have so completely forgotten his name. Atramonte? Why, I’m sure it’s a very singular name for an American gentleman–at least it seems so to me–but I don’t know much about them, you know. Tell me, darling, who is he?” “He–he saved my life.” “What! saved your life? Why, my precious child, what are you talking about? It was the Italian that saved your life, you know, not this one.” “Oh, but he did too,” said Minnie, despairingly. “I couldn’t help it. He would do it. Papa was washed away. I wish they all wouldn’t be so horrid.” Lady Dalrymple looked in an equally despairing manner at Mrs. Willoughby. “What is it, Kitty dear? Is the child insane, or what does she mean? How could this person have saved her life?” “That’s just what distracts me,” said Minnie. “They all do it. Every single person comes and saves my life. And now I suppose I must go down and see this person.”

It seems possible to read the utterance q Oh but he did too with either the additive or the refutational use, although most of the English speakers

90

Scott A. Schwenter and Richard Waltereit

we asked preferred the additive one. Under the additive interpretation, q is parallel to the utterance p The Italian saved your life, assuming that he in q is co-referential with the American: (21)

Additive interpretation: p: The Italiani saved your life. q: But hej [the American] did too. … Every single person comes and saves my life.

However it is also possible to read q as a denial of the foregoing It was the Italian that saved your life, not this one [the American.], more precisely of the Quantity-implicature ‘no one else other than the Italian’ arising from the it-cleft: (22)

Refutational interpretation: p: the Americanj saved my life. r: It was the Italiani that saved your life, you know, not this onej. q: But hej did too.

Note that the refutational interpretation is still additive as well, i.e. even under the refutational reading, the proposition q is topically parallel with the proposition p The Italian saved your life. This example makes it clear what the link between the additive and the refutational too may be. In particular, the particle too may be attached to an utterance q which is parallel with p (i.e. additive) but denies some third proposition r. Another quite nice example showing the same kind of ambiguity, though much later chronologically, can be found in the following extract from the novel Gone with the Wind: (23)

From Gone with the Wind, by Margaret Mitchell (1936) “Because your own dear mother is dead and Miss Pitty, not being a matron, is not qualified to–er, well, to talk to you upon such a subject, I feel that I must warn you, Scarlett, Captain Butler is not the kind of a man for any woman of good family to marry. He is a–” “He managed to save Grandpa Merriwether’s neck and your nephew’s, too.” Mrs. Merriwether swelled. Hardly an hour before she had had an irritating talk with Grandpa. The old man had remarked that she

Presupposition accommodation and language change

91

must not value his hide very much if she did not feel some gratitude to Rhett Butler, even if the man was a Scallawag and a scoundrel. “He only did that as a dirty trick on us all, Scarlett, to embarrass us in front of the Yankees,” Mrs. Merriwether continued. “You know as well as I do that the man is a rogue. He always has been and now he’s unspeakable. He is simply not the kind of man decent people receive.” “No? That’s strange, Mrs. Merriwether. He was in your parlor often enough during the war. And he gave Maybelle her white satin wedding dress, didn’t he? Or is my memory wrong?” “Things are so different during the war and nice people associated with many men who were not quite–It was all for the Cause and very proper, too. Surely you can’t be thinking of marrying a man who wasn’t in the army, who jeered at men who did enlist?” “He was, too, in the army. He was in the army eight months. He was in the last campaign and fought at Franklin and was with General Johnston when he surrendered.” The too in the last utterance He was, too, in the army is additive – it binds together a battery of arguments Scarlett O’Hara is offering for the conclusion ‘I want to marry Rhett Butler’, namely ‘He managed to save Grandpa Merriwether’s neck’, ‘He was in your parlor often enough during the war’ and finally ‘he was in the army’, each of which is countered by an argument brought forward by Mrs. Merriwether militating against that same conclusion. At the same time, that too in Scarlett’s utterance denies the presupposition of Mrs Merriwether’s immediately preceding turn ‘Rhett Butler was not in the Army’. The too in Scarlett’s last turn can be read as additive or as refutational. We have an interdigitating series of arguments for and against one conclusion, the last one being the direct denial of the second last one. Schematically, the relevant parts of the exchange can be distilled as in (24): (24)

r1: Captain Butleri is not the kind of a man for any woman of good family to marry. p1: Hei managed to save Grandpa Merriwether’s neck and your nephew’s, too. r2: Hei is simply not the kind of man decent people receive. p2: Hei was in your parlor often enough during the war.

92

Scott A. Schwenter and Richard Waltereit

r3: Surely you can’t be thinking of marrying a man who wasn’t in the army? q: Hei was, too, in the army. In this and the preceding example, the presuppositional structure of additive TOO is still satisfied, but backgrounded by a series of other phenomena: for one, the proposition q, while still topically parallel with its conjuncts p1 – p3, is acting counter-argumentatively mainly upon the immediately preceding r. Furthermore, and in fact a direct consequence of the first observation, the proposition p is not immediately preceding q in discourse; rather, there is a certain textual distance between p and TOO q. Finally, the presupposition-triggering quality of TOO q is functionally subordinated to its counter-argumentative discourse function. In other words, the presupposition, while still present, becomes unimportant in the light of the more salient counter-argumentative function in dialogue. This may lead hearers to reanalyse the item with its new meaning, along the lines of the principle “Avoid pragmatic overload”, discussed in Section 1. Generalizing from these examples, the link between additive and refutational too becomes readily apparent – and even quite natural: it is not really surprising that constructions conveying high argumentative strength, like additive too, occur in dialogual contexts where speakers are arguing against an interlocutor’s opposing viewpoint. The very use of high argumentative strength requires a context of utterance where one speaker is arguing against the other. At the same time, those contexts are not unlikely to contain denials of the type associated with refutational too. Challenging the truthfulness of the interlocutor’s utterance in such a blunt way as that particle does might be more likely to occur in contexts where interlocutors are arguing anyway than in other ones, not only because of the relative rudeness of the situation but also because contradictory viewpoints often differ not only with respect to the conclusions they favour but also with respect to how they judge the truth of a given proposition. An unambiguous example which does not support an additive reading, providing evidence that the additive particle has been reanalysed as having a new polysemy involving refutational meaning/function, is (25):

Presupposition accommodation and language change

(25)

93

From Penrod and Sam, by Booth Tarkington (1916) “My uncle don’t PLAY on it!” Roddy shrieked. “It’s an ole wore– out horn nobody wants, and it’s mine, I tell you! I can blow on it, or bust it, or kick it out in the alley and leave it there, if I want to!” “No, you can’t!” “I can, too!” “No, you can’t. You can’t PROVE you can, and unless you prove it, I got a perf—” Roddy stamped his foot. “I can, too!” he shrieked. “You ole durn jackass, I can, too! I can, can, can, can—”

The two interlocutors are arguing about whether Roddy is able to play the horn, with Roddy finally denying his interlocutor’s claim to the contrary. Schematically: (26)

r: You can’t [play on the horn]! q: I can, too!

Summarizing so far, the early examples of English refutational too in bridging contexts suggest that the argumentative strength implicated by TOO q is taken advantage of in order to deny an interlocutor’s proposition r. TOO in these cases is a means for lending one’s utterance more rhetorical strength than it would have otherwise, which is an especially important discourse function in the face of a contradictory viewpoint. Note as well that too with refutational meaning receives the primary prosodic focus of the utterance in which it appears, thereby acoustically highlighting its role as a marker of greater rhetorical strength. The change just discussed illustrates the complex dialogue-related structure of TOO. Additive TOO links two propositions p and q that share topicality. This feature makes it likely for additive particles to be assigned to the same speaker in dialogue, as in the examples discussed, though it is of course possible for one speaker to complement another one’s argument by a “q TOO” utterance: (27)

A: Emma has a lot of teaching experience. B: And she has a good publication record, too. (ĺ so we should hire her)

94

Scott A. Schwenter and Richard Waltereit

Contrariwise, the refutational too must refer dialogually to a different speaker. In fact, this constraint seems to be even stronger than the converse constraint on additive too being used by the same speaker. While it is possible to share an additive too-structure between two speakers (see 26), it would be the very idea of obvious and pragmatically unacceptable selfcontradiction to refute one’s own assertions within the same turn. Indeed, refutational too in a monologue would be tantamount to immediately denying one’s own prior assertion: (28)

A: ?Emma doesn’t have a lot of teaching experience. She has, too!

Interacting with these dialogual constraints, an additional feature is important for the rise of refutational too: there may be material intervening between the two conjuncts p and q of additive too. If an utterance r, said by the interlocutor, inserted between p and q is the negation of q, as in (20) and (23), then we have a configuration compatible both with the additive and the refutational reading – a bridging context. The argumentative properties of additive too in dialogue make such a configuration in fact quite natural. 3.3. Spanish: Independent tampoco We now turn to the Spanish additive particle tampoco. Just as with English too, the additive meaning of this particle is much older than the independent one. However it is much less clear when the independent reading arose, partly because of the lack of studies on the subject and the relative neglect of that latter variant by dictionaries and lexicographers. The additive meaning has been attested since the 14th century (Espinosa Elorza 1989), as in the following example found in the Corpus del espaĖol: (29)

Libro de las maravillas del mundo. AUTHOR: Juan de Mandevilla; Anónimo tr. DATE: [__] SOURCE: [ADMYTE] Valencia 1524– 10–13. ... mas no puede esser porque assi como nosotros no caemos agora contra el cielo. assi tampoco caen los que estan dela otra parte. ‘That is not possible because just as we do not fall against the sky those on the other side do not fall either.’

Presupposition accommodation and language change

95

In order to find out how the independent use may have arisen from the additive-presuppositional one, we proceed in the same way as with English too, namely by looking for suitable bridging contexts. Such examples seem to be offered by adversative contexts, where the additive interpretation of tampoco, often preceded by pero ‘but’, is employed in order to limit the inferential conclusions derivable from an immediately preceding proposition; similar to a concessive. One of the earliest examples we have found of this configuration is the following: (30)

From: Tablas poéticas, Francisco Cascales (Spain), 1603. También el poema brevíssimo no es agradable ni gallardo; porque en él la especulación se acaba en breve espacio. Y por ser las partes tan pequeñas, se confunde en ellas el entendimiento, y apenas puede distinguir unas de otras. Por tanto, es conveniente que el poema tenga un cuerpo grande, cuyas partes sean conocidas y distinctas; de manera que halle la vista dónde reparar y hazer su especulación. Pero tampoco a de ser sobre modo y excessivamente grande. ‘As well the very short poem is neither pleasant nor elegant; because in it the speculation ends in a short space. And because the parts are so small, understanding them becomes confusing, and one can barely distinguish the parts among themselves. Therefore, it is advisable that the poem have a large body, whose parts are known and distinct; so that [the reader’s] sight can find where to observe and make its speculation. But it should not (tampoco) be overblown or excessively long.’

In this context tampoco has a particular argumentative function. The author offers advice to poets regarding the suitable length of poems. He advises to avoid extremes in length – the poem should be neither too short nor too long. He does so in a particular way, namely not by saying “Avoid extremes” but by discussing disadvantages of the too short poem, praising the qualities of the longer poem, and then however arguing against the possible inference from that praise that the poem should be as long as possible. This is why tampoco has a three-way anaphoric structure in example (30) – it links three propositions p, r, and q: p: r:

the very short poem is neither pleasant nor elegant it is advisable that the poem have a large body, whose parts are known and distinct

96

Scott A. Schwenter and Richard Waltereit q:

But it should not (tampoco) be overblown or excessively long.

On the one hand, the link between p and q is the traditional presuppositional connection between a TOO and its first conjunct. Tampoco takes up anaphorically the negative proposition no es agradable ni gallardo ‘[the very short poem] is neither pleasant nor elegant’, adding a third negative proposition. All three propositions share parallel topicality – they count as answers to the discourse-salient question under discussion ‘what is the appropriate size of a poem?’ On the other hand, the proposition tampoco q apparently argues against the inference ‘the bigger the poem is, the better’ arising from the proposition r: es conveniente que el poema tenga un cuerpo grande, cuyas partes sean conocidas y distinctas ‘it is advisable that the poem have a large body, whose parts are known and distinct’. This proposition is not part of the traditional presuppositional structure of tampoco as a TOO element. It could be left out without making the discourse deviant or unacceptable. However canceling the inference to the effect of ‘the bigger, the better’ seems to be the main point of that part of the utterance, to the point of the anaphoric link between p and q being barely recognizable in contemporary Spanish. In fact, the combination of tampoco with the connective pero ‘but’ makes it clear that the proposition is argumentatively related (anti-oriented) to the immediately preceding proposition (cf. Anscombre and Ducrot 1977). Again, the presupposition of an additive TOO (tampoco) requiring a first conjunct is present, but this constraint is backgrounded to the benefit of the argumentative effect of the second conjunct acting upon a third proposition r intervening between p and q. This is why context (30) is compatible with both the additive and the independent readings of tampoco. And just as with the English too, the argumentative strength associated with the second conjunct seems to have transferred, as it were, to a different proposition. A context like (30) may have led hearers to reanalyse additive tampoco as having a separate, non-additive polysemy, in line with the “Avoid pragmatic overload” principle cited earlier. The additive particle tampoco, requiring a first conjunct, is reanalysed as a concessive particle. The following example is even closer to the contemporary use of independent tampoco, since it lacks a first (additive) conjunct that can be interpreted as argumentatively co-oriented with the one signalled by tampoco:

Presupposition accommodation and language change

(31)

97

From: Viaje al reino del Perú, Antonio de Ulloa (Peru) (1748) Assimismo, se hacen allí algunas minas de esta piedra pero tampoco se labra nada con ella ni se aprecia… ‘Likewise, they do some mining there of this stone, but they do not carve anything out of it or hold it in esteem.’

Here tampoco has apparently already been reanalysed, in combination with the adversative sentence conjunction pero, as an inference-denying connective. From the assertion that the people in question do some mining of the stone in question the reader might infer that they also have the habit of doing some carving out of it. That anticipated inference (on the part of the reader) is canceled in the following proposition, in a similar way to how proposition q acts upon proposition r in example (30): r: q:

they do some mining there of this stone they do not carve anything out of it or hold it in esteem

A slightly different use of independent tampoco is the following one, where it does not cancel an inference but rather denies the entailed content of a proposition directly derivable from the prior speaker’s utterance: (32)

From http://www.el-mundo.es/encuentros/invitados/2003/11/883/ (Interview with José Apezarena, expert on the Spanish Royal Family, about the royal wedding in 2003.) – No entiendo las tremendas críticas realizadas contra Letizia Ortiz por su agitada vida sentimental e íntima ¿No cree usted que está a la altura de la vida sentimental e íntima de la familia Borbón? ‘I don’t understand the strong criticism leveled against Letizia Ortiz because of her rocky private and lovelife. Don’t you think that it’s on a par with the private and lovelife of the Borbón family?’ – Tampoco he escuchado críticas tan tremendas, la verdad. ‘I haven’t heard such strong criticisms, truthfully.’

Summarizing, the pathway yielding independent tampoco appears to be as follows: the argumentative strength implicated by ~TOO q is reinterpreted so as to de-realize an inference arising from an immediately preceding proposition r. Continued evolution via understatement produces a marker of attenuated denials, as in (32).

98

Scott A. Schwenter and Richard Waltereit

3.4. German: Independent auch As our last example, we would like to sketch a discussion of the German modal particle auch, which emerged from the additive auch ‘too’. Again, we see how the rich anaphoric and argumentative structure of TOO is used to create a new kind of interpersonal use. The additive auch is attested from very early on, since the 800s. We may assume that in German as well, the anaphoric and thematic structure of TOO confers higher argumentative strength to the second conjunct: (33)

Ich mag dieses Haus, und Erna gefällt es auch. ‘I like this house, and Erna likes it as well.’ (ĺ that’s why I want to live in it)

Independent auch is attested since much later, namely since the late 1400s according to the DWN. Independent auch is a modal particle, i.e. a particle that has a particular function within a speech act. Specifically, it appears in rhetorical questions with the function of implicating a rejection of the propositional content: (34)

Warum auch hatte sie mein wichtigstes Wort nicht akzeptiert? (DWN) ‘Why only didn’t she accept my most important word?’

Another function of the modal particle auch is its independent use in exclamations since the late 1800s (still according to DWN): (35)

Was der Kerl auch für Einfälle hat! (DWN) ‘Hell, what sort of ideas does this guy have!’

Now, both independent uses in (34) and (35) convey that the state of affairs in question is an “extreme” one. In (34), the superlative ‘my most important word’ implies an extreme position on a pragmatic scale. The exclamative in (35) implies that the ideas in question are either very brilliant or very far-fetched or aberrant, at any rate uncommon. This can be linked to the higher argumentative strength associated with the second conjunct of the additive use in (33). In the appropriate context of utterance, an uncommon state of affairs may be associated with higher argumentative strength. For example, a particularly low price of a certain good provides

Presupposition accommodation and language change

99

higher argumentative strength for the conclusion of buying it than a higher price does. Ducrot and his associates have highlighted in their work (e.g. Anscombre and Ducrot 1980; Carel and Ducrot 1999) how argumentative force may be linked to the coded content of lexical expressions. One can therefore gather the way the high argumentative strength conversationally implicated by additive auch may have been reinterpreted as being conventionally associated with pragmatically extreme states of affairs. 4. Conclusion The data presented in this paper illustrate how rather similar counterargumentative uses of cross-linguistic additive particles with the meaning TOO, when used in novel discourse contexts, can override the old presuppositional constraint associated with such particles. This novel meaning/function of TOO liberates it of the requirement of a parallel first conjunct that is (typically, but not always) argumentatively co-oriented with the conjunct containing TOO. As we have noted above, TOO (and especially English too) is without a doubt the most famous example of a presupposition-triggering expression that typically does not permit accommodation of the presupposed proposition. Both speakers and hearers presumably have recognized this fact, thereby leading to innovative uses in non-presuppositional contexts—or at least, not in the same kind of presuppositional contexts as before, as we have illustrated above. Accommodating the additive presupposition associated with TOO becomes too costly and leads to reanalysing the TOO element as a new form–function pairing. This particular case of semantic change, then, provides clear evidence for hearer-based change. The initial diachronic examples in bridging contexts are plausibly interpretable as conveying the additive meaning of TOO, but the adversative properties of the dialogual discourse context— prototypically in the form of an argument between two interlocutors— appear to have led hearers to reanalyse TOO as expressing a new, rhetorically-strategic meaning with strong counter-argumentative force. The trajectory of change thereby produces a clear path from the ideational/textual meaning of additive TOO to the more clearly interactionally-bound interpersonal meanings associated with non-additive TOO.

100

Scott A. Schwenter and Richard Waltereit

References Anscombre, Jean-Claude and Oswald Ducrot 1977 Deux mais en français? Lingua 43: 23–40. Anscombre, Jean-Claude and Oswald Ducrot 1980 L’Argumentation dans la langue. Brussels: Mardaga. Atlas, Jay and Stephen C. Levinson 1981 It-clefts, informativeness, and logical form: radical pragmatics (revised standard version). In Radical Pragmatics, Peter Cole (ed.), 1– 6. New York: Academic Press. Beaver, David and Henk Zeevat 2004 Accommodation. In Oxford Handbook of Linguistic Interfaces, Gillian Ramchand and Charles Reiss (eds.), 503–539. Oxford: Oxford University Press. Carel, Marion and Oswald Ducrot 1999 Le Problème du paradoxe dans une sémantique argumentative. Langue française: 6–26. Croft, William 2000 Explaining Language Change: An Evolutionary Approch. London: Longman. Detges, Ulrich and Richard Waltereit 2002 Grammaticalization vs. Reanalysis: A semantic-pragmatic account of functional change in grammar. Zeitschrift für Sprachwissenschaft 21: 151–195. Ducrot, Oswald 1995 Les modificateurs déréalisants. Journal of Pragmatics 24: 145–165. DWN = Grimm, Jacob and Wilhelm Grimm 1983 Deutsches Wörterbuch. Neubearbeitung. Leipzig: Hirzel. Eckardt, Regine to appear Grammaticalization and semantic reanalysis. In Semantics: An International Handbook of Natural Language Meaning, Claudia Maienborn, Klaus von Heusinger, and Paul Portner (eds.). Berlin/New York: Mouton de Gruyter. Espinosa Elorza, Rosa María 1989 También y tampoco: origen y evolución. Anuario de estudios filológicos 12: 67–79. Givón, Talmy 1995 Functionalism and Grammar. Amsterdam/Philadelphia: Benjamins. Grice, Paul 1981 Presupposition and conversational implicature. In Radical Pragmatics, Peter Cole (ed.), 183–198. New York: Academic Press.

Presupposition accommodation and language change

101

Hansen, Maj-Britt Mosegaard 2002 La polysémie de l’adverbe ‘encore’. Travaux de linguistique 44: 143–166. Hansen, Maj-Britt Mosegaard 2008 Particles at the Semantics/Pragmatics Interface: Synchronic and Diachronic Issues. A Study with Special Reference to the French Phasal Adverbs. Amsterdam: Elsevier. Haspelmath, Martin 1999 Why is grammaticalization irreversible? Linguistics 37: 1043–68. Heim, Irene 1992 Presupposition projection and the semantics of attitude verbs. Journal of Semantics 9: 183–221. Heine, Bernd 2002 On the role of context in grammaticalization. In New Reflections on Grammaticalization, Ilse Wischer and Gabriele Diewald (eds.), 83– 101. Amsterdam/Philadelphia: Benjamins. Keller, Rudi 1994 On Language Change. The Invisible Hand in Language. Trans. by Brigitte Nerlich. London: Routledge. Kripke, Saul 1990 Presupposition and Anaphora: Remarks on the Formulation of the Projection Problem. Ms., Princeton University. Lambrecht, Knud 1994 Information Structure and Sentence Form: Topic, Focus, and the Mental Representation of Discourse Referents. Cambridge: Cambridge University Press. Lewis, David 1979 Scorekeeping in a language game. Journal of Philosophical Logic 8: 339–359. Morgan, Jerry L. 1978 Two types of convention in indirect speech acts. In Syntax and Semantics (vol. 9), Peter Cole (ed.), 261–280. New York: Academic Press. Roberts, Craige to appear Only, presupposition and implicature. Journal of Semantics. Schwenter, Scott A. 2003 No and tampoco: a pragmatic distinction in Spanish negation. Journal of Pragmatics 35: 999–1030. Schwenter, Scott A. 2005 The pragmatics of negation in Brazilian Portuguese. Lingua 115: 1427–1456.

102

Scott A. Schwenter and Richard Waltereit

Schwenter, Scott A. 2006 Fine-tuning Jespersen’s Cycle. In Drawing the Boundaries of Meaning: Neo-Gricean Studies in Pragmatics and Semantics in Honor of Laurence R. Horn, Betty J. Birner and Gregory Ward (eds.), 327– 344. Amsterdam/Philadelphia: Benjamins. Schwenter, Scott and Iker Zulaica-Hernández 2003 On the contextual licensing of tampoco. In Linguistic Theory and Language Development in Hispanic Languages: Papers from the 5th Hispanic Linguistics Symposium and the 4th Conference on the Acquisition of Spanish and Portuguese, Silvina Montrul and Francisco Ordóñez (eds.), 62–80. Somerville, MA: Cascadilla Press. Traugott, Elizabeth Closs and Richard B. Dasher 2002 Regularity in Semantic Change. Cambridge: Cambridge University Press. von Fintel, Kai 2000 What is presupposition accommodation? Ms. MIT. Waltereit, Richard 2001 Modal particles and their functional equivalents: A speech-act theoretic approach. Journal of Pragmatics 33: 1391–1417. Zeevat, Henk 2004 Particles: Presupposition triggers, context markers or speech act markers. In Optimality theory and pragmatics, R. Blutner and H. Zeevat (eds), 163–199. Basingstoke: Palgrave-Macmillan.

How prosody reflects semantic change: A synchronic case study of of course Anne Wichmann, Anne-Marie Simon-Vandenbergen and Karin Aijmer Abstract Although discourse markers have recently received a fair amount of linguistic attention their prosodic realization has been relatively neglected. Since discourse markers have been found to be more typical of spoken than of written language, the scarcity of prosodic studies is regrettable. This article has two aims: firstly, to provide a detailed description of the functions and prosodic realizations of of course in present-day spoken British English, on the basis of corpus data. The study seeks to answer the question to what extent of course fulfils the criteria for discourse marker status. Secondly, on a more general level, the article contributes to the discussion on the relation between grammaticalization and prosody. The findings relate to three main areas of inquiry: structure, meaning and use, and prosody. In structural terms the data showed a clear preference for of course to occur in initial position as part of the thematic material. In terms of meaning, strong evidence was found of grammaticalization, with more literal meanings occurring alongside subjective and intersubjective developments. In a number of cases, the tokens have acquired a routinized pragmatic function. From a rhetorical perspective, the subjective and intersubjective meanings of of course are exploited dialogically. Finally, on the basis of theories of intonational meaning, it was predicted that semantic change involving a loss of semantic weight in favour of pragmatic meaning will also involve a loss of prosodic prominence. This is borne out by the data, which shows a statistically significant association between prosodic prominence and meaning. This study also has broader implications for the study of grammaticalization. These implications are discussed in the article.

1. Introduction Discourse markers (DMs) have recently received a fair amount of linguistic attention, both from synchronic and diachronic perspectives, in a range of

104

Anne Wichmann, Anne-Marie Simon-Vandenbergen and Karin Aijmer

languages as well as across languages in contrastive studies (see e.g. two recently published edited volumes, Fischer 2006 and Aijmer and SimonVandenbergen 2006). However, one aspect which has received little attention in this area is the prosodic realization of discourse markers. Since they are generally accepted to be more typical of spoken than of written language (Brinton 1996), the neglect of this aspect is regrettable. In this article we have two aims. One is to provide a detailed description of the functions and prosodic realizations of of course in present-day spoken British English (PDE), on the basis of corpus data. How of course functions in spoken language is an interesting question in itself, since it appears to be an extremely frequent phrase (see Simon-Vandenbergen and Aijmer 2002/2003, Note 1). With the analysis we want to answer the question to what extent of course fulfils the criteria for discourse marker status. On a more general level, however, we aim to contribute to the discussion on the relation between grammaticalization and prosody. 2. Discourse markers 2.1. Discourse markers and grammaticalization Before we can begin to examine the question of the status of of course in PDE it is necessary to take a position with regard to the status of the historical changes which generally lead to the emergence of the class of discourse markers. Even though we do not engage in a diachronic study, we will show that the various uses of of course in PDE reflect varying degrees of closeness to and distance from its earlier meaning as a fully lexical item and that these degrees can be related to a decrease of semantic weight and a correlated increase of pragmatic functions. This means that the variation we find in PDE partly reflects past and perhaps ongoing change which must be accounted for by reference to what we know about similar changes in other adverbs. The question of the status of the development into DMs has been answered in four different ways. The four ways are extensively discussed by Traugott ([1995] 1997) and we shall not repeat the arguments for and against the positions here. Suffice it to say that the positions are lexicalization, pragmaticalization, grammaticalization and post-grammaticalization. The reason why there is disagreement over how to label these developments is that there is disagreement over the necessary conditions for talking about grammaticalization. As Traugott explains, those who want to exclude

How prosody reflects semantic change

105

DMs from grammaticalization argue that they are not grammatical items, and that they do not fulfil what are seen as certain essential criteria such as reduction in scope and fixation, criteria given by Lehmann (1995: 164). In fact, DMs show the opposite tendency, namely an increase in scope and a looser syntactic connection to the rest of the sentence. So, as Traugott points out, if we want to claim that DMs are the result of processes of grammaticalization we have to rethink the criteria of grammaticalization as well as the nature of the grammar (Traugott 1997: 5). Traugott argues that of the four solutions proposed in the literature, grammaticalization is the most appealing one. Pragmaticalization, which has been proposed by Erman and Kotsinas (1993) for cases such as you know, captures the increase in pragmatic functions which such items have gained in the course of their development but the concept does not satisfactorily account for formal and structural changes. ‘Pragmatic strengthening’ (Traugott 1989) is a concomitant feature of grammaticalization but it does not explain the whole process. We follow Traugott in opting for the term ‘grammaticalization’ in spite of the need to relax some of the criteria. The reasons are briefly summed up below (for a more extensive account, see Traugott 1997). DMs typically develop from lexical material into items which serve grammatical functions and which occur in well-defined syntactic slots. The development entails desemanticization and increasing pragmaticalization. Typically DMs move to the left periphery of the sentence and acquire new meanings, new syntactic constraints and new prosodic characteristics. The typical diachronic path is described by Traugott (1997: 13) as follows: verbal adverb > sentential adverb > discourse marker

The development into a DM is, as pointed out by Traugott, accompanied by a number of shifts “normally associated with grammaticalisation” (1997: 13–14). Traugott mentions six shifts and shows how these have taken place in the instances under consideration, indeed, in fact and besides. These shifts are (i) decategorialization (the lexical nouns deed, fact, side become fixed in prepositional phrases), (ii) bonding within the phrase (bonding with the prepositions in and by), (iii) phonological reduction (which is possible though not always realized), (iv) generalization of meaning (increase of polysemies), (v) increase in pragmatic function (concrete > epistemic > metatextual elaborator), (vi) subjectification (increasing association with speaker attitude). The problem areas for the grammaticalization stance are scope and disjunction. In contrast with ‘typical’ cases of grammaticalization (e.g. Lehmann 1995: 143), DMs do not have a reduced but an in-

106

Anne Wichmann, Anne-Marie Simon-Vandenbergen and Karin Aijmer

creased scope, and they do not come to occupy fixed syntactic slots in the sense that, say, clitics do. Traugott’s arguments are that “syntactic scope increases must be allowed for in a theory of grammaticalisation” and that the disjunct slot occupied by DMs can be considered as “the typical syntactic site” to which items move in the process in languages such as English. Other languages such as German reserve the middle field for discourse particles with similar functions to DMs (Traugott 1997: 14; see also Diewald 2006, discussed below). If we accept – as we do – that from a formal and structural point of view DMs are instances of grammaticalization, this still leaves us with the question of whether they indeed fulfil ‘grammatical’ functions. Traugott’s position is the following: The view of grammar adopted here is that it structures cognitive and communicative aspects of language. It encompasses not only phonology, morphosyntax and semantics but also inferences that arise out of linguistic form, in other words, linguistic pragmatics such as topicalisation, deixis. (Traugott 1997: 5)

In the same vein, Diewald (2006: 405) writes that the pragmatic functions that DMs1 have are “genuine grammatical functions which are indispensable for the organisation and structuring of spoken dialogic discourse.” She, too, rejects the distinction ‘pragmaticalization vs. grammaticalization’ in favour of a view of grammar which includes discourse structuring elements. Diewald refers to various arguments given in the literature to support the ‘DMs as grammaticalization view’, including Günthner (1999), Barth and Couper-Kuhlen (2002), Lima (2002). Diewald (2006: 408) also sees DMs as syntactically non-integrated. This syntactic non-integration, in the sense that DMs have no syntactically fixed position, distinguishes them from modal particles, which appear in the middle field. In Diewald’s view, the functional criteria are primary and are valid cross-linguistically, while the formal and structural ones may well be language specific (2006: 408). With regard to the question whether DMs have an inherent semantic meaning which plays a role in the pragmatics, Diewald’s answer is ‘yes’. She postulates a ‘core meaning’ which is present in all the uses of the DM and the “synchronic polyfunctionality of the particle [including DM] lexemes is 1. Diewald (2006) actually uses the term discourse particles instead of discourse markers and distinguishes them from modal particles. In the discussion of Diewald’s position we shall, however, keep using the term discourse marker for the sake of consistency.

How prosody reflects semantic change

107

due to the reinterpretation of the basic semantic template” (Diewald 2006: 405). We share this view, as shown in Section 6 (see 6.1 and 6.2). The semantic development traceable for DMs follows the tendencies generally attested as typical of grammaticalization. These tendencies have been described as metaphorization from concrete to more abstract (Heine, Claudi and Hünnemeyer 1991), for example typically along the following path: local > temporal > abstract (e.g. causal; adversative; copulative) (Diewald 2006: 410)

In addition, Traugott distinguishes three recurrent diachronic stages (1989: 34–35) typical of the development of DMs, which are well-known and need not be further discussed: the functional development from meanings based in the external situation to meanings based in the textual situation to meanings increasingly based in the speaker’s subjective belief. This scheme was later refined (e.g. Traugott and Dasher 2002) to include the further step from subjectification to intersubjectification, to account for social deixis (e.g. for such functions as the expression of politeness or solidarity). DMs have by definition a discourse function, which entails indexing the utterances to the surrounding discourse, both in terms of structuring the ongoing discourse and in terms of signalling to the addressee how he/she should interpret the speaker’s stance. Aijmer, Foolen and SimonVandenbergen (2006) mention three functions as important in this respect: reflexivity (metacommunicative function), indexicality (pointing to the speaker’s position with regard to persons and situations) and heteroglossia (positioning the speaker’s voice within a context of other voices, see e.g. White 2003 for an account). These functions are all non-propositional and abstract. Summing up, there are cogent reasons for considering the development of DMs as one of grammaticalization, and it is the purpose of this article to examine the extent to which of course in PDE is a DM according to the generally accepted criteria prosody, discourse function and semantic change.

108

Anne Wichmann, Anne-Marie Simon-Vandenbergen and Karin Aijmer

2.2. The prosody of discourse markers 2.2.1. Previous studies: inconclusive results Despite numerous claims in the literature that prosodic features contribute to the identity of discourse markers, there is to date no comprehensive overview of what these features are and how they operate. Early quantitative studies based on corpus data (Meyer 1986; Altenberg, 1987, 1990; Stenström 1990) are inconclusive. They are in fact studies of the prosodic behaviour of the broad class of ‘adverbials’, but the definition of adverbials follows the categories of Quirk et al. (1985), which include ‘disjuncts’, and ‘conjuncts’, both of which categories contain items that we would now refer to as discourse markers. Stenström points out, rightly, that adverbials are distinguished prosodically not only by segmentation (presence or absence of a boundary) but also by tonicity (presence or absence of nuclear tone) and by tonality (choice of tone). Nonetheless, the generalizations offered are more a matter of expedience for text-to-speech systems than linguistically revealing. Altenberg concludes that prediction rules (i.e. for text-to-speech synthesis) for adverbials are complicated by “their formal and functional diversity” (1990: 283). The reason for variable prosodic patterning may in part be morphological. There seems to be some agreement, for example, that the longer and more complex the adverbial, the more likely it is to be punctuated (and prosodically separate). According to Altenberg, “initial adverbials, which generally have a grounding, connective or attitudinal function (as adjuncts, conjuncts or disjuncts), are normally set off in a separate tone unit if they are polysyllabic” (1990: 283–284). Speed of delivery also plays a part – the slower the delivery, the more likely a tone unit boundary. This of course is in part determined by the context of situation, which constrains speech style (e.g. formal–informal, dialogue–monologue, prepared–spontaneous), and in turn constrains aspects of articulation, e.g. tempo, degrees of reduction and other connected speech processes. More recently there have been studies of the prosody of individual discourse markers, e.g. in English you know (Holmes 1986), well (Bolinger 1989), now (Hirschberg and Litman 1993), anyway (Ferrara 1997, see also Wennerstrom 2001); in Swedish men (Horne et al. 2001). The results have identified more nuances than in earlier studies, but there is still no clearer view of the extent to which discourse markers exhibit unique sets of prosodic patterns. We are left with an incomplete picture of what to expect

How prosody reflects semantic change

109

prosodically of discourse markers, making it difficult in turn to use prosodic features to identify function in cases of ambiguity. We believe nonetheless that it is possible to identify a number of patterns that systematically co-occur with discourse markers, and we will argue that these are underpinned by the processes of lexical change. This argument is based firstly on the fact that discourse markers share functions with other grammatical classes, and secondly on a theory of intonational meaning that relates propositional content to prosodic prominence. 2.2.2. Prosody and discourse function The extreme variability reported by e.g. Stenström is the result of generalization on the basis of lexical items (which a speech synthesizer can identify) rather than on the basis of discourse functions (which it cannot). If we re-analyse Stenström’s findings according to individual subcategories of adverb in the London–Lund Corpus (1990: 261), we find greater systematicity. All ‘disjuncts’ (they include epistemic stance adverbs, e.g. apparently, of course, clearly, fortunately) are prosodically separate in her data, with either a falling tone or a fall–rise. The distribution of these tones is not given, but may be consistent with Cruttenden’s observation that “adverbials which limit the main clause take a rise, while those that reinforce take a fall” (1997: 95). ‘Conjuncts’ (adverbials with a connective function2), on the other hand, are either separate with a falling tone, or unstressed and integrated. The distribution is not given. Thus we would predict (invented examples): Disjuncts / stance adverbials (1) A\/pparently | he’s not coming after \all ||3

2. It is therefore not surprising that the same prosodic patterns can be found on ‘normal’ conjunctions (co-ordinating, subordinating). The idea that discourse particles can have uses in common with other grammatical categories has been noted by Aijmer (2002: 27). 3. The prosodic symbols throughout are based on those used in the British system of intonation. The \ indicates a falling pitch contour, associated with an accented syllable and any unstressed material up to the next accent or a boundary. The / indicates a rising contour, and the \/ indicates a falling–rising contour. This is marked before the accented syllable on which the contour begins, but the rising part of the contour may occur on subsequent unstressed ma-

110

Anne Wichmann, Anne-Marie Simon-Vandenbergen and Karin Aijmer

Or \Certainly| we’re going to have to \deal with this || Conjuncts/ discourse markers (2) Now `let’s give a round of a/pplause | to …. Or \Now | `let’s give a round of a/pplause | to … ‘fact is he’s not \coming The fact \is | I’ve been having some \trouble recently | and .. Well we \/might go | but we’re `not \sure || Or \Well || it’s a `long \story || In their unstressed form, these conjuncts behave just like other grammatical conjunctions (because, and, etc.). In their stressed form, however, they behave as other grammatical elements do, whether NP, adverbial or main clause, when a new topic or title is being announced – namely by treating the item as a complete utterance (cf. Wichmann 1998). The effect is to widen the scope of the conjunct to relate to larger domains. This increased scope of conjuncts is a possibility that has already been noted by Quirk et al.: …conjuncts can relate units much larger than sentences: nonetheless at the beginning of a paragraph or section of a text will indicate a conjoining contrast with the whole preceding paragraph or section. (Quirk et al. 1985: 632n)

In speech, this phenomenon of linking across larger or smaller domains is signalled by prosody. By assigning stress to a conjunctive DM, the size of the shift is exaggerated. The same effect is created by stressing a conjunction.

terial. The symbols | and || represent minor and major tone group boundaries. The symbol ` indicates the first accent in a tone group (normally the ‘onset’).

How prosody reflects semantic change

111

2.2.3. How does this relate to semantic change and intonation theory? Because it is non-propositional, intonational meaning is elusive. Most accounts suggest very broad abstract meanings associated with pitch movements, such as ‘finality’ or ‘non-finality’, ‘closed’ or ‘open’ (Cruttenden 1997; Wichmann 2000). Because the meanings are so abstract, there is no one-to-one relationship between prosody and lexical item, or between prosody and word class. If a discourse marker is realized in a way typical of major shifts in the discourse, this is not the prosody of the discourse marker, it is the prosody of a major structural shift. If there is a tendency for certain co-occurrences, this is because there is a tendency for certain words to be used for a particular discourse function, not because the prosody is the property of the word itself. The theory of intonational meaning in English proposed by Pierrehumbert and Hirschberg (1990) tries to account not only for pitch movement (tone choice) but also for the effect of intonational prominence (presence or absence of an accented syllable). They claim that the degree of prosodic prominence is directly related to the informativeness or propositionality of the associated word or phrase. This accounts for the fact that in speech, lexical words have much greater potential for prominence (stress) than grammatical or function words. The theory of semantic change (see Section 2.1 above) proposes that, over time, certain high frequency items are subject to a gradual loss of semantic weight, a process of ‘bleaching’ through habituation, and acquire a grammatical or pragmatic role. Discourse markers often derive, via this process, from verbal adverbs (see Section 3.2). In line with the theory of intonational meaning, we can predict that this process of semantic change will go hand in hand with a loss of potential for prosodic prominence. Those items that have a purely interpersonal function, highly routinized and with very little residual core meaning, we would expect to be least prominent, while those that still express some propositional meaning, e.g. epistemic or evidential stance towards a proposition, might have greater prominence. If it is possible to consider the distinction made by some grammarians between disjunct and conjunct (see also Section 3.1.1 below) to be a distinction between (subjectified) propositional meaning and discourse meaning, then we can predict that elements used as expressions of stance towards the proposition are more likely to be stressed than those used in a routinized way for cohesive or interpersonal purposes. This would account for the variability reported in earlier studies, given that they tried to gener-

112

Anne Wichmann, Anne-Marie Simon-Vandenbergen and Karin Aijmer

alize over a much broader category of adverbials. It would also account for the systematic realization of subcategories of adverbials: Stenstrom’s disjuncts are typical examples of elements which express stance towards the proposition and have obvious residual propositional meaning themselves. The conjuncts, however, behave frequently like other grammatical conjunctions – unstressed and integrated into a larger tone group. Confusion arises only when one of these conjuncts is stressed. While we do not confuse a stressed conjunction (\And |, \But |) with any other word class, but simply understand it as a ‘bigger’ and or but, a stressed conjunct (actually just a ‘bigger’ well, now or anyway, indicating a greater scope) is prosodically identical to a stressed stance adverbial, and if the same word functions as both, the prosody may introduce ambiguity. For this reason, it is not possible to use prominence alone to disambiguate the time adverb now from the conjunct now. The unstressed version – behaving normally as a grammatical word – is indeed more common, but an emphatic \Now || is not a time adverb but a conjunct with wider scope. 3. Of course: previous work 3.1. Of course in present-day English 3.1.1. The grammatical status of of course Of course is classified as an adverb in Quirk et al. (1985: § 9.5), where it is to be found in the sections on subjuncts, disjuncts and conjuncts. As a subjunct, it is relatively integrated into the clause, occurring next to the clausal element it emphasizes. Quirk et al. also make a comment on the prosodic features of subjuncts, saying that they are not separated intonationally (1985: 584). However, the group of so-called ‘emphasizers’ to which of course as a subjunct belongs, “mainly consists of items that can also function as disjuncts expressing the comment that what is being said is true” (1985: § 8.100). As a disjunct, of course occurs in the periphery of the clause, usually initially. Thirdly, Quirk et al. recognize a conjunct use of of course. Conjuncts, like disjuncts, have a “superordinate” role in the clause (i.e. they have wider scope), and are syntactically peripheral. In contrast with disjuncts, they have a linking function. Quirk et al. distinguish two conjunctive uses, or semantic subtypes, viz. as a ‘resultative’ and ‘concessive’ conjunct. These are illustrated by examples (3) and (4) respectively:

How prosody reflects semantic change

(3) (4)

113

She arrived late, gave answers in an offhand manner and of course displeased the interviewing panel. (Quirk et al. 1985: § 8.140) Of course he’s a bit lazy; all the same I’d like to give him the job. (Quirk et al. 1985: § 8.141)

Summing up, Quirk et al. correlate different functions of of course with the different syntactic options (degrees of clausal integration), as well as with contextual factors which distinguish between disjunctive and conjunctive uses. Prosody is mentioned in passing, where it is claimed that subjuncts are not separated by intonation or punctuation. The fact that disjuncts are often followed by commas would seem to suggest that they are separated intonationally as well, though in the conjunctive uses illustrated by examples (3) and (4) above there is no comma and no comment on the prosody is made. In other words, prosodic realization is not linked explicitly to syntactic status or semantic type. Quirk et al.’s classification is primarily based on position in the clause, disjuncts and conjuncts being mostly initial. Holmes (1988) uses the term pragmatic particle to characterize all uses of of course, regardless of the syntactic slot it occupies and in all contextual functions. Simon-Vandenbergen and Aijmer (2002/2003) take the same position and refer to of course as a pragmatic marker “in all its occurrences, regardless of its position, syntactic integration, prosody and realisation as a full or reduced form” (2002/2003: 20). Holmes does not mention the concept of grammaticalization but states that “of course is a discourse particle or verbal filler like you know and I think” (1988: 49). She also mentions the phonologically reduced form of of course “in rapid or casual speech”. Phonological reduction is, as pointed out (Section 3 above), a feature associated with grammaticalization. In sum, the grammatical status of of course has been given little attention in the literature. While Quirk et al. (1985) distinguish three grammatical types, the assignment of of course to one of these types is based on a mixture of positional and contextual criteria. Holmes (1988) and SimonVandenbergen and Aijmer (2002/2003) classify of course as a pragmatic particle/marker but base this classification purely on functional criteria. What is lacking at the moment is an account which brings together formal, structural and functional features. In Section 3.1.2 we give a brief survey of the functions as discussed in the literature, in Section 3.1.3 the scanty information on prosodic forms of of course is summarized.

114

Anne Wichmann, Anne-Marie Simon-Vandenbergen and Karin Aijmer

3.1.2. The meaning, functions and prosody of of course Both Holmes (1988) and Simon-Vandenbergen (1992) recognize a core meaning of of course which remains invariant in the different contextual functions. In her assessment of the core meaning of of course, Holmes proposes that it “acts as an overt signal that the speaker is assuming that the hearer accepts or is already familiar with the propositional content of her or his utterance, and functions to emphasise the validity of that content” (1988: 53). The definition in Simon-Vandenbergen (1992: 215) is very similar: “…of course combines the meanings of certainly (‘there is no doubt that…’), which expresses a probability judgement, and naturally (‘it was to be expected that’), which conveys a judgement on the extent to which something was expected”. In other words, both studies see the meanings of certainty and expectation as part of the core. In addition to that core meaning, of course has various contextually determined pragmatic functions. Holmes (1988) distinguishes two main functions which she labels “impersonal” and “confidential”. These tend to occur in different registers, the former in formal and public discourse, the latter in casual and intimate discourse. The difference has to do with the nature of the content which is presented as shared and the reasons for presenting it as such. In the impersonal use, of course marks generally shared knowledge, while in the confidential use it marks knowledge shared by intimates. Both types of of course can also have politeness functions and can be used manipulatively. Basically, the impersonal type is authoritative and potentially patronising, while the confidential type signals solidarity and can act as a positive politeness device. Simon-Vandenbergen (1992) also links types of functions with types of contexts. The study compares uses of of course in casual conversation and in political interviews, and the conclusions are comparable to Holmes’s. While in both registers speakers use of course to mark shared knowledge, the interactional aims differ. The meanings may be ‘everybody knows that…’, ‘we both know that…’, ‘I know what you are telling me’, ‘you know what I’m telling you’, ‘you should know what I’m telling you’ (Simon-Vandenbergen 1992: 215). Politeness is also said to operate differently in different genres. The manipulative uses of of course in political colloquy are the focus of attention in Simon-Vandenbergen, White and Aijmer (2007). From a heteroglossic perspective, of course is seen as a strategy which is dialogic but

How prosody reflects semantic change

115

at the same time restricts the possibility for disagreement by expressing consensus or “concurrence” (White 2003: 269). In political discourse it is further frequently used as a rhetorical ‘put-down’, signalling that the opponents’ arguments are not impressive. The distinctive pragmatic functions as discussed in the literature provide a basis for the functional classification of the data in this study. There is little specific reference in the literature to the prosody of of course. We are aware of only two main sources of comments, the first a passing mention by Halliday and Hasan (1976) in relation to cohesion, and the second in a detailed analysis of the meanings and uses of of course in New Zealand speech (Holmes 1988). In their very brief account of of course, Halliday and Hasan identify two prosodic realizations, stressed (in their terms ‘tonic’) or unstressed (in their terms ‘reduced’). In each case, the meaning ascribed is related to shared knowledge (‘you should have known that already’), rather than any epistemic meaning such as ‘naturally’, ‘predictably’. We may therefore infer that Halliday and Hasan classify of course as a grammatical item (discourse marker), and not as a stance adverbial. In her detailed corpus-based analysis, Holmes takes more parameters into account than stress alone, describing the prosody of of course in terms of position in tone group, stress, and choice of pitch contours. She envisages both integrated and prosodically separate versions – a separate tone-unit containing a stressed syllable (rising or falling), or, in initial position, as an unstressed pre-head. Despite the claims made above, our own data suggests that, just as the degree of syntactic integration is probably a matter of degree, there are also degrees of prosodic integration that make such categorical observations problematic: It is uncertain whether we should regard discourse markers as part of the clause or as extra-clausal units (as applies also to parentheticals in writing). It is probably correct to say that there are degrees of integration, as expressed by prosody and the type of orthographic marker. When there is clear prosodic or orthographic separation, they are best treated as independent nonclausal units. (Biber et al. 1999: 140)

3.2. The history of of course An account by Lewis (2003) of the historical development of of course shows that its meaning derives from a structure containing the noun

116

Anne Wichmann, Anne-Marie Simon-Vandenbergen and Karin Aijmer

‘course’ (French/Middle English cours) meaning ‘the path taken e.g. by a river’, which coalesced with of and came to mean ‘in the natural order of things’, ‘predictable’ or ‘to be expected’ (OED). Early use shows that it functioned both as an adjective and as an adverb. Adjectival use was predicative and postmodifying, the latter persisting into the 19th century, e.g.: (5)

1580 1813 1862

The friendship between man and man as it is common so it is of course (OED) You must give me leave to flatter myself, my dear cousin, that your refusal of my addresses is merely words of course. (Jane Austen: Pride and Prejudice) … a matter of course (OED)

In PDE this usage remains only in the fixed expression a matter of course. Until the 19th century, of course was also used as an adverb, e.g.: (6)

1548 1813

A thing which is graunted [=granted] of course. (OED) Mr Collins had only to change from Jane to Elizabeth – and it was soon done – done while Mrs Bennett was stirring the fire. Elizabeth, equally next to Jane in birth and beauty, succeeded her of course. (Jane Austen: Pride and Prejudice)

As a sentence-internal adverbial, of course appeared in the 17th century meaning ‘as usual’ and underwent a further development in the 18th century to a sentence adverbial (‘in accordance with the natural ways of the world’): (7)

1752

…if a poor child is to be whipped equally for telling a lie, or for a snotty nose, he must of course think them equally criminal. (cited in Lewis 2003)

By the end of the 19th century of course had acquired a meta-textual, discourse-oriented function, pointing to a ‘dispreferred argument, contrasting it with a preferred argument’ (concession): (8)

1885

Of course, I am naturally a partial judge of my father’s character; but this I may say, that during my experience of over seventy years I have never known a more incessantly

How prosody reflects semantic change

117

industrious man. (Nasymygh, Autobiography, quoted from Lewis 2003) To sum up, the historical data show that the function of of course has changed over time, some usage now being obsolete, except in fixed expressions. The development is one from more lexical to more grammatical meaning, from more objective to more subjective, which are the characteristics associated with grammaticalization. The parallel with words such as in fact (Traugott 1997, discussed in Section 2.1 above) is obvious: it, too, started as a noun, became a fixed collocation with a preposition (coalescence), developed into a verbal adverb, and then acquired more subjective and discursive meanings. 4. Data and methodological considerations 4.1. The data This study of of course is based on naturally-occurring data taken from the International Corpus of English, British English (ICE-GB) compiled at the Survey of English Usage at University College London (Nelson, Wallis and Aarts 2002). The value of this particular corpus is that the sound files are linked to the text, and can be accessed directly, meaning that sections of text, resulting from an automatic search and displayed on screen, can be listened to individually. The corpus contains 1 million words in all, of which 600,000 words are (transcribed) speech. This study is based on 200 examples of of course taken from a total of 552 occurrences in the spoken section of the corpus. As the corpus contains a variety of speech genres, the examples are taken to represent both dialogue and monologue, and different degrees of formality across those two broad dimensions. The dialogue examples (102) thus include both informal conversation (52) and dialogue in a public setting (50) (e.g. broadcast interviews), while the monologue section includes highly prepared or scripted speech (41) (e.g. broadcast talks) and more informal spontaneous speech (57) (e.g. spontaneous commentary). The tokens in dialogue were selected to represent as many text types as possible, and in cases where there were more tokens than needed, starting at the beginning of the group of texts, excluding inaudible examples, until an appropriate number had been collected. There was no selection other than for quality.

118

Anne Wichmann, Anne-Marie Simon-Vandenbergen and Karin Aijmer

4.2. Analysis The sound files were analysed auditorily by the first author4 and the text samples were prosodically annotated. For each token the accentual status (stressed, unstressed or indeterminate) was noted. The tokens were also categorized according to their position in a tone group, and then according to the pitch contours associated with stressed tokens (fall, fall–rise, rise, etc.). Each token was then categorized as being prosodically integrated or separate, with indeterminate cases marked as such in a third category. All examples were examined in context for their semantic content (see Section 6.2). We assume that of course retains in all cases some core meaning, but that each example will be at a different point on a transparency scale. We found examples with a clear meaning of ‘naturally’ or ‘the natural course of things’ at one end of the scale, and at the other end we found examples signalling shared knowledge (‘you/we know this’) where little of the core meaning remained. Inevitably there were many where both meanings co-existed or that were difficult to determine, and these were categorized separately. We thus had three broad levels of meaning: (1) epistemic/ evidential – glossed as ‘naturally’, (2) interpersonal – glossed as ‘shared knowledge’, and (3) indeterminate. In addition to the semantic and prosodic analysis, the tokens were annotated for their syntactic position, initial, medial or final, and for common collocates (e.g. but + of course, and + of course; see Section 5). 5. Structural patterning of of course 5.1. Utterance positions Of course occurs in various positions in the utterance, which can be roughly described as initial, medial and final. Table 1 gives an overview of the frequency of each of these positions in the data.

4. Instrumental analysis is not possible for all of the tokens, given the noisiness of much speech collected in natural surroundings. In any case, extracted pitch contours (more properly F0 contours) cannot replace an auditory analysis of this kind, since speech software is unable to identify phonological categories such as pitch accents (AM system) or falls, rises, etc. (British system).

How prosody reflects semantic change

119

Table 1. Utterance positions of of course initial 101 (50.5%)

medial 72 (36%)

final 27 (13.5%)

total 200

Table 1 shows that initial position occurs in just over half the cases. It should be noted that cases where of course follows a conjunction (such as but of course, and of course) or a discourse marker (such as well of course, further of course) are also counted as initial because such conjunctions and discourse markers take obligatory initial position and thus force of course into second position in the utterance. Both elements must be considered as part of the thematic material (see Section 5.4 for a definition of Theme) in such cases (Thompson [1996] 2004: 157). A comparison with the results reported by Holmes (1988: 51) shows that the general trends are the same: initial position is by far the most frequent one, followed by medial position, and then final position. The difference is that Holmes found an even higher percentage of initial instances (60% in her New Zealand speech data, 86% in the British data). This difference may be due to the composition of the corpora. For instance, Holmes did not have any monologic data and the ICE-GB covers a wider range of genres. Holmes found a very different trend in written data, where medial position is most frequent. The sample examined in this study further shows that in initial as well as in medial positions, of course recurs in a number of identifiable structural and pragmatic environments. In the following sections we take a closer look at the recurrent structural environments. 5.2. Conjunction followed by of course Of course is frequently preceded by a conjunction, as Table 2 shows. The percentages are calculated on the total number of instances of initial of course (see Table 1): Table 2. Frequency of of course in initial position preceded by and, but or a subordinating conjunction and of course

but of course

35 (34.5%)

15 (15%)

subordinating conjunction + of course 13 (13%)

total 63 (62.5%)

The figures show that in initial position of course occurs in combination with a conjunction in over 60% of the cases. In the total of 200 instances

120

Anne Wichmann, Anne-Marie Simon-Vandenbergen and Karin Aijmer

the combination accounts for nearly one in three cases. If such combinations are frequent they tell us something about the profile of of course. The combination but of course and the subordinating conjunctions (which are mostly contrastive/concessive) show the adversative functioning of of course (see Section 6.4 below). The combination with and, which is even more frequent, can have different functions. One is to flag that a piece of information which is self-evident is to follow, after the production of some not so evident information. Example (9) illustrates this. (9)

F: Well I have decided for a long time uh and the answer is I am very keen indeed to see Douglas Hurd as the next leader of the Conservative Party and the next Prime Minister E: Why F: Uh because he has had a considerable experience the most experience of any of the candidates in very high office I saw him at work when I was in the Cabinet both in Northern Ireland where I had been before and of course at the Home Office where I had also been [broadcast news]

5.3. Relative pronoun followed by of course Like conjunctions, relative pronouns take up obligatory first position in the clause, so that of course following a relative pronoun can also be classified as thematic. There are seven instances. Example (10) illustrates this position. (10)

They suddenly see it as a run up to another nineteen sixty-seven which of course was the great big set-back in modern Arab history when Israel emerged as a sort of imperial power and occupied large tracts of Arab territory which it s of course it still is it still holds [broadcast news]

The occurrence of of course in a relative clause enhances the backgrounding of the information: not only is it put in a syntactically subordinate clause but it is also presented as self-evident.

How prosody reflects semantic change

121

5.4. Of course in post-thematic position One important medial position of of course is post-thematic. We are using the term Theme in the Hallidayan sense (Halliday and Matthiessen 2004: 64ff.) to refer to the first ideational element in the clause, possibly preceded by textual and interpersonal elements. It appears that of course occurs in post-thematic position in 29 instances, which accounts for 34.5% of all medial cases (14.5% of all 200 instances in the database). Most frequent as themes are subjects and adverbials, which are unmarked and fairly unmarked themes respectively (Thompson 2004: 144–145). A special type of subject preceding of course is the demonstrative pronoun, with an anaphoric function. The effect of of course in this position is that it highlights the theme, thus making it more prominent. Marking the thematic structure is one of the discourse functions that discourse markers typically fulfil (Diewald 2006: 406). Examples (11) and (12) illustrate of course following the thematic subject:5 (11)

(12)

Sadly Old Hushwing is no longer part of the great meadows of our countryside One of the problems of course is that the new agricultural methods have decreased grassland areas which were the birds’ feeding grounds and removed the old barns and hollow trees in which they could roost and nest [broadcast talks]

Now the law Members of jury the first point is this This of course as you know is a civil case it’s not a criminal case [legal presentations]

Of course also follows special thematic subjects which are attitudinal or metadiscursive comments such as it’s very important of course…, the question of course….Thompson (2004: 152) refers to them as ‘thematized comments’. They, too, receive extra focus in this way. Another favoured position of of course (11 instances) is after the subject and the finite verb, or, in Hallidayan terms, after the Mood element 5. In the transcription of the ICE-GB the symbols < , > and stand for a short and a long pause respectively. A long pause is a substantive break rather than a gap equivalent to a few syllables.

122

Anne Wichmann, Anne-Marie Simon-Vandenbergen and Karin Aijmer

(Halliday and Matthiessen 2004: 111 ff.). This is also a typical position of sentence adverbs and it has the same effect of highlighting as the postthematic one. Example (13) illustrates this. (13)

And Dick basically said he didn’t like it because it was there were questions like were you given enough guidance and were your personal tutors good enough And these are of course the the sort of questions that Dick doesn’t really want answered because if someone says no the guidance wasn’t good enough and our personal tutors weren’t good enough then Neil’s going to say to Dick [direct conversations]

5.5. Of course in parenthetical remarks In parenthetical remarks of course enhances the backgrounding of the information because it presents it as ‘known’. Examples are (14) and (15). (14) (15)

And the government thanks to the tax payers it represents of course coughs up a lot for various imperial ceremonies and functions [broadcast talks] Her father is a man of incredible moral principle who will never do any work on a Sunday because it’s forbidden as he sees it in the Bible etc uhm Exactly the sort of person that Rockefeller who also by the way of course is very religious but who he is going to trample underfoot and push out of the way [classroom lessons]

In (14) the parenthetical remark is thanks to the tax payers it represents of course, in which of course has final position. In (15), the parenthetical comment is who also by the way of course is very religious, in which of course occurs after the obligatorily initial relative pronoun and after the adverbial also and the discourse marker by the way. These examples show that of course can occupy different positions in parenthetical remarks.

How prosody reflects semantic change

123

5.6. Responses In responses of course is the new element which is given as a confirmatory answer to a polar question. In such contexts it is emphatic in comparison with a simple affirmative yes. Example (16) illustrates this use.

(16)

R: Do you want him to become Prime Minister S: Of course I do I think it would be wonderful Yes I’d be terribly proud of him Wouldn’t anybody [broadcast news]

The reply Of course I do has only one element which is ‘new’ information, viz. the positive response to the polar question, signalled by of course. 6. The semantics and pragmatics of of course 6.1. Introduction The multifunctionality of of course in PDE is the result of its development from a full lexical item (prepositional phrase) into a sentence internal adverb, into a sentence adverb and finally into a discourse marker. Not only is it a characteristic of discourse markers that they are polysemous and multifunctional (see e.g. Aijmer 2002: 19 ff.), but it is also a characteristic of grammaticalized items that they exhibit what Hopper and Traugott ([1993] 2003: 49) refer to as ‘layering’ – the presence of different layers of meaning at the same time. This means that some of the former propositional meaning is still retained as a relic, while newly acquired meanings, some conventionalized, some as pragmatic inferences, are overlaid on the underlying meaning. The results are that the item occurs in different functions in different context, and that the different meanings of the polysemous item cannot always be sharply delineated. In some of its uses the item in question will have a meaning which is close to its propositional one, in other uses its meaning will be far removed from it, with a grey area in the middle, giving a cline of more lexical to more grammatical meanings, or more propositional to more textual or interpersonal ones.

124

Anne Wichmann, Anne-Marie Simon-Vandenbergen and Karin Aijmer

In trying to grasp the multifunctionality of of course it has become clear to us that several factors play a role and that it is the interplay of all these factors that accounts for the complexity of its workings. The reason why of course is apparently, judging by its frequency, so useful, especially in speech, is precisely that it does very different things in different contexts, and its description must therefore be based on contextualized data. In the following sections we discuss the parameters which play a role in the semantic-pragmatic profile of of course. Four parameters are crucial for its description: its semantic meaning, its discourse status, its heteroglossic functioning, and the pragmatic stance it expresses. 6.2. Semantic meaning The lexical meaning of of course ‘as a natural consequence’ (as in as a matter of course) is present in some of its occurrences. This sense is the epistemic-evidential sense, in which the speaker’s assessment of the truth of the proposition is based on the fact that it follows as a natural result, and hence becomes inevitable. In this sense then of course has still a clear residue of lexical meaning. Here are some examples which illustrate this use. (17)

D: The effects of nerve gas would be even worse E: Well the nerve agents act basically by blocking the message from the nerves to the muscles so the muscles go into spasm and of course the muscles that control your breathing are in spasm and people asphyxiate and that can happen very quickly [broadcast news]

In example (17) the propositions in the scope of of course are presented as the natural result of the preceding proposition that ‘the muscles go into spasm’. A paraphrase ‘as a result’ would be possible in this case. (18)

A: You always read poetry presumably B: Yes always I lectured in English at Durham University and uh Well I studied English all my life so of course I love poetry Poetry’s lovely to read in bed at night I think it uhm it quietens your mind and flattens your spirit out [broadcast interviews]

How prosody reflects semantic change

125

In (18) also the proposition ‘I love poetry’ is presented as following naturally from the proposition ‘I studied English all my life’. If an event is presented as following naturally from some other event or state-of-affairs, the mental step to the meaning that the event was or is predictable is a small one. In many contexts the paraphrase for of course is ‘predictably’. The development is one from more objective to more subjective meaning, since it is people who make predictions and voice these. Example (19) illustrates this sense: (19)

Inside the medieval church is a sacred statue of the Marys sailing in a little wooden boat how they came to be here The legend tells us that after the crucifixion the two saints and their Egyptian servant Sarah were set adrift by the Jews of Jerusalem in a boat with no sails no oars and no food But with the divine protection of the Lord the boat was guided ashore to the foot of the church walls Of course over the years the pilgrims have improved this story so that the boat contained Mary Magdalene Lazarus and his sister Martha and various other saints and the whole lot were greeted by Saint Trophemus The reason that this is such a sacred place is because the two Marys remained in the Camargue with Sarah the servant and it is she who has become the most important person for the gypsies After her miraculous escape from the sea she wandered throughout the Mediterranean region and the gypsies identify with her nomadic existence and her life as an oppressed servant [broadcast talks]

In (19) the proposition which is the ‘cause’ of the event modified by of course is not expressed, as was the case in (17) and (18) above. The proposition that ‘over the years the pilgrims have improved this story’ is nevertheless presented as somehow referring to something which ‘took place in the natural course of events’, which was predictable. The causing factor is left implicit because it is assumed to be known: the prediction is based on the speaker’s knowledge of the world, which tells him that improving stories is something pilgrims tend to do. Hence past facts lead to present predictions. A further step in the meaning development is that towards a more intersubjective meaning involving the hearer and ‘predictably’ becomes ‘as

126

Anne Wichmann, Anne-Marie Simon-Vandenbergen and Karin Aijmer

you may expect/may have expected’. Example (16) given above and repeated here with more context illustrates this sense: (16)

R: Do you want him to become Prime Minister S. Of course I do I think it would be wonderful Yes I’d be terribly proud of him Wouldn’t anybody [broadcast news]

Example (16) is an extract from an interview with Mrs Heseltine about her husband, Michael Heseltine. She uses of course to present her positive reply to the question as ‘to be expected’. This is made explicit in her following utterance ‘wouldn’t anybody’. In this way the speaker positions herself in a shared world of common values. This sense can be taken one step further into ‘as everybody knows’, and from there to the more intimate world of the interactants (‘as you and I know’). In this sense of course acquires a predominantly interpersonal meaning. Example (20) illustrates ‘as everyone knows’, while (21) illustrates ‘as you and I know’. (20)

The British obsession with class is quite remarkable We persuade ourselves quite against all the available evidence that it perverts our social political and economic life Foreigners believe this too Here for example is Helmut Schmidt’s famous quote of nineteen seventy-five He said as long as you maintain this damn class-ridden society of yours you will never get out of your mess But Britain is more or less the same class structure as most advanced industrial societies a declining working class an expanding middle class although we display more social mobility than most others if the leading sociologists not usually Conservatives are to be believed Now in spite of this the British class system is regarded as peculiar And it certainly would be peculiar if all the myths about it were actually true but of course they’re not For example in nineteen eighty-four according to a Gallup poll seventy per cent of the British population claimed to be working class

How prosody reflects semantic change

127

claimed not only that they were working class but their parents were working class According to sociologists the percentage was at least twenty per cent less [broadcast talks] In (20) the speaker presents the proposition that the myths about the British class society are untrue as shared knowledge. (21)

B: And and somebody else said well maybe that was a bit of a mouthful and you know try just an evening on your own together A: Like skiing or something Or a day trip B: Yeah Or something like that uhm so you know so far none of those suggestions A: The other thing is uhm do you confide in her Does she feel excluded because you don’t exactly confide in her B: Well I don’t But then uh I haven’t I mean I never have and I’m I’m rather scared that you know that would seem rather artificial to her and as an attempt to win her over and of course you know she’s terribly alive to things like that uhm [direct conversations]

In this casual conversation (21), the knowledge that ‘she’s terribly alive to things like that’ is not generally known but belongs to the world of intimates. The meaning has evolved into ‘as you and I know’. As a further step this intersubjective meaning gets further eroded into a marker of speech as interaction, where of course functions like you know (see Holmes 1988: 69). It seems to be developing this function in contexts where it co-occurs with other discourse markers and with disfluency markers. Example (22) illustrates this use: (22)

C: They th think different things are important and at the end of the day that makes the atmosphere of the very different to a boys’ school

128

Anne Wichmann, Anne-Marie Simon-Vandenbergen and Karin Aijmer

D: Yes But I mean C: They’re they’re very upper class I mean some of the things that get lost in boys’ schools are are highlighted in a girls’ school B: Uhm D: Like B: Uhm C: Uhm B: Pit bulls and anthems I mean yeah C: Oh yeah I mean it’s probably cos I mean you know they don’t advocate the fact of course that you’re actually dealing with at the end of the day young girls or young boys And they’ve got different aspirations on what’s important as well B: Uhm uhm [direct conversations] Without wanting to claim that of course has become eroded and merely functions as a filler, we feel that it is important that it can occur in a context of hesitation markers and interpersonal hedges such as oh, yeah, I mean, probably, cos, you know, actually. What we see then is that the polysemy of PDE of course reflects a range of senses which can be put on a cline from more to less propositional, or from more to less objective. The comparative terms ‘more to less’ indicate that no instances are purely propositional or purely intersubjective: they are all layered. But some instances can be classified as predominantly one or the other, ‘epistemic/evidential’ or ‘intersubjective’. Other instances are somewhere in between and indeterminate. Example (23) illustrates what we mean by ‘indeterminate’: (23)

I come back to uh Dumas every so often and read The Count of Monte Cristo or The Three Musketeers Things like that Things that have been with me I suppose since I was in my teens

And and every so often I come back to them and uh and of course they get dog-eared and so on

How prosody reflects semantic change

129

so then then I scour the secondhand bookshops looking for decent copies of them to uh bring my uh collection back to uh some decent state [direct conversations] In (23) there is a clear sense of resultative of course: if one handles books very often they get dog-eared. A paraphrase with ‘as a natural consequence’ is possible here. On the other hand there is also a very clear interpersonal element, an appeal to the hearer to recognize the type of situation, to place it in what she knows about the world. The sense of inevitability is used to draw the hearer into a world of shared expectations, ‘as you may expect’, ‘you know what I’m talking about’. We would therefore classify such instances as indeterminate. A special type of usage which also combines the meanings ‘inevitably’ and ‘shared expectations’ is illustrated in (24) below. (24)

And literally she followed him down to and the romance blossomed somehow And another fellow who we cruelly nicknamed uh foetus Foetus that’s a very dreadful word When he went over on a month’s trip to the Soviet Union He was at Liverpool Poly to begin with but he was far too intelligent Apart from that he tried to smuggle this girl back Vera you know in the train com compartment where you’re supposed to shove the luggage And of course Leda came up This ha happened when my friend was with him on a Sept September trip And of course Leda came in and said have you seen Richard And she said these two little heads were pro peeping down from the top of the carriage [direct conversations]

The fact that ‘Leda came in’ is presented as an inevitability because of what ‘we know of the world’: when things can go wrong they do go wrong. The sense of inevitability is not the result of some objective causal factor but is

130

Anne Wichmann, Anne-Marie Simon-Vandenbergen and Karin Aijmer

based on a subjective assessment of how things happen in this world. The tone of irony is the same as in the predictability sense which the modal would has developed in expressions such as Well he would say that, wouldn’t he? In both cases a modal item has acquired a tone of irony to signal inevitability as a shared world-view.6 6.3. Discourse status A second parameter which leads to two types of usage is ‘discourse status’. This distinguishes between cases where of course confirms a previously brought up proposition (typically as a response marker) and where it modifies new information. Example (16) above is an instance of a response: the only new element in the speaker’s utterance Of course I do is the confirmation of course, given in answer to a polar question. The ‘given’ information need not be present in a question but may also be in a previous statement, as in (25): (25)

B: Now I think that over the years a breadth of experience outside politics and inside politics has equipped me to understand many of those concerns so I believe I can do it But my colleagues will decide I’m not going to breast beat about it They will make a decision and I will abide by that decision A: Of course But of But of course of course of course they will and of course you will too it’s self-evident [broadcast interviews]

In most cases, however, the speaker uses of course in an utterance which conveys new information. For example in (24) above the fact that ‘Leda came in’ has not been mentioned before in the conversation and is introduced by the speaker as new information.

6. The same development seems to have taken place in the case of no doubt in some contexts, where the speaker uses it as an ironic afterthought or response commenting on a proposition which s/he presents as to be expected (see Simon-Vandenbergen 2007).

How prosody reflects semantic change

131

6.4. Heteroglossic functioning The term ‘heteroglossia’ has been borrowed from White (e.g. 2003), who uses the Bakhtinian term to refer to the view that all utterances take place in a context of other utterances, and that speakers in using language inevitably take a stance towards other viewpoints. White (2003) classifies semantic resources according to the type of positioning they convey. The function of the word of course is classified as ‘concurrence’, in that it represents “the textual voice as taking up some generally held position and thereby as concurring with the reader”. The rhetorical function can also be termed “dialogical alignment” (White 2003: 269). A first basic distinction we noted in the data from this perspective is that between what we would call ‘interior’ and ‘exterior’ dialogue. In the former type the speaker is as it were weighing arguments for him/herself, and using of course to self-correct, to modify, to add another aspect, generally to signal a ‘dip’ in the argument. Again, there are various possibilities, which the following examples illustrate. Example (26) is a clear instance of engagement with an exterior voice: (26)

B: I am making the point that the Conservative Party in my judgement has three excellently qualified candidates and they will make a judgement as to which of those candidates most suit their preferences to be leader of their party A: Sure but the reason why I invited you is because I thought you wanted to be at Morden and felt that the party ought to have you rather than them but uh of course it’s up to them [broadcast interviews]

In (26) speaker A expresses agreement with a viewpoint (‘they will make a judgement’) voiced by speaker B. Example (27) illustrates of course in a counter-argument against a viewpoint which the speaker does not share but which she herself voices in her discourse: (27)

A: The other thing is I suppose uhm uh that we this extraordinary notion that adolescents should have their family unit as their centre of their life which of course children who go to boarding school don’t have but of course presumably if she is saying no I don’t want to go to boarding school

132

Anne Wichmann, Anne-Marie Simon-Vandenbergen and Karin Aijmer

and yet you see she’d have such a lot of fun there because she wouldn’t have to think about you and Gavin and Bernard which is actually rather boring for her She would have all her friends Couldn’t you send her uhm to a mixed school at sixteen like King’s Canterbury where Fran’s going or weekly B: uhm uhm Oh I see that yes uhm uhm uhm uhm uhm [conversations] Speaker A is here weighing pros and cons of sending B’s daughter to boarding school as a solution for a problem B has brought up. The argument introduced by but of course is a counter-argument against the viewpoint expressed in A’s own preceding utterance. The speaker thus positions herself against an alternative viewpoint which she introduces first to then argue against it. In example (28) there is no alternative voice which the speaker engages with except his/her own preceding statement. In such cases of course may announce a major or minor shift of topic. (28)

B: I’ve never had one D: Who you Nor’ve I B: No you wouldn’t Stuart because you don’t pay tax But of course now they’re going to send everyone one aren’t they regardless of whether you pay tax or not E: Uh I don’t know [direct conversations]

One heteroglossic function which is frequently fulfilled by of course is ‘countering’. ‘Countering’ is a general cover term for contexts in which speakers put propositions against other propositions. Arguments given by others or foreseen as possible arguments are ‘disclaimed’ (White’s term 2003: 271) in various ways. White subsumes the strategy of disclamation and its subtype countering under the resources which contract the dialogue. His definition is the following: They [i.e. proclamations] stand beside a second grouping of resources which are even more contractive in that they entail the direct rejection or

How prosody reflects semantic change

133

countering of a dialogically contrary position. They are the resources which I group under the general heading of DISCLAMATION which operate either as DENIAL (negation in the broadest sense) or as COUNTERING (various types of concessives, adversatives, and counter-expectancy). (White 2003: 271)

The occurrences of of course in structural environments which express countering have been grouped into the following types: (i) the collocation but of course: this is an instance of initial of course preceded by a coordinating conjunction; (ii) but … of course: the occurrence of of course in a clause introduced by but where but and of course are separated by intervening constituents; (iii) whereas/although/while/without … of course: the occurrence of of course in a subordinate clause of contrast or concession; (iv) of course … but: the occurrence of of course in a clause which is followed by another clause introduced by but.

Table 3 gives the frequency of these patterns in the total sample of 200 instances. Table 3. Frequency of occurrence of (i) but of course; (ii) but … of course; (iii) whereas/although/while/without … of course; (iv) of course … but type (i)

type (ii)

type (iii)

type (iv)

total

15

4

6

5

30

% in the sample 15%

Table 3 shows that 15% of all instances of of course occur in adversative or countering contexts.7 The likelihood of certainty markers in such contexts has been noted in various studies (Holmes 1988; Downing 2001; SimonVandenbergen and Aijmer 2002/2003), and can be explained in terms of flouting of the Gricean maxim of Quantity (Grice 1975): since one is supposed to say only what one has evidence for, the addition of certainty markers is superfluous under ‘normal’ circumstances. If they are added they often signal the need for the speaker to persuade, to counter, to chal7. In fact this is an underestimation of the real frequency because type (iv) is here limited to instances where the contrastive but-clause follows the of courseclause immediately, but such contrastive clauses may also be separated from the of course-clause by intervening clauses or the contrastive clause may not be introduced by an explicit but but nevertheless express a counter-argument. Such instances have, however, not been counted.

134

Anne Wichmann, Anne-Marie Simon-Vandenbergen and Karin Aijmer

lenge, to dismiss. Certainty markers in general, then, are likely to be frequent in argumentative contexts where consensus cannot be taken for granted. Of course has the additional meaning of expressing shared knowledge, on top of its modal meaning of certainty, and this adds the ‘complication’ that we need to explain why and in what contexts speakers want to convey information which is already known and mark that information as such by the use of of course. Here are some examples of types (i) to (iv) in that order: (29)

(30)

(31)

(32)

George Bush in his inauguration speech said A president is neither prince nor Pope It was his way of saying that his approach to the job was not going to be grandiose or imperious But of course the American people who love British Royals perhaps more than the Britons do rather want that kind of figure and in the main they don’t mind paying for it At least they get to see some of what their money buys [broadcast talks] And it’s an example of the uh hopelessness of the Europeans when faced with a crisis which is nothing to do with Europe but is outside the area of Europe because of course traditionally they have always been concerned with what happens within the Continent and left what happens outside the Continent to others including ourselves But it’s of course one which doesn’t alter the fundamental objections to political and economic and monetary union which uh have always been there [broadcast news] And again my honourable friend doesn’t understand the meaning of morality Last week on a visit to Israel I found that the Israeli government was well aware of the dangers of her becoming militarily involved in the Gulf crisis Uh while she does of course have every right to defend herself will my honourable and learned friend urge the government of Israel to continue to show the considerable constraint which she has so far shown [parliamentary debates] Royalists say that the country makes a profit out of the monarchy Andrew Morton explains how this idea is mistaken

How prosody reflects semantic change

135

Those people were interviewed outside Buckingham Palace Like many others they believe that having a Royal Family is good for Britain’s tourist industry I couldn’t find any real evidence that it is Anyway other countries in Europe such as Austria and Switzerland make far more money out of tourism than we do and they are republics Of course republicanism isn’t about the Royal Family failing as a tourist attraction but even so it is interesting to discover that it isn’t a very great attraction The British Tourist Authority’s own figures show that the main tourist attractions have very little to do with living royalty [broadcast talks] It can be seen that there are two main types of immediate contexts here, viz. contrastive and concessive. In both cases some proposition which is presented as shared knowledge is juxtaposed to another one. The types are repeated here for convenience: (i) the collocation but of course: this is an instance of initial of course preceded by a coordinating conjunction. This type is illustrated in example (29). (ii) but … of course: the occurrence of of course in a clause introduced by but where but and of course are separated by intervening constituents. This type is illustrated in example (30). (iii) whereas/although/while/without … of course: the occurrence of of course in a subordinate clause of contrast or concession. This type is illustrated in example (31). (iv) of course … but: the occurrence of of course in a clause which is followed by another clause introduced by but. This type is illustrated in example (32).

In types (i) and (ii) the speaker contrasts his/her statement, which is presented as shared knowledge, with a preceding one. In types (iii) and (iv) he/she backgrounds as self-evident an argument which is contrasted with his/her main argument. While the specific functions of of course may contextually vary (from authority to solidarity and from ‘predictably’ to ‘as we all know’), the overall rhetorical function of of course is to express a consensual knowledge of some things in order to strengthen one’s position in the dialogue. The meaning is very often a concessive one, with the speaker dismissing a proposition as ‘true but irrelevant’ or ‘true but not the main point’. This is especially the case in type (iv), but the other types can also

136

Anne Wichmann, Anne-Marie Simon-Vandenbergen and Karin Aijmer

have the function of focusing on the ‘main issue’, as is very clearly and explicitly the case in example (30) (with reference to the fact that it does not ‘alter the fundamental objections’). One important difference between the examples is the degree of foregrounding of the of course utterance. In some instances it is foregrounded (as in 29 and 30), in others it is backgrounded (as in 31 and 32). In types (i) and (ii) it is necessarily foregrounded because the but-clause follows the one it contrasts with; in type (iv) it is necessarily backgrounded because it precedes the but-clause. 6.5. Pragmatic stance As pointed out by Holmes (1988), referred to in Section 2.1.2 above, of course may signal both authority and solidarity. By presenting a viewpoint as ‘generally known’ or ‘self-evident’ one may project an image of superiority as well as create an egalitarian relationship, depending on the heteroglossic function of the utterance and on contextual factors such as relationship between speakers and genre. In casual conversation the stance is generally one of solidarity and intimacy, with of course signalling a shared world, while in political colloquy the stance is generally authoritarian, with of course signalling superior knowledge. Example (21) above (from a conversation) illustrates the solidarity stance. Example (20) (from a broadcast talk) illustrates how of course projects authority. These examples are repeated here for convenience’ sake. (20)

The British obsession with class is quite remarkable We persuade ourselves quite against all the available evidence that it perverts our social political and economic life Foreigners believe this too Here for example is Helmut Schmidt’s famous quote of nineteen seventy-five He said as long as you maintain this damn class-ridden society of yours you will never get out of your mess But Britain is more or less the same class structure as most advanced industrial societies a declining working class an expanding middle class although we display more social mobility than most others if the leading sociologists not usually Conservatives are to be believed

How prosody reflects semantic change

(21)

137

Now in spite of this the British class system is regarded as peculiar And it certainly would be peculiar if all the myths about it were actually true but of course they’re not For example in nineteen eighty-four according to a Gallup poll seventy per cent of the British population claimed to be working class claimed not only that they were working class but their parents were working class According to sociologists the percentage was at least twenty per cent less [broadcast talks] B: And and somebody else said well maybe that was a bit of a mouthful and you know try just an evening on your own together A: Like skiing or something Or a day trip B: Yeah Or something like that uhm so you know so far none of those suggestions A: The other thing is uhm do you confide in her Does she feel excluded because you don’t exactly confide in her B: Well I don’t But then uh I haven’t I mean I never have and I’m I’m rather scared that you know that would seem rather artificial to her and as an attempt to win her over and of course you know she’s terribly alive to things like that uhm [conversations]

7. Prosodic analysis 7.1. Prosodic prominence and separation If, as intonation theory claims, prosodic prominence is directly related to semantic weight, we would expect grammaticalized items to be infrequently stressed, just as function words are usually unstressed in English. Where layering occurs, i.e. where the grammaticalized form and the lexical form of a word or phrase co-exist in PDE, we would expect the latter to be more frequently stressed than the former. If our assumption is correct, namely

138

Anne Wichmann, Anne-Marie Simon-Vandenbergen and Karin Aijmer

that of course has gone further in the process of grammaticalization, and now has discourse functions in which the lexical content has bleached further in comparison with its epistemic/evidential function, we would expect stressed tokens to be less frequent than unstressed tokens (other factors being equal). The distribution (Table 4) shows that this is indeed the case.8 Table 4. Distribution of of course tokens that are stressed, unstressed or indeterminate in status indeterminate 16 (8%)

stressed 63 (31.5%)

unstressed 121 (60.5%)

total 200 (100%)

On this evidence, then, we have prosodic grounds for considering of course to have undergone a process of semantic weakening, and to be functioning at least in many cases as a discourse marker. Secondly, we consider the prosodic separateness or integration of of course. There are claims in the literature that discourse markers have a tendency to prosodic separation, reflecting their decreased syntactic integration in the utterance, but empirical studies show great variability. Our overview of the literature (2.2) shows that these opposing views are not necessarily mutually exclusive, but that the picture is more complex than would appear. In our data we find, predictably, both separation and integration. There are, however, far more cases of integration than separation (79.5% of all tokens) (see Table 5). This is, of course, consistent with the strong tendency for tokens to be unstressed, since separate tone units are assumed to contain at least one stressed syllable. It is not, however, consistent with the view that discourse markers tend towards increased prosodic separation. Table 5. Distribution of of course tokens that are prosodically integrated, separate, or where this is indeterminate integrated 159 (79.5%)

separate 35 (17.5%)

indeterminate 6 (3%)

total 200

8. While stress is a binary feature in theory, the phonetic cues in naturallyoccurring speech are not always clear-cut. Research has shown (Rietveld and Gussenhoven 2003) that a high utterance-initial item can be perceived either as having a high pitch accent (H*) or a high boundary tone (%H) (the former is maximally prominent and the latter minimally prominent). In the British system of intonation this distinction would be between a high pre-head syllable and a high onset. In some utterances in our data it was not possible to decide.

How prosody reflects semantic change

139

Theoretically (i.e. according to most models of intonational phonology) only the stressed tokens have the potential to have their own tone unit, and we would for the same reason expect all unstressed tokens to be integrated into a larger tone unit. The stressed tokens in this data are evenly divided between integration and separation. It must be remembered that there is potential for indeterminacy in identifying both prominence and boundaries, the choice of one often determining the choice of the other. If we examine the co-occurrence patterns of the two parameters stress and separation (Table 6), we can say that more than half (56%) of the tokens in the data are both unstressed and integrated. This contradicts suggestions that discourse markers tend to be prosodically separate, but supports the view that of course has a grammatical function rather than a lexical one. We will examine these patterns again below in relation to the various discourse functions described in Section 6. Table 6. Distribution of of course tokens according to the parameters stress and integration integrated % of integrated % of all tokens separate % of separate % of all tokens indeterminate total

stressed 33 (21%) (16.5%) 26 (74%) (13%) 4 63

unstressed 112 (70%) (56%) 8 (23%) (1%) 1 121

indeterminate 14

total 159

1

35

1 16

6 200

To summarize, these overall results are consistent first with the notion that of course has undergone semantic bleaching and acquired a grammatical function (as reflected in the tendency to be prosodically non-prominent), and secondly with earlier observations that discourse markers can be both prosodically separate and prosodically integrated. In the following we illustrate from the corpus how the parameters of stress and separation interact, dealing in turn with the four main cooccurrences: unstressed and integrated, stressed and integrated, unstressed and separate, stressed and separate. Over half of the tokens (56%) are unstressed and prosodically integrated, including initial, medial and final positions (33a–c).

140 (33)

Anne Wichmann, Anne-Marie Simon-Vandenbergen and Karin Aijmer

Unstressed and integrated: a. How you \/fashion them of course | is the big big problem

b. And of course \nobody will own \/up c. a::::nd uh \they’re very exciting of course

Those tokens that are stressed and integrated (34a–c) tend to have a falling tone.9 (34)

Stressed and integrated: a. Uhm a::nd of \course the \other part of the joke is that there is a sort of uhm suggestion that the unity of the two is a mask as well a doubtful unity b. And what \/I would do | would `not10 be of \course| at the same level in any sense at all c. In the \nicest possible \way of \course

Stressed and separate tokens in initial position (35) are typical of the prosodic separation frequently observed with other discourse markers. In initial position (35a–e) they are followed, in most cases, by a pause (including uh). The tone choice is typical of reinforcing adverbials (e.g. obviously, clearly, definitely, certainly, without a doubt) – the sense of ‘finality’ or ‘closure’ being consistent with the exclusion of other possibilities entailed in the expression of certainty or reinforcement. (35)

Stressed and separate: (initial) a. Of \course | republicanism isn’t about the Royal Family failing as a tourist attraction b. But of \course | Alexander came along after classical ballet

c. But of \course uh | `foreign \/governments uh | warning their citizens have to take wider things into ac into account

9. Only 7 of a total of 63 stressed tokens have non-falling tones: 4 rise, 2 fall–rise and 1 low level. 10. The ` symbol indicates here a high onset that displays little pitch movement.

How prosody reflects semantic change

141

d.

Of \course | you are entitled to draw conclusions from the uh evidence e. Now uh of \course uh `you’ve \heard uh these uh uh words read out al already uh more than once (final) f. Well we’re /\not going to force them to /take it | of \course

g. No not a \word| of \course (medial) h. As soon as it is possible to say something more concrete about that | then of \/course | we will do so i. The buildings as we have them | of \course | \/are later Unstressed and separate tokens are anomalous: according to the rules of prosodic phonology, unstressed tokens could not be other than prosodically integrated, given that the presence of a tone (pitch accent) is a defining characteristic of a tone unit.11 However, a number of initial unstressed tokens in our data are separated by a pause from the rest of the utterance (36a–c). (36)

Unstressed and separate: a. And of course | a late date is suggested by the fact that Malachi is the last of the prophets b. but of course pr\/esumably | if she is saying c. Of `course it’s it’s en\tirely possible to waste your ent en\tire working \/day on it d. and of course you know | she’s \terribly \/alive to things like that uhm

This anomaly is already well documented in the intonation literature (e.g. Cruttenden 1997). A pause is assumed to be a strong indication of a prosodic boundary, and by this definition these tokens are clearly separate, but a separate tone unit is required to contain an accented syllable, which is not present here. The conflict thus lies between external (boundaries) and inter11. Variously called tone groups, tone units, intonation units, intonation phrases etc.

142

Anne Wichmann, Anne-Marie Simon-Vandenbergen and Karin Aijmer

nal (accented syllables) definitions of tone units. If this is a problem, however, it is more a problem for generative phonology than for the study of discourse.12 Such pauses are quite systematic in some contexts, even in formal read-aloud prose.13 It is entirely consistent with observations of other discourse markers (well, now, so) that they can be followed by a pause despite being short and unstressed. From a Conversation Analysis perspective (see Sacks et al. 1974: 719–720) these can be seen as projecting a turn and claiming the floor while allowing time to formulate an utterance.14 This phenomenon does highlight, however, the lack of specificity in some comments on prosody in the discourse literature. The binary distinction between integration and separation does not take into account indeterminacy nor the fact that temporal and melodic separation/integration have to be specified separately. In other words, when a discourse marker is referred to as ‘prosodically separate’, we do not know whether it is unstressed followed by a pause, or stressed and carrying a tonal contour followed by a boundary (optionally including a pause). Formal intonational phonology can account for the latter but not for the former. 7.2. Prosodic realization and semantic meaning In this section we consider the extent to which prosodic realization correlates with meaning. We have already suggested that the high frequency of unstressed tokens suggests a high degree of grammaticalization. Here we attempt to relate the accentual status of tokens to an independently judged degree of semantic weight. Despite retaining a consistent underlying core meaning, of course can convey anything from this is a natural consequence to you/we know this (Section 6.2). In other words, some uses are clearly epistemic in that they express the speaker’s stance towards the truth of the proposition, while in others the meaning has become more opaque15 to con12. An unaccented discourse marker followed by a pause was observed often enough by Chafe (1994) for him to see this as what he calls a ‘regulatory’ tone unit. By distinguishing between different kinds of unit he dispenses with the anomalies arising from more formal intonational phonology. 13. We refer e.g. to the pause that often precedes a final reporting clause (… he said) even though it contains no accented material. 14. We are grateful to an anonymous reviewer for drawing our attention to this. 15. House (1989: 104) uses ‘opaque’ to mean ‘non-transparent and having highly negotiable illocutionary force’.

How prosody reflects semantic change

143

vey little more than ‘shared knowledge’. In some cases of course seems to perform some “routine (organizational) task in interaction” which discourse markers such as I think, I mean, and you know also perform in some contexts (Kärkkäinen 2003: 172). In between these two extremes there are many shades and gradations, and these will be discussed later, but for practical purposes we have divided this continuum into three broad categories: epistemic, shared knowledge and indeterminate. Intonation theory, as discussed above, would predict that those items at the transparent end of the spectrum, and therefore having greater residual semantic weight, are more likely to be stressed than those at the other extreme. The data confirms this prediction (see Table 7). A chi-squared test was carried out to ascertain whether there was an association between meaning and accentual status. The tokens whose accentual status could not be decided (first column) were omitted from the analysis. The test showed a very strong association between meaning and accentual status (chi-squared = 22.3, df=2, p ‘indeterminate’ > ‘epistemic’) the probability increases that the token will be stressed, and conversely, as the literal meaning fades, the probability increases that the token will be unstressed. Table 7. Distribution of of course tokens according to accentual status and meaning stress

indeterminate

unstressed observed/expected 47 38.3

total

3

stressed observed/expected 11 19.7

shared knowledge indeterminate epistemic total

9 5 17

10 41 62

38 36 121

57 82 200

16.3 26.1

31.7 50.9

61

These broad overviews of the data do not take into account utterance position, nor do they distinguish between the various discourse functions of of course. We also acknowledge the inevitable difficulty of forcing gradient data into fixed categories, but by basing our interpretations on the written transcripts alone we have at least endeavoured to avoid circularity. The results support claims that of course has undergone grammaticalization, functioning chiefly as a discourse marker (with both textual and interpersonal functions), while still having the potential to express more transparent, epistemic meaning, closer to the original core meaning. However,

144

Anne Wichmann, Anne-Marie Simon-Vandenbergen and Karin Aijmer

despite the evidence for a relationship between stress and semantic weight, we will see from the examples below that there are other parameters, e.g. information status and pragmatic function, that make it impossible to expect a one-to-one relationship between prosodic realization and any one parameter. 7.3. Prosody and discourse status The status of the information modified by of course, as set out in Section 6.3, can be either given or new in conjunction with previously given (or otherwise readily accessible) information and the of course serves to emphasize its certainty or self-evident truth. In these cases the of course carries by default nuclear tone, since the given information is either omitted or de-accented, and there is no other material available to carry the nucleus (37a–c). (37)

a. b. c.

\/win | of /\course I’m running to win Of \course I need a longer break from Darryl Well of \course it is | [but uh]

To treat the given proposition as ‘natural’ or ‘predictable’ in this way undermines to some extent the previous speaker’s justification in making it in the first place, having the sense of ‘you should know that’. It is therefore not simply a function of information structure, but has a marked attitudinal meaning, the nuances of which will depend on other contextual parameters. 7.4. Prosody and heteroglossia The attitudinal effect ascribed to the above examples also reflects their dialogic function (heteroglossia). In the simplest cases, the of course utterance is a response to an actual previous question or statement. Sometimes, however, the exterior ‘voice’ being responded to is hypothetical, allowing the speaker to pre-empt and dismiss an unspoken objection. In this case the hypothetical objection must be introduced to the discourse by the (real) speaker, but at the same time treated as ‘given’ (and thus already dismissed). This occurs in the following example:

How prosody reflects semantic change

(38)

145

A: Liberal Democrat leader Paddy Ashdown says that civilian deaths are regrettable but he says it would be wrong to condemn the U S military commanders for what appears to have been an error B: It is not as some of us have perhaps more comfortably believed a transcontinental video game in which nobody gets hurt This was clearly a mistake It’s not part of the policy And these mistakes happen in war Of course if there are lessons to be learnt we should be learning them But it would be wrong for politicians to second-guess the military who are taking the tough decisions to win this war from four thousand miles away [broadcast news]

In extract (38) the speaker is countering the condemnation of the US military commanders for civilian deaths, by first of all stating that these deaths were not “part of the policy” and that such mistakes are inevitable (these mistakes happen in war). The hypothetical objection that ‘there are lessons to be learnt’ is then treated as leading to a self-evident conclusion (Of course … we should be learning them). By presenting the conclusion as self-evident, it is backgrounded as ‘non-new’. The main point of the speaker’s answer is in the following clause, introduced by but, which denies that the military should be condemned. The prosodic challenge here is to simultaneously introduce a new proposition ‘learning lessons’ and treat it as given (a mutual belief). The speaker realizes the utterance as follows: (38ƍ)

Of \course if there are lessons to be /learnt | we should be \learning them

The absence of any accent on lessons suggests that the notion is already accessible. The high terminal on learnt is consistent with grammatical subordination of an if-clause, but of possible choices (level, fall–rise, high rise, low rise) at this point, the low rise is psychologically the least salient (Cruttenden 1997: 43), adding to the impression of ‘given information’.16 Such 16. An anonymous reviewer points out, rightly, that a rise on a subordinate clause in this position is predictable. However, this subordinate clause does not have its own tone group, as one would expect, but is integrated prosodically with the

146

Anne Wichmann, Anne-Marie Simon-Vandenbergen and Karin Aijmer

countering appears more effective and authoritative when of course is stressed, as if responding energetically to a real dissenting voice, and thus has a stronger dialogic effect. This is a feature of a persuasive rhetorical style, typical of political debate, and sets up an asymmetrical power relationship between speaker and hearers (or silent interlocutors). The third type of heteroglossic response is ratiocinative, in other words it is part of the process of developing one’s own argument in real time, and the ‘voices’ reflect different aspects of one’s own reasoning. This kind of thinking aloud is common in unplanned discourse, reflected in a number of connective strategies (but on the other hand, but then again, mind you, on balance though). The prosodic strategy is different from the cases described earlier, because the function is contrastive rather than countering, and may simply signal a shift in the discourse. We therefore find a number of cases where the of course is unstressed (39a–b). (39)

a. b.

But of course \/now | they’re going to send everyone one aren’t they regardless of whether you pay tax or not

but of course pr\/esumably | if she is saying

If a stressed token occurs in this context it suggests less an emphatic countering of an objection (as in the previous example) and more a strong shift in the discourse – either a temporary digression, in other words a move to a different level in the hierarchical structure of the discourse, or a linear move from one (sub)topic to another. Example (23), repeated in part as (40), is an instance of this: so of course they get dog-eared on the one hand simply expresses the natural consequence of frequent reading, but at the same time it creates a link to a new part of the narrative (so then I scour the second-hand bookshops). The and uh and sequence is symptomatic of a hesitant planning phase, and is typical of topic transitions in spontaneous narrative where a slowing down is followed by an acceleration into the next section – a change in tempo reflected in the reduced form of of course (‘course). This further supports our assumption that the stress on of course signals a shift rather than emphasizing the fact that books get dog-eared if read frequently.

preceding of course. This enhances the impression that it is being treated as given information.

How prosody reflects semantic change

(40)

147

And and every so often I come back to them and uh and (of?) \course | they get \/dog-eared and so on so then then I scour the secondhand bookshops looking for decent copies of them to uh bring my uh collection back to uh some decent state [direct conversations]

7.5. Pragmatic stance We have shown that a common function of of course is to express shared knowledge, either with the purpose of exerting power or in order to express solidarity. Where an authoritative tone is intended, we often find emphatic use of of course. Where symmetry and closeness are the goal, however, there appears to be little need to focus on the of course since there is no argument to be countered and no dissenting voice to be answered. Thus we find tokens used in this way are usually minimally prominent, reflecting also the fact that the core meaning is only minimally present. The expression of solidarity has different functions in different contexts, depending on such factors as the roles of the interactants and the aims they want to achieve in the interaction. In a teaching situation the word can function as a way of including the learners and their knowledge. In example (41) this attempt at inclusion is conveyed both by of course and by the tag question doesn’t he. (41)

And O\thello of course | calls on uh Desdemona’s \father doesn’t he | and then he tells her the story of his life [direct conversations]

In a football commentary, example (42), the of course serves to suggest to a radio audience that they are experts, sharing detailed knowledge with the commentator, who is filling in with what ‘everyone knows’: (42)

United on a \/good run at the moment | whereas \/Rangers of course | have been \struggling in the First Di/vision [spontaneous commentaries]

In the special context of a radio interview, the interviewee is often obliged to ‘tell’ the interviewer what they already know, because the interview has been prepared in advance. What is new to the radio audience is shared

148

Anne Wichmann, Anne-Marie Simon-Vandenbergen and Karin Aijmer

knowledge between the interlocutors. The dilemma for the speaker can be reflected in the use of of course, signalling complicity with the interviewer while performing the information as if new for the audience. In example (43) the late Archbishop of Canterbury, Robert Runcie, talks about his mother, and the fact that she had been a hairdresser. There is little residual meaning here of ‘a natural course of events’, since our knowledge of the world around us, or at least that part of the world that contains Anglican archbishops, does not normally permit us to predict the occupation of their mothers: (43)

Uh but uh she \had been of course uhm a a \hairdresser on an ocean liner

A similar interview situation elicited examples (44a–b), in which the information is in no way predictable but is marked as knowledge shared by the interlocutors: (44)

a. b.

hhh of course I’ve /\been a congregational /rabbi but \then of course | I \went into secular phi /losophy

In this special situation, where a speech situation involves multiple addressees, the use of of course is a way of providing information that is new to one part of the audience and known to the other. 8. Conclusion Our findings relate to three main areas of inquiry: structure, meaning and use, and prosody. In structural terms our data showed a clear preference for of course to occur in initial position as part of the thematic material, followed by medial position, where of course is post-thematic and serves to highlight the theme. In terms of meaning, we found strong evidence of grammaticalization with more literal meanings (‘as a natural consequence’) occurring alongside subjective (‘predictably’) and intersubjective developments (‘as you may expect’ or ‘as you and I know’). In a number of cases, the tokens have acquired a routinized pragmatic function and operate as an interactive marker. From a rhetorical perspective, the subjective and intersubjective meanings of of course are exploited dialogically, including con-

How prosody reflects semantic change

149

ceding or countering other points of view. Depending on context, this function can reinforce solidarity, as for example in casual conversation, but also express power, as for example in political debate, where dissenting voices are dismissed in favour of one’s own argument. Finally, on the basis of theories of intonational meaning, we predicted that semantic change involving a loss of semantic weight in favour of pragmatic meaning will also involve a loss of prosodic prominence. This is borne out by the data, which shows a statistically significant association between prosodic prominence and meaning. Of course as a grammaticalized/pragmaticalized item may be said to be unstressed in its unmarked realization, although the constraints of information structure, text structure or dialogic demands can require it to be brought into focus. In this respect the prosodic patterns found in our data, though diverse, are all consistent with a polysemous pragmatic marker. This study also has broader implications for the study of grammaticalization. A usage-based model of language (e.g. Bybee 2001) argues that patterns of change are affected by frequency of use: high frequency items can be vulnerable to phonological change involving attenuation of form, and frequency-induced habituation can lead to a loss of semantic meaning. Our corpus-based approach, therefore, acknowledges the importance of quantitative (frequency) information to complement more qualitative analyses. Secondly, since the processes associated with grammaticalization are in part phonological, it makes sense to focus on speech rather than writing. Phonological reduction typical of grammaticalization has so far only been described in segmental terms, and prosodic phenomena ignored, but the reduction of articulatory gestures is clearly a function of the prosodic realization of the word or phrase in question: a word that is unstressed has shorter duration than its stressed counterpart, leaving less time for careful articulation. Thus articulatory gestures may overlap or fail to meet targets. There are obvious problems associated with studying prosody from a historical perspective, but we believe that synchronically co-occurring prosodic patterns may reflect different stages in the grammaticalization process. If semantic change and phonological change are related, our view of such processes can only be enhanced by the study of contemporary speech and its prosodic characteristics. As we reported, there has been little systematic study of the prosody of discourse markers in general. While we, too, have focused on a single specific marker, we believe that our study has broader implications for the study of discourse markers in general. Most importantly, it is clear that there can be no single prosodic pattern that is ‘the prosody of discourse

150

Anne Wichmann, Anne-Marie Simon-Vandenbergen and Karin Aijmer

markers’. We know, for example, that tone choice is determined inter alia by position in an utterance and by meaning. Thus, falling tones conveying ‘finality or closure’ are not surprisingly more commonly associated with initial adverbials that reinforce rather than limit a proposition (compare certainly, of course with possibly, apparently), thus overriding another constraint, namely to use a non-final, or ‘open’ tone (e.g. a rise) to indicate that the utterance is incomplete. Accent status is also constrained partly by context, partly by morphology, partly by information structure (given–new) and by text structure. Finally, phrasing (status in a tone group) may depend on speech style, situation, or on the role of the speaker in terms of power, knowledge or rhetorical goals. In sum, prosodic choices – segmentation, accent placement and tone choice – convey abstract meanings that can be related only indirectly to lexical items, and are motivated in part by convention, but largely by the often conflicting demands and constraints of the semantic, pragmatic and discoursal functions that discourse markers fulfil. References Aijmer, Karin 2002 English Discourse Particles: Evidence from a Corpus. Amsterdam/ Philadelphia: Benjamins. Aijmer, Karin and Anne-Marie Simon-Vandenbergen (eds.) 2006 Pragmatic Markers in Contrast. Amsterdam: Elsevier. Aijmer, Karin, Ad Foolen and Anne-Marie Simon-Vandenbergen 2006 Pragmatic markers in translation: a methodological proposal. In Approaches to Discourse Particles, Kerstin Fischer (ed.), 101–114. Amsterdam: Elsevier, Altenberg, Bengt 1987 Predicting text segmentation into tone units. In Corpus Linguistics and Beyond, Willem Meijs (ed.), 49–60. Amsterdam: Rodopi. Altenberg, Bengt 1990 Automatic text segmentation into tone units. In The London–Lund Corpus of Spoken English: Description and Research, Jan Svartvik (ed.), 287–323. Lund: Lund University Press. Anon 1989 Oxford English Dictionary. 2nd ed. Oxford: Clarendon Press. Barth, Dagmar and Elizabeth Couper-Kuhlen 2002 On the development of final though: a case of grammaticalization? In New Reflections on Grammaticalization, Ilse Wischer and

How prosody reflects semantic change

151

Gabriele Diewald (eds.), 345–361. Amsterdam/Philadelphia: Benjamins. Biber, Douglas, Stig Johansson, Geoffrey Leech, Susan Conrad and Edward Finegan 1999 Longman Grammar of Spoken and Written English. London: Longman. Bolinger, Dwight 1989 Intonation and its Uses. London: Edward Arnold. Brinton, Laurel J. 1996 Pragmatic Markers in English: Grammaticalization and Discourse Function. Berlin/New York: Mouton de Gruyter. Bybee, Joan 2001 Phonology and Language Use. Cambridge: Cambridge University Press. Chafe, Wallace 1994 Discourse, Consciousness and Time. Chicago: University of Chicago Press. Cruttenden, Alan 1997 Intonation. 2nd ed. [1st ed., 1986.] Cambridge: Cambridge University Press. Diewald, Gabriele 2006 Discourse particles and modal particles as grammatical elements. In Approaches to Discourse Particles, Kerstin Fischer (ed.), 403–425. Amsterdam: Elsevier. Downing, Angela 2001 ‘Surely you knew!’ Surely as a marker of evidentiality and stance. Functions of Language 8: 251–282. Erman, Britt and Ulla-Britt Kotsinas 1993 Pragmaticalization: the case of ba’ and you know. In Studier i Modern Språkvetenskap, Johan Falk, Kerstin Jonasson, Gunnel Melchers and Barbro Nilsson (eds.), 76–93. Stockholm: Almqvist and Wiksell International. Ferrara, Kathleen W. 1997 Form and function of the discourse marker anyway: Implications for discourse analysis. Linguistics 35: 343–378. Fischer, Kerstin (ed.) 2006 Approaches to Discourse Particles. Amsterdam: Elsevier. Grice, Paul 1975 Logic and conversation. In Syntax and Semantics 3: Speech Acts, Peter Cole and Jerry Morgan (eds.), 41–58. New York: Academic Press.

152

Anne Wichmann, Anne-Marie Simon-Vandenbergen and Karin Aijmer

Günthner, Susanne 1999 Entwickelt sich der Konzessivkonnektor obwohl zum Diskursmarker? Grammatikalisierungstendenzen im gesprochenen Deutsch. Linguistische Berichte 180: 409–446. Halliday, M. A. K. and Ruqaiya Hasan 1976 Cohesion in English. London: Longman. Halliday, M. A. K. and Christian Matthiessen 2004 An Introduction to Functional Grammar. 3rd ed. London: Arnold. Heine, Bernd, Ulrike Claudi and Friederike Hünnemeyer 1991 Grammaticalization: A Conceptual Framework. Chicago: University of Chicago Press. Hirschberg Julia and Diane Litman 1993 Empirical studies on disambiguation of cue phrases. Computational Linguistics 19: 501–530. Holmes, Janet 1986 Functions of you know in women’s and men’s speech. Language in Society 15: 1–22. Holmes, Janet 1988 Of course: a pragmatic particle in New Zealand women’s and men’s speech. Australian Journal of Linguistics 2: 49–74. Hopper, Paul J. and Elizabeth Closs Traugott 2003 Grammaticalization. 2nd rev. ed. [1st ed., 1993.] Cambridge: Cambridge University Press. Horne, Merle, Petra Hansson, Gösta Bruce, Johan Frid and Marcus Filipsson 2001 Cue words and the topic structure of spoken discourse: the case of Swedish men. Journal of Pragmatics 33: 1061–1081. House, Juliane 1989 Politeness in English and German: The functions of please and bitte. In Cross-Cultural Pragmatics: Requests and Apologies, Shoshana Blum-Kulka, Juliane House, and Gabriele Kasper (eds.), 96–119. Norwood NJ: Ablex. Kärkkäinen, Elise 2003 Epistemic Stance in English Conversation. Amsterdam/Philadelphia: Benjamins. Lehmann, Christian 1995 Thoughts on Grammaticalization. Munich/Newcastle: LINCOM EUROPA. Original publication in Arbeiten der Kölner UniversalienProjektes 48, University of Cologne, 1982. Lewis, Diana M. 2003 Rhetorical motivations for the emergence of discourse particles, with special reference to English of course. In Particles, Ton van der Wouden, Ad Foolen and Piet Van de Craen (eds.), 79–91. Belgian Journal of Linguistics vol. 16. Amsterdam/Philadelphia: Benjamins.

How prosody reflects semantic change

153

Lima, J. Pinto de 2002 Grammaticalization, subjectification and the origin of phatic markers. In New Reflections on Grammaticalization, Ilse Wischer and Gabriele Diewald (eds.), 363–378. Amsterdam/Philadelphia: Benjamins. Meyer, Charles F. 1986 Punctuation practice in the Brown Corpus. ICAME news 10: 80–95. Nelson, Gerald, Sean Wallis and Bas Aarts 2002 Exploring Natural Language: Working with the British Component of the International Corpus of English. Amsterdam/Philadelphia: Benjamins. Pierrehumbert, Janet and Julia Hirschberg 1990 The meaning of intonational contours in the interpretation of discourse. In Intentions in Communication, Philip R. Cohen, Jerry Morgan and Martha E. Pollack (eds.), 271–311. Cambridge Ma., London: MIT Press. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech and Jan Svartvik 1985 A Comprehensive Grammar of the English Language. London: Longman. Rietveld, Toni and Carlos Gussenhoven 2003 The perception of preheads as accents. In Proceedings ICPhS, Barcelona, 759–762. Sacks, Harvey, Emanuel A. Schegloff, Gail Jefferson 1974 A simplest systematics for the organization of turn-taking for conversation. Language 50: 696–735. Simon-Vandenbergen, Anne-Marie 1992 The interactional utility of of course in spoken discourse. Occasional Papers in Systemic Linguistics 6: 213–226. Simon-Vandenbergen, Anne-Marie 2007 No doubt and related expressions: A functional account. In Structural-Functional Studies in English Grammar: In honour of Lachlan Mackenzie, Mike Hannay and Gerard J. Steen (eds.), 9–34. Amsterdam/Philadelphia: Benjamins. Simon-Vandenbergen, Anne-Marie and Karin Aijmer 2002/03 The expectation marker of course. Languages in Contrast 4: 13–43. Simon-Vandenbergen, Anne-Marie, Peter White and Karin Aijmer 2007 Presupposition and ‘taking-for-granted’ in mass communicated political argument: An illustration from British, Flemish and Swedish political colloquy. In Political Discourse in the Media: Cross-cultural Perspectives, Anita Fetzer and Gerda Lauerbach (eds.), 31–74. Amsterdam/Philadelphia: John Benjamins.

154

Anne Wichmann, Anne-Marie Simon-Vandenbergen and Karin Aijmer

Stenström, Anna-Brita 1990 Adverbial commas and prosodic segmenatation. In The London– Lund Corpus of Spoken English: Description and Research, Jan Svartvik (ed.), 253–266. Lund: Lund University Press. Thompson, Geoff 2004 Introducing Functional Grammar. 2nd ed. [1st ed., 1996.] London: Arnold. Traugott, Elizabeth Closs 1989 On the rise of epistemic meanings in English: an example of subjectification in semantic change. Language 65: 31–55. Traugott, Elizabeth Closs 1997 The role of the development of discourse markers in a theory of grammaticalization. Paper presented at ICEHL XII, Manchester 1995. Version of 11/1997 accessed on the internet at http://www.stanford.edu/~traugott/papers/discourse.pdf. Traugott, Elizabeth Closs and Richard B. Dasher 2002 Regularity in Semantic Change. Cambridge: Cambridge University Press. Wennerstrom, Ann 2001 The Music of Everyday Speech: Prosody and Discourse Analysis. Oxford: Oxford University Press. White, Peter 2003 Beyond modality and hedging: A dialogic view of the language of intersubjective stance. Text 23: 259–284. Wichmann, Anne 1998 Using intonation to create conversational space: Projecting topics and turns. In Explorations in Corpus Linguistics, Antoinette Renouf (ed.), 217–232. Amsterdam: Rodopi. Wichmann, Anne 2000 Intonation in Text and Discourse. London: Longman.

The semantic-pragmatic development of well from the viewpoint of (inter)subjectification Tine Defour Abstract Previous research on the semantic-pragmatic evolution of well has mainly focused on the functional development of manner adverbs towards pragmatic markers placed in utterance-initial position (Finell 1989; Jucker 1997). The complex relationship between well as a pragmatic marker and its propositional source has been subject to a gradual non-straightforward diversification (e.g. Traugott and Dasher 2002). According to Traugott’s hypotheses (1999), adverbial well could be expected to have followed a historical development from propositional, over textual, to interpersonal meanings – with an increase in subjectivity and intersubjectivity. In order to gain a more complete understanding of how the functional development of well has occurred, and how various propositional uses have influenced the development of subjective meanings, this paper aims to examine the semanticpragmatic uses of well in the historical data from the Helsinki Corpus of English texts (HC) and the Corpus of Early English Correspondence (Sampler) (CEECS). More specifically, different shades of meaning of propositional well will be considered – as a source for historical speaker-based and hearer-oriented pragmatic development. Secondly, two frequently occurring collocations of well will be examined against the background of increasing subjectivity and epistemicity.

1. Introduction 1.1. Aims and research material The historical development of propositional adverbs into pragmatic markers has received a fair amount of attention in recent years (see Schiffrin 1992 on then; Powell 1992 on stance adverbs; Aijmer 2002; and Traugott and Dasher 2002 among others). Diachronic hypotheses state that pragmatic markers which have evolved out of adverbials, such as well and now, are still coloured by their original propositional meanings. According to various hypotheses, manner adverbs are predicted to undergo an evolution from

156

Tine Defour

propositional meanings to more expressive and subjective meanings, i.e. from referential meanings to abstract discourse meanings with a wider scope (see Traugott 1995b) and a greater speaker-oriented perspective. It has been shown, however, that there is no such thing as a straightforward development from adverb to pragmatic marker (Schourup 2001; Traugott and Dasher 2002 among others). Jucker (1997 – see also Schourup 2001; Traugott and Dasher 2002) has attested that adverbial well has undergone a gradual functional diversification, and that the many pragmatic meanings of the present-day marker still show ties with adjectival and adverbial meanings of well. On the basis of a historical corpus study, this paper examines to what extent the pragmatic diversification of well can be explained against the background of hypotheses of grammaticalization, subjectification and intersubjectification, and whether we can trace diverging developments in the marker’s semantic-pragmatic history. My aim is to look at the historical development which well has undergone, and to examine to what extent the marker’s syntactic position has had an influence on the semantic-pragmatic meanings of well. A second focus will be on the pragmatic influence which well fulfils in two frequently occurring collocations, viz. with cognitive verbs (e.g. you know well that…) and with modal auxiliaries (e.g. you may well see that…). Various contextual factors will be taken into account, such as sentence position, relative frequencies and the relationship between speaker and addressee. We will see that the additional use of well can serve a subjective function in contexts where speaker and addressee have diverging opinions, and that the adverbial use of well can be used pragmatically to fulfil intensified and modal meanings. The data will attest to what extent well presents a valuable element in the positioning of the speaker and the acknowledgement of the addressee. For each example dated before 1500 used in this paper, a translation of the entire utterance or of relevant parts is given. The findings discussed in this paper result from research on two historical corpora, namely the diachronic part of the Helsinki Corpus of English Texts (HC) and the Corpus of Early English Correspondence (Sampler) (CEECS). The HC consists of a collection of extracts of continuous text, taken from a wide range of different genres and covering the period between Old English (c. 850) and c. 1710. The corpus contains approximately 1.6 million words. The CEECS is a letter corpus which runs from 1417 to 1681. The corpus contains 23 letter collections and a total of 450,000

The semantic-pragmatic development of well

157

words.1 The HC and CEECS are important methodologically because they contain “speech-related” genres. The HC for instance includes a certain percentage of dramatic constructions of speech (e.g. plays or sermons) or reconstructed recordings of speech (e.g. trial proceedings or witness accounts) – balanced by less speechlike, written genres (e.g. law documents). As a letter corpus, the CEECS also offers a valuable source of genre data. Letters, particularly private letters “contain many interactional features such as address terms, directives, politeness markers, apologies, and so on” (Jacobs and Jucker 1995: 8). Because pragmatic markers are more frequent in oral discourse and it has only recently been made possible to record spoken data (see Culpeper and Kytö 1997, 2000), the text types in these historical corpora come as close as possible to “natural” spoken interaction of previous centuries. Although synchronic and diachronic corpora are “assumed to be representative” of a certain language or language period (Francis 1982: 7), the linguistic domain they cover remains limited in the sense that it is hardly possible to achieve a complete level of equal distribution with regard to time periods, dialect, or text type. If this is difficult for synchronic data, it poses even more difficulties for historical periods. The compilers of the HC and CEECS have aimed to achieve different kinds of representativeness – “chronological, regional, sociolinguistic and generic”, with an emphasis on sociolinguistic coverage for the CEECS. We need to take into account that frequencies of pragmatic markers will be influenced by differences in genre representation. Palander-Collin (1999) points out, for example, that the phrase I think has a speech-based nature and was found to be much more frequent in private letters than in non-private letters. Inevitable differences in representativeness should therefore be kept in mind when considering quantitative or qualitative corpus results. Genre-related differences are not the main focus in this paper – the examples that are selected belong to a variety of (speech-related) genres, and only when relevant will the genre of the illustration be discussed. Apart from a reference to the corpus, the date and abbreviated file name, the genre of the excerpt is added between brackets after each example from the Helsinki Corpus. Examples from the CEECS all belong to the genre of (private) letters. The influence of genre 1. Both historical corpora are taken from the ICAME cd-rom (the International Computer Archive of Modern and Medieval English). Further information on the HC and the CEECS can be found on (last accessed on 1/04/2009). For the CEECS: also see Nevalainen and Raumolin-Brunberg 1996.

158

Tine Defour

differences on semantic-pragmatic meanings of well will be discussed more extensively in future publications. 1.2. Terminology Many different terms have been used to refer to the elements of speech known as pragmatic markers (Brinton 1996), discourse markers (Jucker 1997; Fraser 1999), discourse particles (Aijmer 2002) or discourse operators (Schourup 1999). In this paper I will use the term pragmatic marker (see Brinton 1996 for a list of defining features) as a general, encompassing term for those items which do not contribute to the propositional contents of an utterance, and which predominantly occur utterance-initially, more particularly in the “pre-front” field (Auer 1996). Subclassifications can be made on the basis of formal or functional criteria, which can clarify the differences between for instance (modal) adverbs (e.g. certainly, surely), pragmatic expressions or particles (Hoye 1997: 212; Aijmer, Foolen and Simon-Vandenbergen 2006). A general denotation with subdivisions allows for a more specific delineation of various forms and uses. Discourse markers, for instance, are considered to be a specific subclassification focusing on text-structuring functions, i.e. on how interactants structure discourse. In this article, the term particle is reserved for reference to modal particles. 2. Synchronic functions of well The discourse functions of well in present-day English have been the subject of a number of studies (Lakoff 1973; Svartvik 1980; Watts 1986; Schiffrin 1987; Schourup 2001 among others). Research on the semanticpragmatic uses of well shows that the marker is often used in the context of a confrontation between speaker and addressee and that well can be applied to create or restore coherence. According to Jucker’s division (1997), synchronic well can be classified according to four main pragmatic uses (table 1).

The semantic-pragmatic development of well

159

Table 1. Functional classification of well (Jucker 1997: 92) function adverb, adjective discourse marker discourse marker discourse marker discourse marker

level propositional textual interpersonal interpersonal interpersonal

label – frame marker face-threat mitigator qualifier pause filler

The main textual function of well as a frame-marker allows the speaker to create textual coherence by indicating topic changes or by introducing direct reported speech (see examples 1 and 2). In example (1), well forms a frame between two subtopics, uttered by the same speaker (Aijmer and Simon-Vandenbergen 2003: 1142). (1) (2)

“Well, where’d you sleep last night?” the woman said softly. “You get kicked out?” (Aijmer and Simon-Vandenbergen 2003: 1142) And he said well tell me something about rickets. (Jucker 1997: 93)

On an interpersonal level, well expresses attitudes or creates common ground between speaker and addressee. Jucker distinguishes three interpersonal meanings for well. As a face-threat mitigator, the marker can be used in situations where the addressee’s face (or that of the speaker) is threatened. In this case, well serves as a means to prevent face loss. (3)

A: I didn’t know there was such a job going B: [mhm] well there was about a year ago now they had a first batch… – and then a second batch … (Jucker 1997: 93)

Secondly, well can be used as a qualifier when a respondent is unable to give a reply which is – in the speaker’s or the addressee’s eyes – “sufficient” or when he or she cannot seem to give an answer which is “optimally coherent with the preceding question” because the respondent cannot supply the requested information (Schiffrin in Jucker 1997: 94).

160 (4)

Tine Defour

A: Are you from Philadelphia? B: Well I grew up uh out in the suburbs. And then I lived for about seven years up in upstate New York. And then I came back here t’go to college. (Schiffrin 1987: 106)

A third interpersonal function according to Jucker’s research is the use of well as a pause filler, which enables a speaker to ‘bridge interactional silence’ in situations where he or she wants to keep the floor. (5)

yes … quite … well there you are … you see … em – it’s an obvious – x application (Jucker 1997: 95) [adapted]

3. Historical development of the pragmatic marker 3.1. Grammaticalization and (inter)subjectification Previous studies have shown that the present-day multifunctionality of well can be traced back to propositional (adverbial or adjectival) functions with the meaning “in a good way or manner, according to a certain standard” (Finell 1989; Jucker 1997; Schourup 2001 among others). However, Schourup stresses that the “marker well has […] properties that cannot be predicted on the view that the marker is simply an adverb called into illocutionary service” (2001: 1038). The relationship between adverbial well – as in “He plays the guitar well” – and a pragmatic use of well – e.g. “Well, let me see” – is “far more tenuous” (Schourup 2001: 1038). An explanation for the present-day relationship between form and functions, and between original propositional meanings and present-day pragmatic functions of well may be found in the historical process of grammaticalization. Traugott (1995a: 15) defines grammaticalization as “the process whereby lexical material in highly constrained pragmatic and morphosyntactic contexts becomes grammatical, and already grammatical material become[s] more grammatical” (see also Traugott 1995b, 1997). Grammaticalization involves changes through which an element can gain increased scope or increased syntactic freedom, while at the same time the element’s position becomes more restricted in the sense that its position becomes “more clearly defined in its syntagmatic relation to other sentence elements” (Palander-Collin 1999: 50). Apart from grammatical or syntactic

The semantic-pragmatic development of well

161

changes, grammaticalization can also cause linguistic elements to lose semantic meaning and to gain pragmatic strength – evolving from propositional meanings, over text-structuring to increasingly interpersonal functions according to a unidirectional cline. In recent studies (Traugott 1999), more emphasis has been placed on an increase in expressiveness and subjectivity. According to these hypotheses, a propositional manner adverb such as well can develop abstract pragmatic meanings which are increasingly grounded in the speaker’s perspective. Grammaticalization can be motivated by the speaker’s need to communicate attitudes and personal opinions. The processes that support this need are called subjectification and intersubjectification and entail the recruitment of propositional material for subjective and intersubjective purposes. Subjectification (Traugott 1995b, 1997) causes meanings to “shift toward greater subjectivity [and] become increasingly associated with speaker attitude” (Traugott 1995a: 2) and can be defined more specifically as “the process whereby meanings come over time to index, encode, and externalize the speaker/writer’s perspectives and attitudes” (Schwenter and Traugott 2000: 10). Intersubjectification, which is generally preceded by subjectification and cannot occur without it (Traugott and Dasher 2002: 97), highlights the other end of the speech event and is “a mechanism whereby meanings become more centered on the addressee” (Traugott 1999: 3). As such, intersubjectification is described as “the explicit expression of SP/W’s attention to the ‘self’ of AD/R in both an epistemic sense, paying attention to their (likely) attitudes to the content of what is said, and in a more social sense (paying attention to ‘face’ or ‘image needs’)” (Traugott 1999: 2). Traugott (this volume) stresses that both subjectification and intersubjectification are processes that are independent of the process of grammaticalization, but that there is definitely a close interaction between subjectification and grammaticalization because the latter involves the development of (pragmatic) markers towards the ideational component of discourse structuring and/or towards text-structuring functions. An example of pragmatically motivated change is the evolution of indeed (Traugott 1995a: 7–9; Aijmer 2002: 16f.). In origin a prepositional phrase, indeed (‘in deed’) moved away from a clause-internal adverbial form towards a (modal) sentence adverbial – with the epistemic (modal) meaning “certainly” (Traugott 1995a: 8). Through further development on the adverbial cline, indeed shifted to clause-initial position and acquired meanings expressing “elaboration and clarification of the [the speaker’s] discourse intent” (Traugott 1995b: 11).

162

Tine Defour

Semantic-pragmatic change and (inter)subjectification are dynamic processes. As historical phenomena, these processes nevertheless result in synchronic layers of functional split for one particular item. The coexistence of different variants [A > A ~ B > B], which may or may not replace older meanings, is called “layering” (Hopper 1991). By tracing different steps in the semantic-pragmatic evolution of well we are able to gain a clearer picture of how different layers of functional split can be compared, and how propositional meanings have evolved towards a diversity of pragmatic and (inter)subjective uses. 3.2. The historical evolution: Previous research Although previous studies by Finell (1989), Jucker (1997), Schourup (2001) and Traugott and Dasher (2002) have established a connection between the semantic-pragmatic development of well and a propositional source, the marker’s evolution is not transparent. According to Finell (1989), who studied well in responses, early propositional uses of well as a predicative adjective can be seen as precursors for later interpersonal functions. She considers the following use of adjectival well to be the earliest context for the marker’s further pragmatic evolution. (6)

And where as they saye that the Gospell must be taught after the interpretations approued by the churche (that is very well) but all the stryfe is, which is the trewe church. (OED, well – 1560. taken from Finell 1989: 655).

In the following example, well can be seen as an abbreviated form of the phrase that is very well. (7)

Cloten: Nay, come, let’s go together. Second Lord: Well, my lord. (Shakespeare, Cymbeline, qtd. Schourup 2001: 1049)

Schourup (2001: 1049) states that these meanings are now obsolete but that their use expressed consent or agreement, and that they illustrated a particularly strong semantic connection between propositional and pragmatic functions. Placed in utterance-initial position, adjectival well allows the speaker to express “acceptance of a situation” or acceptance of a pre-

The semantic-pragmatic development of well

163

vious speaker turn. As in example (7), the central positive meaning of propositional well can then be used as a starting point for additional disagreeing or concessive comments (that is very well, but…). It is this combination of acceptance and concession which recurs in present-day interpersonal uses of well, as in example (8). (8)

Are you coming to the lecture? – Well, I’d like to, but I’m afraid I can’t today. (Finell 1989: 655)

In conformity with Traugott’s hypotheses on shifts of meaning and subjectification, Finell finds that well evolved from a predicative adjective to an expressive, independent part of the sentence. Jucker expands this point of view by stating that propositional (adjectival or adverbial) uses are not “sufficiently transparent as being the origins of the discourse meanings of well” (1997: 107). Finell’s study does not take into consideration the possibility that interpersonal pragmatic uses in present-day English may have a more textual function as a precursor. All Jucker’s early examples from Middle English place well in a context of direct reported speech, accompanied by a verb of speaking (see examples 9 and 10). (9)

(10)

“Ye sey well,” seyde the kynge. “Aske what ye woll and ye shall have hit and hit lye in my power to gyff hit.” “Well,” seyde thys lady, “than I aske the hede of thys knyght that hath wonne the swerde,…” (HC Cmmalory 1470 [Romances] – Jucker 1997: 99). ‘… “Ask what you will and you shall have it if it lies in my power to give it.” “Well”, said this lady, “then I ask the head of this knight who has taken the sword,…”’ “Howe say you, mistris Alice,” quoth he, “is itt not so?” “Bone deus, bone deus, man, will this geare neuer be lefte?” quoth shee. “Well then, mistris Ales, if it be so,” quoth he, “it is very well…”. (HC, Cebio1 [Biography]: 1500–1570)

Jucker’s historical data illustrate a “continuous [pragmatic] diversification” from the Middle English period onwards, starting with this predominantly textual function. This frame-marker use becomes more versatile in later periods, occurring also without communicative verbs. In late Early Modern English, pragmatic well is then seen to adopt new meanings – e.g. as an interpersonal face-threat mitigator. The functions of qualifier and pause

164

Tine Defour

filler were unattested in Jucker’s data and are therefore likely to have originated in later, post-Shakespearean periods.2 Although Jucker’s and Finell’s data validate Traugott’s evolution from propositional – via textual – to interpersonal functions, the Old English form wella suggests that an interpersonal element may lie at the basis of further developments of well. Research by Van Herreweghe (2003) states that the development of well as a discourse marker can to a certain extent be related to Old English wella or wel la. Wella is an Old English discourse marker which mainly served as an indicator of positive appraisal or as an attention-getting device, i.e. an interpersonal function. Although wella did not continue to exist in this particular form, Van Herreweghe suggests that a possible blending of forms may connect adverbial or adjectival meanings of well with Old English wella, as well as with other forms such as Old English weg la or wa la (“alas”).3 This possible connection again draws attention to the fact that already in its earliest developments, well essentially had an interpersonal element in its core meaning. Secondly, in considering the early textual functions of well – in the context of a quotation – we need to keep in mind that calling something well, i.e. good, in a good manner, is in itself already a subjective, evaluative act because it entails that the speaker is matching the situation against a certain norm or standard. It is clear that [w]ell functions as a frame marker and text-sequencing device, but in many cases it may also indicate an acceptance of a situation that has been expressed or indicated, and thus it may already have some interpersonal significance besides its mainly textual function. (Jucker 1997: 99)

Initially captured in a textual frame (quotation), well signals that the speaker finds that “the interlocutor has got a point” (Finell 1989: 655).

2. As pointed out by one of the anonymous referees, the absence of the function of pause filler might partly be attributed to the fact that spoken language is often rendered into written language in an idealized way. 3. Jucker (1997) and Van Herreweghe (2003) found that certain uses of wella have been translated as ‘alas’, which may seem at odds with the generally positive semantics of well(a). Jucker suggests that translators perhaps did not have an appropriate word for the translation of wella, while Van Herreweghe mentions a possible connection with Latin (e)heu, which is an interjection of pain or grief. A related expression of sorrow still exists today – though obsolete – in the form of wellaway (i.e. originally a combination of wel(l) and away).

The semantic-pragmatic development of well

165

A quantitative and qualitative analysis of the diachronic data from the HC and CEECS indicates that the historical evolution of well largely follows Jucker’s findings in the sense that most of the corpus tokens before 1500 (9 out of 10) are found in a textual frame of direct reported speech. After further examination of all utterance-initial tokens of well in the historical corpora, paying specific attention to the ways in which well interacts with the speaker’s need to express personal stance and create interpersonal ties with the interlocutor, we can see that the sense of acceptance signalled by well is increasingly followed by a personal – concessive or diverging – point of view, indicating that the speaker “is not prepared to completely comply with the interlocutor” (Finell 1989: 655) (examples 11 and 12). (11)

(12)

Miss.: That’s my best Nurse, do as you wou’d be done by; trust us together this once… Nurse: Well, this once I’ll venture you; but if you disparage me… (HC, Ceplay3a [Drama, Comedies]; 1640–1710) Leaue her company ketha? Alas poore soule, this reward she hath for her good will. I wis I wis, she is more your friend, then you are your owne. Well let her be what she will sayd her husband: but if shee come any more in my house, shee were as good no. And therefore take this for a warning I would aduise you: and so away he went. (HC, Cefict 2b [Fiction]: 1570–1640)

The data from the HC and CEECS attest that well is increasingly used in contexts where speaker and addressee have diverging views or need to reach a source of common ground – as in the examples above. Aijmer and Simon-Vandenbergen (2003) underline the fact that the central positive sense of well makes it a valuable element in interaction and “interactively useful in cases where speakers are aware of possibly divergent interpretations” (2003: 1129). Whereas early examples of well mainly indicated the acceptance of a situation (as an indication of a “positive judgement”), it seems that most tokens of well in corpus examples before ca. 1500 are increasingly followed by a personal elaboration, thought or modification (often differing from the point of view of the addressee). The sense of “acceptance” expressed in earlier, more transparent and propositional uses of well now seems to become oriented towards acknowledging a previous speaker turn for the benefit of the addressee, as in (13) and (14). These examples show

166

Tine Defour

that acceptance not necessarily implies actual belief of what is said. Both examples are excerpts from trial proceedings, in which the speaker’s consideration of preceding information is central, and forms the basis for a continuation of the dialogue. The interrogees’ answers are acknowledged, but not recognized as indisputable. (13)

(14)

L. C. J.: Did you lie with them? Dunne: No, my Lord, I did not. L. C. J.: Well, I see thou wilt answer nothing ingenuously, therefore I will trouble my self no more with thee: Go on with your Evidence, Gentlemen. (HC, Cetri3b [Proceedings, trials]; 1640–1710) A: Nay (quoth [the earl]) it is now not tyme to answere vppon thinkinge, did you indeede so counsell me? (B:) he answered: “I did”. A: The E. pausing as it were in a wonder replyed thus. Well, let his lyfe and my death witnes howe truely he speakes. Then was agayne vrged the Consultation at Drewery house,…. (HC, Cetri2a [Proceedings, trials]; 1570–1640)

The positive adverbial sense is thus used strategically as a defence against possible face loss, and allows the speaker to develop a possible divergent argument. Schourup (2001) has called the present-day use of well “epistemic” in the sense that the marker indicates that the speaker looks back and expresses consideration of a previous utterance – “actively taking into account what is already known or assumed” (2001: 1043) and “[granting] what is (though not necessarily approving of it)” (Schourup 2001: 1049). To some extent, this is already noticeable in example (13), where well results from speaker A’s evaluation of the previous speaker turn. The historical corpus data illustrate an evolution where well initially has a propositional significance (that is well; If this is so, then…). This layer of meaning continues to be used in later data, but well also evolves towards a semantically weakened use which can serve to acknowledge the addressee and which is increasingly positioned in the speaker’s subjective perspective and in a personal discourse dialectic.

The semantic-pragmatic development of well

167

4. Adverbial well: Increase in speaker-perspective 4.1. Layers of adverbial meaning Jucker (1997: 99) notes that several senses of the adverb well were already firmly established in Middle English. Although these cannot be called direct, transparent origins of present-day interpersonal functions, they may have served as precursors for later – textual – functions. Jucker lists three meanings (1997: 99–100): (i) in a way appropriate to the facts and circumstances; fittingly, properly (OED2, well adv., 5) (ii) with good reason; naturally; as a natural result or consequence (OED2, well adv., 8a) (iii) clearly, definitely, without any doubt or uncertainty (OED2, well adv., 14a)

Interestingly, as Jucker points out, meaning (i) occurs mainly with verbs of saying or speaking, as in (15)

Ĉæt is wel cweden swa gewritu secqað, þæt… (a900 Cynewulf Christ 547 – Jucker 1997: 99). ‘That is well said, as the book says, that…’

The importance of this co-occurrence lies in the fact that the early textual functions of well – for which these adverbial meanings can be seen as predecessors – were all accompanied by a communicative verb (e.g. “‘Well,’ seyde thys lady, ‘than I aske the hede of…’”), as attested in Jucker’s data and confirmed in the historical material from the HC and CEECS. We can relate this to the fact that the utterance-initial use of “Well,…” can in some cases be paraphrased as “that is well” or “that is well said”, and in that sense functions as an acceptance of the previous speaker turn. It appears that the interpretation of the semantic-pragmatic value of well depends at least partly on the marker’s verbal collocates and on the positional relationship between verb and adverb. Before considering the idea that there may be a correlation between grammatical context (position, verb type) and level of (inter)subjectivity, we need to take into account the fact that adverbial well in itself displays a wide range of meanings – clearly propositional, but also more subjective and even epistemic meanings. The following examples (taken from the Middle English Dictionary Online [MED]) define well as a manner adverb indicating that when some-

168

Tine Defour

thing is “done well”, it is “done in accordance with a good or high standard of conduct or morality” (MED wel, adv. 1.a; example 16 below), “in a way appropriate to the facts or circumstances” (e.g. with verbs of speaking; example 17 below), or in accordance with the standards of an art, a craft or a profession (MED wel, adv. 5a; example 18 below). These propositional meanings are largely contextual. (16)

(17) (18)

Do(n) well, live(n) well (Act/live virtuously, in accordance to God’s will): E.g. For when they doe well or ill the praise or blame will be laid there. (HC, Ceeduc3a, 1640–1710) Speak well (= speak eloquently) Make…well (Make…in accordance with the standards of a craft): E.g. Now also I schall speke of…the Gerneres Joseph that he leet make for to kepe the greynes for the perile of the dere yeres. And thei ben made of ston, full wel made of Masounes craft. (HC, Cmmandev, 1350–1420) ‘…And they are made of stone, full wel made through the skills of a mason.’

Apart from the “standard” manner meaning, i.e. “in a good manner, according to a standard”, propositional well can be used in a more emphatic or intensified use, indicating degree (rather than manner). (19) (20) (21)

So wel she loved clennesse and eke trouthe. (MED wel, adv. 14a – c 1430) (= greatly, devotedly) ‘So greatly she loved cleanliness and also faithfulness.’ Wite thou wel that in the last dayes schal come perilous tymes. (MED wel, adv. 14a – c ?1387) (= assuredly) ‘Be assured that in the last days perilous times shall come.’ Knowen well (= in depth, with great familiarity): E.g. I wot wel thi werkes. (MED wel, adv. 11a – 1400) ‘I know your doings well.’

A third type of meaning for well was already mentioned as one of the established meanings of well in Middle English (Jucker 1997 – see above: meaning iii). This use of well, i.e. “clearly, definitely, without any doubt or uncertainty” (OED2, well adv., 14a) represents a more subjective, epistemic use of the adverb – as illustrated in the following examples (22–24).

The semantic-pragmatic development of well

(22)

(23)

(24)

169

He seeþ turmentoures on euery syde bisette to do hym woo, as he haþ wel deserued. (MED wel, adv. 14a (d) – a1450) (with verbs denoting obligation, deserving: duly, indeed) ‘He sees torturers [placed] on every side to do him harm, as he has well deserved.’ Swich a noble theatre as it was I dar wel seyn in this world ther nas. (MED wel, adv. 14a – c1385) ‘Such a noble theatre as it was, I dare well say there was none (i.e. not a similar one) in this world.’ Well may he be a kyngys son, for he hath many good tacchis. (MED wel, adv. 15a – a1470) (= possibly, likely – intensified sense of likelihood) ‘He may well be a king’s son, for he has many good characteristics.’

The notion of epistemicity is “concerned with knowledge and belief (as opposed to fact)” (Traugott and Dasher 2002: 106). More specifically, epistemic expressions illustrate the speaker’s point of view with regard to the truth of a proposition. Epistemicity can be expressed by means of pragmatic markers (e.g. in fact), adverbs of modality (probably) or modal verbs (You must be tired.). If we consider these three levels of meaning, we see that the degree of speaker-intrusion or subjectivity remains fairly limited in the “clear” propositional uses. In epistemic uses of the adverb however, a greater level of speaker-attitude or interpersonal expression becomes visible. The dividing line between “clear” propositional uses, intensifying forms, and epistemic meanings of well is fairly thin and seems to depend mainly on the accompanying verbs. In order to attest whether there is a correlation between different forms of adverbial well and an increase in (inter)subjectivity, the next sections will discuss two of the most frequently occurring correlations in the historical corpus data, namely the co-occurrence of well with mental verbs and with epistemic modals. 4.2. Semantic field types: Mental verbs As will be shown in this section, adverbial well illustrates a high level of co-occurrence with verbs of cognition in particular. In order to attest to what extent this collocation influences a possible evolution towards in-

170

Tine Defour

creased subjectivity, the correlation between contextual factors (e.g. semantic verb fields; subject forms; positional differences) and levels of (inter)subjectivity in the historical corpus data is discussed here. 4.2.1. Classification Lexical verbs can be categorized according to a number of different semantic fields. In order to establish quantitative and qualitative distinctions between different semantic patterns, all verbs co-occurring with adverbial well in the historical corpus data were classified accordingly. A quantitative and qualitative study of typical – verbal – patterns in which well is used in the corpus data points out that adverbial well (modifying a verb) accounts for up to 60% of all tokens of well in the HC and CEECS. Well appears 602 times as a Verbal Adverb (VAdv) in the CEECS (on a total of 1064 welltokens, i.e. 56.6%) and 1909 times in the HC (on a total of 3088, i.e. 61.8%). The verbs in this category were divided according to the semantic verb field they belong to based on a classification into seven categories taken from Biber et al. (1999: Section 5.1.).4 Illustrations of each category are given below: (i)

Activity:5 e.g. make, buy, work, carry, wear, open. E.g. … he shall now pay your mastership well and suffisauntly (CEECS, Stonor; 1424–1483) ‘… he shall now pay your mastership well and sufficiently’

4. Many lexical verbs can be called ambiguous in the sense that they can be placed into more than one semantic verb category. Verbs such as resist, obey or follow (a command or law), for instance, seemed to fit both into the cognitive and activity verb categories. Moreover, it is possible for one single verb to display different meanings, consequently fitting into different categories (e.g. deserve can be categorized as either cognitive or existential, depending on context; see Biber et al. 1999: 367–371). In order to create a consistent categorisation, all entries were classified on the basis of their individual contexts. 5. Activity verbs are verbs which “primarily denote actions and events that could be associated with choice”, e.g. come, bring, leave, run, take, work (Biber et al. 1999: 361).

The semantic-pragmatic development of well

(ii)

171

Mental:6 e.g. think, love, know, want, believe, read. E.g. Whan Melibee hadde herd the wordes of his wyf Prudence, he seyde thus:“I se wel that the word of Salomon is sooth.” … (HC, Cmctpros; 1350–1420) ‘When Melibee had heard the words of his wife Prudence, he spoke thus: “I see well that the word of Salomon is true.”’

(iii)

Communication: e.g. say, tell, shout, suggest, thank, write. E.g. Yif me gold and oþer fe, þat Y mowe riche be, … for þu ful wel bihetet me þanne I last spak with þe! (HC, Cmhavelo; 1250–1350) ‘Give me gold and other (lit.) property, (so) that I may be rich, … for you full well promised me when I last spoke with you.’

(iv)

Existence: e.g. represent, include, involve, indicate, seem, live. E.g. … and very true it was they did all acknowledge, that her highnes had shewed herself a most loving princesse and neighbour to them, as did well appeer to their embassadors in England … (CEECS, Leycester; 1585–1586)

(v)

Occurrence: e.g. change, grow, develop, occur, become, happen. E.g. … þei takeþ cold water & salt to-geder & waschuþ ouer-al & froteþ him wel. … & so wol þe swellynge aswage wel [enough]. (HC, Cmhorses; 1350–1420) ‘… they take cold water and salt together and wash (him) all over and rub him well … and so the swelling will be well [enough] reduced.’

6. Mental verbs “denote a wide range of activities and states experienced by humans; they do not involve physical action and no not necessarily entail volition. … They include both cognitive meanings (e.g. think or know) and emotional meanings expressing various attitudes or desires (e.g. love, want), together with perception (e.g. see, taste) and receipt of communication (e.g. read, hear)” (Biber et al. 1999: 362–363). Mental verbs include dynamic verbs (examine, discover) but also verbs that are more stative in meaning (e.g. believe, know, remember, understand).

172 (vi)

Tine Defour

Causative: e.g. allow, cause, enable, help, require, let. E.g. His Highnes also well allowed that your Grace noteth not onely remisse dealing, but also some suspitione in that the Lord Dacre so litle estemede the mynde and opinion of the Kings sister … (HC, Ceoffic1; 1500–1570)

(vii)

Aspectual: e.g. start, keep, begin, continue, stop, finish. E.g. … I blles the Lord, that He has ouer-ruled the harts of men, and I hope they goo now on well, to doo that greate worke they haue in hand. (CEECS, Harley; 1625–1666)

According to Scheibman (2002), studying the conventionalization of different patterns and of subjective forms requires us to look at “a range of combinations of grammatical and discursive elements and constructions that appear frequently in conversation” (Scheibman 2002: 60). The historical data from the HC and CEECS (see table 2 – raw figures are added in brackets) show that well is most likely to occur with the semantic field category of cognitive verbs. The HC and CEECS show a correlation with mental verbs amounting to 41.3 and 39.8 % respectively, compared to smaller percentages of correlation with activity verbs (36.4 and 31.4%).7 An additional comparison with present-day data (of verbs collocating with adverbial well) from the British National Corpus (BNC), however, presents a stronger correlation with activity verbs (56.1%) than with mental verbs (21.2).

7. The separate category of Verbal Adjectives (VAdj), although not included in the discussion, can be mentioned as a complementary set of data. VAdjs (e.g. a well-ground powder; They are well pleased with their bishop; …of those wellaffected people) can be considered as Predicative Adjectives, with verbal elements however. As such, they can also be categorized according to verb type. For the HC and CEECS respectively, VAdjs take up 36.5 % (72/197) and 72 % (90/125) of mental verbs – compared to 34 % (67/197) and 21.6 % (27/125) of activity verbs.

The semantic-pragmatic development of well

173

Table 2. Verbs modified by well: Semantic Field distribution in the HC, CEECS and BNC Semantic Field Activity Mental Communication Existence Occurrence Causative Aspectual Indefinite/Implicit Total

HC % 36.4 (695) 41.3 (789) 11.3 (215) 7.1 (135) 3.3 (64) 0.3 (6) 0.1 (2) 0.2 (3) 100 (1909)

CEECS % 31.4 (189) 39.8 (239) 13.5 (81) 9.1 (55) 5.1 (31) 0.2 (1) 0.5 (3) 0.3 (2) 100 (601)

BNC % 56.1 (37) 21.2 (14) 9.1 (6) 7.6 (5) 3 (2) – – 3 (2) 100 (66)

In table 3, a second comparison is presented with more general presentday data (i.e. figures of verb types without well-collocates) from Biber et al. (1999: 365). As in the BNC data, these percentages display the strongest correlation with activity verbs (also see Defour 2005 for a comparison between synchronic and diachronic corpus data). Table 3. General semantic field distribution of verbs in synchronic data (Biber et al. 1999: 365) Semantic Field Activity Mental Communication Existence Occurrence Causative Aspectual Indefinite/Implicit Total

Biber et al. % 49 19 13 8 5 4 3 1018

The fact that historical uses of adverbial well collocate with mental verbs in such high percentages may be connected with the nature of the corpus texts. It is assumed that a higher level of interaction (for one particular genre) will result in a higher use of “subjective” linguistic items. Scheibman states that 8. No raw figures were given in Biber et al. (1999). In total, the percentages do not add up to an even hundred.

174

Tine Defour

“language – in particular, spontaneous conversation – is subjective in that it is fundamentally used by speakers to express their perceptions, feelings, and opinions …” (2002: 61). She establishes a connection between the communication of subjective attitudes and verb types by referring to the fact that, in her conversational data, verbs of cognition are the most frequently occurring verb class, in combination with first person singular subjects. This ties in with a general hypothesis saying that linguistic elements that commonly appear in conversation should be those that participate in subjective expression …. There should also be greater cooccurrence of items whose combinations lend themselves to conveying speaker point of view than those whose combinations do not (e.g. after Benveniste 1971, verbs of cognition would more frequently appear with a first person singular subject than with a third person singular). In other words … there should be associations between commonly occurring conversational material and semantic and pragmatic expression of subjectivity. (Scheibman 2002: 61)

Mental verbs, expressing “opinions, wants, and feelings, and [those] of other people” (Biber et al. 1999: 365), are a suitable medium for communicating perception and cognitive processes and may therefore serve a relevant function in the interactional frame. Biber et al. state that “mental verbs, especially know, think, see, want, and mean, are particularly common in conversation” (Biber et al. 1999: 378), and confirm that mental verbs seem to have a natural correlation with first and second person subject forms. Finally, Fitzmaurice (2004) also stresses that mental and modal verbs in particular are common grammatical resources for the expression of speaker self-expression – as in e.g. you see, you know, or I believe. It should be noted that, although the use of first person subject forms can be relevant for the development of subjective meanings, this does not necessarily indicate an evolution towards increasing subjectivity (cp. Scheibman 2002: 167; Traugott, this volume). The historical corpus data confirm that the mental verbs with the highest frequency, i.e. know, understand and remember,9 occur most frequently

9. In both historical corpora, know has the highest frequency, followed by do. Apart from know, understand and remember, mental verbs like, love and see also appear in the list of most frequent verbs. However, these are verbs of emotion and perception, and the three cognitive verbs with the highest frequency were chosen because they are more likely to reflect the speaker’s cognitive activities.

175

The semantic-pragmatic development of well

with I and you in collocation with well, as can be seen in table 4 below. Raw figures are added in brackets.10 Table 4. Percentages per corpus: Subjects of mental verbs modified by well: a) Know b) Understand c) Remember KNOW CEECS

(73)

HC

(280)

UNDERSTAND CEECS

(20)

HC

(53)

REMEMBER CEECS

(15)

HC

(15)

I % 35.6 (26) 33.9 (95)

You % 32.9 (24) 25.4 (71)

They % 8.2 (6) 8.2 (23)

He % 11 (8) 18.6 (52)

She % 4.1 (3) 2.5 (7)

It % –

We % –

0.7 (2)

2.5 (7)

I % 55 (11) 11.3 (6)

You % 35 (7) 35.8 (19)

They % –

He % –

She % –

7.5 (4)

24.5 (13)

–

I % 46.7 (7) 66.7 (10)

You % 40 (6) 20 (3)

They % –

He % 13.3 (2) 13.3 (3)

–

Rest % 8.2 (6) 8.2 (23)

We % 5 (1) 9.4 (5)

Rest % 5 (1) 11.3 (6)

She % –

We % –

Rest % –

–

–

–

In contrast to cognitive verbs, activity verbs and communicative verbs from the HC and CEECS – such as do (table 5) or greet – present a more varied picture with regard to subject forms. Apart from these quantitative results, which show that historical uses of adverbial well mainly occur with cognitive verbs and first and second person subject forms, a qualitative study of the contexts in which this collocation occurs may attest whether well provides an additional pragmatic value in its co-occurrence with mental verbs and “interactive” subject forms. 10. The large differences between masculine and feminine third person singular forms may be context-dependent or depending on genre differences. Because of the small numbers of occurrence, more contexts would need to be examined in order to make further generalisations.

176

Tine Defour

Table 5. Percentages per corpus: Subjects of activity verbs modified by well: Do DO CEECS

(61)

HC

(102)

I % 8.2 (5) 2.9 (3)

You % 54.1 (33) 29.4 (30)

They % 3.3 (2) 15.7 (16)

He % 18 (11) 22.5 (23)

She % 3.3 (2) 4.9 (5)

It % -

We % -

3.9 (4)

6.9 (7)

Rest % 13.1 (8) 13.7 (14)

4.2.2. Mental collocations Although the meaning of well in phrases such as as you well know or you well know that… is mainly referential, Quirk et al. (1985) state that this type of formula (e.g. as you (well) know; as you may know) does not function as a mere reference to the hearer’s cognition or to his or her knowledge of a fact or event, but rather serves as an appeal to the addressee, indicating that “he or she is not being underestimated” and that it is probable that he or she already knows the facts that are referred to. In that sense, a phrase such as as you know can serve as a help to acknowledge the hearer. Its use can prevent face loss and can keep the argumentational flow of a conversation going by integrating the addressee in the discourse frame. The collocation is not limited to second person subject forms. Examples in which a first person subject is combined with well and a verb of cognition (e.g. I understand well that…) can equally help to stress the speaker’s assertion. The following examples may help to attest what exactly the added interactional value of adverbial well is in a collocation of first or second person subject forms with mental verbs. Examples with a first person subject form will be discussed first (examples 25 and 26). (25)

A: I have norysshed in thys place a grete whyle a serpente whych pleased me much…. And yestirday ye slew hym as he gate hys pray. Sey me for what cause ye slew hym, for the lyon was nat youres. B: Madam, I know well the lyon was nat myne, but for the lyon ys more of jantiller nature than the serpente, therefore I slew hym … (HC, Cmmalory [Romances]; 1420–1500) [adapted] A: ‘For a long time I have nourished a serpent in this place, which pleased me much…. And yesterday you killed it as it was trying

The semantic-pragmatic development of well

(26)

177

to catch its prey. Tell me for what cause you slew it, for the lion was not yours.’ B: ‘Madam, I know well the lion was not mine, but because the lion is of a more noble nature than the serpent, therefore I slew it …’ Her majestie I do remember well indeade, and so may you, howe before all my lords she seamed to mislyke that I should take any other charge then as her generall…. (CEECS, Leycester; 1585– 1586)

It appears that the contexts in which these formulas are used can all be explained in the larger frame of an argumentation. In example (25), for instance, the second speaker has to justify himself for killing a snake – which belonged to the first speaker. The situation poses a possible risk of face loss for speaker A. What we can see is that speaker B therefore starts by acknowledging his interlocutor and her claim (I know well the lyon was nat myne) before putting forward his own point of view. The reflection of the accusation diminishes a possible face-threat and allows the speaker to develop his line of reasoning. The additional use of well may play a role in this, in the sense that well presents an intensification of the speaker’s assertion and seems to express a personal viewpoint of the speaker on the truthvalue of his or her claim. Well as it appears in I know well that… can still be seen as a propositional adverb, modifying a lexical verb (as “in a good manner”), but the adverb seems to come close to a more intensified, epistemic meaning, signifying indeed or certainly. This meaning is even explicitly added in example (26) (I do remember well indeade), indicating the speaker’s judgement on the truth-value of the utterance. In examples (27), (28) and (29), the use of a second person subject form explicitly acknowledges the addressee and similarly occurs in contexts where speaker and addressee have divergent views or where the speaker attempts to bring across his or her personal perspective. At the same time, these examples illustrate that the speaker wants to express a form of politeness. In each of the three excerpts, the writer addresses someone who can be considered socially high-ranked (cp. “your maystership”, i.e. Thomas Stonor; “your Ladyship”, i.e. Lady Jane Cornwallis; “Madam”, i.e. Lady Bacon). The relationship between speaker and addressee will necessarily be affected by this. Positive attention is therefore given to the addressee’s face, by presenting the given information as something that is shared already at the start of the utterance.

178 (27)

(28)

(29)

Tine Defour

…, which as your maystership knoweth well was right shorte warnyng, remembring þat þe more parte of my lordes servauntes were sente into Suffolk … (CEECS, Stonor; 1424–1483) ‘… which, as your mastership knows well, was right short notice, bearing in mind that the majority of my lord’s servants were sent to Suffolk …’ I confesse I am not without some regrette that, eyther by [your Ladyship’s] election or my misfortune, it falls out at such a time when I am not soe much at liberty, as [your Ladyship] well knows, nor soe much master of myself, … (CEECS, Cornwall; 1613–1644) Surely, Madam, there is great reason, you very well know, that you should strayne yourself for the effecting of this mach; for, as I have often expressed… (CEECS; Cornwall; 1613–1644)

A phrase such as as your Ladyship well knows… (28), for instance, urges the addressee to consider the validity of the speaker’s utterance. This ties in with Östman’s suggestions on the use of you know, saying that the speaker “does not indicate by you know that he wants the addressee to accept the truth of his proposition, but he wants the addressee to PRESUPPOSE the tenability of what he is saying” (Östman 1981: 18). While you know well can be used by a speaker to emphasize and attain a sense of common ground, certain examples seem to add a more critical sense to the proposition. In a context where speaker and addressee appear to have diverging expectations or points of view, the difference between I/you know that… and I/you very well know that… can illustrate an underlying implication (made by the speaker) which can be paraphrased as “I am not trying to pretend that I don’t know this” or “don’t (you) try to pretend that you don’t know this”. We can compare this with two phrases in Dutch in which a hearer is addressed, namely the neutral phrase “Je weet dat hij in Gent woont” (“You know he lives in Ghent”) versus “Je weet goed genoeg dat hij in Gent woont” (“You know well (enough) he lives in Ghent” – said when the addressee denies). The latter use of you know allows a speaker to express (presumed) certainty (“as you know”): the speaker invokes a prior agreement as the source for the trustworthiness of what he is saying; and the addressee might be expected to believe the speaker. This use of you know is primarily a speaker-oriented, Face-Saving you know: by using the declarative you know, the speaker does not want to be argued against. (Östman 1981: 18)

The semantic-pragmatic development of well

179

The intensified positive meaning of adverbial well can in this sense be used to anticipate possible diverging opinions – which could undermine the truth-content of the utterance, to validate the utterance of the speaker, and to serve as a basis on which the speaker can tackle diverging opinions without creating face loss for the addressee (“I see what you’re saying. However…”). More so, the speaker creates an assumption of shared knowledge (or at least expresses the hope that the hearer will be brought to the same level of understanding). Although well is not an indispensable element in this mental collocation, the use of the adverb does provide an additional intensifying factor, and seems to play an important role in the positioning of the speaker, as well as in establishing common ground. 4.2.3. Positional shift: You know well that… ~ as you well know The historical corpus data illustrate that the relationship between well and collocating verbs (e.g. with to know) is also subject to differences in position. The positional division used in this paper is taken from Quirk et al. (1985: § 8.15–8.23) and can be summarized as follows. An Initial position (I) places the adverbial before all other clause elements (e.g. “Suddenly, the driver started the engine”). In contrast, an adverbial can follow all other obligatory elements in End Position (E); e.g. “The light was fading rapidly”), or can be placed after Subject and Verb but still followed by an obligatory element, i.e. Initial End position – as in “She kept writing in feverish rage long, violent letters of complaint”. Medial position (M) (Quirk et al. 1985: 492) places the adverb between Subject and (finite lexical) verb (“The driver suddenly started the engine”) or after the first auxiliary or the verb to be (“She hadn’t really delighted her audience”). Two variants of M are the Initial Medial position (iM; “She really hadn’t delighted her audience”) and End Medial (eM; “The room must have been quite carefully searched”). The variation between you understand well, with well in Final position, and you well understand, with well in Medial position, may lead us to wonder whether this difference has a correlated influence on the meaning of the collocation, and whether the position which well has in relation to the main verb can serve as a syntactic criterion for determining pragmatic meaning. Tables 6 and 7 below indicate the historical evolution of the position of well in the HC and CEECS – as a verbal modifier of mental verbs, categorized according to three main time periods.

180

Tine Defour

Table 6. CEECS: know + well: percentages (and raw figures) of the position of well CEECS Initial Medial Final Initial End Total

1400–1500 2.1 (1) 16.7 (8) 39.6 (19) 41.7 (20) 100.1 (48)

1500–1600 – 60 (6) 30 (3) 10 (1) 100 (10)

1600–1680 – 43.75 (7) 25 (4) 31.25 (5) 100 (16)

Table 7. HC: know + well: percentages (and raw figures) of the position of well HC Initial Medial

850 – 950 100 (1) –

950 – 1050 –

End Medial Final

–

100 (1) –

–

–

Initial End Rest

–

–

–

–

100 (1)

100 (1)

Total

1150 – 1250 22.7 (5) 9.1 (2) –

1250 – 1350 29.2 (7) 8.3 (2) –

1420 – 1500 9.6 (8) 8.4 (7) 1.2 (1) 16.9 (14) 63.9 (53) –

1500 – 1570 2.8 (1) 38.9 (14) –

1570 – 1640 –

1640 – 1710 –

42.9 (6) –

21.1 (4) –

29.2 (7) 33.3 (8) –

1350 – 1420 18.75 (15) 11.25 (9) 1.25 (1) 15 (12) 53.75 (43) –

27.3 (6) 36.4 (8) 4.5 (1) 100 (22)

38.9 (14) 19.4 (7) –

28.6 (4) 28.6 (4) –

52.6 (10) 26.3 (5) –

100 (24)

100 (80)

100 (83)

100 (36)

100 (14)

100 (19)

Hoye (1997: 149) noted that Medial position is more usually associated with modality and degree – but whereas we can observe a tentative shift from Final (you know well) towards “intensified” Medial position (you well know) in the CEECS, the HC presents more varied results (table 7) and shows an increase for both Medial and Final position. The distinction between activity verbs and mental verbs, collocating with well in the historical data, made no relevant difference in terms of positional restrictions or evolutions. A comparison with present-day data from the BNC shows that, while the co-occurrence of pragmatic well with know does not undergo a strict positional shift from Final to Medial position, the combination as a whole

The semantic-pragmatic development of well

181

becomes subject to a structural change which can be situated on a broader syntactic level. A data sample from the BNC illustrates that well in collocation with know frequently (43%) and increasingly occurs as a parenthetical remark, introduced by as (e.g. as you well know; as you know right well). The development of such parenthetical remarks has been dealt with quite extensively in the literature. Three hypotheses have been suggested for the grammaticalization of epistemic parentheticals such as I mean, you know, I think or I guess. For the evolution of I think/methinks, for instance, Palander-Collin suggests a development from a following that-clause (I think that he will win) to a that-less clause (I think he will win) and finally to a structure where I think can be postponed as an epistemic parenthetical, after the proposition (He will win, I think) (Palander-Collin 1999: in Brinton 2003: 9). An alternative development has been suggested for I think and I guess and related parentheticals (Brinton 1996), starting from a relative structure (e.g. as/so/which I mean) and evolving towards deletion of the relative pronoun and to “a change in status…from adjunct to disjunct” (Brinton 2003: 10). A third hypothesized evolution – suggested for the development of I mean (Brinton 2003) – entails that I mean initially “governs a phrasal element {NP, VP, AP, PP, AdvP}” (2003: 12). The connection with this element is gradually loosened, and allows for I mean to be postponed and reanalysed as an independent element. For the development of you know well, our historical data show that the combination of you and know is initially followed by a noun phrase (with pronoun: example 30), a that-clause (example 31), or a subordinated interrogative clause (example 32). (30)

(31)

þurrh þatt Godd wass wurrþenn mann forr ure miccle nede, þurrh þatt wass he, þatt witt tu wel, all wiþþ hiss lefe wille niþþredd & wannsedd wunnderrlig … (HC, Cmorm [Homilies]; 1150–1250) [adapted] ‘because God had become man for our great need, through that he was, you know [that] well, humbled and greatly diminished al through his will …’ Þou wost ful wel, yif þu wilt wite, þat Ætelwold þe dide site on knes and sweren on messe-bok … þat þou hise douhter sholdest yelde … (HC, Cmhavelo [Romances]; 1250–1350) ‘You know full well, if you want to know, that Ætelwold made you sit on your knees and swear on the missal … that you would (lit. should) hand over his daughter…’

182 (32)

Tine Defour

“Guode womman,” seide þe holie man: “þou wost wel hou it is, þat þat chief louerd habbe þe beste aygte: …”. (HC, Cmseleg [Biographies, saint’s lives]; 1250–1350) ‘“Good woman”, said the holy man: “you know well how it is, that the highest ruler has the most valuable possession: …”.’

In these early sentences, know well has a concrete referential meaning, and appears to have limited scope. In later periods, we can see an increase in the use of you know well as a parenthetical construction, which may or may not be introduced by as (examples 33–36).11 (33)

(34)

(35)

(36)

… ant sire Iohan Abel, mo y mihte telle by tale, boþe of grete ant of smale, ye knowen suyþe wel. (HC, Cmpoemh [History]; 1250– 1350) ‘… and sir Iohan Abel, I might tell more through story, both of great and of small, you know very well.’ Þu hast a garnement wel iweue adoun to þi foot, in whiche þyn husbounde Crist wil haue gret lykyngge to fynde þe icloþed in. An hemme, as þu wost wel, is þe laste ende of a cloþ … (HC, Cmaelr3 [Rules]; 1350–1420) ‘You have a garment well woven down to your foot, in which your husband Christ will be pleased to see you dressed. A hem, as you know well, is the last end of a piece of cloth …’ It is well done ye remembre hym off them ffor dyverse consederacions, as ye know bothe right well. (HC, Cmpriv [Private letters]; 1420–1500) ‘It is well done (if) you remind him of them for diverse reasons, as you both know right well.’ He that doth procure another Man to commit a Felonie or a Murther, I am sure you know well ynough, the Law doth adjudge the Procurer there, a Felon or a Murtherer. (HC, Cetri1 [Proceedings, trials]; 1500–1570)

These examples illustrate the fact that the combination of a mental verb with well and a first or second person subject form appears to become in11. As can be seen in examples (33) to (36) among others, the combination of know and well is frequently complemented by a premodifying or postmodifying element which lends added strength to the assertion. Examples are you know very well, you know right well, or I know well enough.

The semantic-pragmatic development of well

183

creasingly detached from the internal clause structure, and gradually becomes more fixed as a parenthetical collocation with greater positional freedom. A preliminary comparison with contemporary material from the BNC shows that this construction has become even more fixed in presentday use. However, more research is necessary in order to be able to map exact figures and possible evolutions. The combination of well with mental verbs and with what can be called “interactional” subject forms diachronically forms a variety of different layers of meaning – which range from propositional uses to more intensifying meanings that can be applied when speaker and addressee have diverging points of view and want to find a source of shared understanding. In addition, this is structurally reflected in the co-existence of parenthetical uses with sentence-integrated structures. 4.3. Modal collocations 4.3.1. Well with epistemic modals A second collocation occurring frequently in the historical data from the HC and CEECS is the combination of well with modal auxiliaries, as illustrated in the examples below. (37) (38)

(39) (40)

Mr. Coosin, you may well thinke that I wonder at these proceedings in the College. (CEECS, Cosin, 1617–1669) I cannot but bee extremly troubbled at my one misfortune, in that it appears to you (and I confesse it may verie well appeare so) that I am the worst of children to the best of mothers. (CEECS, Cornwall, 1613–1644) My wit is short, ye may wel understonde. (HC, Cmctvers [Fiction], 1350–1420) ‘My wit is short, you may well understand.’ I fere me he cannot well shew them to your mastership. (CEECS, Stonor, 1424–1483) ‘I fear (that) he cannot well show them to your mastership.’

The frequent appearance of the combination of well with modal auxiliary verbs in the historical data is reminiscent of an Old English example mentioned by Traugott and Dasher (2002), in which the use of well is shown to illustrate the subjectification of adverbial well.

184 (41)

Tine Defour

Cwæð he: Wel þæt swa mæg, forþon hi englice ansyne habbað. ‘He said: Well that may (be) so, because they have angelic faces.’ (?900 Bede, ii.i. [Schipper] 110 [Jucker 1997: 100], in Traugott and Dasher 2002: 175)

Example (41) portrays well as an element which can hardly be seen as a propositional modifier but rather as an element which has evolved towards a more epistemic meaning – which can be paraphrased as ‘indeed’ or ‘certainly’. Placed in utterance-initial position, and as early as the Old English period, well can be attributed with an increased level of epistemic modality,12 which is, according to Hoye’s definition (1997: 42–43), concerned with matters of knowledge or belief on which basis speakers express their judgements about states of affairs, events or actions. [The speaker is] not making statements of fact or categorical assertions but conveying his subjective view of the world.

The combination of well with epistemic may is by far the most frequent modal collocation in the historical material from the HC and CEECS, taking up more than 75% (HC) and 60% (CEECS) respectively. Table 8 presents an overview of the epistemic modals collocating with well in the respective corpora. Table 8. Most frequent modals in collocation with well: percentages per corpus Modals May Might Can Shall Ought (to) Would Birde (= ought to) Total

CEECS 61.1 (22) 22.2 (8) 13.9 (5) 2.8 (1) – – – 100 (36)

HC 75.3 13.4 1 6.2 2.1 1 1 100

(73) (13) (1) (6) (2) (1) (1) (97)

12. Both Goossens (1982) and Bybee (1988) state that “clear epistemic meanings are hard to find out in Middle English” (Goossens 1982: 78). In many cases, the interpretation depends on semantic context – mark, for instance, the ambiguity in “…he must extraordinarily well know paintings”. Because of the ambiguity with regard to modals in historical data, all entries were classified on a case-by-case basis in order to distinguish between tokens illustrating deontic and epistemic modality.

The semantic-pragmatic development of well

185

May well has turned into a fixed idiom in present-day English. Shibasaki states that it is “the most frequently used expression in the modal verb– adverb construction, synchronically and diachronically” (2003: 400). The collocation is invariable and an example of “semantic harmonization”, which entails that may and well are semantically within the same scope of modality, and that their co-occurrence creates a stronger level of epistemicity than the sum of the two individual elements. Hoye (1997: 240) indicates that the combination of well and may not only creates an intensification of the modal, but a transformation of its meaning. Whereas the unmodified modal signifies possibility (example 42), the combination with well conveys probability (43). (42) (43)

it may / might / can / could ‡ be true that he beat her it may / might / can / could well be true that he beat her (Quirk et al. 1985: 588)

The value of well in this type of restricted modal environment (may well) lies in the fact that well signifies a transformation in the epistemic value of the auxiliary, which … alters the status of the speaker’s attitude and commitment towards the ‘known facts’. (Hoye 1997: 210)

Whereas the use of may represents the speaker’s subjective view on the probability or truth-value of a propositional fact or an utterance, the supplementary use of well “additionally lends weight to the force of the speaker’s argument” (Hoye 1997: 144), creating an environment in which not only the contents of the utterance are given a greater truth-value but also the subjective view of the speaker is credited with additional authority. 4.3.2. Position The Old English example mentioned by Traugott and Dasher (example 41) places well in utterance-initial position (Well that may be so). Although this use needs to be distinguished from pragmatic marker-uses in which well is also placed utterance-initially (as in Well, that may be so…), the fact that well has initial position in this example may tell us something about the historical evolution and positional value of well in this modal collocation. Well may…(Initial; e.g. example 44) needs to be differentiated from the related collocational string may well (Medial; e.g. 45).

186 (44)

(45)

Tine Defour

For out of a mans hert es broght al-kins euil, als idell thoght of man-slaghter and avowtri, fornicaciowns and felony …. Wele ge may vmthink gow þan þat þise er þai þat files a man. (HC, Cmnorhom [Homilies]; 1350–1420) ‘For out of a man’s heart is brought all kinds of evil, such as vain thoughts of manslaughter and adultery, fornications and treachery, …. Well you may consider then that these are the things that render a man morally corrupt.’ … layeth a very great obligation upon me to returne you my most thankfull acknowledgment of your speciall kindnes …. It may well be that I am in this particular likewise beholden to Mr. Gayers, of whose generous freedome and bonte I have had divers testimonies heretofore. (CEECS, Cosin; 1617–1669)

Hoye stresses that, in present-day English, the former (with well preceding the modal auxiliary) breaks the unity of meaning that may well has, and “more clearly carries the independent meanings of its constituent items and denotes ‘indeed possible’” (1997: 233). This kind of correlation between syntax and meaning can also be seen in can’t possibly versus possibly…can’t. Both may well and can’t possibly respectively indicate probability and impossibility (negative certainty) and form a collocational unit where the adverb strengthens the sense of (un)certainty expressed by the modal auxiliary. In the case of possibly can’t (as with well may), the adverb adds propositional meaning and changes the content of the collocational string. The positive sense of adverbial well seems intensified when placed in initial position, and in the Old English example (example 41), the position of well provides additional strength to the speaker’s personal assertion and [renders] the speaker’s assertion all the more forceful and … can be used in a manipulative sense ‘to seduce the addressee into believing the content of the proposition’ (Hoye 1997: 213).

Historically, the position of well with regard to the collocating modal auxiliary undergoes an evolution towards an increased use of Medial position. In the material from the CEECS, all 22 tokens of well (in collocation with may) are placed in Medial position (including one eM token in the period 1566–1638). The HC-data (table 9 and figure 1) show a more varied picture.

187

The semantic-pragmatic development of well Table 9. HC: Well + may: positions of well (per period) Period 850–950 950–1050 1150–1250 1250–1350 1350–1420 1420–1500 1500–1570 1570–1640 1640–1710 Total

Total/ period 3 2 9 2 17 19 9 9 3 73

I

M

iM

eM

iE

33.3 (1) – 33.3 (3) 50 (1) 11.8 (2) 10.5 (2) – – – (9)

66.7 (2) 50 (1) 44.4 (4) 50 (1) 82.3 (14) 89.5 (17) 88.9 (8) 100 (9) 100 (3) (59)

– 50 (1) 22.2 (2) – – – – – – (3)

– – – – 5.9 (1) – – – – (1)

– – – – – – 11.1 (1) – – (1)

120 100

I

80

M

60

iM

40

eM

20

iE

85 09 95 50 01 11 050 50 -1 12 250 50 -1 13 350 50 -1 14 420 20 -1 15 500 00 -1 15 570 70 -1 16 640 40 -1 71 0

0

Figure 1. HC: Well + may: positions of well (per period)

We can see that there is a general – although not absolute – tendency for well to occur in Medial position (88.8% of all historical tokens), and that there are increasingly fewer collocations in which well appears utteranceinitially. It must be kept in mind that these tendencies are based on a relatively small number of occurrences, and that percentages therefore indicate general directions rather than straightforward theoretical evolutions. The graph illustrates that well and may evolve towards an increasingly fixed collocation, in which well follows and specifies the modal verb head. The use of adverbial well as it appears in this modal collocation (e.g. as you may well know) shows an evolution towards semantic reduction, a high

188

Tine Defour

level of dependency on the modal verb head – strengthening the epistemic character of well – and a high degree of integration in the clause – which indicates an advanced progression in the process of grammaticalization. These factors are indications that well could be treated as a modal particle (see Hoye 1997: 209). Modal particles (MP) are related to pragmatic markers and work on similar levels. Modal particles and pragmatic markers both have an evaluative and interpersonal significance (e.g. Weydt 1969; Hansen 1998). Although there is no clear consensus about their exact definitions, we can say that modal particles focus more on the speaker’s stance, the speaker’s relationship with the addressee, pragmatic politeness strategies (see Aijmer and Simon-Vandenbergen 2003), background assumptions and the concern of creating shared knowledge in conversation. MPs can be seen as a subset of pragmatic markers and are said to “‘sit in between’ propositional content and interaction-related functions of language” (Waltereit 2005). Hoye states that modal particles help to “characterize and promote the speaker’s version of the world which is then offered to or imposed on others” (1997: 67). In English, modal particles are uncommon and according to most researchers even non-existent. However, in other languages we can find modal particles which behave similarly to well (at least in particular restricted contexts), for instance the German MPs ja and dann (Abraham 1984), the Swedish väl (‘surely’) or ju (‘as you know’) – also see Aijmer and Simon-Vandenbergen 2003 – and vel in Norwegian13 (see Johansson 2006). Johansson refers to examples where well is translated by vel in its function as modal particle and has the meaning of ‘I suppose’. In contrastive research, ju, for instance, is found as a translation of well in the context of an ongoing argumentation. Ja…ju not only lexically shares the positive meaning of well, but is also considered to be an “obviousness particle (‘as you know’) with a rhetorical, argumentative character” (Aijmer and Simon-Vandenbergen 2003: 1140) – which can be compared with our findings that, in certain restricted contexts, well can be used by the speaker to defend a particular viewpoint and at the same time to acknowledge a possible “conflict of interests”. In this historical collocation of well and may it becomes apparent that the propositional adverb well (in a good manner) has evolved to a meaning that is weakened semantically but strengthened pragmatically and epistemically. The use of well as the “satellite” of 13. In initial position, Norwegian vel and English well can both express agreement, disagreement or qualified agreement. Both can be found in dialogues as well as in monologues – representing a person’s thought. Vel, however, covers a more restricted area of use than well (Johansson 2006: 118–121).

The semantic-pragmatic development of well

189

a modal auxiliary such as may allows the speaker to give additional strength to a subjective evaluation and supports the speaker’s authority to claim that something may or may not be the case. 5. Conclusion From a historical perspective, various propositional uses of well can be seen as a source for further semantic-pragmatic development. Placed in utterance-initial position, the inherent positive meaning of well can indicate the acceptance of a situation – and thus serves as a starting-point for the historical development of increasingly pragmatic and (inter)subjective meanings. Diachronic material from the HC and CEECS illustrates that non-initial uses of well can similarly display different levels of delexicalization or intensification, depending on collocating subjects and/or verbs. In collocation with cognitive verbs and a first or second person subject (e.g. as you well know), adverbial well can serve as a means to strengthen the positioning of the speaker and to establish a source of common ground with the addressee in contexts where both interactants have diverging opinions. Syntactically, the combination of well with mental verbs seems to have evolved towards an independent parenthetical structure, with a broader scope (as/which you know well). Well displays an advanced level of delexicalization and syntactic dependency in collocation with modal auxiliaries. In combination with epistemic may in particular, well conveys an increased level of modality and subjectivity, allowing the speaker to give additional strength to his or her opinion on the truth of an utterance (as you may well know). Propositional meanings of well remain visible in later pragmatic and epistemic uses of the marker, even in advanced stages of grammaticalization and (inter)subjectification. Although we cannot establish a straightforward connection between intensified collocations in which well plays a defining role and utterance-initial pragmatic uses of well, we can, however, consider the possibility that these various uses – with different levels of epistemic or subjective strength – are the result of diverging developments. As such, different polysemous uses of well co-exist, but each has a specific semantic-pragmatic context in which they can be applied. As a common denominator, the element of acceptance inherent in adverbial well can be seen to be increasingly “recruited” in contexts where speaker and addressee need a

190

Tine Defour

means to express speaker attitude, and to establish a greater interactional understanding between different interactants. Acknowledgements The research reported on in this article was carried out in the framework of the Project “Grammaticalization and (Inter)Subjectification” (Project Number P6/44 Interuniversity Attraction Poles), funded by the Belgian Science Policy Office. Sources of data BNC CEECS

CED

HC

MED

OED

The British National Corpus. (last acessed 01/04/2009) The Corpus of Early English Correspondence (Sampler) (on cdrom). 1998. International Computer Archive of Modern and Medieval English (ICAME). Compiled by Terttu Nevalainen and Helena Raumolin-Brunberg. Department of English, University of Helsinki. (last accessed 01/04/2009) A Corpus of English Dialogues 1560–1760. 2006. Compiled under the supervision of Merja Kytö (Uppsala University) and Jonathan Culpeper (Lancaster University). (last accessed 01/04/2009) The Diachronic Part of the Helsinki Corpus of English Texts, Third Edition (1996) on cd-rom (International Computer Archive of Modern and Medieval English [ICAME]). Matti Rissanen. Compiled by Merja Kytö. Department of English, University of Helsinki. (last accessed 01/04/2009) The Middle English Dictionary. Hans Kurath, Sherman Kuhn et al. (eds.). 1952–. Ann Arbor, Michigan: Michigan University Press. The Middle English Dictionary Online: (last accessed 01/04/2009) The Oxford English Dictionary. James A. H. Murray, Henry Bradley, William A. Craigie and Charles T. Onions (eds.) 1989 [1961]. Second edition. Oxford: Clarendon Press. The Oxford English Dictionary Online: (last accessed 01/04/2009)

The semantic-pragmatic development of well

191

References Abraham, Werner 1984 De betekenis en de functie van het Nederlandse wel – een vergelijking met het Duits [The meaning and the function of Dutch wel – a comparison with German]. In Studies over Nederlandse Partikels [Studies on Dutch particles]. Antwerp Papers in Linguistics 35, Johan van der Auwera and Willy Vandeweghe (eds.), 17–46. Antwerp: University of Antwerp. Aijmer, Karin 2002 English Discourse Particles: Evidence From a Corpus. Amsterdam/Philadelphia: Benjamins. Aijmer, Karin and Anne-Marie Simon-Vandenbergen 2003 The discourse particle well and its equivalents in Swedish and Dutch. Linguistics 41: 1123–1161. Aijmer, Karin, Ad Foolen and Anne-Marie Simon-Vandenbergen 2006 Pragmatic markers in translation: a methodological proposal. In Approaches to Discourse Particles, Kerstin Fischer (ed.), 101–114. Amsterdam: Elsevier. Auer, Peter 1996 The pre-front field in spoken German and its relevance as a grammaticalization position. Pragmatics 6: 295–322. Benveniste, Emile 1971 Problems in General Linguistics. Trans. by Mary Elizabeth Meek. Coral Cables, FL: University of Miami Press. Original French publication, 1958. Biber, Douglas, Stig Johansson, Geoffrey Leech, Susan Conrad and Edward Finegan 1999 The Longman Grammar of Spoken and Written English. London: Longman. Brinton, Laurel J. 1996 Pragmatic Markers in English: Grammaticalization and Discourse Functions. Berlin/New York: Mouton de Gruyter. Brinton, Laurel J. 2003 I mean: the rise of a pragmatic marker. Paper presented at the Georgetown University Round Table on Languages and Linguistics (GURT), Georgetown, Washington, DC, February 15–17 2003. Bybee, Joan L. 1988 Semantic substance vs. contrast in the development of grammatical meaning. In Proceedings of the fourteenth annual meeting of the Berkeley Linguistics Society, Shelley Axmaker, Annie Jaisser and Helen Singmaster (eds.), 247–264. Berkeley, CA: Berkeley Linguistics Society.

192

Tine Defour

Culpeper, Jonathan and Merja Kytö 1997 Towards a Corpus of Dialogues, 1550–1750. In Language in Time and Space. Studies in Honour of Wolfgang Viereck on the Occasion of His 60th Birthday (Zeitschrift für Dialektologie und Linguistik – Beihefte 97), Heinrich Ramisch and Kenneth Wynne (eds.), 60–73. Stuttgart: Franz Steiner Verlag. Culpeper, Jonathan and Merja Kytö 2000 Data in historical pragmatics. Spoken interaction (re)cast as writing. Journal of Historical Pragmatics 1: 175–199. Defour, Tine 2005 (Inter)subjectification and pragmatic change in the development of well: A corpus-based comparison. Phrasis 46: 37–61. Finell, Anne 1989 Well now and then. (squib) Journal of Pragmatics 13: 653–656. Fitzmaurice, Susan 2004 Subjectivity, intersubjectivity and the historical construction of interlocutor stance: from stance markers to discourse markers. Discourse Studies 6: 427–448. Francis, W. Nelson 1982 Problems of assembling and computerizing large corpora. In Computer Corpora in English Language Research, Stig Johansson (ed.), 7–24. Bergen: Norwegian Computing Centre for the Humanities. Fraser, Bruce 1999 What are discourse markers? Journal of Pragmatics 31: 931–952. Goossens, Louis 1982 On the development of the modals and of the epistemic function in English. In Papers from the fifth ICHL, Ander Ahlqvist (ed.), 74–84. Amsterdam/Philadelphia: Benjamins. Hansen, Maj-Britt Mosegaard 1998 The Function of Discourse Particles. Amsterdam/Philadelphia: John Benjamins. Hopper, Paul J. 1991 On some principles of grammaticization. In Approaches to Grammaticalization, Vol. I, Elizabeth Closs Traugott and Bernd Heine (eds.), 17–35. Amsterdam/Philadelphia: Benjamins. Hoye, Leo 1997 Adverbs and Modality in English. London/NY: Longman Jacobs, Andreas and Andreas H. Jucker 1995 The historical perspective in pragmatics. In Historical Pragmatics, Andreas H. Jucker (ed.), 3–33. Amsterdam/Philadelphia: Benjamins. Johansson, Stig 2006 How well can well be translated? On the English discourse particle well and its correspondences in Norwegian and German. In

The semantic-pragmatic development of well

193

Pragmatic Markers in Contrast, Karin Aijmer and Anne-Marie Simon-Vandenbergen (eds.), 115–137. Amsterdam: Elsevier. Jucker, Andreas H. 1997 The discourse marker well in the history of English. English Language and Linguistics 1: 91–110. Lakoff, Robin 1973 Questionable answers and answerable questions. In Issues in Linguistics, Braj B. Kachru, Robert B. Lees, Yakov Malkiel, Angelina Pietrangeli and Sol Saporta (eds.), 453–467. Urbana: University of Illinois Press. Nevalainen, Terttu and Helena Raumolin-Brunberg 1996 The Corpus of Early English Correspondence. In Sociolinguistics and Language History. Studies based on the Corpus of Early English Correspondence, Terttu Nevalainen and Helena Raumolin-Brunberg (eds.), 39–54. Amsterdam/Atlanta: Rodopi. Östman, Jan-Ola 1981 You Know: A Discourse-Functional Approach. Amsterdam/ Philadelphia: Benjamins Palander-Collin, Mina 1999 Grammaticalization and Social Embedding: I THINK and METHINKS in Middle and Early Modern English. Helsinki: Société Néophilologique. Powell, Mava Jo 1992 The systematic development of correlated interpersonal and metalinguistic uses in stance adverbs. Cognitive Linguistics 3: 75–110. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech and Jan Svartvik 1985 A Comprehensive Grammar of the English Language. London: Longman. Scheibman, Joanne 2002 Point of View and Grammar: Structural Patterns of Subjectivity in American English Conversation. Amsterdam/Philadelphia: Benjamins. Schiffrin, Deborah 1987 Discourse Markers. Cambridge: Cambridge University Press. Schiffrin, Deborah 1992 Anaphoric then: aspectual, textual and epistemic meaning. Linguistics 30: 753–792. Schourup, Lawrence 1999 Discourse Markers. Tutorial overview. Lingua 107: 227–265. Schourup, Lawrence 2001 Rethinking well. Journal of Pragmatics 33: 1026–1060.

194

Tine Defour

Schwenter, Scott A. and Elizabeth Closs Traugott 2000 Invoking scalarity: The development of in fact. Journal of Historical Pragmatics 1: 7–25. Shibasaki, Reijirou 2003 Patterns of semantic harmonization in English: The case of may well. Paper presented at the Berkeley Linguistics Society Annual Meeting, California, USA, February 2003. Svartvik, Jan 1980 Well in conversation. In Studies in English linguistics for Randolph Quirk, Sidney Greenbaum, Geoffrey Leech, and Jan Svartvik (eds.), 167–177. London: Longman. Traugott, Elizabeth Closs 1995a The role of the development of discourse markers in a theory of grammaticalization. Paper presented at ICHL XII, Manchester 1995. Version of 12/1997. Traugott, Elizabeth Closs 1995b Subjectification in grammaticalisation. In Subjectivity and Subjectivisation, Dieter Stein and Susan Wright (eds.), 31–54. Cambridge: Cambridge University Press. Traugott, Elizabeth Closs 1997 The rhetoric of counter-expectation in semantic change: A study in subjectification. In Historical Semantics and Cognition, Andreas Blank and Peter Koch (eds.), 177–196. Berlin/New York: Mouton de Gruyter. Traugott, Elizabeth Closs 1999 From subjectification to intersubjectification. Paper presented at the Workshop of Historical Pragmatics, Fourteenth International Conference on Historical Linguistics, Vancouver, Canada, July 1999. Traugott, Elizabeth Closs This vol. (Inter)subjectivity and (inter)subjectification: A reassessment. In this volume, 29–71. Traugott, Elizabeth Closs and Richard B. Dasher 2002 Regularity in Semantic Change. Cambridge: Cambridge University Press. Van Herreweghe, Mieke 2003 Well as pragmatic marker in Old English. Paper presented at the 3rd Workshop of the Research Community on Interpersonal and Ideational Grammar: Grammar and Discourse, Ghent University, 27–29 November 2003. Waltereit, Richard 2005 On distinguishing modal particles and discourse particles. Paper presented at the 9th International Pragmatics Conference, Italy, 10– 15 July 2005.

The semantic-pragmatic development of well

195

Watts, Richard J. 1986 Relevance in conversational moves: a reappraisal of ‘well’. Studia Anglica Posnaniensia 19: 37–59. Weydt, Harald 1969 Abtönungspartikel: Die deutschen Modalwörter und ihre französischen Entsprechungen. Bad Homburg: Gehlen.

Paths in the development of elaborative discourse markers: Evidence from Spanish Teresa Fanego Abstract This chapter examines the development of the Spanish manner adverbial de hecho ‘in practice’ into a DM roughly synonymous with English in fact and indeed, and addresses two main research questions, namely: a) whether the history of Spanish de hecho can help to confirm some of the hypotheses put forward by Traugott regarding the trajectories followed by DMs cross-linguistically (cf., among others, Traugott 1982, 1995, 2003; Traugott and Dasher 2002); and b) whether there is any morphosyntactic evidence correlating with the semantic-pragmatic changes undergone by de hecho which might justify an analysis in terms of grammaticalization. The analysis shows that, as expected, the historical development of de hecho is analogous to that of DMs like in fact, indeed, or actually, all of which started as VPAdvs and evolved into DMs via a sentence adverbial stage. With respect to the second research question, the evidence from two large corpora of historical and contemporary Spanish indicates that, when de hecho evolves from a VPAdv into a SAdv, the shift has both semantic and morphosyntactic consequences, such as desemanticization, context generalization (use in new contexts) and decategorialization (loss in the morphosyntactic properties characteristic of the source forms), all of which are indeed indicative of increased grammaticalization. By contrast, there are no grounds to consider the further development of the SAdv de hecho into a DM as also falling within the field of grammaticalization, since there is no evidence of the structural and morphosyntactic changes which are criterial for the definition of that process, as this is understood in Traugott’s work generally.

1. Introduction Since the 1980s, grammaticalization has taken up an important place in linguistic research, and the term grammaticalization has come to be applied to a very large number of linguistic changes which include, apart from prototypical instances of grammaticalization such as the development of function words from content words, other more peripheral or even controversial

198

Teresa Fanego

cases, such as changes in word order (Heine, Claudi and Hünnemeyer 1991: 25; Haspelmath 2004: 26, 38), the development of discourse markers (DMs), or categorial changes like those undergone by the English toinfinitive (Los 2004, 2005) and gerund (Tabor and Traugott 1998: 240– 244; Fanego 2004: 45–49), two forms which started as verbal nouns of action and evolved into parts of the verb system. The focus of this paper1 is on the development of the Spanish manner adverbial de hecho ‘in practice’ into a DM roughly synonymous with English in fact and indeed. The view that the historical development of DMs has similarities with the developments taking place within the domain of grammaticalization was first put forward by Elizabeth Traugott in her ground-breaking article on the regularity of semantic change (1982) and was explored in greater detail in Traugott (1995, 2003), Tabor and Traugott (1998), Schwenter and Traugott (2000) and Traugott and Dasher (2002). In the two decades since Traugott’s initial work, research on DMs from the perspective of grammaticalization has multiplied; witness Jucker (1995), Brinton (1996), Onodera (2004), and Hansen and Rossari (2005), among many others. Yet with exceptions such as Garachana Camarero (1998), Pons Bordería and Ruiz Gurillo (2001) or Company Company (2004, 2006, 2008), the diachrony of Spanish DMs has attracted very little attention to date. One of the aims of this paper, therefore, is to examine the history of a Spanish DM whose English cognate, in fact, has been studied from both the synchronic (Oh 2000; Smith and Jucker 2000; Aijmer and SimonVandenbergen 2004) and diachronic (Schwenter and Traugott 2000) perspectives; synchronic descriptions of French en fait/de fait and Italian infatti are also available (cf. Danjou-Flax 1980; Roulet 1987; Rossari 1992; Iordanskaja and Mel’þuk 1995; Brutti 1999). A second aim is to check wheth1. Some of the material in this paper was originally prepared for a plenary lecture presented at the International Conference From ideational to interpersonal: Perspectives from grammaticalization (FITIGRA), held at the University of Leuven in February 2005; I would like to thank the participants in the conference for helpful discussion and the conference organizers for kindly inviting me to speak at such an intellectually rewarding event. Thanks are also due to Carla Dechant, Bruce Fraser, Tomás Jiménez Juliá, Salvador Pons Bordería, Elena Rivas, María José Rodríguez Espiñeira, Scott Schwenter, Victoria Vázquez Rozas, and the editors and anonymous reviewers for this volume. Last but not least, I am grateful for generous financial support to the Spanish Ministry of Education and Science and the European Regional Development Fund (grant HUM2007–60706/FILO), and the Autonomus Government of Galicia (INCITE grant 08PXIB204016PR).

Paths in the development of elaborative discourse markers

199

er the history of Spanish de hecho can help to confirm some of the hypotheses put forward by Traugott and her associates regarding the trajectories followed by DMs cross-linguistically. The paper is organized as follows. Section 2 defines DMs; Sections 3 and 4 give an overview of research on DMs from the point of view of semantic change and grammaticalization respectively; Section 5 introduces the research questions that this study intends to answer and examines the functions and development of Spanish de hecho; Section 6 closes the paper. 2. Discourse markers defined It is notorious that DMs constitute a very heterogeneous group rather than a well defined word class, as is clear from the great number of different definitions and descriptive terms found in the literature, such as discourse particles (Schourup [1982] 1985; Zwicky 1985; Fischer 2000, 2006; Aijmer 2002), discourse markers (Schiffrin 1987; Blakemore 2002), discourse connectives (Blakemore 1987: 104ff) or pragmatic markers (Brinton 1996). In this paper the label discourse marker will be understood in the restrictive sense it has in authors such as Fraser and Malamud-Makowski (1996), Traugott (1995, 2002: 154ff) or Fleischman and Yaguello (2004: 143), rather than in the broader sense initially proposed by Schiffrin (1987); in this restrictive use the term DM refers to a subset of a far larger set of pragmatic markers2 and comprises expressions whose function necessarily involves a relationship between two segments of discourse: [DMs] are usually in initial position, although medial and final position are possible for many of them, and they signal how the utterance following (U2) is to be interpreted, given the first utterance (U1). … The meaning of a discourse marker is procedural rather than representational, which means that it provides instructions to the hearer about how to interpret U2 rather than designating a specific concept. (Fraser and Malamud-Makowski 1996: 864– 865)

As Traugott and Dasher (2002: 152) put it, DMs are subjectified expressions

2. Among Hispanic linguists, a similar distinction between pragmatic markers in general and a smaller subset of DMs or conectores (‘connectives’) has been made most explicitly by Fuentes Rodríguez (1994, 2001, 2003).

200

Teresa Fanego

signal[ling] an aspect of the speaker’s rhetorical stance toward what he or she is saying, or toward the addressee’s role in the discourse situation. They have little conceptual semantics, and do not contribute significantly to the truth-conditional meaning of propositions … they mark the speaker’s view of the sequential relationship between units of discourse, that is, they serve as connectives between utterances.

Some examples of DMs, all taken from Traugott’s work, include so in some of its meanings (e.g. “So, what’s for lunch?”); anyway as used in “anyway, probably the enemy is surviving on birds and squirrels”, where it is used to signal a return to a prior topic (cf. Tabor and Traugott 1998: 255; Traugott and Dasher 2002: 95); or indeed as used in (1), where it signals that what follows “is not only in agreement with what precedes, but is additional evidence being brought to bear on the argument” (Traugott and Dasher 2002: 164): (1)

1630 Taylor, Penniless Pilgrimage, 131.C1 (Traugott and Dasher 2002: 165): any a one that is not well, comes farre and neere in hope to be made well: indeed I did heare that it had done much good, and that it hath a rare operation to expel or kill diuers maladies.

The essential property of DMs, as defined above, is therefore that they have primarily procedural meanings (cf. Blakemore 1987, 2002).3 In this respect, they are to be distinguished from sentence adverbials (SAdvs), which include, among others, modality or epistemic adverbials (probably, certainly), evaluatives (fortunately, regrettably), and speech act adverbials (frankly, briefly). Unlike DMs, sentence adverbials have conceptual meaning, in the sense of Blakemore (1987, 2002); witness (2): (2)

A: Sinceramente, es una buena película. B: Eso no lo dices sinceramente, lo dices porque sabes que a mí me gusta. A: ‘Sincerely, that’s a good movie.’

3. An anonymous reviewer wonders whether I hold “that DMs have purely procedural meaning and absolutely no conceptual meaning”. It is apparent throughout this chapter, however, that I adhere to the view that grammaticalized items, whether DMs, SAdvs or other, exhibit what Hopper and Traugott ([1993] 2003: 49) refer to as layering, i.e. the presence of different layers of meaning at the same time.

Paths in the development of elaborative discourse markers

201

B: ‘You are not saying that sincerely, you are saying it because you know that I like it.’ The Spanish speech act adverb sinceramente does not impose additional truth conditions on its own clause, but its meaning is conceptual, rather than procedural, as shown by the fact that it is possible to negate that one is speaking “sincerely” (cf. Wilson and Sperber 1993; Martín Zorraquino and Portolés Lázaro 1999: 4058–4059).4 Apart from the semantic difference just mentioned, two additional differences between DMs and sentence adverbials have sometimes been pointed out in the relevant literature, one pertaining to morphosyntax, the other to intonation; they will be discussed in Section 4 below. 3. Cross-linguistic regularities in the semantic development of DMs In her seminal paper on the regularity of semantic change, Traugott (1982) put forward the hypothesis that there is a strong cross-linguistic tendency for semantic-pragmatic change to proceed along the path in (3), but not in the reverse direction: (3)

propositional > (textual >) expressive

This cline, which was based on Halliday and Hasan’s (1976: 26ff) proposal that there are three functional-semantic components or dimensions of language, namely the ideational, textual and interpersonal (= Traugott’s expressive) dimensions, has given way in Traugott’s more recent work to a more complex model of semantic change, labelled the Invited Inferencing Theory of Semantic Change (IITSC; see in particular Traugott and Dasher 2002: 34ff, also Traugott this volume: 54ff). This assumes several corre-

4. The contrast between conceptual and procedural meanings is analogous to the distinction established by Sweetser (1990) between interpretation in the content domain and interpretation in the speech-act domain, as illustrated in the following examples of causal constructions: (i) Since John wasn’t there, we decided to leave a note for him. (His absence caused our decision in the real world.) (ii) Since we’re on the subject, when was George Washington born? (The fact that we’re on the subject enables my act of asking the question.)

202

Teresa Fanego

lated paths of unidirectionality (see Table 1) and views pragmatic implicatures as playing a crucial bridging role in semantic change. Table 1. Correlated paths of directionality in semantic change (Traugott and Dasher 2002: 40) truth-conditional 5

>

non-truth-conditional

content

> content/procedural

> procedural

scope within proposition

> scope over proposition

> scope over discourse

nonsubjective

> subjective

> intersubjective

Traugott and Dasher illustrate these recurring patterns of change by examining changes in four different semantic domains, namely modal verbs, DMs, performative verbs, and social deictics. It is hypothesized that the regularities observed in these four domains “are major tendencies … available for any kind of change that involves semantic development of at least those lexemes that are recruited to the target domains” (2002: 282) in question. With regard to DMs in particular, Traugott and Dasher (see also Traugott 1995, 2003; Tabor and Traugott 1998: 254ff; Schwenter and Traugott 2000) note that many of them have their source in VP adverbials which evolve into DMs, often via a sentence adverbial stage in which the adverbial has an adversative meaning or function. More specifically, based on the analysis of Japanese sate ‘so, well’ and of the English markers indeed, in fact, actually and well, Traugott and Dasher (2002: 187, 281) propose that the development of DMs takes place along the various stages of development reflected in Table 2, each stage involving increased subjectivity and, ultimately, intersubjectivity, in the sense that the expressions in question gradually develop a semantic or pragmatic meaning primarily indexing speaker attittude or viewpoint (subjectivity) and speaker’s attention to addressee self-image (intersubjectivity).

5. Following Sweetser (1990), Traugott and Dasher (2002: 10) use the labels “content meaning” and “contentful meaning” for Blakemore’s (1987, 2002) “conceptual meaning”, as defined in Section 2 above.

Paths in the development of elaborative discourse markers

203

Table 2. Correlated paths of directionality in the development of DMs (Traugott and Dasher 2002: 187) ADVmanner

> ADVadversative

content

> content/procedural

scope within proposition

> scope over proposition

nonsubjective

> subjective

> ADVelaboration

> ADVhedge6 > procedural

> scope over discourse > intersubjective

Thus, in the case of English indeed, in fact and actually, three elaborative7 DMs closely related to the Spanish adverbial which is the concern of this paper, Traugott and Dasher (2002: 157–175) show that all three items started as clause-internal adverbials, came then to function as epistemic sentence adverbials, and ended up as DMs, though indeed (< Middle English in dede ‘in action, in practice’) developed DM uses around 1600, some two hundred years earlier than the other two items (cf. Traugott and Dasher 2002: 171). These various stages are outlined in the following paragraphs. A) Stage I: indeed1 / in fact1 / actually1 = VP adverbials of respect or manner meaning ‘in action, in practice’. Certain discourse contexts could invite the inference that the event was observable, and the fallacy “seeing is believing”, which draws on the view that what is physically/empirically accessible is true, allowed the adverbials to be endowed with evidential (epistemic) modal meanings: ‘in action/practice’ > ‘in actuality, certainly’ (Traugott and Dasher 2002: 160). E.g.:

6. An anonymous reviewer enquires about the term “adverbial hedge” used by Traugott and Dasher and how it differs from the term DM. Though the distinction is by no means clear-cut, hedges are used to convey imprecision or approximation; they encode the speaker/writer’s anticipation of the hearer/reader’s expectations and are aimed at reducing the hearer’s belief in what has just been said. In this sense, they are markers of intersubjectivity. An example would be the use of well in the sentence Yeah, we allow dogs in here; well you’ve managed to get one in anyway. For discussion of hedges see for instance Fleischman and Yaguello (2004: 134) and Lewis (2006: 46), from whom the example is taken. 7. The terms “elaboration” and “elaborative” go back to Halliday ([1985] 1994: 324). In fact and related DMs are members of the semantic field of expectation, and the label “DMs of expectation” is also often applied to them.

204 (4)

(5)

Teresa Fanego

c. 1300 Fox and Wolf, p. 34 (Traugott and Dasher 2002: 159): “Vuolf,” quod þe vox him þo, “All þat þou hauest her before I-do, In þohut, in speche, and in dede, In euche oþeres kunnes quede, Ich þe forвeue.” ‘“Wolf,” said Fox to him then, “All that you have done before this, in thought, in speech, and in action, in evil of every other kind, I forgive you for it.”’ 1690 Locke, Ess. Hum. Und., Bk. 2, Ch. 21, Sec. 36 (Schwenter and Traugott 2000: 15): If we enquire into the reason of what experience makes so evident in fact, and examine why …

B) Stage II: indeed2 / in fact2 / actually2 = epistemic adverbials with sentential scope often “serving a polyphonic adversative function” (Traugott and Dasher 2002: 171) and expressing commitment to the truth of the proposition (i.e. “It is true that…”, “the truth is that…”). E.g.: (6)

(7)

1531 Elyot, The boke named the gouernour, p. 21 (Traugott and Dasher 2002: 163): [teachers] somtyme purposely suffring [‘allowing’] the more noble children to vainquysshe [‘succeed, get the prizes’], and, as it were, gyuyng to them place and soueraintie, thoughe in dede the inferiour chyldren haue more lernyng. 1739–40 Hume, Treat. Hum. Und. Foot. 75, p. 510 (Schwenter and Traugott 2000: 18): since the transition is in that case very easy from the small object to the great one, and should connect them together in the closest manner. But in fact the case in always found to be otherwise.

C) Stage III: indeed3 / in fact3 / actually3 = DMs with an elaborative function. At this stage the prime function of the adverbials is to signal additivity: what follows is not only in agreement with what precedes, but is additional evidence being brought to bear on the argument. (8)

1630 Taylor Penniless Pilgrimage, p. 131.C1 (Traugott and Dasher 2002: 165): any a one that is not well, comes farre and neere in hope to be made well: indeed I did heare that it had done much good, and that it hath a rare operation to expel or kill diuers maladies.

Paths in the development of elaborative discourse markers

(9)

205

1886 Sidgwick, Outlines of the History of Ethics, chapter 1, p. 5 (Traugott and Dasher 2002: 168): Thus in various ways ethical questions lead inevitably to psychological discussions; in fact, we may say that all important ethical notions are also psychological.

We can conclude this brief outline of Traugott’s approach to semantic change by drawing attention to its implications for current hypotheses about language change in general. Unlike generative models, which view language change as triggered by child language acquisition (see e.g. Lightfoot 1991), Traugott’s model assumes that many linguistic changes are not initiated by children, but rather by teens and adults, because of the complex inferences involved and the discourse functions in structuring text: the model of acquisition assumed is one that does not privilege young children, but rather young adults … and, especially in earlier times, those carrying the authority of education, law, or political or ecclesiastical and educational power. However, they may well be spread by children and younger adults, who pick up and extend innovations. (Traugott and Dasher 2002: 41–42; see also Traugott this volume: 55–56) To have a full understanding of the changes discussed here and their impact on language-users, it will be important to know, among other things, in which text types particular changes are favored, and among which groups of people. For example, it has been suggested (Macaulay 1995) that English speakers in authoritative position, or those who position themselves as having authority, may favor stronger epistemic adverbials (e.g. in fact, surely) than others in less authoritative position. (Traugott and Dasher 2002: 283– 284)

4. The development of DMs in relation to grammaticalization In addition to the attention she has given to DMs within her IITSC model, Traugott has also looked at DMs from the point of view of grammaticalization studies. In her (1995) address to the 12th International Conference on Historical Linguistics (Manchester), as also in several later publications (Tabor and Traugott 1998; Traugott 2003; Brinton and Traugott 2005: 136– 140), she has put forward the question of whether the development of DMs could or could not be included within the domain of grammaticalization. She aptly notes that SAdvs and DMs appear to undergo several of the morphosyntactic and semantic changes associated with grammaticalization (as outlined, for instance, in Heine, Claudi and Hünnemeyer 1991 or Lehmann

206

Teresa Fanego

[1982] 2002: 108ff), such as decategorialization (loss in morphosyntactic properties characteristic of the source forms) or shift from more to less referential meanings. They do not, however, exhibit one of the syntactic parameters identified by Lehmann as also criterial for grammaticalization, namely condensation in structural scope, that is condensation in the extent of the construction which they enter or help to form, since both SAdvs and DMs have wider scope and relate to larger stretches of discourse than the VP internal adverbials which often are their sources, as becomes evident if one compares the VP adverbial in fact with the adversative SAdv (cf. 11) and the elaborative DM (cf. 12) that have evolved from it: (10) (11)

(12)

United Airlines Magazine, May 1997 (Tabor and Traugott 1998: 11): Humanity is absent in fact but everywhere present in feeling. 1739–40 Hume, Treat. Hum. Und. Foot. 75, p. 510 (Schwenter and Traugott 2000: 18): since the transition is in that case very easy from the small object to the great one, and should connect them together in the closest manner. But in fact the case in always found to be otherwise. 1886 Sidgwick, Outlines of the History of Ethics, chapter 1, p. 5 (Traugott and Dasher 2002: 168): Thus in various ways ethical questions lead inevitably to psychological discussions; in fact, we may say that all important ethical notions are also psychological.

Tabor and Traugott (1998) and Traugott (2003; also this volume: 41) therefore challenge Lehmann’s notion of decrease in structural scope as criterial for grammaticalization, and propose instead that in some change episodes, such as those involving the development of SAdvs and DMs from earlier VP adverbs, grammaticalization may involve increase in structural scope, rather than decrease.8 From this perspective, therefore, it is appropriate to consider the development of DMs and SAdvs from other parts of speech as

8. Since my concern in this paper is with the question of whether there exist structural differences between SAdvs and DMs, as claimed by Traugott in many places and as more fully discussed in the next paragraphs, I will not go here into the problems posed by Tabor and Traugott’s definition of scope, which they characterize in terms of the generative notion of C-command. As Fischer (2007: 313) aptly notes, a definition of scope in such formal terms is “bound up with all kinds of theory-internal motivations”, and may therefore not be the best tool to measure scope by in research on grammaticalization.

Paths in the development of elaborative discourse markers

207

“the legitimate object of study in terms of grammaticalization” (Traugott 2003: 643). A different issue is how exactly SAdvs and DMs differ from each other. As noted above, both SAdvs and DMs exhibit increase in structural scope; they differ semantically, however, in that the former encode primarily conceptual meanings and the latter procedural meanings (see Section 2 above), but this criterion alone does not seem to be enough to argue that they involve different degrees of grammaticalization, since grammaticalization is a process generally understood to imply not just semantic/pragmatic changes but also structural and morphosyntactic ones.9 Witness in this respect various definitions of grammaticalization by Traugott and her colleagues: In sum, we take the limits of the field of inquiry [of grammaticalization] to be gradual morphosyntactic and semantic change which results in grammatical reanalysis. (Tabor and Traugott 1998: 236; emphasis added) Grammaticalization is in essence a morphosyntactic phenomenon, most crucially the development of functional categories (auxiliary, case, preposition, subordinator, etc.) out of constructions including lexical categories (main verb, nominal in adposition, etc.); it also involves intraconstructional fusion. (Traugott and Dasher 2002: 283; emphasis added)

In order for SAdvs and DMs to be recognized as separate formations, each worthy of consideration from the grammaticalization perspective, it should be possible, therefore, to identify in them distinctive structural properties. That such properties can be found has been claimed by Traugott: most analyses of adverbials and pragmatic markers conflate the IPAdv [= SAdv] and DM functions. For example, Biber and Finegan in their seminal (1988) paper on “stance adverbs” conflate them as IPAdvs, Fraser (1988, 1990) conflates them as DMs. However, they are different syntactically, semantically, pragmatically, and intonationally. (Traugott 1995: 6)

Statements to the same effect can be read in Tabor and Traugott (1998: 253–257), Schwenter and Traugott (2000: 13, 21), Traugott and Dasher (2002: 158–159) and Traugott (2003: 639–640). More specifically, Traugott notes two main structural differences between SAdvs and DMs, one 9. An apparent exception to this prevailing view is Himmelmann (2004: 33). In a rather confused account of the differences between lexicalization and grammaticalization, he argues that semantic/pragmatic changes are “the core defining feature of grammaticization processes” and that “often, but not necessarily” it will be possible to show that they are accompanied by syntactic changes.

208

Teresa Fanego

having to do with morphosyntax, the other with intonation. From the point of view of morphosyntax, based on work by Kiparsky (1995) on ProtoGermanic and Aissen (1992) on Mayan, Traugott claims that in generative terms SAdvs like perhaps or adversative in fact and indeed are sisters of IP (Inflection Phrase), whereas DMs occupy an outer position and “have syntactic properties in common with the left-most ‘E-node’ posited for … external topics in Mayan”10 (Tabor and Traugott 1998: 256): (13)

E PP

CP IP AdvP

In fact3

IP

perhaps

To illustrate these differences, Traugott draws attention to examples like (14), where the SAdv indeed “is found in clause-initial postComplementizer position as a contrastive adverb”, and (15), where the elaborative DM indeed occurs “in clause-initial, pre-Complementizer position” (Traugott 2003: 640): (14)

(15)

1531 Elyot, The boke named the gouernour, p. 21: [teachers] somtyme purposely suffring [‘allowing’] the more noble children to vainquysshe [‘succeed, get the prizes’], and, as it were, gyuyng to them place and soueraintie, thoughe in dede the inferiour chyldren haue more lernyng. 1665 Hooke, Micrographia, 135: 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.

10. Aissen’s E-node stands for “Expression-node”, following Banfield (1973).

Paths in the development of elaborative discourse markers

209

As is clear from (15), Traugott adopts the X-bar analysis of phrase structure which assumes that main, non-dependent clauses contain an empty Complementizer constituent. Examples which reflected more clearly the different structural positions occupied by SAdvs and DMs would have been desirable, and the fact that Traugott does not adduce them, neither for indeed itself nor for its close synonym in fact, suggests that none could be found in her corpus material. On the other hand, she has acknowledged in other places that “the syntactic positions of in fact2 and in fact3 overlap” (Schwenter and Traugott 2000: 13), and recent analyses of in fact in Present-day English (Oh 2000; Aijmer and Simon-Vandenbergen 2004) show its great positional mobility as both a SAdv and a DM, and also its very frequent occurrence in medial position when used in DM function. More specifically, Aijmer and Simon-Vandenbergen raise doubts concerning Schwenter and Traugott’s analysis of in fact2 (the adversative adverb) and in fact3 (the DM) as “distinct polysemies, rather than contextuallybound uses of a monosemous lexical item” (2000: 21); they argue instead that from a synchronic point of view in fact2 and in fact3 can best be seen “as pragmatic implicatures which are conventionalised to a greater or lesser extent, as some contextual meanings are more frequent and more conventionalised than others” (Aijmer and Simon-Vandenbergen 2004: 1788). All this suggests, I believe, that unless there exists clear supporting evidence, one should be wary of claiming that all SAdvs and DMs differ in terms of syntactic scope and distribution, though it is true that in the case of some DMs deriving from earlier SAdvs this claim appears to be justified, as happens with the topic-resuming DM anyway versus the concessive SAdv anyway ‘nonetheless’. These two uses can be kept apart not only semantically and pragmatically but also positionally (see Ferrara 1997): the DM is always sentence-initial, whereas the SAdv appears to be restricted to the right periphery of the VP; witness (16–17): (16) (17)

They got up early. That’s rare for them. Anyway, they left at noon. It was ugly but he wanted to buy the dog anyway.

The fact that anyway, in fact or indeed do not behave alike syntactically need not be considered surprising if we recall that, as was pointed out in the opening lines of Section 2, DMs constitute a very heterogeneous class, not only in terms of their sources – many English DMs, such as yes, oh, ok, as you know, I mean, hark ye, etc., do not originate in other adverbials – but also in terms of their functional properties.

210

Teresa Fanego

Turning now to the question of whether there exist prosodic differences between DMs and the corresponding SAdvs, Traugott (1995: 6) and Tabor and Traugott (1998: 255) rely on Ferrara’s above-mentioned analysis of anyway (1997: 356), which shows that in Texas English this adverb carries a special intonation contour when used as a DM, to argue that SAdvs and DMs differ intonationally; see in this connection the quotation from Traugott (1995: 6) cited earlier in this section. In more recent work, however, Traugott appears to have changed her views, to judge at least from the following observation: Out of context, the written form In fact, humanity is usually absent is ambiguous: it could evoke either adversativity to or an elaboration of something that preceded. In speech, in fact in both meanings [i.e. SAdv and DM] may have a typical disjunct intonation with a sharp rise and fall (Traugott and Dasher 2002: 158).

In addition, it should be pointed out that Wichmann, Simon-Vandenbergen and Aijmer (this volume) have reached inconclusive results regarding the extent to which the different prosodic realizations of English of course correlate with its uses as epistemic adverbial and DM. Thus they note (p. 144) that “despite the evidence for a relationship between stress and semantic weight, … there are other parameters, e.g. information status and pragmatic function, that make it impossible to expect a one-to-one relationship between prosodic realization and any one parameter”. 5. Spanish de hecho: historical development and present-day usage 5.1. Research questions Bearing in mind the various issues mentioned in Sections 3 and 4 with regard to the cross-linguistic development of DMs and their relation to grammaticalization, the present study attempts to answer the following research questions: (i)

Is the development of de hecho ‘in fact, indeed, actually’ analogous to that of English in fact and other related DMs? Which functions, if any, do they have in common? (ii) Is there any morphosyntactic evidence correlating with the semanticpragmatic changes undergone by de hecho which might justify an analysis in terms of grammaticalization?

Paths in the development of elaborative discourse markers

211

(iii) What kinds of text-types favour the use of de hecho? What is the role of generic conventions in change? Recall in connection with this Traugott and Dasher’s observations (2002: 41–42), quoted at the end of Section 3 above, on the role played in language change not only by children but also by teens and adults.

5.2. The corpus The diachronic analysis of de hecho is based on data from the Real Academia Española’s CORDE (“Corpus Diacrónico del Español” [‘Diachronic Corpus of Spanish’]). This database is accessible at http://www.rae.es and consists of approximately 196 million words of European Spanish covering from c. 1200 till 1975; eighty-five per cent of this material is prose and the rest verse. The synchronic data were retrieved from CREA (“Corpus de Referencia del Español Actual” [‘Reference Corpus of Present-day Spanish]’); this too is accessible at http://www.rae.es and comprises both written (85 million words) and spoken (4.5 million words) European Spanish dating from the period 1975–2000. Varieties of Spanish used outside Spain are also extensively represented in both CORDE and CREA, but have not been included in the analysis. Throughout the rest of the discussion the ascription of linguistic examples to a given text type is based on the classifications provided by CORDE and CREA. 5.3. De hecho in Present-day Spanish11 Though synchronic approaches to Spanish pragmatic markers are very numerous, not much has been written on de hecho, apart from a few passing comments in Martín Zorraquino and Portolés Lázaro (1999: 4141–4142) 11. Though there is no complete agreement among scholars as to when exactly one period in the history of Spanish ends and another begins, I will here adopt the more or less traditional divisions of Medieval Spanish (up to 1492, coinciding with the publication of Antonio de Nebrija’s Gramática castellana, the earliest vernacular grammar in Europe), Classical Spanish (1492–1726, where the year 1726 marks the publication of the Diccionario de Autoridades), Modern Spanish (1726 to about 1975), and Present-day Spanish. For discussion and different proposals see Eberenz (1991), Girón Alconchel (2004), and references there.

212

Teresa Fanego

and a short article by Fuentes Rodríguez (1994) where she compares de hecho with the SAdvs en efecto and efectivamente ‘indeed’. Like its English (in fact), French (de fait/en fait) and Italian (infatti) cognates, de hecho is today an idiomatic unit. It can often be found functioning at the VP level as an adverbial of manner (henceforth de hecho1), with the meaning ‘in practice, de facto’; in this use it tends to occur in opposition to de derecho ‘according to law, with legal sanction, rightfully’, but other collocations are also possible, as is the case in (19), where de hecho contrasts with con garantías ‘with assurance’: (18)

(19)

CREA (written) 1980 El País, 10/07/1980 (03. Politics, economy, business and finance [Politics]): “Quiero expresar aquí públicamente”, dijo más adelante, “en nombre de mis compatriotas y en el mío propio, nuestro reconocimiento por la ayuda y el aliento que siempre hemos encontrado en Luxemburgo para formar parte, de derecho, de esta Europa a la que, de hecho, ya pertenecemos”. ‘“I would like to publicly express here,” he later stated, “on behalf of my fellow countrymen and myself, our appreciation for the support and encouragement that we have always found in Luxemburg to form part, by right, of this Europe to which, de facto / in practice, we already belong.”’ CREA (written) 2000 Ángel Calle Collado, Ciudadanía y solidaridad. 183 (03. Social development): Pienso que el problema de la Cooperación, y más aún el de la Solidaridad sigue siendo de la finalidad objetiva … Poder asegurar que se da, de hecho y con garantías, la Solidaridad Internacional es, también, un problema de medición de resultados en el plazo medio y largo. ‘I think the problem with Cooperation, and even more so with Solidarity, continues to be with the ultimate objective …. To be able to ensure that International Solidarity is extended, in practice and with assurance, is also a problem of measuring the results in the short and long run.’

In addition, de hecho occurs very frequently as an elaborative DM (henceforth de hecho3) which signals, like English in fact3 or some uses of French en fait (Rossari 1992: 153–154), that “what follows is a stronger argument than what precedes, with respect to the speaker’s rhetorical purpose at that point in the discourse” (Schwenter and Traugott 2000: 12); its prime function is to signal additivity (i.e. ‘what’s more’) and to

Paths in the development of elaborative discourse markers

213

elaborate on the previous utterance. Examples of this use, by far the most common one in the spoken component of CREA, are given in (20–21):12 (20)

(21)

CREA (spoken) 1996 Estadio dos, 06/10/96, TVE 2 (1. TV and Radio [Sports Report]): El partido fue igualadísimo hasta el descanso, de hecho al descanso se llegó con empate a dos. ‘The game was neck and neck up to half-time, in fact at half-time the score was two all.’ CREA (spoken) 1996 Grupo G 4: Sociología (2. Other recordings): A: ¿Y en el trabajo has aprendido algo en especial? B: Sí, bueno, muchos idiomas, bastantes, de hecho, hablo seis y de los cuales cuatro más o menos, bastante bien, A: ‘And have you learned anything in particular at work?’ B: ‘Yes, well, many languages, lots, in fact, I speak six and at least four of them pretty well,’

As pointed out by Fuentes Rodríguez (1994: 7; see also Fuentes Rodríguez 1996: 43, 2003: 67–68), de hecho can also be employed as “un adverbio oracional con valor modal … cercano a en realidad” (‘a sentence adverb with modal meaning … very similar to in reality’). In this use de hecho is roughly synonymous with English in fact2 and most uses of French en fait (Danjou-Flaux 1980: 133), has strongly epistemic meaning and occurs in adversative contexts of various kinds, with “adversativity” being understood as “contrast between different points of view as these are constructed in language use” (Schwenter 1999: 127). See (22–23): (22)

CREA (written) 1998 Several authors, Filosofía. 1º bachillerato. Anaya, 160 (0.2 Social sciences [Philosophy]): Se conoce como determinismo la teoría que niega la realidad de la libertad: aunque nos pueda parecer, a causa de un conocimiento insuficiente, que nuestras acciones son libres, de hecho no lo son. ‘Determinism is understood as the theory that disputes the existence of free will: even though it may seem to us, because we lack sufficient knowledge, that our actions are free, in reality they are not.’

12. The technical restraints of CORDE and CREA make it difficult to give exact frequencies. But on average the DM use of de hecho represents about 70 percent of all instances in the spoken component of CREA.

214 (23)

Teresa Fanego

CREA (spoken) 1996 Esta noche cruzamos el Mississippi, 21/10/96, Tele 5 (1. TV and Radio): … se especula sobre otra cuestión, que dentro de un año vuelven otra vez a casarse. … parece ser … fue el propio Rainiero quien ratificó los papeles, fue él quien presionó a las altas estancias [sic] también jurídicas de ahí de Mónaco, para que el día cuatro … pues se ratificase directamente este divorcio. Y de hecho, ya digo, es un divorcio de cara a la galería, puesto que ellos se siguen viendo. ‘There is speculation about something else, that within a year they will remarry each other. … it seems … it was Rainiero himself who pressured the high courts there in Monaco, to on the fourth … directly ratify the divorce. And [i.e. ‘but’] in fact, you’ll see, it’s a divorce playing to the gallery since they are still seeing each other.’

I will refer to this strongly epistemic use of de hecho as de hecho2Adv, so as to distinguish it from the sentence adverbial exemplified in (24); this will be identified henceforth as de hecho2Conf (for ‘confirmatory’): (24)

CREA (written) 1997 El País, 24/09/1997: Termina en Zaragoza el conflicto del transporte escolar (0.2 Social sciences [Education]): La ministra de Educación, Esperanza Aguirre, ya vaticinó el martes en Radio Nacional que los problemas de Zaragoza se solucionarían en breve y, de hecho, así ha sido. ‘The Minister of Education, Esperanza Aguirre, already predicted on Tuesday on National Radio that the problems in Zaragoza would soon be solved and so they have been indeed.’ [or: ‘and indeed that has been the case.’]

In this type of context, which is not adversative or contrastive, de hecho cannot be felicitously replaced by en realidad ‘in reality’; its function is rather to signal agreement and confirm a preceding utterance, so that its closest synonyms in Spanish would be en efecto and efectivamente ‘indeed’. These are weakly epistemic adverbials used “for confirmation” (Bolinger 1950: 349; see also Barrenechea 1979: 54–55; Fuentes Rodríguez 1994: 12; Martín Zorraquino and Portolés Lázaro 1999: 4148–4149); cf. (25): (25)

Prometió que llegaría temprano y en efecto llegó a las seis. ‘He promised to come early, and he did show up around six.’

Paths in the development of elaborative discourse markers

215

According to Brutti (1999: 520–521, 529), “displaying agreement with what has been stated before” is precisely the main use of Italian infatti.13 English in fact, by contrast, cannot signal agreement to a prior utterance or implicature (cf. Traugott and Dasher 2002: 171) and thus differs from indeed, which, as noted by Blakemore (2002: 97) and Traugott and Dasher (2002: 164, 171), is often confirmatory. E.g.: (26)

A: Do you agree? B: Indeed I do.

This section has looked at the main uses of de hecho in Present-day Spanish; they were found to have much in common with the functions reported in the relevant literature for its English, French and Italian cognates, as might have been expected in view of the fact that all of them go back, ultimately, to the Latin noun factum ‘deed, action’. The stages in the development of de hecho from its original function as a VP adverbial of manner into a DM are the concern of the next section, while the question of what is the best analysis, from a synchronic point of view, for its various uses today will be discussed in Section 6 below. 5.4. Historical development of de hecho The noun hecho ‘deed, action’ is recorded in Spanish from the earliest written documents. Its ancestor, Lat. factum, was sometimes used to denote ‘a real happening, a fact’, as opposed to fiction, as in (27); the evidential meaning that marks some uses of Sp. de hecho, Eng. in fact, Fr. en fait/de

13. With respect to French en fait and de fait, Roulet (1987: 125–126) points out that “Bien que certains dictionnaires, comme le Dictionnaire Larousse du français contemporain, affirment que de fait marque une relation de conformité alors que en fait marque une relation d’opposition, les deux connecteurs paraissent toujours mutuellement substituables. … Les interpretations de confirmation et d’opposition que donnent les dictionnaires ne sont que des effets indirects, qui ont leur source dans le contenu des constituants articulés” (‘Even if certain dictionaries, including the Larousse Dictionary of Contemporary French, claim that de fait indexes conformity whereas en fait indexes opposition, the two connectives always appear to be mutually exchangeable. … The interpretations of confirmation and opposition given in dictionaries are only indirect effects, which find their source in the contents of phrases uttered’).

216

Teresa Fanego

fait and It. infatti was, therefore, already a feature of their common Latin source. (27)

Petronius Arbiter 76.4 (OLD s.v. factum 3): omnes naues naufragarunt, factum, non fabula. ‘All ships sank, in fact, not in fiction.’

5.4.1. De hecho as a VP adverbial of manner From around the middle of the thirteenth century the noun hecho can be found preceded by the preposition de ‘of’ functioning as a VP adverbial of manner meaning ‘in action, in practice, in actuality’. It may occur on its own, as in (28), or in combination with other prepositional phrases more or less formulaic in character, such as de voluntad ‘in thought’, de dicho ‘in word’, or de palabra ‘in word’: (28)

(29)

CORDE 1293 Anonymous, Castigos, fol. 116R (14. Didactic prose [Gnomic literature]): los omnes quando los lleua el prinçipe con braueza o del todo quebrantan contra el prinçipe o nonle Son bien obedientes E asi lo prouamos muchos vezes de fecho que por que algunos prinçipes quesieron mucho abaxar sus subditos ouieron de ser Rebeldes contra los prinçipes ‘Men, when their prince wields his rage at them, either rebel entirely against him or are disobedient to him. We have proven this many times in practice observing that because some princes wished to completely humiliate their subjects [the latter] had to become rebels against their princes.’ CORDE 1256–1263 Alfonso X, Siete Partidas, fol. 19r (10. Documents and laws): todos los xpistianos le deuen onrrar and amar en estas tres maneras. de uoluntad. and en dicho. and en fecho. En la primera que es de uoluntad …. La segunda que es por palabra; quel deuen onrrar. llamandol padre sancto and sennor. La tercera que es de fecho; que quando algunos uinieren a el; quel besen el pie. and quel onrren en todas cosas ‘all Christians must honour and love him [i.e. the Pope] in these three ways: in thought, and in word, and in action. The first way is in thought …. The second is in word, that they must honour him calling him Holy Father and Lord. The third is in action, that when

Paths in the development of elaborative discourse markers

217

they come close to him, they must kiss his foot and honour him in every respect’ In the kind of context shown in (29) de hecho continues some of the uses of Lat. factum, which was often opposed to words; cf. ACC.trag.47 “dicta factis discrepant” (OLD s.v. factum n. 1.b). A similar use has been reported for Middle English in dede (> PDE indeed), as can be seen in (4) above, repeated here for ease of reference: (4)

c. 1300 Fox and Wolf, p. 34 (Traugott and Dasher 2002: 159): “Vuolf,” quod þe vox him þo, “All þat þou hauest her before I-do, In þohut, in speche, and in dede, In euche oþeres kunnes quede, Ich þe forƢeue.” ‘“Wolf,” said Fox to him then, “All that you have done before this, in thought, in speech, and in action, in evil of every other kind, I forgive you for it.”’

However, by far the most frequent collocation of de hecho in Medieval Spanish is forming a contrast set with de derecho ‘de jure, according to law, with legal sanction, rightfully’, as in (30) below. The earliest occurrence of de hecho in this use is in the Siete Partidas, the work where King Alphonso X (1252–1284) codified Spanish law following the principles of Roman law and, specifically, of the Corpus Iuris Civilis of Emperor Justinian (527–565) (cf. Kabatek 2001). The Siete Partidas formed part of a vast programme of translations of legal, historical and scientific works which were supervised by the Royal Chancery and marked the beginnings of the use of the vernacular for subjects that formerly had been written in Latin or Arabic (Fisher 1986; Fernández-Ordóñez 2004: 386ff). (30)

CORDE 1256–1263 Alfonso X, Siete Partidas, fol. 97r (10. Documents and laws): E ninguna destas cosas spiritales sobredichas … no las pueden uender de derecho cuemo quier que algunos las compran and las uenden de fecho, que es symonia ‘And the spiritual things above-mentioned … cannot be sold de jure though some buy them and sell them de facto, which is simony’

In this legal use de hecho can be glossed as ‘in action, in practice, in actuality’, just like in (28–29) above, but carries the strong additional implication

218

Teresa Fanego

that the action being referred to is done without legal sanction and/or by force. This polysemy allowed de hecho to occur in coordination not only with manner adverbs having evidential meaning such as realmente ‘in a real manner’,14 but also with adverbs and adverbial expressions like ynjustamente ‘unjustly, illegally’ or por fuerça ‘by force’. Examples of such coordinate structures in CORDE include realmente and de fecho ‘in a real manner and de facto’, de fecho y sin ficçion alguna ‘in actuality and not fictitiously’, de fecho e ynjustamente ‘de facto and unjustly’ and de fecho o por fuerça e non por via de juysio ‘de facto or by force and not through legal means’; see also (31–32). In most of these cases it is difficult to tell whether the intended meaning was primarily ‘in actuality’ or ‘illegally’ or both. (31)

(32)

CORDE 1398 Carta de entrega de posesión (10. Documents and laws): … que agora ellos ponían and posieron luego al dicho abat, en nonbre de si and del dicho prior and conuento del dicho monesterio, en posesión paçífica, realmente and de fecho de los dichos sus bienes todos and de cada vno dellos ‘… that now they placed and did place the abbot in question, on his behalf and that of the said prior and convent of the said monastery, in the undisputed possession, in reality and de facto of each and everyone of the said goods’ CORDE 1483 Petición de traslado (10. Documents and laws): E puesto que lo oviera prendido dentro en el termino e jurediçion del dicho valle dicho tenia que, sy lo fiziera para lo llevar e lo llevara a la dicha villa de Castro se fazia e fiziera de fecho e ynjustamente, e que no cabsara derecho alguno ‘And even though he had detained him within the boundary and jurisdiction of the said valley, it was considered that, if he did it to take him and he took him to the said town of Castro, that was done and would be done de facto [i.e. illegally] and unjustly, and would entitle him to nothing’

14. Note that in Medieval Spanish realmente was used exclusively as a manner adverb and could hence occur in coordination with other manner adverbs like bien ‘well’ (e.g. CORDE 1382 Anonymous [10. Documents and laws]: dar galardón a aquéllos que les bien e realmente sirvieron ‘reward those who served them well and in a real manner’); its use as a sentence adverb is a later development. See Corominas and Pascual (1981).

Paths in the development of elaborative discourse markers

219

In early Medieval Spanish de hecho usually follows the verb, as is the case in the above instances, where the higher verbs are respectively ponían and posieron and se fazia e fiziera. From the fourteenth century, however, we come across examples in which de hecho is given front position in the clause; thus between 1350 and 1400 CORDE contains 33 occurrences of de hecho on its own, that is, not coordinated to another adverbial; 23 of these occurrences follow the verb group and 10 precede it (two ex. in verse), as in (33–34): (33)

(34)

CORDE 1325–1335 Juan Manuel, El Conde Lucanor, 81 (12. Imaginative narration /14. Didactic prose): pues sabedes que este omne que a vós vino es muy adebdado con aquel vuestro enemigo et naturalmente él et todo su linage son vuestros enemigos, conséjovos yo que en ninguna manera non lo trayades en vuestra compaña, ca cierto sed que non vino a vós sinon por engañar et por vos fazer algún daño. Pero si él vos quisiere servir seyendo alongado de vós, de guisa que vos non pueda enpescer nin saber nada de vuestra fazienda, et de fecho fiziere tanto mal et tales manzellamientos a aquel vuestro enemigo con quien él ha algunos debdos, que veades vós que non le finca logar para se poder nunca avenir con él, estonce podredes vós fiar en él ‘since you know that this man who has come to you is bound [by feudal law] to your enemy and that as a consequence he and all his lineage are your enemies, I advise you not to accept him in any way as a member of your retinue, because it is evident that he has come only to deceive you and to cause you harm. But if he is willing to serve you [even though] being far away from you, in such a way that he can neither harm you nor know anything about your affairs, and de facto [illegally?] causes so much harm and such dishonour to your enemy to whom he has feudal obligations, so that you plainly see that he no longer has occasion to reconcile with him, in that case you can trust him’ CORDE 1379–1384 Juan Fernández de Heredia, Vidas Paralelas (Agesilao), Plutarco II, fol. 10v (19. Historical prose [Biography]): los laçedemonjos … se tenjan por agraujados de Agissilao, porque tal ciudat and tan grant qu’ellos possedian la perdieron en el tiempo que Agissilao regnaua. Et por esto Agissilao contrastaua por no dar jurisdicçion a aquellos qui de fecho la tenjan. Mas no la pudo hauer, antes fue en periglo de perder toda la Spartia.

220

Teresa Fanego

‘the Lacedaemonians … were offended by Agesilaus, because such a city [i.e. Messenia] and so important which they possessed in years’ past was lost during Agesilaus’ reign. And for that reason [although the Thebans offered peace] Agesilaus resisted giving jurisdiction to those [i.e. the Thebans] who de facto had it. But he was unable to recover it [Messenia], and was in danger of losing all of Sparta.’ A common assumption is that leftward movement, manifested for instance in the increased use of a linguistic element in preposing and topicalization, constitutes evidence of subjectification; thus Torres Cacoullos and Schwenter (2006: 356–357) suggest that the development of Medieval Spanish pesar de ‘regret of’ into the concessive connective a pesar de que ‘in spite of (that)’ is manifested, inter alia, in the increased use from the eighteenth century onwards of the a pesar de phrase in a position preceding the main verb (e.g. A pesar de estas reflexiones, no estaba tranquilo ‘in spite of these reflections, he wasn’t tranquil’); see also Fischer (2007: 259 ff) and Traugott (this volume) for similar observations. The preposing of de hecho from about 1330 might thus be a genuine indication of widening of predicational scope and hence of the acquisition of epistemic meanings. Yet it should be noted that in most of my early examples the location of de hecho clause-initially seems to respond, primarily, to a desire of underlining the contrast between what is done according to law, and what is done without legal sanction or by force; note in this connection the choice of legal terminology like adebdado ‘bound by feudal law’, debdos ‘feudal obligations’, or jurisdicçion ‘jurisdiction’, which all suggest that in (33) and (34) de hecho largely retains its original, source meaning (for this label see Heine 2002: 84). In addition to frequent preposing, another interesting development taking place from about the late fourteenth century is the very frequent use of de hecho in the second of two clauses linked by e/y ‘and’ or como ‘as’; the second clause contains the same verb as the first and confirms or repeats its information: (35)

CORDE 1464 Anonymous, Carta de apeamiento. Documentación medieval de la cuadrilla de Salvatierra, 354 (10. Documents and laws): … nin podiesen pidir anulaçion nin reçesion deste dicho contracto … nin de cosa alguna de lo en el contenido …. E caso que lo contrario dello quisiesen o tentasen fazer e fiziesen de fe-

Paths in the development of elaborative discourse markers

(36)

221

cho, que les non valiese nin fuesen … oydos sobrello en juyzio nin fuera de juyzio ante algund juez eclesiastico nin seglar. ‘nor could they ask for the cancellation or rescission of the contract in question … nor of any thing contained therein …. And in the event that they wanted or tried to do the contrary and did so de facto, that it would not be of any use to them or be admitted in court before any ecclesiastical or lay judge.’ CORDE 1480–1484 Hernando del Pulgar, Crónica de los Reyes Católicos, I, 45 (19. Historical prose): en el año siguiente del Señor de mill e quatroçientos and setenta e dos años, el prínçipe y la prinçesa partieron … para la villa de Alcalá de Henares. Estando allí con el arçobispo de Toledo, algunos prinçipales de la villa de Aranda de Duero, que era de la rreyna doña Juana, ovieron acuerdo de la entregar al prínçipe e a la prinçesa, and de fecho se la entregaron. Y luego se apoderaron della, …, e fueron echados quinientos de la villa: la justiçia e todos los otros ofiçiales que estauan puestos por la rreyna doña Juana. ‘In the year of our Lord 1472, the prince and princess left … for the town of Alcalá de Henares. While there with the archbishop of Toledo, some noblemen from the town of Aranda de Duero, which belonged to Queen Juana, agreed to hand it over to the prince and the princess, and de facto [illegally? / indeed?] they handed it over to them. Then they took control of it …, and five hundred [people] were thrown out of the town: the administrators of justice and all of the other officials that had been appointed by Queen Juana.’

This type of confirmatory structure15 is a prominent characteristic of some medieval prose styles. Though research on the history of Spanish genres is 15. Confirmatory sentences not containing de hecho are also very common, as in the following example: (i) CORDE 1434 Anonymous, Autos de ejecución de sentencias del Concejo de Segura (Guipúzcoa), II, 231 (10. Documents and laws): E luego los dichos Lope Ybannes e Lope Sanches e Juan Martines, alcaldes de la dicha Hermandat, dixieron … que ellos commo alcaldes de la dicha Hermandat, mandavan e mandaron al dicho Merino que non executase la dicha sentençia e mandamiento del dicho Pero Sarmiento ‘And afterwards the said Lope Ybannes and Lope Sanches and Juan Martines, representatives of the said citizens, said … that as representatives of the said citizens they had ordered and did order the said judge to not execute the said sentence and ruling against the said Pero Sarmiento’

222

Teresa Fanego

still much needed, it is generally agreed that the vast programme of translations of historical, scientific and statutory works initiated under the auspices of King Alphonso X (1252–1284) not only contributed to language standardization proper, but also to the establishment of the discoursal features of a number of genres which up to that point had not existed as conventionalized text types simply because prose was not produced in the vernacular, but in Latin or Arabic.16 By the fifteenth century, however, if not earlier, legal and historical writings already showed clear genredistinctive characteristics, including, among others: (i)

Frequent use of connective sentence adverbs (e.g. otrosí ‘also, in addition’, demás ‘besides’) and other items making for increasing cohesiveness and referential accuracy (see Eberenz 1994: 18, Cano Aguilar 2002: 499). Note for instance in (31–32) above the repeated use of dicho ‘said’: al dicho abat ‘to the said abbot’, del dicho prior ‘of the said prior’, el dicho valle ‘the said valley’, la dicha villa ‘the said town’, etc. (ii) Pairing of synonyms or near synonyms: termino e jurediçion ‘boundary and jurisdiction’ (cf. 32), anulaçion nin receçion ‘cancellation or rescission’ (cf. 35), la dicha sentençia e mandamiento ‘the said sentence and ruling’ (cf. footnote 15). (iii) Related to (ii), the resort to confirmatory sentences such as those exemplified in (35–36).

5.4.2. De hecho as an epistemic sentence adverbial In an example like (36) clause-initial de hecho appears to have wide (sentential) scope, so that a modern reader would feel inclined to take it in its modern, subjectivized sense of ‘indeed’ (i.e. ‘and indeed they handed it over to them’), that is, as expressing the speaker’s/writer’s commitment to the truth of the utterance. Yet the legal overtones of the passage – note especially the reference to ‘the administrators of justice’ – do not allow us to rule out an interpretation of de hecho in its original source meaning, that is, ‘and illegally they handed it over to them’. From the fifteenth century, however, we come across examples where de hecho is found in confirmatory sentences in which the older, legal meaning of de hecho is incompatible with the overall context. This is the case, for instance, in (37–38) below; 16. For analogous developments in other European vernaculars see Fisher (1986) and Rissanen (2000).

Paths in the development of elaborative discourse markers

223

both passages can be identified as switch (Heine 2002) or isolating (Diewald 2002) contexts, that is, specific linguistic contexts that favour the target, more grammaticalized, meaning to the exclusion of the source meaning, so that the adverbial can only be understood as having wide scope over the clause and signalling that what was announced in prior discourse is true: (37)

(38)

CORDE 1487 Fray Hernando de Talavera, Católica impugnación del herético libelo maldito y descomulgado, 189 (17. Religious prose): También ha este necio por inconveniente, que se diga que la imagen ríe y que llora y que suda. Verdad es que puede haber y de hecho hay en esto muchas burlas y mucho sacadinero, pero bien es posible que ría y llore y sude y hable. ‘Also this ignorant man considers it an error saying that the statue laughs and cries and sweats. It is true that there may be and in fact there is much mockery and fraud in this matter, but it may be that [the image] smiles and cries and sweats and speaks.’ CORDE 1492–1493 anonymous, Diario del primer viaje de Colón, 173 19. (Historical prose [Diaries]): Así que tomó por la mano el señor al escrivano y lo llevó a su casa …, y les hizo dar de comer, y todos los indios les traían muchas cosas …. Después que fue tarde, … ellos mismos porfiavan de traellos a cuestas, y de hecho lo hizieron por algunos ríos y por algunos lugares lodosos. ‘So the chief took the official by the hand and took him to his house …, and made certain that they were fed, and all of the Indians brought them many things. When it got late, … they [i.e. the Indians] themselves insisted on carrying them on their backs, and they indeed did so crossing some rivers and some muddy places.’

Occurrences of this kind, where de hecho is unambiguously the sentence adverb I referred to earlier as de hecho2Conf, are very frequent in late Medieval and Classical Spanish. The adversative sentence adverb (de hecho2Adv) illustrated in (22–23) above can also be found over the same period, but is much rarer: only three or four reasonably clear instances are recorded in the fifteenth and sixteenth centuries. The following are two of them: (39)

CORDE 1440–1460 Vasco Ramírez de Guzmán, Jugurtha, fol. 87r (19. Historical prose): Jugurta … vido quelo tentaua por las sus

224

(40)

Teresa Fanego

artes que el solia vsar Ca el le dezjan palabras de paz and de fecho veya que era guerra muy aspera. ‘Jugurtha realized that he [Metellus] was tempting him with the stratagems he used to employ for they spoke words of peace and [i.e. ‘but’] in fact he [Jugurtha] saw that it was very fierce war.’ CORDE 1535–1557 Gonzalo Fernández de Oviedo, Historia general y natural de las Indias, I, 297 (15. Biology): en las sierras de Capira … hay unos árboles hermosos e grandes que los cristianos llaman perales. Y de hecho, la fructa que llevan son peras en el talle y en la color, e no en más, porque el cuero es tan gordo como de un borceguí de cordobán … e el cuesco es grande, que ocupa todo lo demás ‘In the Capira mountain range … there are beautiful and large trees that the Christians call pear trees. And [i.e. ‘but’] in fact, the fruit that grows on them looks like pears in shape and colour, and in nothing else, because the skin is as thick as a shoe made of cordovan leather, … and the stone is so large that it takes up the whole fruit’

In these passages, as also in those adduced earlier to illustrate the confirmatory use of de hecho, the conjunction y ‘and’ comes at the front of the de hecho clause, but its meaning appears to be adversative rather that additive (see also example 23 in Section 5.3), as it introduces an explicit rejection of the preceding proposition. This determines the contextual interpretation of de hecho itself as a strongly epistemic adverbial that could be replaced by en realidad ‘in reality’, but not by the more weakly epistemic en efecto ‘indeed’. The adversative use of de hecho has become more frequent in modern times (see 42 for an eighteenth-century example). This, I suspect, could have to do with the fact that between about 1430 and 1650 the strengthened adverbial phrase en hecho de verdad, lit. ‘in true fact’ (< Lat. re uera; OLD s.v. res 6.b), was used very frequently with adversative function, at first in legal documents (see CORDE and Autoridades), later more generally, as in (41). Its disuse from the second half of the seventeenth century – the last example in CORDE dates from 1653 – may have facilitated the increased use of de hecho adversatively: (41)

CORDE 1612-1625 Fray Juan Márquez, El gobernador cristiano (14. Didactic prose [Other texts]): Como hazía Demetrio de Efeso,

Paths in the development of elaborative discourse markers

(42)

225

que predicando san Pablo contra el ídolo de Diana, en que él tenía todo su aprovechamiento, porque labrava con gran primor la imagen, y andas de la Diosa, … movió una gran sedición en la ciudad, condenando a gritos la dotrina de san Pablo, con zelo al parecer de la honra de Diana, y en hecho de verdad por miedo de perder la grangería ‘As Demetrius of Ephesus did, who, when St. Paul preached against the statue of Diana, in which he [Demetrius] had all of his economic benefits, because he carefully worked the statue, and the portable platform that carried the goddess, provoked a great public uprising in the city, condemning St. Paul’s teachings, apparently driven by his concern over the cult of Diana, but really out of fear of losing his trade’ CORDE 1734 Benito Jerónimo Feijoo, Teatro crítico universal, 342 (14. Didactic prose [Treatises and essays]): Esta religiosa, que se llamaba doña Eulalia Pérez, … cayó en una fiebre, que pareció al médico peligrosíssima (aunque de hecho no lo era), por lo qual fui llamado para administrarla el socorro espiritual ‘This nun, whose name was Eulalia Pérez, … fell sick with a fever that the physician thought extremely dangerous (though in fact it wasn’t), on account of which I was summoned to bring her spiritual relief’

5.4.3. De hecho as an elaborative DM Because of its confirmatory function with respect to a prior proposition, the sentence adverb (de hecho2Conf) illustrated in (37–38) above can be said to denote “a relationship across rather than within utterances” (Fraser and Malamud-Makowski 1996: 864; cf. Section 2), and has thus both conceptual and procedural meanings. It is therefore a much more likely source than adversative de hecho for the use of de hecho as a DM signalling additivity and elaborating on the previous utterance: (43)

CORDE 1604–1618 Fray Prudencio de Sandoval, Historia de la vida y hechos del Emperador Carlos V, XLVI, 407 (19. Historical prose): … y se juntaron todos a tratar dello. Pero estaban tan soberbios, y por otra parte, temían tanto dejar los cargos que tenían, principalmente los capitanes, que no se pudo acabar con ellos que

226

(44)

(45)

Teresa Fanego

quisiesen tregua ni paz (si bien algunos de la Junta la votaron). El principal de los cuales fue don Pedro Laso de la Vega, que desde allí los dejó y se apartó de aquel mal propósito, y de hecho se vino a Tordesillas para los gobernadores. ‘… and everyone met to discuss the matter. But they were so arrogant, and also, they feared so much giving up the posts they held, mainly the captains, that they could not reach an agreement concerning a truce or peace (although some [members] of the ‘Junta’ voted in favour of it). The main one of which was Pedro Laso de la Vega, who from there left them and withdrew from that bad purpose, and in fact came to Tordesillas [to be] with the governors.’ CORDE 1659–1664 Anonymous, Noticias de la Corte (18. Press): [El Rey de Inglaterra] no corre bien con el Parlamento bajo, el cual le contradice con teson la libertad que desea entablar en sus reinos para la religion católica. El mismo Parlamento trata de fortificar á Tánger, y de hecho quieren enviar 10 fragatas para el efecto ‘[The King of England] does not get on well with the lower Parliament, which tenaciously opposes the freedom that he wishes to establish in his kingdoms for the Catholic religion. This same Parliament is trying to fortify Tangiers, and in fact they want to send 10 frigates for that purpose’ CORDE 1726 Benito Jerónimo Feijoo, Teatro crítico universal, I, 110 14. (Didactic prose [Treatises and essays]): … la regla de el Apóstol, que en la Epístola primera a los Corinthios prohibe a las mugeres hablar en la Iglesia, admite algunas excepciones: como las admite la prohibición, de que enseñen, en la Epístola primera a Thimoteo: pues de hecho Priscila, compañera de el mismo Apóstol, enseñó, e instruyó a Apolo Pontico, en la Doctrina Evangélica ‘… the Apostle’s rule, who in the First Letter to the Corinthians prohibited women from speaking in Church, allows for some exceptions: as it does the prohibition that women teach in the First Letter to Timothy: for in fact Priscilla, companion of the Apostle himself, taught, and instructed Apollo Ponticus, in Evangelical Doctrine’

Unambiguous instances of this new type are recorded from the beginning of the seventeenth century. Initially they typically occur in the by now familiar structure where y ‘and’ introduces the de hecho clause, but from 1700

Paths in the development of elaborative discourse markers

227

onwards, when the DM becomes much more frequent, it is no longer restricted to that kind of discourse context. 6. Summary and conclusions This section summarizes the findings of the present study in terms of the research questions introduced in Section 5.1. The most obvious finding is that the historical development of de hecho is analogous to that of DMs like in fact, indeed, or actually, all of which, as pointed out earlier in this paper, started as VPAdvs and evolved into DMs via a sentence adverbial stage. A number of other items have been shown to have similar trajectories; see for instance Brinton (1999) on the development of whilom from a VPAdv meaning ‘at times’ into a SAdv meaning ‘formerly’, and from this into a DM marking the initiation of a story or episode, or Lenker (2000) on the Old English adverbs soþlice and witodlice ‘truly’, which shifted from manner adjuncts to style disjuncts and then to “indicators of thematic discontinuity” (p. 243). In Present-day Spanish the confirmatory SAdv efectivamente ‘indeed’, originally a VPAdv meaning ‘in a real manner’,17 seems to be acquiring elaborative uses comparable to those of de hecho, as in the following attested example quoted by Fuentes and Alcaide (1996: 123): (46)

[Spoken Andalusian Spanish]: Esta Navidad ha sido más cara. Efectivamente, todo ha subido descaradamente. ‘This Christmas has been more expensive. In fact, prices have risen exorbitantly.’

With respect to de hecho in particular, its chronological development is a mirror image of the development of English indeed: in both cases the earliest uses as epistemic SAdvs date from the second half of the fifteenth century, and the earliest uses as DMs from about 1600 (see examples 37 and 43 above, and Traugott and Dasher 2002: 171). Also noteworthy is the fact that as SAdvs de hecho and indeed can be employed either adversatively, to reject a prior proposition, or to signal agreement. I showed above, however, 17. Cf. Autoridades (s.v. Efectivamente), quoting Nieremberg, Varones ilustres de la Compañía de Jesús (1643–1647), lib. 2, cap.3: Que efectivamente hace en el alma lo que el alma en el cuerpo ‘Which truly affects the soul as the soul does the body’.

228

Teresa Fanego

that in terms of pragmatic motivation the confirmatory reading of de hecho is a much more likely source for its later use as an elaborative DM than the adversative reading, and it seems plausible that further research might reveal a similar trajectory for indeed, since this is attested with confirmatory meaning from a very early date.18 I would suggest, therefore, that elaborative DMs will often follow the correlated paths of directionality shown in Table 3, which modifies slightly Table 2 above as proposed by Traugott and Dasher (2002). Table 3. Paths of directionality in the development of Spanish de hecho ADVmanner

> ADVconfirmatory

> ADVelaboration

content

> content/procedural

> procedural

scope within proposition

> scope over proposition

> scope over discourse

nonsubjective

> subjective

> intersubjective

> ADVhedge

The second research question in Section 5.1 addressed the issue of whether there is any morphosyntactic evidence correlating with the semanticpragmatic changes undergone by de hecho which might justify an analysis in terms of grammaticalization. Clearly, when de hecho evolves from a VPAdv into a SAdv the shift has semantic, phonological and morphosyntactic consequences, as follows: (i) (ii) (iii) (iv) (v)

There is semantic/pragmatic change from more to less referential meaning. A special intonation contour comes to be used for the SAdv. The SAdv tends to appear in positions (e.g. front position in the clause) not normally occupied by VPAdvs in Spanish. Unlike the VPAdv (see 47), the SAdv is no longer eligible as the focus of a cleft sentence and cannot be questioned by cómo ‘how, in what way’. Coordination with other manner adverbials, as in de hecho o de derecho (see 47), or de fecho e ynjustamente ‘illegally and unjustly’ (see 32), remains possible as long as de hecho is a VPAdv, but is disallowed once it has become a SAdv.

18. Traugott and Dasher (2002: 164) claim that the confirmatory use of indeed is found only from the sixteenth century, but some of their fifteenth-century examples, such as (14b) on p. 161, are clearly confirmatory rather than adversative.

Paths in the development of elaborative discourse markers

(47)

229

¿Cómo formamos parte de Europa? ¿De hecho o de derecho? How do we belong in Europe? In practice or by right?

Most of the above changes are indicative of increased grammaticalization. Thus (i) illustrates desemanticization, (iii) extension or context generalization, that is use in new contexts, and (iv) and (v) decategorialization, or loss in the morphosyntactic properties characteristic of the source forms (see Heine 2003: 578ff for an overview of these and other interrelated mechanisms in grammaticalization). By contrast, if we now consider the further development of the SAdv de hecho into a DM, things are far less clear. The semantic/pragmatic shift from conceptual/procedural to procedural meaning is indeed typical of grammaticalization processes, but the SAdv and DM uses of de hecho are not distinguished by intonation, and, even more importantly, they share syntactic positions. In other words, both the SAdv and the DM can occur initially, medially, or finally, and in exactly the same discourse contexts, as can be seen if we look at some of the examples quoted earlier in these pages; note in particular (20), (22), (23), (24) and (43–45) above, which all show that whether functioning as a SAdv or a DM, de hecho continues to have scope over the proposition rather than over larger discourse chunks. There are no grounds, therefore, to assume that the DM occupies a structural position distinct (i.e. further left) from that which would be assigned to the SAdv in a phrase-structural analysis, as has been suggested for the various meanings of in fact, indeed, actually or anyway by Traugott and her associates (see 13 above). In conclusion, as long as grammaticalization is understood as it is understood here and in Traugott’s work more generally, that is, as a process involving not only semantic/pragmatic changes but also structural and morphosyntactic ones, the confirmatory, adversative and elaborative uses of de hecho are better interpreted as generalized conversational implicatures (GCIs; see Levinson 2000: 11), that is, default inferences and conventions of use in languagespecific communities that can be exploited to imply/insinuate certain meanings, but may however be cancelled.19 With regard to the third research question about the kinds of text-types favouring the use of de hecho, it has been shown in the previous pages that generic conventions may play an important role in linguistic change. Originally, Spanish de hecho was chiefly used in legal texts (see Section 5.4.1), and the various changes affecting it are clearly associated with statutory 19. Traugott and Dasher (2002: 16) refer to them as generalized invited inferences (GIINs).

230

Teresa Fanego

and legal writings, and, in general, with official or semi-official documents of various kinds, including chronicles and historical prose. In the eighteenth century, when the use of de hecho as a DM is already well established, it is commonly found in treatises and essays such as Feijoo’s Teatro crítico universal (1726; see example 45) or Ignacio de Luzán’s Defensa de España (1742). In terms of Biber’s multi-dimensional approach to register variation, all these kinds of texts score high for features such as informational (vs. involved) production (Dimension 1; see Biber 1988: 107), or elaborated (vs. situation-dependent) reference (Dimension 3; see Biber 1988: 110), so that, overall, Traugott and Dasher’s (2002) hypothesis (cf. Section 3) that certain kinds of semantic change are not initiated by children, but by adults ‘in authoritative position’ appears to be correct. Finally, the history of de hecho suggests, too, that the widespread assumption that “many (indeed, presumably most) discourse markers occur first in spoken language and only gradually make their way into written texts” (Hansen and Rossari 2005: 181) is by no means applicable to all types of markers. Sources of data Autoridades REAL ACADEMIA ESPAÑOLA. 1726. Diccionario de autoridades. Edición facsímil. Madrid: Gredos, 1990. CORDE REAL ACADEMIA ESPAÑOLA: Banco de datos [on-line]. Corpus diacrónico del español. [Last date of access: 1 February 2005] CREA REAL ACADEMIA ESPAÑOLA: Banco de datos [on-line]. Corpus de referencia del español actual. [Last date of access: 1 February 2005] OLD Glare, P.G.W. et al. (eds.), 1996. Oxford Latin Dictionary. Oxford: Oxford at the Clarendon Press.

References Aijmer, Karin 2002 English Discourse Particles: Evidence from a Corpus. Amsterdam/Philadelphia: Benjamins. Aijmer, Karin and Anne-Marie Simon-Vandenbergen 2004 A model and a methodology for the study of pragmatic markers: The semantic field of expectation. Journal of Pragmatics 36: 1781–1805.

Paths in the development of elaborative discourse markers

231

Aissen, Judith 1992 Topic and focus in Mayan. Language 68: 43–80. Banfield, Ann 1973 Narrative style and the grammar of direct and indirect speech. Foundations of Language 10: 1–39. Barrenechea, Ana María 1979 Operadores pragmáticos de actitud oracional: Los adverbios en – mente y otros signos. In Estudios lingüísticos y dialectológicos: Temas hispánicos, Ana María Barrenechea et al. (eds.), 39–59. Buenos Aires: Hachette. Biber, Douglas 1988 Variation across Speech and Writing. Cambridge: Cambridge University Press. Blakemore, Diane 1987 Semantic Constraints on Relevance. Oxford: Blackwell. Blakemore, Diane 2002 Relevance and Linguistic Meaning: The Semantics and Pragmatics of Discourse Markers. Cambridge: Cambridge University Press. Bolinger, Dwight L. 1950 En efecto does not mean in fact. Hispania XXXIII: 349–350. Brinton, Laurel J. 1996 Pragmatic Markers in English: Grammaticalization and Discourse Functions. Berlin/New York: Mouton de Gruyter. Brinton, Laurel J. 1999 ‘Whilom, as olde stories tellen us’: The discourse marker whilom in Middle English. In From Arabye to Engelond: Medieval studies in honour of Mahmoud Manzalaoui on his 75th birthday, A. E. Christa Canitz and Gernot R. Wieland (eds.), 175–199. Ottawa: University of Ottawa Press. Brinton, Laurel J. and Elizabeth Closs Traugott 2005 Lexicalization and Language Change. Cambridge: Cambridge University Press. Brutti, Silvia 1999 In fact and infatti: the same, similar or different? Pragmatics 9: 519– 533. Cano Aguilar, Rafael 2002 Elementos de ilación textual en castellano medieval (época postalfonsí). In Actas del V Congreso Internacional de Historia de la Lengua Española, María Teresa Echenique Elizondo and Juan Sánchez Méndez (eds.), 489–502. Madrid: Gredos.

232

Teresa Fanego

Company Company, Concepción 2004 ¿Gramaticalización o desgramaticalización? El reanálisis y subjetivización de verbos como marcadores discursivos en la historia del español. Revista de Filología Española 84: 29–66. Company Company, Concepción 2006 Subjectification of verbs into discourse markers. Semantic-pragmatic change only? In Topics in subjectification and modalization (Special issue of Belgian Journal of Linguistics 20), Bert Cornillie and Nicole Delbecque (eds.), 97–121. Amsterdam/Philadelphia: Benjamins. Company Company, Concepción 2008 The directionality of grammaticalization in Spanish. Journal of Historical Pragmatics 9: 200–224. Corominas, Joan and José A. Pascual 1981 Diccionario crítico etimológico castellano e hispánico. Madrid: Gredos. Danjou-Flaux, Nelly 1980 A propos de de fait, en fait, en effet et effectivement. Le français moderne 48 : 110–139. Diewald, Gabriele 2002 A model for relevant types of context in grammaticalization. In New Reflections on Grammaticalization, Ilse Wischer and Gabriele Diewald (eds.), 103–120. Amsterdam/Philadelphia: Benjamins. Eberenz, Rolf 1991 Castellano antiguo y español moderno: Reflexiones sobre la periodización en la historia de la lengua. Revista de Filología Española LXXI: 79–106. Eberenz, Rolf 1994 Enlaces conjuntivos y adjuntos de sentido aditivo del español preclásico: Otrosí, eso mismo, asimismo, demás, también, aún, etc. IberoRomania 39: 1–20. Fanego, Teresa 2004 On reanalysis and actualization in syntactic change: The rise and development of English verbal gerunds. Diachronica 21: 5–55. Fernández-Ordóñez, Inés 2004 Alfonso X el Sabio en la historia del español. In Historia de la lengua española, Rafael Cano (ed.), 381–422. Barcelona: Ariel. Ferrara, Kathleen 1997 Form and function of the discourse marker anyway: Implications for discourse analysis. Linguistics 35: 343–378.

Paths in the development of elaborative discourse markers

233

Fischer, Kerstin 2000 From Cognitive Semantics to Lexical Pragmatics. The Functional Polysemy of Discourse Particles. Berlin/New York: Mouton de Gruyter. Fischer, Kerstin (ed.) 2006 Approaches to Discourse Particles. Amsterdam: Elsevier. Fischer, Olga 2007 Morphosyntactic Change: Functional and Formal Perspectives. Oxford: Oxford University Press. Fisher, John H. 1986 European chancelleries and the rise of standard languages. Proceedings of the Illinois Medieval Association 3: 1–33. Fleischman, Suzanne and Marina Yaguello 2004 Discourse markers across languages. Evidence from English and French. In Discourse Across Languages and Cultures, Carol Lynn Moder and Aida Martinovic-Zic (eds.), 129–147. Amsterdam/Philadelphia: Benjamins. Fraser, Bruce and Monica Malamud-Makowski 1996 English and Spanish contrastive discourse markers. Language Sciences 18: 863–881. Fuentes Rodríguez, Catalina 1994 Usos discursivos y orientación argumentativa: De hecho, en efecto, efectivamente. Español Actual 62: 5–18. Fuentes Rodríguez, Catalina 1996 Los adverbios modales. In La expresión de la modalidad en el habla de Sevilla, Catalina Fuentes Rodríguez and Esperanza R. Alcalde Lara, 17–259. Sevilla: Servicio de Publicaciones del Ayuntamiento de Sevilla. Fuentes Rodríguez, Catalina 2001 Los marcadores del discurso, ¿una categoría gramatical? In Indagaciones sobre la lengua: Estudios de Filología y Lingüística Españolas en memoria de Emilio Alarcos, Elena Méndez, Josefa Mendoza and Yolanda Congosto (eds.), 323–348. Sevilla: Secretariado de Publicaciones de la Universidad de Sevilla. Fuentes Rodríguez, Catalina 2003 Operador/conector, un criterio para la sintaxis discursiva. RILCE. Revista de Filología Hispánica 19: 61–85. Fuentes Rodríguez, Catalina and Esperanza R. Alcaide Lara 1996 La expresión de la modalidad en el habla de Sevilla. Sevilla: Servicio de Publicaciones del Ayuntamiento de Sevilla. Garachana Camarero, Mar 1998 La evolución de los conectores contraargumentativos: La gramaticalización de no obstante y sin embargo. In Los marcadores del dis-

234

Teresa Fanego

curso: teoría y análisis, María Antonia Martín Zorraquino and Estrella Montolío Durán (eds)., 193–212. Madrid: Arco Libros. Girón Alconchel, José Luis 2004 Cambios sintácticos en el español de la Edad de Oro. Edad de Oro XXIII: 71–93. Halliday, M. A. K. 1994 An Introduction to Functional Grammar. 2nd ed. [1st ed., 1985.] London: Arnold. Halliday, M. A. K. and Ruqaiya Hasan 1976 Cohesion in English. London: Longman. Hansen, Maj-Britt and Corinne Rossari (eds.) 2005 The Evolution of Pragmatic Markers. Special issue of Journal of Historical Pragmatics 6 (2). Haspelmath, Martin 2004 On directionality in language change with particular reference to unidirectionality in grammaticalization. In Up and Down the Cline – The Nature of Grammaticalization, Olga Fischer, Muriel Norde and Harry Perridon (eds.), 16–44. Amsterdam/Philadelphia: Benjamins. Heine, Bernd 2002 On the role of context in grammaticalization. In New Reflections on Grammaticalization, Ilse Wischer and Gabriele Diewald (eds.), 83– 101. Amsterdam/Philadelphia: Benjamins. Heine, Bernd, Ulrike Claudi and Friederike Hünnemeyer 1991 Grammaticalization: A Conceptual Framework. Chicago: The University of Chicago Press. Himmelmann, Nikolaus P. 2004 Lexicalization and grammaticization: opposite or orthogonal? In What Makes Grammaticalization? A Look from Its Fringes and Its Components, Walter Bisang, Nikolaus P. Himmelmann and Björn Wiemer (eds.), 21–42. Berlin/New York: Mouton de Gruyter. Hopper, Paul J. and Elizabeth Closs Traugott 2003 Grammaticalization. 2nd rev. ed. [1st ed., 1993.] Cambridge: Cambridge University Press. Iordanskaja, Lidija and Igor Mel’þuk 1995 Traitement lexicographique de deux connecteurs textuels du franÇais contemporain: en fait vs. en réalité. In Tendances recentes en linguistique française et générale: Volume dedie a David Gaatone, Hava Bat-Zeev Shyldkrot and Lucien Kupferman (eds.), 211–236. Amsterdam/Philadelphia: Benjamins. Jucker, Andreas H. (ed.) 1995 Historical Pragmatics: Pragmatic Developments in the History of English. Amsterdam/Philadelphia: Benjamins.

Paths in the development of elaborative discourse markers

235

Kabatek, Johannes 2001 ¿Cómo investigar las tradiciones discursivas medievales? El ejemplo de los textos jurídicos castellanos. In Lengua medieval y tradiciones discursivas en la Península Ibérica: Descripción gramatical – pragmática histórica – metodología, Daniel Jacob and Johannes Kabatek (eds.), 97–132. Madrid: Iberoamericana. Kiparsky, Paul 1995 Indo-European origins of Germanic syntax. In Clause Structure and Language Change, Adrian Battye and Ian Roberts (eds.), 140–169. Oxford: Oxford University Press. Lehmann, Christian 2002 Thoughts on Grammaticalization. 2nd rev. ed. as Arbeitspapiere des Seminars für Sprachwissenschaft der Universität Erfurt 9. [1st ed., 1995, Munich/Newcastle: LINCOM EUROPA.] Erfurt: Universität Erfurt. Original publication in Arbeiten der Kölner UniversalienProjektes 48, University of Cologne, 1982. Lenker, Ursula 2000 Soþlice and witodlice: Discourse markers in Old English. In Pathways of Change: Grammaticalization in English, Olga Fischer, Anette Rosenbach and Dieter Stein (eds.), 229–249. Amsterdam/Philadelphia: Benjamins. Levinson, Stephen C. 2000 Presumptive Meaning: The Theory of Generalized Conversational Implicature. Cambridge, Mass.: The MIT Press. Lewis, Diana M. 2006 Discourse markers in English: A discourse-pragmatic view. In Approaches to Discourse Particles, Kerstin Fischer (ed.), 43–60. Amsterdam: Elsevier. Lightfoot, David 1991 How to Set Parameters: Arguments from Language Change. Cambridge, MA: MIT Press. Los, Bettelou 2004 To: from preposition to nonfinite modal. Paper presented at the Thirteenth International Conference on English Historical Linguistics (13 ICEHL), University of Vienna, 23–29 August 2004. Los, Bettelou 2005 The Rise of the To-Infinitive. Oxford: Oxford University Press. Macaulay, Ronald K.S. 1995 The adverbs of authority. English World-Wide 16: 37–60. Martín Zorraquino, María Antonia and José Portolés Lázaro 1999 Los marcadores del discurso. In Gramática descriptiva de la lengua española, Ignacio Bosque and Violeta Demonte (eds.), 4051–4213. Madrid: Espasa Calpe para la Real Academia Española.

236

Teresa Fanego

Oh, Sun-Young 2000 Actually and in fact in American English: A data-based analysis. English Language and Linguistics 4: 243–268. Onodera, Noriko 2004 Japanese Discourse Markers: Synchronic and Diachronic Discourse Analysis. Amsterdam/Philadelphia: Benjamins. Pons Bordería, Salvador and Leonor Ruiz Gurillo 2001 Los orígenes del conector de todas maneras: fijación formal y pragmática. Revista de Filología Española LXXXI: 317–351. Rissanen, Matti 2000 Standardisation and the language of early statutes. In The Development of Standard English 1300–1800. Theories, Descriptions, Conflicts, Laura Wright (ed.), 117–130. Cambridge: Cambridge University Press. Rossari, Corinne 1992 De fait, en fait, en réalité : trois marqueurs aux emplois inclusifs. Verbum 1992/3: 139–161. Roulet, Eddy 1987 Completude interactive et connecteurs reformulatifs. Cahiers de Linguistique Française 8: 111–140. Schiffrin, Deborah 1987 Discourse Markers. Cambridge: Cambridge University Press. Schourup, Lawrence C. 1985 Common Discourse Particles in English Conversation. New York: Garland. Original publication in Ohio State University Working Papers in Linguistics 28 (1982). Schwenter, Scott A. 1999 Pragmatics of Conditional Marking: Implicature, Scalarity, and Exclusivity. New York/London: Garland. Schwenter, Scott A. and Elizabeth Closs Traugott 2000 Invoking scalarity: The development of in fact. Journal of Historical Pragmatics 1: 7–25. Smith, Sara W. and Andreas H. Jucker 2000 Actually and other markers of an apparent discrepancy between propositional attitudes of conversational partners. In Pragmatic Markers and Propositional Attitude, Gisle Andersen and Thorstein Fretheim (eds.), 207–237. Amsterdam/Philadelphia: Benjamins. Sweetser, Eve 1990 From Etymology to Pragmatics: Metaphorical and Cultural Aspects of Semantic Structure. Cambridge: Cambridge University Press.

Paths in the development of elaborative discourse markers

237

Tabor, Whitney and Elizabeth Closs Traugott 1998 Structural scope expansion and grammaticalization. In The Limits of Grammaticalization, Anna Giacalone Ramat and Paul J. Hopper (eds.), 229–272. Amsterdam/Philadelphia: Benjamins. Torres Cacoullos, Rena and Scott A. Schwenter 2006 Towards an operational notion of subjectification. In Proceedings of the 31st Annual Meeting of the Berkeley Linguistics Society: General Session and Parasession on Prosodic Variation and Change, Rebecca T. Cover and Yuni Kim (eds.), 347–358. Berkeley, CA: Berkeley Linguistics Society. Traugott, Elizabeth Closs 1982 From propositional to textual and expressive meanings; some semantic-pragmatic aspects of grammaticalization. In Perspectives on Historical Linguistics, Winfred P. Lehmann and Yakov Malkiel (eds.), 245–271. Amsterdam/Philadelphia: Benjamins. Traugott, Elizabeth Closs 1995 The role of the development of discourse markers in a theory of grammaticalization. Paper presented at the 12th International Conference on Historical Linguistics, University of Manchester, August 1995 [version of November 1997]. Traugott, Elizabeth Closs 2003 Constructions in grammaticalization. In The Handbook of Historical Linguistics, Brian D. Joseph and Richard D. Janda (eds.), 624–647. Oxford: Blackwell. Traugott, Elizabeth Closs This vol. (Inter)subjectivity and (inter)subjectification: A reassessment. In this volume, 29–71. Traugott, Elizabeth Closs and Richard B. Dasher 2002 Regularity in Semantic Change. Cambridge: Cambridge University Press. Wichmann, Anne, Anne-Marie Simon-Vandenbergen and Karin Aijmer This vol. How prosody reflects semantic change: A synchronic case study of of course. In this volume, 103–154. Wilson, Deirdre and Dan Sperber 1993 Linguistic form and relevance. Lingua 90: 1–25. Zwicky, Arnold M. 1985 Clitics and particles. Language 61: 283–305.

Part 3 (Inter)subjectification and grammaticalization in the noun phrase

The emergence of the definite article: ille in competition with ipse in Late Latin Anne Carlier and Walter De Mulder Abstract This paper is devoted to the emergence of the definite article in the transition from Late Latin to French. It tackles the enigma of the selection of ille as the source of the definite article in French, in spite of its lower frequency in comparison with ipse in Late Latin. Used concurrently with ille as an incipient definite article, the identity marker ipse is more frequent because it conveys a more simple and precise instruction as an anaphoric device. This precision turns out to be a handicap when paradigmatization leads to the selection of one form for the definite article, in binary opposition with the indefinite article. Ille, in virtue of its deictic force, is a more flexible tool for anaphora than ipse and even allows first mention use. Moreover, it takes advantage of its status of demonstrative associated with the third person. Indeed, contrary to the first and second person demonstrative, ille signals that the referent cannot wholly be identified by elements in the context of utterance. It thus suggests that previous knowledge should be activated in order to retrieve the referent. This feature makes ille the preferred candidate to become the definite article. Previous studies have analysed the shift from the demonstrative ille to the definite article as a linear process of desemantization, i.e. loss of deictic meaning, or have accounted for it by invoking subjectification. This paper argues that, though subjectification is involved in the first stage of the process leading to the definite article, the transition to a full-fledged definite article implies that the initial imbalance between speaker and hearer changes into a more symmetric relationship, strengthening the intersubjective dimension.

1. The restructuring of the demonstrative system according to Harris and the role of subjectification It is generally accepted that the demonstratives in Classical Latin are structured in a person-oriented system: hic, iste and ille are linked respectively to the first, the second and the third person. Their meaning has been tradi-

242

Anne Carlier and Walter De Mulder

tionally defined in terms of physical location: hic denotes a referent located near the speaker, iste marks its location close to the hearer, whereas ille situates it in a location remote from both speaker and hearer. Is is unmarked with respect to this distinction and has mainly an anaphoric role. According to Harris (1978: 69), from Classical to Vulgar Latin, the demonstrative system underwent the shift represented in Table 1, whereby the identity marker ipse was recruited to fulfil a typically demonstrative function.1 Table 1. Shift in Demonstratives from Classical to Vulgar Latin according to Harris (1978: 69) DEMONSTRATIVES “1”

“2”

“3”

Anaphoric

Classical Latin

hic

iste

ille

is

Vulgar Latin

iste

ipse

ille

IDENTITIVES “self” “same” ipse

idem

*met-ips-imum

Harris’s proposal is essentially based upon a comparison between the Classical Latin system and the result of the evolution in different Romance languages, rather than upon an empirical analysis of Late Latin texts. For instance, ille is analysed as inheriting the value of the distance-unmarked is, because this hypothesis explains why it can be a forerunner of the definite article and of the personal pronoun in most Romance languages. The elimination of the demonstratives hic and is2 and the identity marker idem is based on the fact that none of them were retained in the Romance languages. As for iste, ipse and ille, their position in the Vulgar Latin demonstrative system is inferred from the fact that they gave rise in Spanish, Portuguese and Catalan to a three-term deictic system, represented in Table 2.

1. In spite of the register-based terminology of ‘Classical’ versus ‘Vulgar Latin’, Harris’s perspective is diachronic. In accordance with this diachronic perspective, instead of ‘Vulgar Latin’, we will use the term ‘Late Latin’ in this article. 2. The disappearance of hic and is is generally explained in terms of phonetic weakness (Harris 1978; Vincent 1997). As to idem, its destiny is linked to that of is, as it contains is as one of its constituents.

The emergence of the definite article

243

Table 2. Demonstratives derived from iste, ipse and ille in Spanish, Portuguese and Catalan (Harris 1978: 71–72) Vulgar Latin Spanish Portuguese Catalan

(accu-)iste “proximal” este este aquest

(accu-)ipse “medial” ese ese aqueix

(accu-)ille “distal” aquel aquele aquell

Traugott puts forward the hypothesis of a cline of intersubjectification (see Traugott, this volume, and Traugott and Dasher 2002: 225): Non-/ less-subjective (Ideational / Contentful

–

Subjective

–

Intersubjective Interpersonal / Procedural)

The restructuring of the demonstrative system as it has been analysed by Harris (1978) is mentioned by Traugott and Dasher (2002) as an example of subjectification: iste and ipse shift in the direction of the deictic centre and become respectively the first and the second person demonstrative. Ipse is reinforced in its function of identity marker by met3 and by the superlative suffix -imum (Harris 1980: 69). Harris’s representation of the restructuring of the demonstrative system, from Classical to Late Latin, raises at least three questions. Firstly, it could give the impression that the demonstratives in Classical Latin have a nonsubjective meaning. However, iste and ille already exhibit a subjectified meaning in Classical Latin: the distal demonstrative ille expresses a respectful attitude of the speaker / writer towards the referent (Socrates ille “the illustrious Socrates”), whereas the proximal demonstrative iste is depreciative (iste homo “this dishonorable man”). In this sense, ille and iste partake of social deixis, at least when they refer to persons (Traugott and Dasher 2002: 226 ff.). The second question is raised by Harris’s analysis of the identity marker ipse. According to Harris (1978), the three-term deictic demonstrative system in Late Latin was rebuilt by the recruitment of the identity marker ipse for the role of the second person demonstrative. However, this representation of the demonstrative system in Late Latin cannot be invoked as the 3. In Classical Latin, -met is a reinforcing suffix combining with personal pronouns, often followed by ipse (e.g. egomet ipse, memet ipsum, cf. Menge 2000: § 57). The formation of the Vulgar Latin form *metipsimum, giving rise to the forms medismo in Italian, mismo in Spanish, même in French and so on, involves a reanalysis (cf. Vaänänen 1967: § 279; Harris 1978: 69).

244

Anne Carlier and Walter De Mulder

common source of the demonstrative system in all Romance languages, since, as is also noted by Harris (1978), several Romance languages do not have a descendant of ipse within their demonstrative system. In Old French, for instance, there is no ternary but rather a binary opposition between the proximal demonstrative cist (< reinforced form ecce istu(m)) and the distal demonstrative cil (< reinforced form ecce illu(m)). (Standard) Italian does not have a demonstrative derived from ipse either. Moreover, ipse provides the source of the definite article in Romance languages such as the Balearic variety of Catalan and Sardinian (Aebischer 1948; Selig 1992; Bauer 2007). Yet the textual function characteristic of the definite article is not directly related to the second person. Thirdly, the semantic shift postulated for ille and its relationship to the disappearance of is deserve some further analysis. According to Harris (1980: 78), “had is survived, it would surely have been the source of both the definite articles and the third person pronouns in Romance”. The remote demonstrative ille inherited this double destiny. This link is however not straightforward. Even though the Classical Latin is, which is unmarked as to distance, is almost restricted to anaphora, it does not have the monopoly on the marking of anaphoric relationships: the distance-marked demonstratives hic, iste and ille and the identity expressions ipse and idem are also used in this role (Menge 2000: § 68–81; Pinkster 1996). Hence, the decrease in frequency of is as an anaphoric device, from Classical to Late Latin, cannot be assumed to automatically and immediately result in its replacement by ille. This paper will be devoted to the emergence of the definite article in the transition from Latin to French, and will reexamine the semantic nature of ille and its relationship to ipse on their pathway to the status of definite article, both from a theoretical and an empirical viewpoint. On the theoretical side, we will investigate whether the notion of (inter)subjectification proposed by Traugott and Dasher (2002) and Traugott (1995; 2003; this volume) is relevant to explain the emergence of the definite article out of the demonstrative in Late Latin. The empirical basis of this study is a historical narrative entitled the Chronicle of the Merovingian Times, written in Gaul during the 7th century by a certain Fredegarius4 and continued during the 8th century. We opted for a narrative because in this type of text discourse participants are introduced and tracked as they are involved in a 4. Fredegarii Chronicurum liber quartus cum continuationibus / The Fourth Book of the Chronicle of Fredegar, translated from the Latin with Introduction and Notes by J.M. Wallace-Hadrill, London: Thomas Nelson and Sons, 1960.

The emergence of the definite article

245

sequence of events. A narrative is thus appropriate to illustrate a variety of anaphoric relationships that could be expressed by the incipient definite article. The very choice of Fredegarius’ Chronicle and its continuations as our corpus is motivated by the period of its composition: it goes back to a pivotal period preceding the divorce between Latin and Romance languages. As has been noted by Banniard (1992) and Renzi (2000: 187), from 650 to 750, learned written style and oral speech are growing more distant without however being perceived as different languages. Hence, the often despised “barbaric” Latin of the Fredegarius Chronicle written during this period and its numerous “errors” against the classical standards should be viewed as an attempt to represent the new sounds, morphology5 and syntax of the emerging Romance language and to keep in this way the text accessible for an illiterate public. These very strong interferences with oral speech will diminish from 750 on, when the Carolingian reform recommends a readjustment to the Classical norms and leads in this way to an autonomous transcription of the oral speech. From a statistical viewpoint, our corpus does not confirm the evolutions sketched in Table 1. Adnominal is is indeed less frequent than in Classical Latin. But contrary to what might be suggested by Table 1, hic is still often used, whereas iste is exceptional. The most striking feature, however, is the high frequency of ipse.6 Given the historiographical nature of the text, the

5. According to Serbat (1975: 124), we find for instance in the Chronicle of Fredegarius the first occurrence of the Romance future composed of the infinitive and the auxiliary habere agglutinated to the infinitive: (i) Et ille respondebat: Non dabo. Iustinianus dicebat: Daras. (daras = dare + -as < habes) ‘He answered: I will not give. Iustinianus said: you will give.’ 6. These tendencies are not idiosyncratic. The numeric importance of hic and / or the relative paucity of iste have also been observed in other corpora by Fischer (1908), Bonnet (1968: 301, cited by Harris 1978), Selig (1992), Abel (1971), Flobert (1974: 152) and by the numerous studies on the Peregrinatio (inter alii Löfstedt 1911; Nocentini 1990; Christol 1994; Vincent 1997; Fruyt 2003). They are actually inherited from Classical Latin (Fischer 1908). As is noted by Abel (1971) and Bauer (2007: 117), the persistence of the high frequency of hic and of the sporadic use of iste is problematic in the light of their ultimate fate in the Romance languages. According to Abel (1971), the high frequency of hic in the Vetus Latina corpus could be ascribed to its frequent use in fixed expressions, which could be interpreted as a phenomenon of fossilization (Lehmann 1991). This interpretation of the statistical data biased by the results

246

Anne Carlier and Walter De Mulder

meaning of ipse cannot be analysed as the second person demonstrative, resulting from a process of subjectification. Rather, ipse extends its frequency as a marker of anaphora and competes in this role with ille. 2. The development from the demonstrative to the definite article in Late Latin In this section, we will examine several hypotheses that have been proposed to account for the grammaticalization process transforming the third person demonstrative ille into the definite article. Firstly, the semantic shift from demonstrative to definite article has been described as a process of semantic weakening or “bleaching” (Section 2.1). According to a second hypothesis, the demonstrative becomes a definite article when it is selected not to identify the referent but to express the speaker’s subjective attitude towards the referent (Section 2.2). The viewpoint that will be defended here is that the driving force that triggers the grammaticalization process leading to the definite article is intersubjectification (Section 2.3). 2.1. Semantic weakening or “bleaching” 2.1.1. The hypothesis According to Harris (1978: 67–68), a demonstrative determiner serves two interrelated functions: to specify one particular item (as opposed to an indefinite number of alternatives) and to mark the proximity of that item, whether that proximity be spatial, temporal, or simply psychological …; a definite article, on the other hand, when fulfilling its original semantic function …,7 has only the first of

in Romance is however contested by Flobert (1974: 152). The long-lasting high frequency of hic in Late Latin is a subject of further study. 7. According to Greenberg (1978), the definite article can lose its specifying function and become merely a sign of nominality. This hypothesis has been exploited by Harris (1978: 76–78) to analyse the Modern French definite article, on the basis of two observations: – In Modern French, nouns have lost their suffixal endings of number and gender. The large expansion of the articles during the history of the French

The emergence of the definite article

247

these two functions, namely to specify … one particular referent as opposed to all others.

In this perspective, the demonstrative turns into a definite article when it loses the information relating to the proximity of the referent of the nominal expression, i.e. its deictic feature (quoted from Harris 1978: 70; also see Hawkins 2004: 84). Lyons (1999: 161, 332) argues that the shift from demonstrative to definite article is initiated in those contexts of use where the referent is either directly perceived in the physical surroundings or straightforwardly recalled from the preceding discourse. Deictic information is redundant in these cases because the referent is immediately accessible and so semantic weakening or “bleaching” can occur. The new definite article is assumed to be restricted to this range of uses in the initial stage and to expand to other contexts of use in a later stage. 2.1.2. Evaluation of the hypothesis It is generally agreed that the definite article, in comparison with its demonstrative source, lacks the deictic meaning component. Several facts remain nevertheless unexplained in the framework of Lyons’s hypothesis. 2.1.2.1. In the contexts of use where the shift from demonstrative to definite article is initiated according to Lyons’s (1999) hypothesis, deictic inlanguage can be explained in the light of this evolution, since they inherit the grammatical function of marking number and gender. – According to Harris, under the pressure of the generalization of an obligatory prenominal determiner in French, the definite article even extended its use to contexts where it cannot be motivated by its fundamental semantic value, which consists in specifying a particular entity. Harris (1978: 76) mentions in this context the generic use of the definite article le (J’aime les livres; la haine provoque la guerre). It has to be noted that these two evolutions are at least partially independent of each other: in other Romance languages, like Spanish, Portuguese or Italian, where zero determination is still widespread, genericity is nevertheless conceptualized in terms of definiteness. Similarly, in Modern French, the use of the definite article in generic interpretation is semantically motivated and is not merely due to the regression of zero determination (cf. Kleiber 1990a). For a general evaluation of Harris’s analysis of the modern French definite article as a pure marker of nominality, see De Mulder and Carlier (2006).

248

Anne Carlier and Walter De Mulder

formation is redundant, so that proximal and distal demonstratives alternate without semantic distinction. This free alternation can be illustrated for the situational use by the French examples (1) and for the textual use by the Latin examples (2). (1)

a. b.

(2)

a.

b.

Ne renverse pas ce seau-ci! Ne renverse pas ce seau-là! ‘Don’t knock over this / that bucket!’ Lucca castrum dirigunt, atque funditus subvertunt, custodes illius castri capiunt (Continuations § 25) ‘They go to the fort of Loches, they raze it to the ground and take prisoner the guardians of that fort.’ Radulfus haec cernens, castrum lignis monitum in quodam montem super Vnestrude fluvio in Toringia construens, exercitum undique quantum plus potuit collegens, cum uxorem et liberis in hunc castrum ad se definsandum stabilibit. (Fredegarius § 87) ‘Observing this, Radulf put up a fort protected with a wooden stockade on a rise above the banks of the Unstrut, in Thuringia, and when he had assembled from everywhere an army as big as he could, he established himself with his wife and children within this fort to withstand a siege.’

As pointed out by Greenberg (1978: 61) and Himmelmann (1997: 96–98) from a typological perspective, the most common source of the definite article is the distal demonstrative or the third person demonstrative. If there is free alternation between proximal and distal demonstratives in the contexts of use where the shift from demonstrative is initiated, how can we account for this widespread tendency? 2.1.2.2. As has been observed by several empirical studies of Late Latin corpora (Renzi 1976: 29; Orlandini 1981; Calboli 1990; Selig 1992: 165; Fruyt 2003: 109), there is one context were ille occurs frequently, without any serious competition of other determiners:8 ille is used when the nomin-

8. On the basis of the analysis of her corpus, Selig (1992: 165) affirms: “Der Bereich der definite Erstnennungen ist die eigentliche Funktionsdomäne von ille in den hier analysierten Texten” (‘The domain of definite first mentions is the proper functional domain of ille in this corpus’). The use of ipse in combi-

The emergence of the definite article

249

al expression refers to a referent not previously mentioned in the text, whose identity is established by a relative clause. This definite first mention use of ille is illustrated by (3). (3)

… homines illos quos Waiofarius ad defendendam ipsam ciuitatem dimiserat clementiam sue pietatis absoluit dimissisque reuersi sunt ad propria. (Continuationes, § 43) ‘Of his goodness he showed mercy to the [lit. those] men that Waiofar left there to defend the city, and dismissed them to go off home.’

The association of ille with a relative clause was so frequent in Late Latin that it led according to Herman (2000) to a morphological contamination of ille by the relative pronouns qui / cui (nominative singular: illi; dative: illui, lui, ilaei, lei). Selig (1992: 169) observes pertinently that the use of ille in this type of context favors the weakening of the “demonstrative force” and that it foreshadows its later article function. Contrary to what can be expected in the framework of Lyons’s hypothesis, there is however in this context of use no highly accessible referent available, neither in the immediate situation nor in the preceding context.9 2.1.2.3. It is widely accepted that the characteristic context of use of the definite article is the associative anaphor (Kleiber 1992; Himmelmann 1996, 1997), where the definite article gives the instruction to retrieve the referent indirectly, by activating a frame of accessible knowledge which the referent is associated with (Kleiber 1992; Epstein 2002). The demonstrative, by contrast, typically focuses the hearer’s attention on the referent directly, without considering its structural links with a frame of accessible knowledge.10 Neither Lyons’s hypothesis nor the hypothesis deriving the nation with a relative clause establishing the identity of the referent is very exceptional and limited to certain types of documents (Selig 1992: 158). 9. The relevance of this first mention use for the emergence of the definite article also calls into question the idea that the definite article originates from anaphoric uses of demonstratives, as proposed e.g. by Diessel (2006: 477). 10. The opposition between both determiners can be illustrated by the following classical examples (cf. Charolles 1990: 128): (i) Nous arrivâmes dans un village. L’église était sur une colline. ‘We arrived in a town. The church was on a hill.’

250

Anne Carlier and Walter De Mulder

definite article from an anaphoric demonstrative referring to a previously mentioned discourse participant can account for this meaning shift from direct reference to indirect reference, which is crucial in the development of the definite article. 2.1.3. Pragmatic strengthening and semantic weakening Although it is true that the definite article, in comparison with its demonstrative source, has lost the deictic meaning component, more attention should be paid to the factors that set off the grammaticalization process to understand the evolutionary pathway from distal demonstrative to definite article. With respect to the early stages of grammaticalization, Hopper and Traugott (2003: 75–76) point out that “meaning changes and the cognitive strategies that motivate them … are crucially linked to expressivity.” According to the same authors (2003: 94), it is only at later stages, as grammaticalization continues and forms become more routinized, that meaning loss or “bleaching” typically occurs. 2.2. Subjectification Traugott (this volume) puts forward the view that there is subjectification when the strengthening of the speaker’s pragmatic viewpoint is encoded as a semantic feature. A study of the first stages of the development of the definite article exploiting the notion of subjectivity is conducted by Epstein (1993, 1994a–b, 1995, 2001, 2002). Adopting the hypothesis of Traugott and König (1991: 191) according to which “especially in the early stages of (ii) Nous arrivâmes dans un village. Cette église, tout de même, quelle horreur! ‘We arrived in a town. That church, really, how awful!’ As Charolles points out, in the first example, the church is identified in virtue of the stereotypical relation between churches and villages (‘each village has a church’) and is presented as ‘the church-of-the-village’. In this example, the definite article cannot readily be replaced by a demonstrative determiner. The demonstrative determiner is, however, appropriately used in the exclamative utterance that is part of the second example. It presents the referent as a recollection of a past experience of the speaker, in which he was directly confronted with the church in question. Thus, the use of the demonstrative reflects a more subjective construal of the referent.

The emergence of the definite article

251

grammaticalization, new meanings may be added to an element in the form of a ‘strengthening of the expression of speaker involvement’” (Epstein 1994b: 77), he argues that the foregrounding of this speaker involvement and of his subjective attitude towards the referent is the crucial factor in the emergence and evolution of the definite article. 2.2.1. The hypothesis Considering that earlier studies on the definite article privilege a purely referential analysis, Epstein (1993, 1994a, 1994b, 1995, 2001, 2002) claims that the use of the definite article is often motivated by expressive factors. The opposition between the referential and the expressive use of the definite article corresponds according to him to a difference of orientation: the referential use is hearer-oriented insofar as the definite article guides the hearer to the intended referent, whereas the expressive use is speakeroriented and “is motivated by the speaker’s desire to portray the referent as discourse prominent” (Epstein 2001: 186). The expressive value of the definite article, in particular in the early stages of its development, arises according to Epstein (1993: 129) from its demonstrative origin: the demonstrative being the linguistic equivalent of the act of pointing, it identifies an entity in the immediate situation of its utterance and acts at the same time as a marker of attention.11 From a diachronic perspective, Epstein (2001: 186) argues that expressivity is responsible for the shift from demonstrative to definite article: the development of the demonstrative ille into a definite article is initiated when speakers begin to choose – for expressive purposes – a demonstrative in contexts where the identification of the referent normally does not require it. This type of use thus allows the demonstrative to lose its referential properties. An example is given in (4), where the content of the nominal expression is sufficient to identify the bishop and where the demonstrative is used purely for emphasis.

11. In the same vein, Diessel (2006: 469) holds that demonstratives serve two basic functions: “1. First, they indicate the location of a referent relative to the deictic centre. 2. Second, they serve to coordinate the interlocutors’ joint attentional focus.” In his view, the fact that demonstratives coordinate the interlocutors’ joint focus of attention, which is a fundamental element of human communication, makes demonstratives a basic, but also a particular word class.

252 (4)

Anne Carlier and Walter De Mulder

Ductus itaque sanctus Eugenius ad regem, cum illo Arrianorum episcopo pro fide catholica decertavit. (Gregory of Tours, quoted by Trager 1932: 173, cf. Epstein 2001) ‘And so the holy Eugenius was led before the king, and disputed with the bishop [lit. “that bishop”] of the Arians in defense of the Catholic faith.’

2.2.2. Evaluation of the hypothesis 2.2.2.1. The Fredegarius corpus confirms the role of ille as a marker of discourse prominence in Late Latin. Ille appears especially for highlighting a protagonist or an important object at the moment when it is involved in an event that marks a turning point in the narration. Witness the use of ille in the following excerpt, where the author tells how, after repeated military successes of the Persians against the Greeks and their territorial expansion, one moment of distraction of a Persian patrician causes the decisive defeat of the Persians and the annexation of their Empire. The mention of this patrician at the very moment when he is involved in this fatal event is preceded by ille, while the other anaphoric expressions referring to the same referent are not preceded by any determiner. (5)

Aeraclius imperatur arma sumens, telam priliae et falange a suis postergum preparatam relinquens, singolare certamen, ut nouos Dauit procedit ad bellum. Emperator Persarum Cosdroes patricium quidam ex suis quem fortissemum in prelio cernere potuerat huius conuenenciae ad instar pro se contra Aeraglio ad priliandum direxit. Cumque uterque cum aequetis hy duo congressione priliae in inuicem propinquarint, Aeraglius ait ad patricium, quem emperatore Persarum Cosdroae stemabat, dixit: “Sic conuenerat, ut singulare certamen priliandum debuissimus confligere: quare postergum tuom alii secuntur?” Patricius ille girans capud conspecere qui postergum eius uenerit, Aeraglius aecum calcaneum uehementer urguens, extrahens uxum capud patriciae Persarum truncauit. (Fredegarius, § 64) ‘The emperor Heraclius armed himself, left behind him his army drawn in fighting array, and advanced to the fray like a new David. But the Persian Emperor Chrosroes honored their pact by sending one of his patricians [lit.: a certain patrician], whose great value he

The emergence of the definite article

253

knew, to fight in his place against Heraclius. Advancing on horseback to each other, Heraclius said to the patrician, whom he took for the Emperor Chrosroes: “Since we have agreed to single combat, why are those other warriors following behind you?” The patrician turned his head to see who was following him, whereupon in a flash Heraclius spurred his horse forward, drew his short sword and cut off the patrician’s head.’ 2.2.2.2. Although ille, in accordance with Epstein’s hypothesis, can have the role of discourse prominence marker, one could ask if this use has to be considered as speaker-oriented, i.e. expressing a purely subjective judgment, whereas the referential use is assumed to be hearer-oriented. In fact, the speaker also plays an active role in acts of reference and seldom uses determiners only to express subjectivity. He rather tries to give the hearer the key to a correct and coherent interpretation of the discourse, not only by signalling the textual importance of a referent at a certain point of the discourse, but also by guiding the hearer to the intended referent. Hence, referential value and expressive value are not disjoint but complementary and most often closely linked. In our view, the emergence of the definite article has to be conceived from the same perspective: what motivates the development of the articles is not the expression of the speaker’s viewpoint on the referent as such but rather an interactional strategy between speaker and hearer to explicitly mark the way the referent has to be identified as well as the discourse status of the referent. A strict separation between referential and expressive use is therefore artificial. 2.2.2.3. According to Epstein (1993: 127), ille’s role of discourse prominence marker originates in the deictic force of its demonstrative source. However, his analysis of the development of the definite article does not mention ipse, an important competitor in the early grammaticalization process leading to the definite article. Though not a demonstrative, ipse can also mark discourse prominence, on the basis of its meaning as identity marker. Hence, in the following example, ipse can be glossed as “himself” or “in person”. (6)

Eo anno uxor Anaulfi imperatores Persarum nomen Caesara uirum relinquens, cum quattuor pueris totidem puellis ad beatum Iohannem episcopum Constantinopule ueniens, se unam esse de populo dixit et baptismi gratiam ad antedictum beatum Johannem expetit.

254

Anne Carlier and Walter De Mulder

Cumque ab ipso pontifice fuisset baptizatus, agusta Maurici imperatores ea de sancto suscepit lauacro (Fredegarius, § 9) ‘In this year Caesarea, wife of Anauld, the Persian emperor, left her husband and came with four male and four female servants to the blessed John, bishop of Constantinople. She said that she was a private person and besought the blessed John to baptize her. She was baptized by the bishop in person, and the illustrious wife of the Emperor Maurice stood godmother to her.’ A similar observation can be made for the early uses of the indefinite singular article uns in Old French: although devoid of deictic meaning, the incipient article uns only occurs in noun phrases denoting referents that have some prominence in the text, typically protagonists, as witnessed in example (7), whereas no determination is used in noun phrases referring to background entities or referents whose individuality is less important than their role or class membership (8). (7)

(8)

Jadis avient que un lu pramist que char ne mangeriet, ceo dist, les quarante jurs de quareme; ‘It once happened that a wolf made the vow not to eat meat – as he said – during the forty days of Lent.’ (Marie de France, Fables, “Le loup et le mouton”, ed. C. Bruckner) Il avint chose que un pere managier, qui n’avoit nul fiz, mais feme espouse soulement, si escrit ensi en son testament: “Se un fiz m’est engendré, ou plusors, cil me soit hoir.” (J. d’Antioche, La Rettorique de Marc Tulles Cyceron, LXV) ‘It happened that a head of the family, childless although married [lit.: who didn’t have a son, but only a wife], wrote the following words in his testament: “If one or several sons are born to me, they should be my heirs.”’

In view of these facts, it is difficult to maintain that ille owes its role of prominence marker to its demonstrative origin alone. A different hypothesis will be put forward here: in a language stage where zero determination is the rule for a certain type of nominal reference, in this case definite reference, the use of a demonstrative or another linguistic expression in an article-like role is a conscious strategy by which the speaker intends to externalize for the hearer the discourse status of a referent. It is normal that

The emergence of the definite article

255

these explicit marks of text cohesion only bear on nominal expressions whose referent is important enough to have a certain textual permanence.12 2.2.2.4. Nevertheless, we agree with Epstein that in the case of ille, its demonstrative origin did contribute to its discourse-based function. The pathway from demonstrative to definite article is, however, conceived differently in the two approaches. In Epstein’s approach, definite articles evolve from demonstratives as markers of speaker involvement. Invoking the following analysis of Trager (1932), It was an emotional, affective usage, which grew upon the language unconsciously until it became firmly fixed as an article … there is nothing at all of a logical analysis; on the contrary, they are wholly non-logical and subjective (Trager 1932: 54–55, quoted by Epstein 2001)

Epstein (2001: 186) maintains that the transition from the demonstrative ille to the definite article boils down to a process in which the referential value of ille resolves into an expressive value, initiated by the fact that some speakers begin to use ille for expressive purposes when the identification of the referent does not require it. In our view, in contrast, definite articles develop as explicit markers regulating speaker – hearer interaction. We will argue in Section 2.3 that the discourse-based function assumed by ipse and ille in Late Latin does not eliminate their referential instruction but develops from this referential instruction by pragmatic strengthening and tends to be superimposed on it rather than to replace it. This assumption not only allows us to explain the constraints imposed upon the use of ille and ipse in their new role in the speaker-hearer interaction (Section 2.3.1), but also provides a basis for understanding why ille, though less frequent than ipse, was eventually selected as the source of the definite article in most Romance languages (Sec12. This highlighting effect of explicit determination, in contrast with the default case of zero determination, is also stressed by Selig (1992: 117–119). She holds, though, that the focalizing effect of determination disappears in the transition from (Late) Latin to Romance, in the context of the grammaticalization of the (in)definiteness opposition. However, as pointed out above, the incipient indefinite singular article un(s) in Old French also bears on nominal expressions referring to protagonists or other important referents, whereas zero determination, being the default case for indefinite reference, is used when the nominal expression refers to a background entity whose precise identity is less important than the category to which it belongs.

256

Anne Carlier and Walter De Mulder

tion 2.3.2). We will finally show how the hypothesis of a cline of intersubjectification sheds a new light on the grammaticalization process leading from the third person demonstrative ille to the definite article (Section 2.3.3). 2.3. Intersubjectification Late Latin innovates with respect to Classical Latin by the increasing use of referential markers to express definiteness and anaphoric relationships. Consider for example the following passage, where the emphasis on the identity of the referent is not motivated by a risk of ambiguity, since no other king is mentioned in the immediate textual environment. (9)

Dum hec ageretur, Remistanius, filius Eudone quondam, fidem suam quod praedicto regi Pippino promiserat fefellit, et ad Waiofarium iterum ueniens, dictioni sue faciens. Quod Waiofarius cum magno gaudio eum recepit, et adiutorem sibi contra Francos uel praedicto rege eum instituit. Superscriptus Remistanius contra praedicto rege et Francos seu custodias quas ipse rex in ipsas ciuitates dimiserat, nimium infestus accessit, et Bitoriuo seu et Limoticino quod ipse rex adquisierat, praedando nimium uastauit, ita ut nullus colonus terre ad laborandum tam agris quam uineis colere non audebant. Praedictus rex Pippinus in Betoricas per hiemem totum cum regina sua Bertradane in palatium resedit. (Cont. § 50) ‘Meanwhile Remistanius, son of the late Eudo, broke the oath of fealty that he had sworn to the aforementioned King Pippin. He went back to Waiofar and became his man. Waiofar was delighted to receive him and to make use of his help against the Franks and the aforementioned king. The aforementioned Remistianus attacked the aforementioned king and the Franks and the garrisons which this same king had left in the cities, and he destroyed the districts of Berry and also the Limousin, that this same king had conquered and he did this so effectively that not a peasant dared work in the fields and vineyards. The aforementioned King Pippin spent the whole winter with Queen Bertrada in his residence at Bourges.’

The emergence of the definite article

257

It is in the context of this global increase of the marking of anaphoric relationships that we should consider the extended use of ipse and ille that will eventually lead to their status of definite article in the emerging Romance languages. In fact, ipse and ille are already attested as markers of anaphora in Classical Latin. When the antecedent and the anaphoric expression do not have the same head noun, ipse and ille alternate in this role with the demonstratives hic and iste and with the identity marker idem. As pointed out by Pinkster (1996: 148), some form of determination is even obligatory with “empty” nouns like tempus, locus, regio. In Late Latin, ipse and ille become common in more types of contexts and they even extend their use to contexts where the head noun of the anaphoric expression is identical to the head noun of the antecedent, at the expense of zero determination. However, ipse and ille do not alternate freely: their distribution as markers of definiteness and anaphor is a function of their specific meaning. 2.3.1. Competition between ille and ipse 2.3.1.1. As has been pointed out by Bertocchi (2000), ipse in Classical Latin can be interpreted in two divergent directions: presupposing a set of other possible referents, ipse can identify a referent by including those other referents (10a: ‘other things than virtue can be despised, but from all of them virtue is the least expected’) or by excluding them (10b: ‘from him and from no one else, even this is unexpected’). (10)

a. b.

A multis uirtus ipsa contemnitur ‘Even virtue is despised by a lot of people.’ Milonem occisum et ex ipso Clodio audirent ‘They heard the death of Milo from the mouth of Claudius himself.’

It is the exclusive interpretation of ipse that will be exploited in Late Latin to express anaphoric relationships. Ipse, in its exclusive interpretation, is a strong marker of identification. It reinforces the continuity of the referential chain when this chain is endangered because the inserted mention of another referent creates a risk of ambiguity (Selig 1992: 159) or because it is for some other reason unex-

258

Anne Carlier and Walter De Mulder

pected that the referent is identical to the previously mentioned one. It is in this way that the use of ipse can be motivated in the following example: (11)

Inde egressus Crisceco uilla ueniens in Pontio, Leudesio sub dolo fidem promitti se simulans fefellit, facto placito ut coniuracione facta cum pacae discederint. Sed Ebroinus fallaciter agens ut solebat, conpatri suo insidias praeparans ipsum Leudesium interficit; … (Continuationes, § 2) ‘Then Ebroin left Baizieux and reached the domain of Crécy in Ponthieu. He there deceived Leudesius by making a false promise that they should arrange a meeting and, after having exchanged loyalty oaths to each other, should part friends. But, as usual, Ebroin acted treacherously. He laid an ambush for his godfather and slew this same Leudesius.’

It is indeed contrary to the expectations that the main character, after having given his word to his godfather Leudesius, kills this very same person. 2.3.1.2. Being a deictic expression, ille conveys the instruction to identify the referent by means of indications present in the context of its occurrence, whether it is the situational or the textual context. Hence, ille implies that the referent is not presented as it was given before or as already salient in the context. The demonstrative ille can nevertheless refer to a previously mentioned referent, but in this use, it brings about a new identification of the referent and creates thus a discontinuity in the referential chain, without however breaking it down (see the analysis of the French demonstrative determiners by Kleiber 1986a, 1986b, 1987a, 1990b and De Mulder 1997). As was illustrated by example (5), it is for instance typically used when the referent is implied in an event that constitutes the turning point of the narration. The deictic force of ille that induces a new identification of the referent by means of the context of utterance also explains why ille, and not ipse, is used when the anaphoric expression contains new information about the previously mentioned referent by means of a recategorization. (12)

Quando Deo conplacuit, Aubedo ligatarius dirictus a Chlodoueo regi causam legationes usque ad Chrotharium regem Langobardorum, Papia coinomento Ticino ciuitatem Aetaliae peruenisset, cernens regina, quam sepius in legationem ueniens uiderat et ab

The emergence of the definite article

259

ipsa benigne semper susceptus fuerat fuisse retrusam, quasi iniunctum habens exinde inter citera Chrothario regi suggessit quod illam parentem Francorum quam reginam habuerat, per quem etiam regnum adsumserat, non dibuisset umiliare; … (Fredegarius § 71) ‘In God’s good time, Aubedo, sent by King Clovis on an embassy to the Lombard King Rothari, reached the Italian city of Pavia, or Ticinum. Realizing that the Queen, whom he had often seen during his missions and who had always received him well, was incarcerated, he put it in the course of the conversation to King Rothari, as if on instructions, that it would have been better not to humiliate that relative of the Franks who had been given to him as Queen and who had been instrumental in obtaining the throne for him.’ The anaphoric use is not the only textual role of ille. As has been pointed out by Selig (1992: 165), ille occurs also in first mention when the existence of the referent is established by a relative clause. This use is coherent with the instruction conveyed by ille, inducing a new identification of the referent. The relative clause allows in this case the hearer to identify the referent. Ipse is practically not used in this configuration (cf. Section 2.1.2.2 and footnote 8, but also Herman 2000: 101; Fruyt 2003: 104). The first mention use of ille is illustrated by the example (3), repeated here for convenience. (3)

… homines illos quos Waiofarius ad defendendam ipsam ciuitatem dimiserat clementiam sue pietatis absoluit dimissisque reuersi sunt ad propria. (Continuationes, § 43) ‘Of his goodness he showed mercy to the [lit. those] men that Waiofar left there to defend the city, and dismissed them to go off home.’

2.3.1.3. The extension of the conditions of use of ipse and ille can be described as a result of pragmatic strengthening: the speaker/writer puts the original meaning of ipse and ille into service to guide the hearer/reader in his interpretation process of identifying the referent and to underline its importance from the point of view of the textual structure. Although ipse and ille are largely conventionalized in Late Latin as markers of this speaker – hearer interaction, there is no divorce from their original meaning. In

260

Anne Carlier and Walter De Mulder

this context, it has to be mentioned that the Fredegarius corpus does not contain reinforced forms of the type ecce ille or met-ipsimus.13 There are however some occurrences where ipse and ille no longer have their full original meaning, inaugurating the desemantization process that will lead to the creation of the definite article. (13)

Burgaris superatis, noue milia uerorum cum uxoris et liberis de Pannonias expulsi, ad Dagoberto expetint, petentes ut eos in terra Francorum manendum receperit. Dagobertus iobit eos ad iaemandum Badowarius recipere, dummodo pertractabat cum Francis quid exinde fierit. Cumque dispersi per domus Baioariorum ad hyemandum fuissent, consilium Francorum Dagobertus Baioariis iobet ut Bulgarus illus (var. Bulgaros illos) cum uxoris et liberis unusquisque in domum suam una nocte Baiuarae interficerint. Quod protinus a Baiouaries est impletum; nec quisquam ex illis remansit Bulgaris nisi tantum Alciocus cum septinientis uiris et uxoris cum liberis, qui in marca Vinedorum saluatus est. (Fredegarius, § 72) ‘The Bulgars were beaten and nine thousand of them were chased out of Pannonia with their wives and their children. They sought asylum of Dagobert, begging him to take them in and give them a home in Frankish territory. Dagobert gave instructions that they might winter among the Bavarians, and in the meantime he would deliberate with the Franks about their future. When they were dispersed among the Bavarian homesteads for the winter, Dagobert took the advice of his Franks and ordered the Bavarians to kill those Bulgars with their wives and families during the night in their homes. The order was at once carried out. Nobody of those Bulgars survived, with the exception of Alzeco, with 700 men and their wives and their families: they found safety in the Wendish Marsh.’

13. Pei (1936: 49) observes that the reinforced forms ecce ille and ecce iste, which appear in the (preclassical) theatre texts of Plautus, are globally not in evidence in the Late Latin texts. See however the often quoted occurrences in the Peregrinatio. As is noted by Renzi (2000: 185) on the basis of a study on Late Latin charts, in the same way, ipse does not alternate with the reinforced form, but combines its classical meaning and its new anaphoric role. The reinforced form is however attested sporadically in other types of texts.

The emergence of the definite article

261

Since the Bulgarians constitute the principal theme of this anecdote, their identification should not be an issue at the end of the excerpt and the instruction conveyed by ille, requiring a new identification of the referent by means of the context of utterance, is too strong. The process of semantic bleaching is even in a more advanced stage for ipse, especially in the more recent part of the corpus. As witnessed in example (14), the identity marker ipse is used as an anaphoric device although the referential chain is not endangered: the referent is resumed not only under the same denomination, viz. its proper name, but moreover as involved in the same event. (14)

ipse praedictus rex cum reliquis Francis et obtimatibus suis persequendum Waiofarium ire perrexit. Cumque praedictus rex ipsum Waiofarium persequente non repperiret, iam tempus hiems erat, cum omni exercito ad Betoricas, …, reuersus est. (Continuationes, § 49) ‘The King himself set out the remaining Franks and nobles in pursuit of Waiofar. But as he persued the same Waiofar without finding him and it was already winter, he returned with his army to Bourges, where he had left Queen Betrada.’

2.3.2. Selection of ille over ipse Ipse and ille initiate concurrently the grammaticalization process leading to the definite article: ipse conveys the instruction to resume the referent identically and strengthens the continuity of the referential chain, whereas ille signals that the referent should be identified anew by means of indications in the context of its occurrence and introduces in this way a discontinuity in the referential chain even in its anaphoric uses. The existence of two competing forms in the beginning stages of the grammaticalization process is not at all exceptional, as is shown by the archetypal example of negation in French (Old French: ne … pas / point / mie / goutte / …> Modern French ne … pas, Meillet [1912] 1958). It is nevertheless paradoxical that ipse, though more frequent in Late Latin, gave way to ille and was eventually not retained as a definite article in most of the Romance languages (cf. Aebischer 1948; Selig 1992: 133; and Bauer 2007 for an overview). Several hypotheses, phonetic (Banniard 1998), socio-linguistic (Aebischer 1948) and structural (Selig 1992), have been put forward to resolve

262

Anne Carlier and Walter De Mulder

this paradox.14 Our hypothesis takes into account the specific referential instruction conveyed by each of these two competing expressions. In comparison with the demonstrative ille, mobilizing the context of utterance for a new identification of the referent, the identity marker ipse conveys a simpler and more precise referential instruction as an anaphoric device. As long as ipse and ille coexisted, this precision of its referential instruction turned out to the advantage of ipse and explains why it was more frequent. The coexistence of several linguistic expressions as incipient definite articles was not to persist, however, especially when a paradigm of articles started to develop, also containing the indefinite singular article, derived from the unity numeral unus. Indeed, when such a tightening of paradigms or “paradigmatization” (Lehmann 1982) takes place, the resulting highly grammaticalized paradigms tend to be structured in terms of binary or two-term oppositions (Lehmann 1982: 136). The development of a paradigm of articles thus led to the selection of one expression as the definite article, in binary opposition with the indefinite article unus. In this process, most Romance languages finally preferred ille over ipse, because ille had several assets to be selected in the grammaticalization process leading to the definite article.15 Firstly, ille is more flexible as an anaphoric device than ipse. Contrary to ipse, ille does not require that the referent be strictly identical to an aforementioned one; it allows, in virtue of its deictic force, that new elements in the context of its occurrence be taken into account for the identification of the referent. This is the reason why ille is not restricted to anaphora, but is also used in the case of first mentions (cf. example 3). Moreover, ille takes advantage of its status of demonstrative associated with the third person. Whereas hic and iste, linked respectively to the first and the second person, i.e. the speech participants, denote a referent that is wholly identified in the context of their occurrence, ille signals that the identification of the referent is not saturated by the immediate context of its occurrence, and thus suggests that information not available in this context 14. Cf. Carlier and De Mulder (2007: Section 2.5) for a critical overview of these hypotheses. 15. As has been shown by Vincent (1997), this evolution is accompanied by the selection of ille as the third person pronoun. In Late Latin, the distinction between pronoun and determiner does not correspond to a difference of morphosyntactic category, but to a difference of syntactic construction. For certain demonstratives, some specialization is observed (e.g. is, cf. Section 1), prefiguring the emergence of separate paradigms.

The emergence of the definite article

263

should be activated to retrieve the referent. In this way, ille can be understood as an invitation addressed to the hearer to mobilize previous knowledge in order to retrieve the referent.16 This intersubjective dimension of the distal demonstrative is foregrounded in its so-called “recognitional” use, illustrated by the following French example: (15)

Cette personne, tu sais, Mlle E... dont je t’ai parlé, avec qui je devais prendre le thé dans la quinzaine chez Mme Chesneau qui la connaît, eh bien, j’ai appris hier qu’elle a un amant, un grand banquier de Paris, qui ne veut pas l’épouser. (Villiers de L’IsleAdam, Correspondance générale) ‘That person, you know, Ms E… I told you about her, with whom I had to have tea within two weeks in the house of Mrs Chesneau who knows her, well, I heard yesterday that she has a lover, an important banker from Paris, who doesn’t want to marry her.’

As is observed by Himmelmann (1996, 1997: 61, 2001) and Diessel (1999), the recognitional use of demonstratives has a distinctive feature with respect to their other uses: the referent is not mentioned in the preceding context or present in the current discourse situation. The demonstrative points to “specific, presumably shared” knowledge (Himmelmann 2001: 833) – often based upon a common experience – that has to be activated by the hearer in order to identify the referent. Moreover, the speaker is not sure that the hearer will be able to activate this knowledge and to identify the referent – witness the insertion of tu sais “you know” in example (15) – and can provide supplementary information that should allow the addressee to find the referent, for instance by adding a relative clause. Examples of the recognitional use of ille in Late Latin are cited by Selig (1992: 166–167). The following example is written on the back of a royal charter of the Saint Denis Abbey: (16)

Hic sunt carctas de illi thellenio de illo mercatho (Les diplômes originaux des Mérovingiens, éd. Ph. Lauer and Ch. Samarin, quoted by Selig 1992: 166) ‘Here are the documents of the toll of the [lit. that] market (=of

16. An analysis along these lines has been proposed by Vuillaume (1980) for the German distal demonstrative and by Kleiber (1987b) for the Old French distal demonstrative.

264

Anne Carlier and Walter De Mulder

our market, of the well-known market organized on the day of the Patron Saint).’ In its recognitional use, the demonstrative ille is similar to the definite article to the extent that it does not require the referent to be available in the context or the immediate situation. However, there also is an important difference: in the recognitional use of the demonstrative, the referent is identified on the basis of specific knowledge that is shared by speaker and hearer; in characteristic uses of the definite article, those where it cannot be replaced by a demonstrative determiner, the identification of the referent is based on stereotypical knowledge, or knowledge that is presumably shared by all members of the relevant speech community.17 As pointed out by Himmelmann (1997; 2001), the recognitional use of the distal demonstrative allows us to understand the transition between the distal demonstrative and the definite article:18 the distal demonstrative becomes a definite article when the presumably shared knowledge necessary to identify the referent is no longer presented as specific to the speaker and the hearer, but as consisting of information presumably shared by all members of the speech community. In sum, the evolution of the Latin demonstrative system, which eventually led to the creation of the definite article, has been shown to be far more intricate than what could be concluded from the schema presented by Harris (1978) that was used by Traugott and Dasher (2002) as an illustration of the process of subjectification. Nevertheless, the hypothesis of a cline of intersubjectification is still relevant to describe the evolution from demonstrative to definite article.

17. Himmelmann (1997, 2001) considers the defining uses of the definite article to be its larger situation uses, where the article refers to entities that are considered to be unique in a given speech community (e.g. the sun, the Queen, the pub), and its associative anaphoric uses, where the uniqueness expressed by the definite article is justified by stereotypical knowledge. 18. As pointed out by Himmelmann (1997: 94–101), this hypothesis is less problematic than the more traditional idea that the definite article develops from the demonstrative’s anaphoric uses (which has recently been resumed by Diessel 2006: 477).

The emergence of the definite article

265

2.3.3. The cline of intersubjectication Traugott (this volume) presents the cline of intersubjectification as follows: Non-/ less-subjective

–

Subjective

–

Intersubjective

As to the identification of the different stages of this evolution, it is an oversimplification to identify the Classical Latin use of demonstratives with the first stage of this evolution: the inclusion or exclusion with respect to the deictic centre is very easily subject to an evaluative interpretation and the first stage should perhaps be conceived rather as a theoretical starting point. In Classical Latin, some demonstratives can already be used as social deictics and encode the subjective attitude of the speaker towards the referent (cf. Section 1). A new evolution is observed in Late Latin, when ipse and ille become involved in a grammaticalization process that will lead to the creation of a definite article. The force that triggers this grammaticalization process is strategic interaction between the writer and the reader:19 the writer aims to offer to the reader the clues needed for a clear and coherent interpretation of the text and for the identification of the referents participating in the narrative. In Late Latin, definite reference to the participants in the narrative is normally marked by zero determination. However, when the writer wants to attract the reader’s attention to the referential act, either because the precise and exact identification of the intended referent is not guaranteed, or because the referent has topical or focal status, he develops explicit markers for this purpose. He makes extended use of ille and, even more markedly, of ipse by exploiting their original meanings: by using the identity marker ipse, he gives the instruction to resume a previously mentioned referent identically; by means of the demonstrative ille, he asks the reader to make a new identification of the referent. Since the use of these expressions is marked with respect to zero determination, it can have as such the effect of attracting attention to their referent and of conveying it topical or focal value. In the case of ille, the role of “attention-getter” relies moreover on its deictic meaning, in virtue of which ille requires the hearer to take into account the specific occurrence of the demonstrative and its context.

19. Traugott and Dasher (2002) use the terms speaker / writer and addressee / reader. Since our corpus is a written text, we will use writer and reader. In so doing, we do not take into account subtle distinctions between writer, author, narrator etc., which do not seem essential to the point we try to make here.

266

Anne Carlier and Walter De Mulder

Traugott (2003: 126) defines subjectification as “the mechanism whereby meanings come over time to encode or externalise the speaker / writer’s perspectives and attitudes as constrained by the communicative world of the speech event, rather than by the so-called ‘real-world’ characteristics of the event or situation referred to”. According to this definition, subjectification is involved in the first stage of the grammaticalization process that leads to the creation of the definite article:20 by the extended use of ipse and ille, the writer verbalizes his perspective on the referent, its discourse status and its role in creating text connectedness. Even if ipse as an identity marker and ille as the distance demonstrative already had this meaning component in Classical Latin, their extended use in Late Latin leads to the foregrounding of this meaning component. In line with this subjectification process, it is not surprising that in these uses, the demonstrative ille implies an imbalance between writer and reader, since it signals that the writer is striving to ensure unambiguous reference, by guiding the reader to the referent. As pointed out above, since the writer is not sure that the reader will be able to identify the referent, he frequently adds, in recognitional uses of the demonstrative, relative clauses that should give the reader more identifying information. In this way, it can truly be said that subjectification involves a centring of meaning on the speaker or writer (Traugott 2003: 129). However, if the use of a demonstrative ille signals that the writer attracts the reader’s attention to the reference act – amongst other things because he is not sure that the hearer will be able to identify the referent –, this implies that the writer has a representation of the reader’s text model and takes into account this representation of the reader’s text model when choosing the referential expression.21 In as far as the demonstrative is an expression of the writer’s awareness of the beliefs of the reader, it is also an expression of intersubjectivity.22 The question might be raised whether the evolution from

20. Epstein also considers subjectification to be involved in the emergence stage of the definite article. In his conception, however, subjectification is restricted to expressive uses, as opposed to referential uses (cf. § 2.2). 21. The term text model is used here as a shorthand for the representation of the meaning of the text that is elaborated by the reader when reading the text. The crucial importance of man’s capacity to understand the other’s mental or intentional states, as well as its close relationship to pointing and to the use of demonstratives, has been pointed out by Diessel (2006: 468). 22. Indeed, “intersubjectivity in my view refers to the way in which natural languages, in their structure and their normal manner of operation, provide for the locutionary agent’s expression of his or her awareness of the addressee’s atti-

The emergence of the definite article

267

demonstrative to definite article can be described in terms of a strengthening of this intersubjective dimension. Intersubjectification can be defined as the mechanism whereby the meaning of an expression, instead of being centred on the speaker as a result of a preceding subjectification, becomes “more deeply centred on the addressee” (Traugott 2003: 129). Traugott (2003: 130) cites the three-step evolution of let’s as a typical example of the cline of intersubjectification: (i) Let us go, will you: “allow us (imp.)” > (ii) Let’s go, shall we: “I propose (hortative)” > (iii) Let’s take our pills now, Roger: “mitigator / marker of ‘care-giver register’”.

The transition from the first to the second step involves subjectification; the transition from the second to the third step represents an instance of intersubjectification, since the speaker “in the newer use positions him- or herself as empathetic to the addressee’s possible objection to the projected activity (in this case, pill-taking) in the here and now of the ongoing discourse” (Traugott 2003: 130). The meaning change between the two last stages can actually also be described as a shift in perspective: whereas in the second stage, the event is presented from the speaker’s perspective and reflects his point of view, in the third stage, the speaker adopts the perspective of the hearer. The transition from demonstrative to definite article shows a similar shift in perspective. For the demonstrative, the role of the writer / speaker is dominant: in the initial stage of the grammaticalization process, demonstratives are used when the writer / speaker feels it is necessary to attract the reader’s / hearer’s attention, because he thinks the hearer / reader will have to make some effort to identify the referent or because the referent has a prominent role in the text or discourse. The definite article signals in its typical uses (cf. footnote 17) that the hearer can identify the referent because the knowledge involved in the identification is stereotypical or presumably shared by all members of the speech community. In other terms, whereas the demonstrative signals that the referent is presented dominantly from the speaker’s perspective, the emergence of the definite article implies a shift to a less speaker-centered and more intersubjectively shared perspective. tudes and beliefs, most especially their face or self-image” (Traugott 2003; this volume: 33).

268

Anne Carlier and Walter De Mulder

The analysis presented above might seem surprising in the light of recent papers by Traugott, where she focuses more exclusively on attitudinal aspects rather than on text coherence as she distinguishes between subjectification and intersubjectification: subjectification is the development of meanings that express speaker attitude or viewpoint, while intersubjectification is the development of meanings that express speaker attention to addressee self-image (Traugott 2003: 126, 130; this volume). In this perspective, honorifics and other items that convey social attitudes of politeness are prototypical examples of intersubjectification. However, even in these articles, some examples allow for a broader conception of subjectification and intersubjectification, which was also present in earlier work, such as Traugott (1995: 47), for example, where subjectification is defined as “the tendency to recruit lexical material for purposes of creating text and indicating attitudes in discourse situations”. In this view, connectives, expressing coherence relations between propositions, or anaphoric or cataphoric pronouns can be considered as instances of subjectification (Traugott 1982: 250), because it is the speaker who decides how to link the propositions in the text to each other and how to link propositions to the context. More generally put, subjectification also includes “the development of meanings with which the speaker ‘creates text’” (Breban 2006: 260). As far as referential expressions are concerned, then, we agree with Breban (2010: 111) that “attitudinal and textual subjectivity are … two different manifestations of the same phenomenon neither of which can be excluded at the expense of the other”; subjectification as well as intersubjectification have according to us both an attitudinal and a textual dimension. 3. Conclusion This paper has offered an account of the emergence of the definite article in the transition from Late Latin to French. It has tried to elucidate the enigma of the selection of ille as the source of the definite article in French and in most Romance languages, in spite of its low frequency in comparison with its competitor ipse in Late Latin. It has been argued that the grammaticalization from the distal demonstrative ille into the definite article cannot be reduced to a linear process of desemantization, consisting only in the loss of its deictic meaning, but that it also involves a stage of pragmatic enrichment: by using an identity marker such as ipse or a demonstrative such as ille, the speaker not only tells the hearer how the referent of the noun

The emergence of the definite article

269

phrase has to be identified; in a language stage where zero determination is the rule, the use of these expressions is also a strategy by which the speaker conveys to the hearer an instruction to pay attention to the identification of the referent, thus highlights it and verbalizes its discourse status. In her production based approach to change, Traugott (this volume) argues that subjectification should precede intersubjectification and that it is likely to occur in the early stages of the grammaticalization process. The current study shows how grammaticalization and subjectification are correlated in the first stage of the process leading to the definite article: this process is initiated when speakers increasingly use existing expressions like ipse and ille in order to facilitate the identification of the referent and indicate its discourse status, strengthening thereby their respective subjective meaning component. In this initial stage the role of the speaker is dominant. Our analysis also brings to light a shift in balance between speaker and hearer as grammaticalization proceeds: the transition to a full-fledged definite article is accompanied by a more symmetric relationship between speaker and hearer and involves in this way a strengthening of the intersubjective dimension. Intersubjectification as it is conceived here is not limited to the encoding of attitudinal aspects; it also concerns more globally items that materialize the strategic interaction between speaker and hearer and reflect the active role of the speaker to orient and to guide the hearer in his interpretational tasks. These reflections raise the question how intersubjectification should be defined, a need for precision that is also hinted at at the end of Traugott’s article. References Abel, Fritz 1971

L’adjectif démonstratif dans la langue de la Bible latine. Tübingen: Niemeyer. Aebischer, Paul 1948 Contribution à la protohistoire des articles ille et ipse dans les langues romanes. Cultura neolatina 8: 181–203. Banniard, Michel 1992 Viva voce: communication écrite et communication orale du IVe au IXe siècle en Occident latin. Paris: Institut des Etudes Augustiniennes.

270

Anne Carlier and Walter De Mulder

Banniard, Michel 1998 Ille et son système: chronologie du développement (IIIe–VIIIe s.). In Latin vulgaire, latin tardif IV, Louis Callebat (ed.), 313–321. Hildesheim: Ols-Weidmann. Bauer, Brigitte L.M. 2007 The definite article in Indo-European: Emergence of a new grammatical category. In Nominal Determination: typology, context constraints, and historical emergence, Elisabeth Stark, Elisabeth Leiss and Werner Abraham (eds), 103–139. Amsterdam/Philadelphia: Benjamins. Bertocchi, Alessandra 2000 Ipse as an intensifier. Papers on Grammar 5: 15–30. Bonnet, Max 1968 Le latin de Grégoire de Tours. Hildesheim: Georg Olms. Breban, Tine 2006 Grammaticalization and subjectification of the English adjectives of general comparison. In Subjectification: Various Paths to Subjectivity, Angeliki Athanasiadou, Costas Canakis and Bert Cornillie (eds.), 241–277. Berlin/New York: Mouton de Gruyter. Breban, Tine 2010 English Adjectives of Comparison: Lexical and Grammaticalized Uses. Berlin/New York: Mouton de Gruyter. Calboli, Gualtiero 1990 Les pronoms démonstratifs latins et la formation de l’article roman. Revue de Philologie 64: 71–88. Carlier, Anne 2001 La genèse de l’article un. Langue française 130: 65–88. Carlier, Anne and Walter De Mulder 2007 Les premiers stades de développement de l’article défini. In Etudes sur le changement linguistique en français, 85–113, Bernard Combettes and Christiane Marchello-Nizia (eds). Nancy: University press. Charolles, Michel 1990 L’anaphore associative: problèmes de délimitation. Verbum 13: 119– 148. Christol, Alain 1994 Ipse: ‘Articloïde’ ou article dans la Peregrinatio? Lalies 13: 143– 153. De Mulder, Walter 1997 Les démonstratifs: des indices de changement de contexte. In Entre général et particulier: les déterminants, Nelly Flaux, Danièle Van de Velde and Walter De Mulder, 137–200. Arras, Artois P.U.

The emergence of the definite article

271

De Mulder, Walter and Anne Carlier 2006 Du démonstratif à l’article défini: le cas de ce en français moderne. Langue française 152: 96–113. Diessel, Holger 1999 Demonstratives. Form, Function, and Grammaticalization. Amsterdam/Philadelphia: Benjamins. Diessel, Holger 2006 Demonstratives, joint attention and the emergence of grammar. Cognitive Linguistics 17: 463–489. Epstein, Richard 1993 The definite article: early stages of development. In Historical Linguistics 1991. Papers from the 10th International Conference on Historical Linguistics, Amsterdam, 12–16 August 1991, Jaap Van Marle (ed.), 111–134. Amsterdam/Philadelphia: Benjamins. Epstein, Richard 1994a Discourse and Definiteness: Synchronic and Diachronic Perspectives. Ph.D., San Diego, Univ. of California. Epstein, Richard 1994b The development of the definite article in French. In Perspectives on Grammaticalization, William Pagliuca (ed.): 63–79. Amsterdam/Philadelphia: Benjamins. Epstein, Richard 1995 L’article défini en ancien français: l’expression de la subjectivité. Langue française 107, 58–72. Epstein, Richard 2001 The meaning of definite articles in cross-linguistic perspective. In Cognition in Language Use, Selected Papers from the 7th International Pragmatics Conference 1, EnikĘ Németh (ed.), 174–189. Antwerp, International Pragmatics Association. Epstein, Richard 2002 The definite article, accessibility, and the construction of discourse referents. Cognitive Linguistics 12: 333–378. Fischer, Anton 1908 Die Stellung der Demonstrativpronomina bei lateinischen Prosaikern. Tübingen: Heckhauer. Flobert, Pierre 1974 Review of Abel (1971), Bulletin de la Société de Linguistique de Paris 69: 149–153. Fruyt, Michèle 2003 Anaphore, cataphore et deixis dans l’Itinerarium d’Egérie. In Latin vulgaire, latin tardif VI: Actes du VIe colloque international sur le latin vulgaire et tardif (Helsinki, 29 août – 2 septembre 2000), Heik-

272

Anne Carlier and Walter De Mulder

ki Solin, Martti Leiwo and Hilla Halla-Aho (eds.), 99–119. Hildesheim: Olms-Weidmann. Greenberg, Joseph H. 1978 How does a language acquire gender markers. In Universals of Human Language 3, Joseph H. Greenberg (ed.), 47–82. Stanford: U.P. Harris, Martin 1978 The Evolution of French syntax: a comparative approach. London: Longman. Harris, Martin 1980 The marking of definiteness in Romance. In Historical Morphology, Jacek Fisiak (ed.), 141–156. The Hague: Mouton. Hawkins, John A. 2004 Efficiency and Complexity in Grammars. Oxford: Oxford University Press. Herman, József 2000 Morphologie pronominale et évolution syntaxique: remarques sur la formation des formes ‘vulgaires’ de ille. Papers on Grammar 5: 95– 107. Himmelmann, Nikolaus 1996 Demonstratives in narrative discourse: a taxonomy of universal uses. In Studies in Anaphora, Barbara A. Fox (ed.), 205–254. Amsterdam/Philadelphia: Benjamins. Himmelmann, Nikolaus 1997 Deiktikon, Artikel, Nominalphrase. Zur Emergenz syntaktischer Struktur. Tübingen: Niemeyer. Hopper, Paul J. and Elizabeth Closs Traugott 2003 Grammaticalization. 2nd rev. ed. [1st ed., 1993.] Cambridge: Cambridge University Press. Kleiber, Georges 1986a Adjectif démonstratif et article défini en anaphore fidèle. In Déterminants: syntaxe et sémantique, Jean David and Georges Kleiber (eds.), 169–185. Paris: Klincksieck. Kleiber, Georges 1986b Pour une explication du paradoxe de la reprise immédiate UN NiĺLE Ni / UN NiĺCE Ni. Langue française 72: 54–79. Kleiber, Georges 1987a L’énigme du Vintimille ou les déterminants à quai. Langue française 75: 107–122. Kleiber, Georges 1987b L’opposition cist / cil en ancien français ou comment analyser les démonstratifs? Revue de linguistique romane 51: 5–35.

The emergence of the definite article

273

Kleiber, Georges 1990a L’article le générique: la généricité sur le mode massif. Genève: Droz. Kleiber, Georges 1990b Celui-ci / celui-là: un marqueur référentiel hybride. In Enonciation et parti pris, Walter De Mulder, Franc Schuerewegen and Liliane Tasmowski (eds.), 237–250. Actes du colloque d’Anvers. Kleiber, Georges 1992 Article défini, unicité et pertinence. Revue romane 27: 61–89. Lehmann, Christian 1982 Thoughts on Grammaticalization: A Programmatic Sketch, Vol. I. Köln, Universität zu Köln, Institut für Sprachwissenschaft (Arbeiten des Kölner Universalien-Projekts 48). Lehmann Christian 1991 Grammaticalization and related changes in contemporary German. In Approaches to grammaticalization II, Elizabeth Closs Traugott and Bernd Heine (eds), 493–535. Amsterdam/Philadelphia: Benjamins. Löfstedt, Einar 1911 Philologische Kommentar zur Peregrinatio Aetheriae. Uppsala. Lyons, Christopher 1999 Definiteness. Cambridge: Cambridge University Press. Meillet, Antoine 1958 Reprint. L’évolution des formes grammaticales. In Linguistique historique et linguistique générale, Antoine Meillet, 130–148. Paris: Champion. Original publication in Scientia (Rivista di Scienza) 12, 1912. Menge, Hermann 2000 Lehrbuch der lateinischen Syntax und Semantik. Darmstadt: Wissenschaftliche Buchgesellschaft. Nocentini, Alberto 1990 L'uso dei dimostrativi nella Peregrinatio Egeriae e la genesi dell'articolo romanzo. In Atti del Convegno Internazionale sulla Peregrinatio Egeriae (Arezzo, 23-25 ottobre 1987), 137–165. Arezzo: Accademia Petrarca di Lettere Arti e Scienze. Orlandini, Anna 1981 Wesen und Entwicklung des Artikels vom Lateinisch zu den romanischen Sprachen. Indogermanische Forschungen 86: 223–247. Pei, Mario A. 1936 Old French Demonstratives. Language 12: 47–51 Pinkster, Harm 1996 Review of Selig (1992). Mittellateinisches Jahrbuch 31: 143–149.

274

Anne Carlier and Walter De Mulder

Renzi, Lorenzi 1976 Grammatica e storia del articulo ialiano. Studii di grammatica italiana 5: 5–42. Renzi, Lorenzi 2000 Storia di ipse. In La preistoria dell’italiano, József Herman, Anna Marinetti and Luca Mondon (eds.), 181–203. Tübingen: Niemeyer. Selig, Maria 1992 Die Entwicklung der Nominaldeterminanten im Spätlatein. Romanischer Sprachwandel und lateinische Schriftlichkeit. Tübingen: G. Narr. Serbat, Guy 1975 Les structures du latin: le système de la langue classique; son évolution jusqu’aux langues romanes. Paris: Picard. Trager, George L. 1932 The Use of the Latin Demonstratives (Especially Ille and Ipse) up to 600 A.D., as the Source of the Romance Article. New York: Publications of the Institute of French Studies. Traugott, Elizabeth Closs 1982 From propositional to textual and expressive meanings. In Perspectives on Historical Linguistics, Winfred P. Lehmann and Yakov Malkiel (eds), 245–271. Amsterdam/Philadelphia: Benjamins. Traugott, Elizabeth Closs 1995 Subjectification in grammaticalization. In Subjectivity and Subjectivisation: Linguistic Perspectives, Dieter Stein and Susan Wright (eds.), 31–54. Cambridge: Cambridge University Press. Traugott, Elizabeth Closs 2003 From subjectification to intersubjectification. In Motives for Language Change, Raymond Hickey (ed.), 124–139. Cambridge: Cambridge University Press. Traugott, Elizabeth Closs this vol. (Inter)subjectivity and (inter)subjectification: A reassessment. In this volume, 29–71. Traugott, Elizabeth Closs and Richard B. Dasher 2002 Regularity in Semantic Change. Cambridge: Cambridge University Press. Traugott, Elizabeth Closs and Ekkehard König 1991 The semantics-pragmatics of grammaticalization revisited. In Approaches to Grammaticalization 1, Elizabeth Closs Traugott and Bernd Heine (eds), 189–218. Amsterdam/Philadelphia: Benjamins. Vaänänen, Veikko 1967 Introduction au latin vulgaire. Paris: Klincksieck.

The emergence of the definite article

275

Vincent, Nigel 1997 The emergence of the D-system in Romance. In Parameters of Morphosyntactic Change, Ans Van Kemenade and Nigel Vincent (eds), 149–169. Cambridge: Cambridge University Press. Vuillaume, Marcel 1980 La déixis en allemand. Thèse d’Etat, Paris-Sorbonne.

On the subjectification and intersubjectification paths followed by the adjectives of completeness Lobke Ghesquière Abstract In the literature, it has been hypothesized that prenominal elements of the English NP may undergo increasing (inter)subjectification and move to the left periphery of the NP structure in the process (Adamson 2000). Such developments are generally assumed to be unidirectional and irreversible. This study presents a case study on the development of the adjectives complete, total and whole and how it fits in with the proposed hypotheses. All three adjectives are shown to have developed (inter)subjectified uses, albeit to different degrees and via distinct pathways. Two possible pathways of semantic change are attested: (i) from objective descriptive modifier to subjective descriptive modifier to emphasizer, and (ii) from objective descriptive modifier to secondary determiner to emphasizer. Whereas the first pathway adheres to the leftward movement hypothesis, the second does not. Moreover, the second pathway challenges the unidirectionality hypothesis, constituting a shift from textually intersubjective to subjective meaning. To account for the attested changes, I propose to incorporate Traugott’s (1982) and (2003) diachronic clines of semantic change into a synchronic segmented model of the NP.

1. Introduction1 The English noun phrase has a very specific functional-structural make up, with a potentially long string of adverbial and adjectival prenominal modifiers intervening between determiners and head noun, as in

1. The research reported on in this article was funded by the Interuniversity Attraction Poles Programme – Belgian State – Belgian Science Policy Office, project P6/44 Grammaticalization and (inter)subjectification and supported by grant no. HUM2007–60706/FILO of the Spanish Ministry of Education and Science and the European Regional Development Fund. I owe a great debt of gratitude to Kristin Davidse for her guidance, input and support in writing this article.

278 (1) (2)

Lobke Ghesquière

the whole repressive feminist, safe-sex, politically correct sexual climate (CB) The same old handsome charming Henry! (CB)

Studies into the structure of the prenominal string have given rise to a number of hypotheses. Synchronically, the prenominal string is often considered to constitute a left-to-right subjective–objective continuum (e.g. Quirk et al. 1972; Halliday and Hasan 1976; Hetzron 1978; Dixon 1982; Halliday 1994; Bache 2000; Vandelanotte 2002). Diachronically, it has been hypothesized that prenominal elements acquire increasingly subjective meanings and at the same time move to more leftward positions in the noun phrase structure (Adamson 2000). This paper investigates the development of three adjectives of completeness as a test case of Adamson’s (2000) leftward movement hypothesis, which for its theorizing refers mainly to Traugott. Three synchronic and diachronic corpus studies on the adjectives complete, total and whole will be presented and their results examined against the above hypothesis. The structure of the paper is as follows. Firstly, the necessary theoretical background will be introduced, briefly discussing the structure of the English noun phrase and some hypotheses on the diachrony of the different prenominal uses (Section 2). Secondly, the main findings of the three databased studies carried out for this paper will be presented (Section 3). Thirdly and finally, a number of conclusions and remarks will be made, based on the results of the case studies (Section 4). 2. Descriptive-theoretical framework 2.1. The structure of the English noun phrase: a subjective  objective continuum The descriptive-theoretical approach used in this study relies mainly on functionalism, complemented with general insights from cognitive grammar. In this framework, the premodifiers and head of the English noun phrase are generally recognized to embody a continuum from subjective to objective meanings (e.g. Quirk et al. 1972; Halliday and Hasan 1976; Hetzron 1978; Dixon 1982; Halliday 1994; Bache 2000; Vandelanotte 2002). As first suggested by Bache (2000) and further developed by Breban (2010), this continuum can be thought of as comprising three main struc-

The adjectives of completeness

279

tural-functional zones, concerned with determination, modification and categorization respectively. In the following paragraphs, these zones and the specific functional slots they contain will be briefly introduced, with specific attention to their objective, subjective, or intersubjective nature. 2.1.1. The determination zone The determination zone is situated at the leftmost end of the noun phrase and comprises from left to right the predeterminers, the core or primary determiners, and the secondary determiners2 (Table 1). These determining elements specify how the instances of a type referred to by the noun phrase can be related to the information about deictically anchored referents which the hearer has built up from the preceding discourse (Langacker 2001). Table 1. The determination zone in the English noun phrase all most of predeterminer

those those the primary determiner

same secondary determiner

problems problems problem head

Primary determiners give either identifying or quantifying information in a broad sense (Langacker 1991; Bache 2000; Davidse 2004). They can signal three basic identifiability statuses (Langacker 1991): (i) instances may be introduced into the text and marked as (not yet) identifiable, (ii) (sets of) instances may be referred to which have already been identified in the discourse, or (iii) instances may be accessible as a portion of an already identified set. Immediately preceding and following the primary determiner there may be predeterminers and secondary determiners, which form a functional unit together with the determiner, identifying the instance referred to more precisely. In other words, they “help single out or quantify the referent” 2. Bache (2000: 160) distinguished the determination zone realized by identifiers such as articles, demonstratives and possessives from the specification zone realized by secondary determiners, which can be numeratives, adjectives, etc. Breban (2010: 24–34), however, argued that a better functional generalization is arrived at if all the elements contributing to the determination of nominal referents are viewed as one structural-functional zone.

280

Lobke Ghesquière

(Bache 2000: 235), as in such a problem and the main problem. Secondary determiners add supplementary “deictic qualifications that are more semantically diverse and more complex than the primary determiners they are bound by” (Davidse, Breban and Van linden 2008: 500). How should the meanings conveyed by determining elements be classified in terms of subjectivity  objectivity? Determiners aid the speaker to create discourse that is so clear and coherent as to allow the hearer or reader to establish mental contact with the referent in question. They serve a textual function in the discourse, “focused on coordinating the interlocutors’ joint attention” (Diessel 2006: 463).3 Diessel’s understanding of deixis stresses the intersubjective nature of the cognitive operations involved in it, as shown by the following quote: In order to communicate, actor and addressee must jointly focus their attention on the same entity or situation. To this end, the actor directs the addressee’s attention to a particular reference object in the surrounding situation … If the communicative act is successful, the communicative partners focus their attention on the same referent. (Diessel 2006: 465)

Traugott (this volume) understands textual meanings as comprising elements that “serve more contentful (and sometimes truth-conditional) purposes of local connectivity (e.g. relativizers, complementizers), whereas others serve the procedural purposes of expressing speaker’s attitude to the text under production (topicalizers, discourse markers)” (Traugott this volume). Textual meanings can thus in Traugott’s view be both objective and subjective. However, following Breban (2010), I argue that textual elements can display a specific type of intersubjectivity. Traugott’s (this volume: 31) definition of intersubjectivity focuses more on meanings relating “to the addressee and addressee’s face”, i.e. “to certain turn-taking contexts that are bracketed by hedging discourse markers like well, tags like clausefinal right?, addressee-oriented expressions such as question-markers, or, more generally, addressee honorifics” (Traugott 2007: 298). However, it would seem that in accordance with Traugott’s own (1995: 47) broad view of subjectification, which encompasses both a text-creating and an attitudinal component, intersubjectivity can be extended to include deictic meanings, by which the speaker negotiates discourse referent tracking for the hearer. Carlier and De Mulder (this volume: 269) arrive at a similar conclusion; they hold that intersubjectification then “concerns more globally items 3. The notion of joint attention is used by Diessel (2006) to define the deictic function of demonstratives.

The adjectives of completeness

281

that materialize the strategic interaction between speaker and hearer and reflect the active role of the speaker to orient and to guide the hearer in his interpretational tasks”. In my view, this understanding of intersubjectivity does not exclude Traugott’s (2007; this volume) and Traugott and Dasher’s (2002) definition. As secondary determiners are a means of expressing deixis, of aiding the identification of the noun phrase referent by the addressee, they are compatible with the notion of intersubjectivity in the Traugottian sense of “crucially involv[ing] SP/W’s [speaker/writer’s] attention to AD/R [addressee/reader] as a participant in the speech event” (Traugott and Dasher 2002: 22). In this study, I will thus consider secondary determiner uses of adjectives as conveying textually intersubjective meaning. 2.1.2. The modification zone To the right of the determination zone, we find the modification zone, which accommodates both adjectival and adverbial modifiers. The right end of the modification zone can be filled by adjectives that describe properties or qualities of the entities referred to by the noun phrase. As pointed out by Quirk et al. (1972: 924), the descriptive modifiers situated within this zone describe either more objectively recognizable properties, as in a tall man, or more subjectively assessible properties, which are more a matter of the speaker’s opinion, as in a beautiful man. Accordingly, they are placed more to the right or more to the left of the modification zone (see also Hetzron 1978: 178). Hence we find splendid old trains rather than old splendid trains (Halliday 1994) and a lovely little example rather than a little lovely example (Adamson 2000). Irrespective of their subjective or objective nature, descriptive modifiers manifest the syntactic behaviour of ‘central’ modifiers (Bache 2000): they can alternately be used in predicative and prenominal position and as they typically depict a gradable property, they can be submodified by intensifiers, e.g. these trains are splendid/old; very splendid/old trains. Intensifiers are typically adverbial and indicate the extent of the property or quality denoted by the descriptive modifiers they precede. Two types of intensifier can be distinguished, viz. bleached and non-bleached.4 4. Bleaching refers to the reduction of semantic complexity of a lexical item and is mostly interpreted as a shift from expressing lexical meaning, which is more concrete and specific, to expressing grammatical meaning, which is concerned

282

Lobke Ghesquière

Bleached intensifiers, like very in a very nice man, have lost most of their original meaning and are often viewed as a type of degree modifiers (Quirk et al. 1972; Paradis 1997), specifying the position of the adjective they modify on an implied quantitative scale. “They come in two basic semantic sets: intensifiers that scale the meaning of the modified element upwards from an assumed norm, and those that scale it downwards” (Nevalainen and Rissanen 2002: 361), viz. amplifiers and downtoners respectively (Quirk et al. 1972: 439). Amplifiers can be further divided into maximizers, such as completely and totally, which “denote the upper extreme of the scale” and boosters, such as very and greatly, which “denote a high degree, a high point on the scale” (Quirk et al. 1972: 444). Non-bleached intensifiers, or value adjectives (Dixon 1982: 15), such as lovely in lovely long legs or good in a good strong box, have not bleached to the same degree as true intensifiers and still have an inherent evaluative meaning. At the extreme left of the modification zone, there can be adjectives which do not describe a property or quality of the entity referred to, but convey strong speaker feelings towards it (Sinclair 1990: 69). These emphasizers are typically used with nouns containing inherently evaluative or scalar notions which they further emphasize, as in complete idiot or whole trains of years respectively. Emphasizing adjectives have scope5 over the entire noun phrase string to their right, whether this is just the head, as in the two previous examples, or the head with one or more intervening elements, as in a complete bloody Nazi (CB Times) and total rock star assholes (CB UK Spoken). They differ from descriptive modifiers in that they do not allow for predicative alternation or for intensification: *the idiot is complete, *a very complete idiot.

with more abstract domains (e.g. intensification) and is applicable to a larger number of contexts. The grammaticalized intensifier very, for instance, has lost most of its original truth semantics and now only intensifies or heightens the meaning of the descriptive modifier over which it has scope. The adverb can consequently be used in a large number of contexts and is no longer restricted to truth-propositional contexts. 5. Scope refers to the ability of a word or unit to apply over a certain domain and leave its mark on the entirety of this domain (McGregor 1997: 210).

The adjectives of completeness

283

2.1.3. The categorization zone In English the dependency structure of the noun phrase is such that the head noun, which designates the type of which the noun phrase referent is an instance (Langacker 1991), occurs at the rightmost end. The general type denoted by the head may be further delineated by classifiers preceding the head noun, as in steam train, electric train, which designate subtypes of the type ‘train’. Like emphasizers and secondary determiners, classifying adjectives do not manifest the syntactic behaviour of central adjectives: they cannot be used as predicates or be intensified: *these trains are electric, *very electric trains. Although classifiers, which typically reflect culturally entrenched taxonomies, do not describe gradable properties, they do allow for a marked form of gradability. Submodifiers of classifiers, such as purely, largely and mainly, indicate to what extent the subcategorization expressed by the classifier applies, as in a purely theoretical discussion (Sinclair 1990: 95). Together, the head and classifying modifiers constitute the categorization zone,6 which is generally assumed to be the most objective zone of the noun phrase,7 providing the categorial specifications for the identification of the referent.

6. In the functional model of the noun phrase proposed by Bache (2000), the categorization zone realized exclusively by the head noun is distinguished from the classification zone realized by adjectives. Breban (2010) proposes to collapse these two into a single categorization zone, adducing analogous arguments as for the determination zone (footnote 2). 7. The alleged objectivity of the categorization zone has to be nuanced in a number of ways. First, as noted in the discussion of emphasizers such as complete in a complete idiot, some head nouns incorporate evaluative, subjective elements. Second, corresponding to the more objective classifier-head units, we also find what Van linden and Davidse (2005) call interpersonal compounds. Interpersonal compounds are fixed collocational units consisting of an adjective and a nominal head, between which a form of “semantic feature copying” (Bublitz 1996) takes place. The adjectives in interpersonal compounds “do not add any new, independent meaning component to their noun. Adjective and noun are co-selected for the one salient feature they share” (Bublitz 1996: 6– 8). For instance, in collocations such as old fogey (i) and old hand (ii), adjective and noun clearly share evaluative, subjective semantic features. (i) old fogey: old in the sense of ‘having the mental or physical characteristics of old age in a negative sense’ (ii) old hand: old in the sense of ‘knowing, experienced’

284

Lobke Ghesquière

2.1.4. A structural-functional model of the English noun phrase The functional slots discussed above can be ordered in the noun phrase syntax as in Figure 1. predeterminer ^ determiner ^ secondary determiner ^ emphasizer ^ intensifier ^ subjective descriptive modifier ^ objective descriptive modifier ^ submodifier of classifier ^ classifier ^ head Figure 1. Syntagmatic positions in the structure of the English noun phrase

These different syntagmatic slots can, following Bache (2000: 31) and Breban (2010: 33), be structured into a structural-functional model consisting of three main zones (Figure 2). I, however, propose a further subdivision of the modification zone into the description zone, containing subjective and objective descriptive modifiers, and the strengthening zone,8 which accommodates bleached and non-bleached intensifiers and emphasizers. The model proposed in this study can then be represented as in Figure 2. Determination Predeterminers Determiners Secondary determiners

Modification Strengthening Description Emphasizers Bleached and non-bleached intensifiers

Subjective and objective descriptive modifiers

Categorization Submodifiers of classifiers Classifiers Head nouns

Figure 2. Structural-functional zones in the English noun phrase

2.2. From synchronic continuum to diachronic development Traugott’s (1982) and (2003) hypotheses concerning the unidirectionality of subjectification predict that semantic change proceeds along a path from objective to subjective (to intersubjective) meaning. A case in point is the ‘NP of NP’-pattern a bit of, which has developed from objective partitive construction over subjective degree modifier into intersubjective hedge (Traugott this volume). Building on Traugott’s earlier work, Adamson (2000) has proposed a hypothesis concerning the subjectification processes 8. The generalizing term ‘strengthening’ uses for emphasizers and intensifiers was proposed by Vandewinkel (2005).

The adjectives of completeness

285

affecting prenominal elements in the English noun phrase that links increasing subjectification to progressive leftward movement. It is the aim of this study to see how the results of the case studies of the completeness adjectives fit in with the hypotheses on the structural-semantic correlates of the diachronic development of prenominal adjectives. 2.2.1. Traugott’s subjectification hypothesis According to Traugott (1989: 35) semantic and syntactic changes affecting linguistic items are often driven by subjectification, i.e. the process whereby “meanings become increasingly based in the speaker’s subjective belief state/attitude toward the proposition”. Over the last twenty-five years Traugott has gradually developed and focused her hypotheses on subjectification. The original subjectification path proposed by Traugott (1982; 1989) refers to Halliday and Hasan (1976), who view the linguistic system as comprising three functional-semantic components, the ideational, the textual and the interpersonal. Traugott (1982) transformed their synchronic analysis into a diachronic hypothesis, arguing that meaning change is essentially a unidirectional process of subjectification which typically proceeds along the following pathway: propositional (> textual) > expressive (Traugott 1982: 256)

The propositional (or ideational) component comprises all “truthconditional relations”, which are “subject to referential verification” (Traugott 1982: 248) and can be further divided into “meanings based in the external described situation” and “meanings based in the internal (evaluative/perceptual/cognitive) described situation” (Traugott 1989: 34). This finer distinction seems to correspond nicely to the characterization of objective and subjective descriptive modifiers proposed in Section 2.1: the latter are still descriptive, but the recognition criteria provided are of a more internal, evaluative nature. “The textual component has to do with the resources available for creating a cohesive discourse”, and “the expressive component bears on the resources a language has for expressing personal attitudes to what is being talked about” (Traugott 1982: 248). Traugott (1995: 47) defines subjectification as “the tendency to recruit lexical material for purposes of creating text and indicating attitudes in discourse situations”. Subjectivity is then understood to encompass both textually and attitudinally subjective meanings.

286

Lobke Ghesquière

Traugott’s (2003: 126) more recent definition of subjectivity as the encoding of the speaker’s attitudes and beliefs seems to focus more exclusively on speaker-attitudes and speaker-stance. Subjectification then entails that “meanings become more deeply centred on the speaker” (2003: 129). Traugott and Dasher (2002) and Traugott (2003) added the notion of intersubjectification to the theoretical apparatus: in this process meanings become more focused on the self of the addressee or reader than on that of the speaker. Intersubjective meanings then “encode or externalise implicatures regarding SP[eaker]/W[riter]’s attention to the ‘self’ of AD[dressee]/ R[eader] in both an epistemic and a social sense” (Traugott 2003: 130). Traugott’s first hypothesis hence came to be reformulated into that of unidirectional (inter)subjectification as represented below. The second line represents the approximate matches between Traugott’s and Halliday and Hasan’s (1976) terminology. non-/less subjective > subjective ideational interpersonal (Traugott this volume: 34)

> intersubjective

Historically, non- or less subjective meaning precedes subjective meaning, which in its turn precedes intersubjective meaning. In other words, “for any lexeme intersubjectivity is historically later than and arises out of subjectification” (Traugott 2003: 130). It is not entirely clear to me where textual meanings fit into this new (inter)subjectification cline. In Section 2.1.1, I have argued that textual meanings expressed by (secondary) determiners, which help coordinate the interlocutors’ joint attention (Diessel 2006), are textually intersubjective. I am therefore inclined to subsume it under the right end of the cline, while stressing that this is not a position that can be attributed to Traugott. The exact position textually intersubjective meanings should take up in Traugott’s (this volume) intersubjectification cline is an issue in need of further reflection. 2.2.2. Adamson’s subjectification hypothesis Adamson (2000) places some restrictions on the idea of unidirectionality of semantic change as presented in the work of Traugott (1982; 1989; 1995; 2003; this volume) and Traugott and Dasher (2002). For the noun phrase, the unidirectionality claim would entail that constituent elements can only undergo semantic changes to more subjective meanings. However, as ar-

The adjectives of completeness

287

gued by Adamson (2000), this idea is challenged by the development of classifier uses out of descriptive modifier uses. The adjective criminal, for example, was originally a descriptive modifier with a (subjective) characterizing function, as in a criminal tyrant, but has developed a classifying use over time, as in criminal law, thus constituting a positional shift to the right of the noun phrase. Such a development can be considered an instance of de-subjectification, 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 … poses a challenge to any hypothesis concerning the unidirectionality of semantic change, such as the “hypothesis of unidirectional increase in subjectification over time” (Traugott 1995: 45). (Adamson 2000: 60)

Consequently, Adamson proposes to restrict the unidirectionality hypothesis to grammaticalization processes and consider the development of classifiers out of descriptive modifiers as an instance of lexicalization. As observed correctly by Adamson (2000), well-entrenched classifier-noun combinations may become so tightly bonded that they closely resemble compounds. Physiotherapist, for instance, has the orthographical status of compound, but still designates a subtype of the type therapist, contrasting with other subtypes like clinical therapist or family therapist (see also Halliday 1994: 185; Vandelanotte 2002: 235). As compounds are generally considered to be single words, the developmental path “descriptive modifier > classifier > compound” could be defined as an instance of lexicalization, i.e. as a type of change resulting in the production of a new fully lexical, contentful item that is stored in the lexicon. Building on Traugott’s unidirectionality hypothesis, Adamson (2000: 42) subsequently puts forward her own hypothesis, namely that of a link between subjectivity and the left periphery in the prenominal string. This entails that elements situated more to the left of the prenominal string are more subjective in meaning than constituents situated more to the right. The underlying idea is that increasing subjectification is paralleled by progressive leftward movement in the noun phrase structure. Such an evolution is illustrated by means of a case study of lovely, which has shifted from descriptive modifier, as in a lovely house, to (non-bleached) intensifier, as in lovely long legs. Although Adamson (2000) focuses on the development of strengthening uses, her leftward movement hypothesis is taken here to also apply to the development of determining meanings. The author, for instance, refers to a number of studies on the development of English possessives and quantifi-

288

Lobke Ghesquière

ers from (adjectival) modifiers to determiners and mentions the developments of the adjectives mere and pure. According to Adamson (2000: 59), these adjectives, which have developed identifying functions, allow her “to posit a pathway characteriser > identifier in which … leftward movement, category shift and subjectivisation are combined”. Traugott (this volume: 60) suggests that the leftward movement hypothesis is more universally valid than proposed by Adamson, as “a growing number of studies have suggested that as they are subjectified linguistic elements are used in increasingly peripheral positions”. As a consequence, changing position in the phrase or clause is a likely structural correlate of (inter)subjectification. 3. Case study of three adjectives of completeness To investigate the validity of the pathways of structural-semantic change proposed by Traugott (1982; 2003) and Adamson (2000), a case study was carried out of three adjectives of the lexico-semantic field of completeness whose diachronic development was expected to be particularly relevant to the proposed pathway: complete, total and whole. These three adjectives have developed descriptive, strengthening and determining uses. The data samples for this study were compiled using five different corpora. The synchronic data were extracted from the contemporary COBUILD corpus (CB), which is available via Collins WordbanksOnline. For this study the Times (5.7 m words) and UK Spoken (9.2 m words) subcorpora were selected, representing formal written and informal spoken British English. The diachronic examples were drawn from the York-TorontoHelsinki Parsed Corpus of Old English Prose (YCOE, 1.5 m words) and the Penn-Helsinki Parsed Corpora of Middle English (PPCME2, 1.2 m words) and Early Modern English (PPCEME, 1.8 m words). For Late Modern English, use was made of the 14.9 m words of the extended version of the Corpus of Late Modern English Texts (CLMETEV), which comprises 18th and 19th century texts that are available through online electronic archiving projects (De Smet 2005). For each of the adjectives I aimed at analysing 100 prenominal examples for each of the different historical periods, provided so many examples were available. All the examples for each of the adjectives were analysed and quantified in terms of the prenominal uses they manifested, which allowed me to draw up the quantitative profiles of complete, total and whole represented in Tables 2 to 4.

The adjectives of completeness

289

Table 2. Quantitative profile of prenominal complete

1640–1710 1710–1780 1780–1850 1850–1920 CB Times CB UK Spoken CB Total

Sec det

Emph

1 50,0% 6 8,2% 9 11,3% 10 8,8% 33 25,8% 22 13,0% 55 18,5%

0 0,0% 3 4,1% 30 37,5% 30 26,3% 39 30,5% 88 52,1% 127 42,8%

Emph/ descr mod 0 0,0% 6 8,2% 4 5,0% 16 14,0% 26 20,3% 18 10,7% 44 14,8%

Subj descr mod 1 50,0% 22 30,1% 18 22,5% 35 30,7% 16 12,5% 10 5,9% 26 8,8%

Obj descr mod 0 0,0% 36 49,3% 19 23,8% 23 20,2% 13 10,2% 27 16,0% 40 13,5%

Subm of class 0 0,0% 0 0,0% 0 0,0% 0 0,0% 1 0,8% 4 2,4% 5 1,7%

Total 2 73 80 114 128 200 328

Table 3. Quantitative profile of prenominal total Sec det 6 6,9% 23 20,0% 38 45,2% 129 CB Times 69,7% 77 CB UK Spoken 43,0% 206 CB Total 56,6% 1710– 1780 1780– 1850 1850– 1920

Subj/ Emph/ Subj obj Emph descr Intens descr descr mod mod mod 16 9 0 6 5 18,4% 10,3% 0,0% 6,9% 5,7% 34 12 0 6 2 29,6% 10,4% 0,0% 5,2% 1,7% 27 4 0 2 0 32,1% 4,8% 0,0% 2,4% 0,0% 21 10 1 3 0 11,4% 5,4% 0,5% 1,6% 0,0% 59 20 2 4 0 33,0% 11,2% 1,1% 2,2% 0,0% 80 30 3 7 0 22,0% 8,2% 0,8% 1,9% 0,0%

Obj descr Class Total mod 45 51,7% 38 33,0% 12 14,3% 19 10,3% 10 5,6% 29 8,0%

0 0,0% 0 0,0% 1 1,2% 2 1,1% 7 3,9% 9 2,5%

87 115 84 185 200 385

290

Lobke Ghesquière

Table 4. Quantitative profile of prenominal whole

850– 950 950– 1050 1050– 1150 1150– 1250 1350– 1420 1420– 1500 1500– 1570 1570– 1640 1640– 1710 1710 – 1780 1780 – 1850 1850 – 1920 CB Times CB UK Spoken Total CB

Wide Wide scope scope emph/ emph Sec det 0 0 0.0% 0.0% 0 0 0.0% 0.0% 0 0 0.0% 0.0% 0 0 0.0% 0.0% 0 0 0.0% 0.0% 0 2 0.0% 9.1% 2 9 1.3% 5.9% 1 0 0.6% 0.0% 13 2 8.1% 1.2% 5 2 6.4% 2.6% 3 3 4.3% 4.3% 9 1 11.8% 1.3% 35 0 20.0% 0.0% 40 0 22.3% 0.0% 75 0 21.2% 0.0%

Sec det 0 0.0% 0 0.0% 0 0.0% 0 0.0% 12 46.2% 12 54.5% 118 77.6% 133 84.7% 117 72.7% 44 56.4% 29 41.4% 28 36.8% 60 34.3% 62 34.6% 122 34.5%

Sec Subj det/ Emph Intens descr Emph mod 0 0.0% 0 0.0% 0 0.0% 0 0.0% 5 19.2% 3 13.6% 14 9.2% 16 10.2% 15 9.3% 14 17.9% 15 21.4% 18 23.7% 19 10.9% 21 11.7% 40 11.3%

Obj descr mod

0 0 1 1 0.0% 0.0% 50.0% 50.0% 0 0 0 0 0.0% 0.0% 0.0% 0.0% 0 0 0 7 0.0% 0.0% 0.0% 100.0% 0 0 2 4 0.0% 0.0% 33.3% 66.7% 0 0 0 9 0.0% 0.0% 0.0% 34.6% 0 0 0 5 0.0% 0.0% 0.0% 22.7% 2 0 1 6 1.3% 0.0% 0.7% 3.9% 3 0 2 2 1.9% 0.0% 1.3% 1.3% 13 0 0 1 8.1% 0.0% 0.0% 0.6% 12 0 0 1 15.4% 0.0% 0.0% 1.3% 20 0 0 0 28.6% 0.0% 0.0% 0.0% 18 1 0 1 23.7% 1.3% 0.0% 1.3% 49 6 0 6 28.0% 3.4% 0.0% 3.4% 51 3 0 2 28.5% 1.7% 0.0% 1.1% 100 9 0 8 28.2% 2.5% 0.0% 2.3%

Total 2 0 7 6 26 22 152 157 161 78 70 76 175 179 354

The Romance loanwords complete and total are found in the data only from Middle and Late Modern English on respectively. Presumably these adjectives have undergone a number of functional shifts in French that cannot be traced in the English data. Unlike complete and total, whole is an

The adjectives of completeness

291

adjective of Anglo-Saxon origin which has attested uses from the Old English period onwards. This study was restricted to prenominal uses of the adjectives, but I have noted that complete, total and whole display considerable differences in syntactic behaviour. Whereas complete and whole are first predominantly used predicatively, their prenominal use gradually become prevalent, so much so that predicative uses of whole have now become extremely rare. The decreasing possibility for whole to occur in predicative position can be considered an instance of decategorialization, a process strongly associated with grammaticalization (Hopper and Traugott [1993] 2003). Decategorialization is “the tendency for relatively prototypical members of Noun, Verb, and Adjective categories to become less prototypical in their distribution” (Hopper and Traugott 2003: 106). Applied to whole, this refers to the gradual loss of the adjective’s ability to be used predicatively, one of the characteristics typically associated with the central descriptive modifier use of adjectives. The syntactic behaviour of total is very different from that of complete and whole, as the adjective is almost exclusively found in prenominal position throughout all historical stages. In the contemporary data some predicative uses are found, but they are still highly uncommon. Whether or not these uses are grammatical is difficult to determine, as “it may well be possible to identify a degree of improbability that is close enough to impossibility to be indistinguishable from it” (Stefanowitsch 2006: 73). The different prenominal uses, as observed in the data, will be discussed in the following sections, largely following the chronology of their occurrence in English. This will allow us to reconstruct the paths of development followed by completeness adjectives. 3.1. Objective descriptive modifier uses For all three adjectives the objective descriptive modifier use, which attributes objectively recognizable, potentially defining properties to the head, appears to be either the most original one or one of the earliest prenominal uses. Descriptive modifiers are characterized by their ability to occur both in attributive and in predicative position (e.g. complete information, the information is complete). Although the first prenominal descriptive modifier uses of complete are found only in the Late Modern English data, I argue that the objective descriptive modifier is nevertheless historically the most

292

Lobke Ghesquière

original function. The late rise of prenominal uses was most likely influenced by the adjective’s postnominal position in French and its verbal origin. Postposed objective descriptive modifiers already occurred in the Middle English period, as in (3), where complete indicates that a certain period of time “has run its full course” (OED 1933: II725). (3)

The fourthe day compleet fro noon to noon. (OED 1933: II725, c1386, Chaucer, The Merchant’s Tale) ‘The fourth day, complete from noon to noon.’

In later periods this use is extended to contexts in which it describes not periods of time as being completed, but processes. In (4), for instance, the manufacturing process has been completed and the finished article is ready to be sold. In this example, complete is readily replaced by the past participle completed. (4)

As it is their interest to sell the complete manufacture as dear, so it is to buy the materials as cheap as possible. (CLMETEV Cook, 1768–71, Captain Cook’s journal during the first voyage round the world)

This type of descriptive modifier use is most frequent at the beginning of the Late Modern English period, in which it predominantly collocates with the head noun manufacture. In later stages this use becomes rare and it is found only occasionally in contemporary data. A second objective descriptive modifier use of complete describes actions or states as being realized to the fullest extent (OED 1933: II725). When used as such, complete often collocates with nominalizations, such as exploration, preparation and isolation, as in (5). The objective descriptive modifier total is semantically similar to this second use of complete, also describing actions or states as being “complete in extent or degree” (OED 1933: XI176) and mostly modifying nominalizations of verbs, such as abolition, separation and exclusion. In this use complete and total semantically resemble their adverbial counterparts completely and totally. The noun phrases in (5) and (6), for instance, could easily be paraphrased as “completely isolated” and “almost totally secluded” respectively. (5)

the horrible cruelties of the Romans, which, during and after the war, might give some cause for the complete isolation of the Jew

The adjectives of completeness

(6)

293

from the rest of the world. (CLMETEV Gibbon, 1776, The decline and fall of the Roman Empire) Ann Yearsley … died many years after, in a state of almost total seclusion, at Melksham. (CLMETEV Cottle, 1847, Reminiscences of Samuel Taylor Coleridge and Robert Southey)

In Present Day English the objective descriptive modifier complete typically indicates that a certain object includes all the parts that are necessary (OALD 2000: 247) or that are normally associated with it, as in (7). Complete is then synonymous with “entire” or “full” rather than with “finished” or “completed”. (7)

The fossil skeleton, the most complete example of its kind, was in danger of being washed out to sea by coastal erosion and winter storms at West Runton near Cromer. (CB Times 1995)9

In Old and Middle English, whole has two distinct objectively descriptive meanings. The first and most recurrent one indicates that a certain body part of a man or animal is in good condition or uninjured (OED 1933: XII89), as in (8). The synonyms skin and flesh are then by far the most common collocates of whole. In a number of cases the adjective refers not to body parts of persons or animals but to the beings themselves, describing them as healthy or in good condition (OED 1933: XII89, 90), as in (9). (8)

(9)

Or ellus if þou wolt þou my3t schere a-wey þe fyke euene bi þe hole skyn & let þe euel blod blede out euerydel þen schalt þou strawe upon pouder of unsleked lyme or of chalke. (PPCME2 a1425, Late Middle English Treatise on Horses, 111) ‘Or else if you want you might cut away the fig [= sarcoid] evenly along the whole skin and let the evil blood bleed out thoroughly, then you shall sprinkle upon it powder of unslacked lime or of chalk.’ Certes is not this lyke a myracle vnto a man that knowethe not, whye that swete thynges agree well to hole folke, and bytter thinges to sycke folk? (PPCEME Colville, 1556, Boethius’ Consolation of Philosophy, 109)

9. The examples from the COBUILD Corpus quoted in this article (marked CB) were extracted via remote log-in and are reproduced with the kind permission of Harper Collins Publishers.

294

Lobke Ghesquière

A second objectively descriptive use of whole indicates that the entity referred to by the head is intact or not divided into parts. The first prenominal examples of this type are found already in the Middle English period, as in (10).10 In Section 3.3, I argue that it is this particular descriptive meaning that gave rise to the secondary determiner reading of whole. Unlike the first objective descriptive modifier use, which is now extinct, this one persists into Present Day English, although it has now become rather rare and its collocational range seems restricted to nouns referring to foods. Although predicative alternation (?the salmon is whole) and gradability (*very whole salmon) do not seem possible, the order of adjectives in example (11) evidences the status of whole as descriptive modifier. Because whole is preceded by the descriptively used adjective fresh, a secondary determiner reading, for instance, is not plausible, as determining elements typically precede all other prenominal elements of the noun phrase (Section 2.1). (10)

(11)

hudeð hare hale clað & doð on alre uueward. fiterokes al to torene (PPCME2 1225–30, Ancrene Riwle, II 244) ‘they hide their whole garment and put rags, all torn up, on top of everything.’ fresh whole turkey £ 2.49 a kg, fresh whole Scottish salmon £ 5.49 a kg, blue Stilton £ 4.49 a kg. (CB Times 1995)

3.2. Subjective descriptive modifier uses Subjective descriptive modifier uses are found among the earliest prenominal attestations of all three adjectives, but in the contemporary data only complete and total still function as such. The subjective descriptive modifier uses of these two adjectives are very similar to one of their objective descriptive uses, which leads me to believe that they originated from them. One of the subjective descriptive modifier uses of complete is semantically comparable to the contemporary objective descriptive modifier which describes something as containing all necessary parts (Section 3.1). Contrary to its objective counterpart, however, the subjective modifier lacks an objectively definable point of reference which can be consulted to determine if an object really includes all necessary parts. Whether or not this is the 10. The OED (1933: XII90) even gives mention of predicative examples dating back to the 9th century.

The adjectives of completeness

295

case is more a matter of opinion and thus subjectively coloured, as in (12). For total as well, the only feature which distinguishes the subjective descriptive modifier is that the quality of totality is attributed to the head on a subjective basis: a personal opinion or attitude of the speaker toward the entity referred to is expressed. In (13), for instance, a subjective, positive evaluation is inherent in the company name due to the adjective total, irrespective of whether all the different facets of sports are indeed covered. (12) (13)

He made his name and fortune when his Complete Book of Running was read by millions. (CB Times 1995) His company is called Total Sports Image, presumably to distance itself from such concepts as total sports substance. (CB Times 1995)

For complete two more subjective descriptive modifier uses are found. A first one describes someone as being “fully equipped or endowed, esp. in reference to a particular art or pursuit” (OED 1933: II725). Complete is then largely synonymous to “well-rounded”, as in (14). A second and final type indicates that a certain object or abstract concept is “perfect in nature or quality; without defect” (OED 1933: II725), as in (15). (14) (15)

Cicero … asserts that a complete orator must be a complete everything, lawyer, philosopher, divine, etc. (CLMETEV Chesterfield, 1746–71, Letters to his son) … four human beings who spent every day of their lives in each other’s society, between whom there was the most complete sympathy and the most cordial good-will. (CLMETEV Disraeli, 1837, Venetia)

For whole, the data suggest that the adjective receives a subjective descriptive reading only through coordination with another subjective descriptive modifier, such as parfytte ‘perfect’ in (16). In this example, the reading of whole as indicating that the entity designated by the head has “no part or element wanting” (OED 1933: XII90) seems to be the result of contextual modulation. “The specifying features of different contextual modulations are, as it were, contributed by the context” (Croft and Cruse 2004: 140), in this case by the adjective perfect. It is the semantically more specific adjective perfect that brings to the fore this particular use of the lexically more general adjective whole.

296 (16)

Lobke Ghesquière

For the nature of thynges toke neuer any begynnynge of thynges dymynished and vnparfitte but procedynge from hole and parfytte thynges, came downe or descended into these lower and baren thynges. (PPCEME Colville, 1556, Boethius’ Consolation of Philosophy, 73)

3.3. Secondary determiner uses Early on in their development all three adjectives developed secondary determiner uses in which they aid identification of the entity referred to by the noun phrase. The secondary determiner whole emerged in the 14th century, and complete and total were used as secondary determiners from their occurrence in English on. Besides their deictic semantics, the secondary determiner uses are characterized by their inability to be graded (*the very whole population) or to be used predicatively (*the population is whole). The determining uses of complete, total and whole indicate that reference is made to the entire entity referred to by the head noun and can thus be described as both quantifying and emphasizing in nature, as in (17) (18) (19)

Numerous though we of Middle England are, we are a minority of the whole population. (CB Times 1995) Section II. … State of the popular mind in Christendom during the complete reign of Popery. (CLMETEV Foster, 1821, An essay on the evils of popular ignorance) The announcement brings the total number of jobs lost by Vernons this year to 200 and reduces the workforce to 425. (CB Times 1995)

“In these examples there is an inherent notion of the ‘whole’ thing, and complete[, total and whole], like a universal quantifier such as all, quantif[y] over this whole thing” (Davidse, Breban and Van linden 2008: 489). The secondary determiners thus emphasize the quantifying semantics of the (definite) primary determiners. The collocational behaviour of the three adjectives further demonstrates their general quantifying meaning, while at the same time pointing to some specific semantic differences between them. All three adjectives take as their heads either ‘measure nouns’ in the

The adjectives of completeness

297

broad sense11 or nouns describing inherently measurable or gradable entities. Complete has developed the most diversified collocational range, as it came to collocate not only with size nouns measuring a specific period of time (years, week), but also with more abstract size nouns (number, amount) and more abstract nouns indicating measurable entities (reign, range). Total and whole are collocationally more restricted than complete. From Middle English on, determining whole always takes measure nouns in the broad sense. Its collocates vary from indications of time  both direct (time, year) and more associative (sickness, dinner) , measures and dimensions (mile, length) to nouns referring to groups of people or animals. The latter can be straightforward collectives such as group or family, but quite often nouns designating buildings, such as house or school, metonymically refer to the people that occupy them. The secondary determiner total generally collocates with nouns indicating “a certain amount or number that is reached after everyone or everything is counted or added together” (OALD 2000: 1373). These entities are thus typically made up of several entities that can be considered either independently or as a whole. In the latter case, total emphasizes that reference is being made to the totality of the different components, as in (19). In noun phrases with definite article, the universal quantification information is rather redundant and the secondary determiner can easily be elided without making the reference less clear or precise, as in (20b). It is generally accepted that the use of the definite article implies that reference is made to the entire potential instantiation of the type designated by the noun phrase (Declerck 1991: 292; Hawkins 1978; Langacker 1991: 102;). Adding a secondary determiner then “merely makes explicit an element that is normally part of definite identification” (Davidse, Breban and Van linden 2008: 491). Complete, total and whole can therefore be viewed as emphasizing secondary determiners (cf. Davidse, Breban and Van linden 2008). (20)

a.

The total amount of capital invested so far in the rail link between London and Paris must be in excess of £ 11 billion. (CB Times 1995)

11. In the strict sense, measure nouns are nouns, such as litre or year, that “measure off a well-defined standard-like portion” (Brems 2003: 284). The notion of measure noun can, however, be extended to include nouns, such as sickness, amount or group, which designate a certain, less strictly defined measure.

298

Lobke Ghesquière

b.

The amount of capital invested so far in the rail link between London and Paris must be in excess of £ 11 billion.

For all three adjectives, I claim that the determining uses developed from specific objective descriptive modifier uses through deictification, i.e. a “semantic shift by which a general relation expressed by the adjective is given a subjective reference point in or relative to the speech event” (Davidse, Breban and Van linden 2008: 475). Hence, the shift from descriptive modifier to secondary determiner can be viewed as an instance of subjectification: a relation with objective reference point is transformed into one with subjective reference point. There is a functional shift from attributing gradable qualities to the entity designated by the noun phrase to expressing, together with the determiner, the specific identifiability status of the referent. The secondary determiner use of complete most likely originates in the objective descriptive modifier use indicating that a certain period of time or process has run its full course (Section 3.1). In the change from descriptive modifier to secondary determiner focus shifts from the final stage of the period of time or process to the whole of the referent, as in (18), and the objective reference point vis-à-vis which something is “completed” is transformed into a subjective reference point in the speech event, viz. the full contextual instantiation of the type in question. Although descriptive modifier and secondary determiner uses are found in equal numbers among the earliest attestation of the adjective, a similar process can be reconstructed for total.12 First, the semantics of the two uses point in the direction of a process of subjectification and deictification. From attributing the gradable quality of totality vis-à-vis some objective reference point, as in (6), total came to specify the referent of the head as the whole contextual instantiation of the referent which speaker and hearer have in mind, as in (19). Second, I observed a steady increase in the proportion of secondary determiner uses together with a gradual decrease in descriptive modifier uses of total in the periods following its occurrence in English (see Table 3).

12. It is unlikely that the secondary determiner use was borrowed independently from French. According to Le Trésor de la Langue Française informatisé, the determining use of the French adjective total(e) developed only in the 19th century. (e.g. Dans sa largeur totale le Nil a plus d’une lieue; une mer [FROMENTIN, Voy. Égypte, 1869, p. 56])

The adjectives of completeness

299

For whole, it is most likely the objective descriptive meaning ‘undivided’ or ‘not missing any parts’, as in (10), that led to the semantically strongly related secondary determiner reading of whole indicating that reference is made to the whole instance which speaker and hearer are aware of, as in (17). As mentioned for total, the relation between the descriptive and determining uses of the three adjectives is apparent from their changing proportions in the data sets. For complete and total, we see that from their emergence the proportion of secondary determiner uses gradually increased over time (from 8% to 19% and from 7% to 57% respectively) at the expense of the objective descriptive modifier uses (see Tables 2 and 3) (from 49% to 14% and from 52% to 8% respectively). In the centuries after its emergence in English the secondary determiner uses of whole always takes up more than a third of all prenominal attestations, but the proportion of fully lexical objective descriptive modifiers radically decreases, taking up only 2% in Present Day English period (see Table 4). 3.4. Emphasizer uses The first emphasizer uses of complete and total can be found in the Late Modern English data, but whereas emphasizing complete becomes more frequent only in the following centuries, the relative proportion of emphasizing uses of total remains stable, taking up roughly one fourth of all prenominal uses in every period (see Table 3). When used as emphasizers, complete and total indicate that something is “as great in extent, degree or amount as it possibly can be” (Sinclair 1990: 303, 1653), as in (21) (22)

… open up any cupboard in the flat and you’ll see complete chaos. (CB Times 1996) All you need is a little nerve and a total disregard for the truth. (CB Times 1995)

The historical data allow us to reconstruct the different meaning extensions leading to the emphasizer uses of complete and total. Although the emphasizer total is found among the earliest prenominal uses of the adjective, possible pathways of change can still be posited on the basis of other uses encountered in the data. One trail that leads to the emphasizers com-

300

Lobke Ghesquière

plete and total proceeds from objective descriptive modifier over subjective descriptive modifier, as illustrated below for complete. (23) (24)

(25)

… he soon obtained a complete victory. A hundred and thirty vessels were destroyed, five thousand men were slain, … (CLMETEV Gibbon, 1776, The decline and fall of the Roman Empire) This winter is to be all ups and downs. The next day (Friday) we had a most complete victory. Mr. Pultney moved for all papers and letters, etc. between the King and the Queen of Hungary and their ministers. (CLMETEV Walpole, 1735–48, Letters) Everybody was surprised, and Darcy, after looking at her for a moment, turned silently away. Mrs. Bennet, who fancied she had gained a complete victory over him, continued her triumph. (CLMETEV Austen, 1813, Pride and Prejudice)

In (23), the collocation complete victory is used in a context of war, in which it is possible to objectively measure the extent of the victory by the number of casualties or destroyed targets. Complete is thus best analysed as an objective descriptive modifier. In example (24), the war is fought not on the battle field, but on a higher, more abstract level. There is a kind of psychological warfare going on in which the committee is the winning party. Here a subjective descriptive modifier reading13 of complete seems the most plausible. In example (25), I opt for an emphasizer reading of complete, stressing the extent of a mental victory in a dispute between two people. Although in the above examples one particular meaning was identified as the predominant one, the different stages in the process cannot always be clearly discerned. In the data I found a number of syntactic blends (cf. Bolinger 1961; Aarts 2007: 188), which display typical properties of both descriptive and emphasizing constructions, allowing the speaker to imply both meanings at the same time. Example (26), for instance, reveals the semantic closeness of the descriptive sense (“the success is complete”) and the strongly subjective emphasizing use (“utter success”), as both readings can be constructed simultaneously. Blended examples are especially frequent in collocations such as complete victory/triumph/ruin/wreck/etc.

13. The use of the periphrastic superlative of complete reveals that the adjective functions as descriptive modifier in this noun phrase, as only these prenominal uses can be graded (Section 2.1).

The adjectives of completeness

(26)

301

The operations conducted by Colonel Parsons thus ended in complete success. Great difficulties were overcome, great perils were encountered, great results were obtained. (CLMETEV Churchill, 1899, The river war, an account of the reconquest of the Sudan)

Besides this first pathway from objective to subjective descriptive modifier to emphasizer, a second path to emphasizing complete can be distinguished. In Section 3.2, a type of subjective modifier was discussed which describes persons as being “fully equipped or endowed, esp. in reference to a particular art or pursuit” (OED 1933: II725). This descriptive use was later extended to nouns other than professions or activities, but still referring to persons, as in (27)

His natural and acquired abilities, continues the Dr. made him very amiable to all who knew and conversed with him, a very few being equal in the becoming qualities, which adorn, and fit off a complete gentleman. (CLMETEV Cibber, 1753, The lives of the poets of Great Britain and Ireland)

Here complete is best analysed as an emphasizer, as it no longer describes a certain quality of the head, but rather emphasizes the positive evaluation inherent in the head. From 1850 onwards, this type of emphasizer use of complete occurs not only with (strongly) positive evaluations, such as gentleman and master, but with (strongly) negative ones as well, such as fool and dweeb. As with the first type of emphasizer, I found a number of descriptiveemphasizing blends in the data. In (28), for example, complete occurs in the superlative form, a characteristic typically associated with descriptive modifiers. On the other hand, the meaning of complete seems largely restricted to emphasizing the evaluation inherent in the head fools. (28)

… an intelligent young man when he was sober; but, the moment the wine began to operate, he was one of the completest fools in christendom (CLMETEV Hunt, 1820–22, Memoirs of Henry Hunt)

Total also developed a second emphasizer use similar to the second emphasizer use of complete in which it stresses the representative values typically associated with the head. Unlike for complete, however, no path from subjective descriptive modifier to emphasizer can be observed nor is there a

302

Lobke Ghesquière

change in semantic prosody. This emphasizer use, which appears in the data from around 1850 onwards, occurs only with heads that have a negative connotation, as in total stranger, total slob and total beginner. Interestingly, total originally only collocates with head nouns with which complete also appears in the data. Possibly, this emphasizer use of total developed by analogy with the second emphasizer use of complete which occurs from the same period on, but which is much more frequent. According to Hopper and Traugott ([1993] 2003: 63), analogy, a process in which extant forms are attracted to existing constructions, is, together with reanalysis, one of the major mechanisms governing language change. Here total gradually attracted all collocates of complete and as such a wider variety of lexemes came to fill the adjectival slot in this specific emphasizer construction. There is thus a generalization in the collocational range of the head nouns in these constructions. The meaning of emphasizing whole is slightly different from that of complete and total, as it stresses how large or important something is (OALD 2000: 1478) and consequently appears in contexts “where there is an implication of an unusually large quantity or number” (OED 1933: XII90). Unlike complete and total, whole takes more neutral nouns which can all be considered measure nouns in the broad sense, such as world, years and group.14 For complete and total we saw that the emphasizer uses most likely originated in descriptive modifier uses. Although the evaluative subjective and emphasizing subjective uses of whole both emerge in the Early Modern English period, there is not such a direct semantic link between them as with complete and total. Instead, the emphasizer whole probably emerged as an extension of the secondary determiner use. As with the emphasizer, 14. In the data, whole is often found in the ‘a/Ø whole N1 of N2’-construction. The noun modified by whole, N1, is typically a size noun in the broad sense and the construction is used to emphasize that there is a very large quantity of the type designated by N2, as in (i). This use of whole was already noted by Brems (2003: 229–230) in relation to bunch. This size noun has developed a grammaticalized quantifier use in which it can be submodified by adjectives expressing quantity, such as whole, which further foreground that quantifying function, as in (ii). (i) Is it not enough that you are one of those whose passions made this cap, and force me through whole trains of years to wear it low upon my brow! (Dickens, 1843, A Christmas carol in prose) (ii) There’s now a whole bunch of studies from different cities that show the same thing. (Brems 2003: 299)

The adjectives of completeness

303

the collocational range of the secondary determiner consists mainly of measure nouns in the broad sense and nouns describing measurable or gradable entities. In Section 3.3, I argued that the secondary determiner whole is emphasizing in nature, stressing that the whole contextual instantiation of the referent is referred to. The line between an emphasizing secondary determiner and a true emphasizer can therefore be rather fuzzy. In (29), for instance, both readings are plausible. The author may be indicating that it had been raining for an entire month, not just for a week for example; but he may also have added whole to the noun phrase to emphasize the length of the period that it had been raining. (29)

It had been raining heavily for one whole month–raining on a camp of thirty thousand men and thousands of camels, elephants, horses, bullocks, and mules all gathered together at a place called Rawal Pindi (CLMETEV Kipling, 1894, The jungle book)

The link between the secondary determiner and emphasizer use of whole is apparent from a construction which was particularly frequent in the Middle and early Modern English period: the “all the whole N”-construction, as in (30), which seems to usher in the rise of the emphasizer whole. (30)

and yn especiall the ende of the seide Fyssh strete enchroched all the hole wey thurt over for a court place to the mancion of the Archideacon of Cornewaill as hyt apperyth openly (PPCME2 Shillingford, 1447–50, Letters and papers) ‘and especially in the end of the said Fish street [they] encroached all the whole way across for a court place to the mansion of the Archdeacon of Cornwall, as it appears openly’

In the above example, the quantifier all functions as predeterminer, further specifying the scope of the definite determiner in terms of quantity: all indicates that reference is made to the whole of the wey talked about. The predeterminer thus fulfils a function similar to that of whole. In my opinion, whole would best be analysed as a secondary determiner in example (30) if all did not fill the predeterminer slot. However, adding all to the noun phrase triggers an emphasizing interpretation of the combination of predeterminer, determiner and completeness adjective, stressing the size of the entity referred to. Gradually whole acquired an emphasizing meaning of its own, no longer needing the presence of a quantifying predeterminer to ex-

304

Lobke Ghesquière

press emphasis. In the data the “all the whole N”-construction is, with the exception of just one contemporary example (31), restricted to the Middle and Early Modern English period, i.e. the period in which the emphasizer use of whole developed. (31)

you get all this whole range of problems stuck in there (CB UK Spoken 1994)

3.5. Wide scope emphasizer uses In the data an adjectival use of whole was found which is perhaps best described as a wide scope emphasizer or “attitudinal scoping use” (Breban and Davidse 2007). The notion of scope, as it is used here, is borrowed from McGregor (1997: 210), who states that expletives, such as fucking, bloody and flaming, are in a scoping relation in which “a unit applies over a certain domain leaving its mark on the entirety of this domain”. Following McGregor, Vandelanotte (2002: 245–251) argued that adjectives expressing strong speaker attitudes, such as damn in this damn spoon, involve scope, as they express “the very private emotive response of one particular person at one particular occasion vis-à-vis” the referent “within a specific, time-bound speech event” (Vandelanotte 2002: 246). In my opinion, uses of whole such as the following are also best analysed as attitudinal scoping adjectives. (32)

(33)

A soldier of the guards … had seen the arrest, and heard the orders given to the coachman. This fellow, accidentally meeting Atkinson, had acquainted him with the whole affair. (CLMETEV Fielding, 1751, Amelia) He likes women who come on to him. He requires a sexual predator because he’s never asked a girl even to dance. The whole idea of being turned down is too distressing. (CB Times 1995)

In the above examples, whole indicates how the situation referred to should be interpreted or evaluated: it expresses the non-commitment of the speaker to the message. In (33), for instance, the distressing idea of being turned down is part of the knowledge base of the person talked about. The speaker accesses this knowledge base but distances himself from it.

The adjectives of completeness

305

As a wide scope emphasizer, whole “which seems to be part of a nominal group actually has scope over an entire situation as expressed by an entire clause” (Vandelanotte 2002: 246). The clause describing this situation can either precede the noun phrase, as in (32), or follow it, as in (33). McGregor (1997: 69–70) recognizes one criterion to distinguish interpersonal relationships, such as scoping relations, from constituency or dependency relationships, viz. expandibility. Expandibility “refers to the potential of being expanded to maximum size … One unit may be expanded into a larger whole  e.g. one filled out with more constituents or dependents  whilst remaining within the confines of the relationship”. This criterion applies to the wide scope emphasizer use as it is possible to add intervening constituents between whole and the head of the noun phrase without breaking the scoping relationship. The entire combination of modifying string and head then falls within the scope of whole, as in (34)

What we are seeing is something of a backlash against the whole repressive feminist, safe-sex, politically correct sexual climate of the past two decades. (CB Times 1996)

3.6. Intensifier uses In the data a few rare intensifier uses of the completeness adjectives were found. As intensifiers, they indicate that the quality denoted by the descriptive modifier they modify is realized to its fullest extent. Intensifier uses of total and complete are restricted to the Present Day English period, while the intensifier whole is found from 1915 onwards. The intensifier slot in the prenominal string is typically filled by adverbs, which probably explains why this adjectival use is so infrequent. In examples (35) and (36), one would expect to find the adverbial forms totally and completely rather than their adjectival counterparts. In relation to whole, the question even poses itself whether the form in (37) is not adverbial rather than adjectival. Unlike for total and complete, the adverbial form of whole is not always formally distinct from the adjective. As the intensifier function is typically realized by adverbs, it is more likely that we are dealing here with the adverb rather than with the adjective whole. (35)

Losses for the 12 months to September 30 were recorded after total exceptional charges of £ 18 million (CB Times 1996)

306 (36) (37)

Lobke Ghesquière

We’ve been able to move limited resources around in [suburb name] but not produce complete any complete new money around. (CB UK Spoken 1993) In a comparatively short time a whole new social class sprang up in the land, and a whole new public opinion. (CLMETEV Carpenter, 1915, The healing of nations and the hidden sources of their strife)

3.7. Classifier uses Total is the only adjective under scrutiny which has a (limited) classifier use. The earliest example in the data, illustrated in (38), features the combination total eclipse, which refers to “an eclipse of the sun or moon in which the whole of the disk is obscured” (OED 1933: XII176) and which can be contrasted with partial eclipses. The other classifier uses of total found in the data are all rather ad hoc classifications, as in (39), where total is used to refer to a specific type of fund-holders. Ad hoc uses are the product of “the speaker’s own creativity [which] leads them to use an adjective in a function that is not normally associated with it” (Breban 2010: 31), in this case the classifier function. Although the basic semantics of total does not invite classifying readings, the adjective can occasionally be used as a classifier to meet a certain expressive need of the speaker in a specific type of context. (38)

(39)

On the 19th of August, 1887, occurred an important total eclipse of the sun, the track of which lay across Germany, Russia, Western Siberia, and Japan. (CLMETEV Bacon, 1902, The dominion of the air) it hasn’t really impinged locally with us er with G P fund-holders because er only a few of them are total fund-holders and even so it’s a paper exercise at the moment. (CB UK Spoken 1995)

4. Conclusions This study set out to investigate to what extent the hypotheses of unidirectional (inter)subjectification and concomitant progressive leftward movement apply to the development of complete, total and whole. To this aim a

The adjectives of completeness

307

functional analysis of these adjectives was carried out, both synchronically and diachronically. The corpus studies allow us to make a number of interesting observations concerning both the adjectives’ diachronic developments and their current behaviour in the English prenominal string. Firstly, I have noted that all three adjectives have grammaticalized and subjectified uses, albeit to different degrees (Section 3). Besides their descriptive modifier uses, complete, total and whole have developed strongly subjective strengthening uses  intensifying and/or emphasizing  and textually intersubjective secondary determiner uses. Secondly, I have observed that the development of the emphasizer function of a particular adjective can proceed along two distinct pathways of change. The emphasizer uses of complete and total most likely developed along a cline of semantic change from objective descriptive modifier to subjective descriptive modifier to emphasizer (e.g. from The fossil skeleton, the most complete example of its kind over a complete orator to a complete gentleman, and from a state of almost total seclusion over Total Sports Image to a total disregard for the truth). As they subjectified, complete and total thus systematically shifted to more leftward positions in the prenominal string, adhering to the hypothesis of progressive leftward movement in the noun phrase. The development of the emphasizer whole, however, does not confirm this hypothesis. As argued in Section 3.4, the more rightwardly positioned emphasizing use of whole probably developed from the secondary determiner use, which occupies a more leftward position in the prenominal string. Such a pathway of change goes against the hypothesis of progressive leftward movement as formulated by Adamson (2000). There are also some chronological issues contradicting Adamson’s (2000) hypothesis. For complete, historical data study showed that secondary determiner uses occur from the 16th century onwards, whereas emphasizer uses, which are positioned more to the right of the noun phrase, appear only from the 18th century onwards.15 In a similar way, the time of emergence of the intensifying uses of the three completeness adjectives, as in total exceptional charges, complete new money and a whole new thing, challenges the leftward movement hypothesis. These intensifiers emerge only in the most recent Late Modern English or Present Day English data. 15. No claims can be made with regard to the chronology of the development of the secondary determiner and emphasizer uses of total. Both uses were attested among the earliest prenominal attestations of the adjective and it was not possible to determine the time at which either of them developed.

308

Lobke Ghesquière

With all three adjectives the more subjective and more leftward emphasizer use as well as the leftmost secondary determiner use developed before their intensifier use. Strictly speaking, no progressive leftward movement has taken place here either. The case studies invite critical reflection of Adamson’s (2000) leftward movement hypothesis, and of Traugott’s (2003) hypothesis of unidirectional intersubjectification, which predicts that processes of semantic change proceed from objective to subjective to intersubjective meaning. In Section 2.2.1, I have argued that the deictic, leftmost functional zone in the noun phrase is textually intersubjective, as it is concerned with speaker-hearer negotiation facilitating the identification of referents in the discourse. If textually intersubjective meanings fit under the heading of intersubjectivity in the cline (Section 2.2.1), the development of the determiner uses of complete, total and whole would appear to go against the unidirectionality hypothesis. The textually intersubjective secondary determiner uses developed prior to and, in the case of whole, from the subjective emphasizer uses. As emerges from the above discussion, two distinct diachronic paths of subjectification can be reconstructed on the basis of the corpus studies of the three adjectives of completeness, viz. (i) objective descriptive modifier > secondary determiner > emphasizer (ii) objective descriptive modifier > subjective descriptive modifier > emphasizer

To grasp the pathways of semantic change as they came forward in this paper, I propose to integrate Traugott’s (1982) and (2003) hypotheses into one, more detailed, model of semantic change. Traugott’s (1982) pathway of semantic change, reproduced below, is semantically more fine-grained than her later cline of intersubjectification and seems to capture better the semantic development of the adjectives of completeness. propositional (> textual) > expressive (Traugott 1982: 256)

This cline nicely captures the two pathways of change attested for complete, total and whole. The first pathway runs through the entire cline as proposed by Traugott (1982). The objective descriptive modifier, which, as argued in Section 2.1, conveys propositional meanings “based in the external described situation” (Traugott 1989: 34), develops into a secondary determiner, mainly serving a textual function. With whole the secondary determiner develops further into an expressive emphasizer, thus covering

The adjectives of completeness

309

all the different stages of Traugott’s (1982) subjectification cline. With complete and total the pathway ends with the textual secondary determiner use, but as Traugott (2001: 3) notes changes “do not have to go to completion; in other words, they do not have to move all the way along a cline, or even continue down it once they start out on it”. The second pathway of change that emerged from the case studies can again be accommodated by Traugott’s (1982) cline. As argued in Section 2.1, the shift from objective to subjective descriptive modifier is a shift within the propositional meaning component. “Meanings based in the external described situation” develop into “meanings based in the internal (evaluative/perceptual/cognitive) described situation” (Traugott 1989: 35). The shift from propositionally subjective descriptive modifier to expressive emphasizer illustrates that “meanings tend to become increasingly based in the speaker’s subjective belief state/attitude towards the proposition” (Traugott 1989: 35). This pathway of change thus proceeds from propositional meaning directly to expressive meaning. The (1982) cline of semantic change thus seems to capture both developmental paths identified for the adjectives of completeness. However, to fully grasp the nature of the changes involved, I propose to add Traugott’s (2003) intersubjectification dimension to her earlier cline, as follows: externally propositional objective

> internally propositional subjective

non-bleached

> textual > expressive subjective and intersubjective bleached

Figure 3. Model of semantic change

The developmental cline above suggests that externally propositional objective meaning typically precedes internally propositional subjective meaning. Propositional meaning in its turn generally precedes textual meaning, which typically precedes expressive meaning. Textual and expressive meanings can be both subjective and intersubjective in meaning. The proposed subjectification cline could also be considered a cline of semantic generalization, in which the elements situated at the rightward end of the cline are semantically more general than those situated at the leftward end. Determining uses, for instance, have retained more of the adjective’s source semantics than the more strongly bleached strengthening uses. Secondary determiners have in their designation of a deictic relation often retained some of the original meaning elements of the objective relation

310

Lobke Ghesquière

expressed by the descriptive uses of these adjectives (Davidse, Breban and Van linden 2008: 497), whereas the meaning of strengthening modifiers is largely restricted to heightening the semantics of the adjectives and/or noun they modify. The model of semantic change proposed above can be translated into a model, which can be argued to visualize better the diachronic development of adjectives within the prenominal string of the English noun phrase. In Section 2.1, I introduced a synchronic functional model of the noun phrase, based on Bache (2000) and Breban (2010), which distinguishes between three main functional zones, viz. the determination or deictic zone; the modification zone, comprising the strengthening and description zone; and the categorization zone. Although the three adjectives of completeness all have their own specific history, they seem to conform to a more general pathway of structuralsemantic change. Based on the observations made above, I suggest a path of subjectification for prenominal adjectives that goes through the different functional zones of the noun phrase in the order represented in Figure 4: description > determination > strengthening. Adjectives may then first develop propositional, descriptive uses, then textual, deictic determining uses and, at yet a later stage, expressive strengthening uses. Determination Predeterminers Determiners Secondary determiners

Modification Strengthening Description Emphasizers Bleached and non-bleached intensifiers

2

Subjective and objective descriptive modifiers

Categorization Submodifiers of classifiers Classifiers Head nouns

1 3

Textual

Expressive

Propositional

Figure 4. Model of semantic change in the prenominal string of the English nounphrase

The pathways of subjectification suggested in this paper, although plausible for the three adjectives of completeness, are only one step to arrive at a better understanding of the diachronic development of prenominal adjectives. It is hoped, however, that this study will be a stimulus for further research into the intricate structure of the prenominal adjectival string.

The adjectives of completeness

311

References Aarts, Bas 2007

Syntactic Gradience: The Nature of Grammatical Indeterminacy. Oxford: Oxford University Press. Adamson, Sylvia 2000 A lovely little example: word order options and category shift in the premodifying string. In Pathways of Change: Grammaticalization in English, Olga Fischer, Anette Rosenbach and Dieter Stein (eds.), 39– 66. Amsterdam/Philadelphia: Benjamins. Bache, Carl 2000 Essentials of Mastering English: A Concise Grammar. Berlin/New York: Mouton de Gruyter. Bolinger, Dwight 1961 Syntactic blends and other matters. In Language 37: 366–381. Breban, Tine 2010 English Adjectives of Comparison: Lexical and Grammaticalized Uses. Berlin/New York: Mouton de Gruyter. Breban, Tine and Kristin Davidse 2003 Adjectives of comparison: The grammaticalization of their attribute uses into postdeterminer and classifier uses. Folia Linguistica 37: 269–317. Breban, Tine and Kristin Davidse 2007 Subjectification and structural movement of prenominal adjectives in the English NP: a diachronic perspective. Paper presented at Colloque des Adjectifs, Lille, 13–15 September 2007. Brems, Lieselotte 2003 Measure Noun Constructions: an instance of semantically-driven grammaticalization. International Journal of Corpus Linguistics 8: 283–312. Bublitz, Wolfram 1996 Semantic prosody and cohesive company: ‘somewhat predictable’. Leuven Contributions in Linguistics and Philology 85: 1–32. Croft, William and D. Alan Cruse 2004 Cognitive Linguistics. Cambridge: Cambridge University Press. Davidse, Kristin 2004 The interaction of quantification and identification in English determiners. In Language, Culture, and Mind, Michel Achard and Suzanne Kemmer (eds.), 507–533. Stanford: CSLI Publications. Davidse, Kristin, Tine Breban, and An Van linden 2008 Deictification: The development of secondary deictic meanings by adjectives in the English NP. English Language and Linguistics 12: 475–503.

312

Lobke Ghesquière

Declerck, Renaat 1991 A Comprehensive Descriptive Grammar of English. Tokyo: Kaitakusha. De Smet, Hendrik 2005 A Corpus of Late Modern English Texts. ICAME Journal 29: 69–82. De Smet, Hendrik and Jean-Christophe Verstraete 2006 Coming to terms with subjectivity. Cognitive Linguistics 17: 365– 392. Dixon, Robert 1982 Where Have All the Adjectives Gone? And Other Essays in Semantics and Syntax. Berlin: Mouton. Halliday, M. A. K. 1994 An Introduction to Functional Grammar. 2nd ed. London: Arnold. Halliday, M. A. K. and Ruqaia Hasan 1976 Cohesion in English. London: Longman. Hawkins, John A. 1978 Definiteness and Indefiniteness: A Study in Reference and Grammaticality Prediction. London: Croom Helm. Hetzron, Robert 1978 On the relative order of adjectives. In Language Universals, Hansjakob Seiler (ed.), 165–184. Tübingen: Narr. Hopper, Paul J. and Traugott, Elizabeth Closs 2003 Grammaticalization. 2nd rev. ed. [1st ed., 1993.] Cambridge: Cambridge University Press. Langacker, Ronald W. 1991 Foundations of Cognitive Grammar. Volume 2: Descriptive Application. Stanford: Stanford University Press. Langacker, Ronald W. 2001 Discourse in cognitive grammar. Cognitive Linguistics 12: 143–188. Le Trésor de la Langue Française informatisé. Available online at http://atilf.atilf.fr/tlf.htm. McGregor, William B. 1997 Semiotic Grammar. Oxford: Clarendon Press. Nevalainen, Terttu and Matti Rissanen 2002 Fairly pretty or pretty fair? On the development and grammaticalization of English downtoners. Language Sciences 24: 359–380. OALD: Hornby, A.S.; Sally Wehmeier, Michael Ashby (eds.) 2000 Oxford Advanved Learner’s Dictionary of Current English. Oxford: Oxford University Press. OED: Murray, James A.H., Henry Brodly, W.A. Craigie and C.T. Onions (eds.) 1933 The Oxford English Dictionary. Oxford: Oxford University Press.

The adjectives of completeness

313

Paradis, Carita 1997 Degree Modifiers of Adjectives in Spoken British English. Lund: Lund University Press. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech and Jan Svartvik 1972 A Grammar of Contemporary English. London: Longman. Sinclair, John 1990 Collins COBUILD English Grammar. London: Collins. Stefanowitsch, Anatol 2006 Negative evidence and the raw frequency fallacy. Corpus Linguistics and Linguistic Theory 2: 61–77. Traugott, Elizabeth Closs 1982 From propositional to textual and expressive meanings. Some semantic-pragmatic aspects of grammaticalization. In Perspectives on Historical Linguistics, Winfred P. Lehmann and Yakov Malkiel (eds.), 245–271. Amsterdam/Philadelphia: Benjamins. Traugott, Elizabeth Closs 1989 On the rise of epistemic meanings in English: An example of subjectification in semantic change. Language 65: 31–55. Traugott, Elizabeth Closs 1995 Subjectification in grammaticalization. In Subjectivity and Subjectivisation: Linguistic Perspectives, Dieter Stein and Susan Wright (eds.), 31–54. Cambridge: Cambridge University Press. Traugott, Elizabeth Closs 2001 Legitimate counterexamples to unidirectionality. Paper presented at Freiburg University, Germany, October 2001. Available at http://www.stanford.edu/~traugott/papers/Freiburg.Unidirect.pdf Traugott, Elizabeth Closs 2003 From subjectification to intersubjectification. In Motives for Language Change, Raymond Hickey (ed.), 124–139. Cambridge: Cambridge University Press. Traugott, Elizabeth Closs 2007 (Inter)subjectification and unidirectionality. In Journal of Historical Pragmatics 8: 295–309. Traugott, Elizabeth Closs this vol. (Inter)subjectivity and (inter)subjectification: a reassessment. In this volume, 29–71. Traugott, Elizabeth Closs and Richard Dasher 2002 Regularity in Semantic Change. Cambridge: Cambridge University Press. Vandelanotte, Lieven 2002 Prenominal adjectives in English: Structures and ordering. Folia Linguistica 36: 219–259.

314

Lobke Ghesquière

Vandewinkel, Sigi 2005 Attitudinal adjectives and category shift in the nominal group: The case of pure. MA Thesis, Department of Linguistics, University of Leuven. Van linden, An and Kristin Davidse 2005 Degrees of subjectivity of prenominal adjectives: diachronic change, linear order, and constructional relations as determining factors. Paper presented at the conference From ideational to interpersonal: perspectives from grammaticalization. Leuven. 10–12 February 2005.

On the rise of (inter)subjective meaning in the grammaticalization of kind of/kinda Hélène Margerie Abstract This corpus-based study investigates the various uses of the pragmatic particle kind of/kinda, ranging from propositional to expressive meanings. An attempt is made at delineating a grammaticalization path, which reveals the extent to which grammaticalization intersects with subjectification and intersubjectification. The paper draws on Denison’s (2002, 2005) constructional analysis but special emphasis is laid on the different determiner patterns the nominal use of kind fits in and on the role they played in the (inter)subjectification process accompanying the grammaticalization of kind of/kinda.

1. Introduction This paper is devoted to the evolution of the form kind of and its phonetically reduced counterpart kinda in terms of an increase in subjectivity “understood as relationship to the speaker and the speaker’s beliefs and attitudes” and intersubjectivity “understood as relationship to the addressee and addressee’s face” (Traugott this volume: 30; see also Traugott 2003). Drawing on Denison (2002, 2005), I wish to depict a tentative grammaticalization cline for the development of the two forms as pragmatic markers starting from the uses of nominal kind modified by the PP [of + NP]. The various stages of the grammaticalization of kind of illustrate a gradual move towards greater subjectivity and intersubjectivity: pragmatic overtones do not arise full-fledged in later stages but already come to the surface of some initial uses of nominal kind before they are deeply entrenched in the more grammaticalized uses of the pragmatic marker. While previous work on the topic essentially emphasized the pragmatic meaning of kind of in its particle uses (Aijmer 1984), my purpose here is to highlight this pragmatic component from the very start of the grammaticalization path through some specific uses at the propositional level.

316

Hélène Margerie

According to Traugott (this volume: 41), “intersubjectification intersects less extensively with grammaticalization. In most languages it is grammaticalized only into some discourse markers and interjections”. The present study shows that intersubjectification can also be involved in the grammaticalization of intensifiers, as it is in the case of the development of kind of/kinda into hedges, for instance. The paper is organized as follows: Section 2 shows the various uses of kind of and kinda in several present-day English corpora, e.g. BNC, COLT, CSAE and MICASE (see sources of data for full details) before an attempt is made in Section 3 to outline a grammaticalization path which illustrates both subjectification and intersubjectification. This last step in the analysis focuses on the different patterns within which nominal kind fits in the pattern [DET (determiner) + kind + (of + N(oun)P(hrase))] (Sections 3.2.1 to 3.2.5). Section 4 provides some concluding remarks. 2. The various uses of kind of/kinda I will first go over the uses of the two forms in order to illustrate their various meanings, ranging from propositional to expressive (Traugott 1989; Traugott and König 1991), with ambiguous meanings related to these two supposedly distinct facets of language (Andersen 2000: 48). Previous analyses of pragmatic markers, especially discourse markers (Brinton 1996: 38; Aijmer 1984), insist on the fact that they convey no propositional meaning at all. On the other hand, they reflect procedural meaning in the sense that they are useful, if not necessary, indications for the addressee on how to process the information contained in the speaker’s discourse. Following Andersen (2000: 48), I argue that the use of kind of/kinda as a pragmatic marker is not exclusive of propositional meaning. Indeed, in a very large number of examples in which [kind + of] is primarily propositional, e.g. in the pattern [DET + kind + (of + N)], or when it is used as an intensifier, it definitely has a pragmatic ring. The uses are classified in three major categories and the tripartite division into i) nominal use ii) intensifier and iii) pragmatic marker is supposed to reflect the different syntactic stages characterising the grammaticalization of kind of (cf. Section 3).

The grammaticalisation of kind of/kinda

317

2.1. The propositional meaning of kind of/kinda 2.1.1. Original meaning The first stage in the grammaticalization cline investigated here is the occurrence of kind as a head noun post-modified by the PP [of + NP] as in examples (1–5): (1) (2) (3) (4) (5)

They wanna look at the kind of college you’re coming from. (MICASE) Again if by definition an elegy is a kind of poem, then it can’t– then poetry can’t be a representation of the elegy. (MICASE) Language as used by Columbus versus the kinda language that we imagine the indigenous people used. (MICASE) They don’t even know what it is like if I’m qualified for what kind of job or not. (MICASE) This kind of test if you’re doing things right, that kinda test is out of the way in an hour’s time. (MICASE)

These examples show the many patterns involving nominal kind and a whole range of determiners in the corpora investigated. Other determiners determining nominal kind are presented in Section 3.2 because they give rise to an ambiguous interpretation, and further syntactic patterns involving the determiners preceding nominal kind in examples (1–5) above, yet in specific contexts, will then be examined in relation to the evolution of kind (of) from propositional to pragmatic meaning (Section 3.2). In (1–5), kind conveys propositional meaning: as a noun, it refers to “a group whose members share certain qualities” (Longman Dictionary of Contemporary English, 2nd ed.: 575) and it is synonymous with “sort” or “type”. Semantically, the noun primarily evokes a cluster of items grouped together on the basis of shared properties, of a likeness. 2.1.2. Derived meaning The idea that nominal kind, in its primary sense, evokes similarity without involving strict identification between the different members of a group may account for other nominal uses which slightly deviate from the original meaning (see Sections 3.1 and 3.2.1 for further detail). In (6),

318 (6)

Hélène Margerie

It’s a kind of reddish-brown colour. (MICASE)

kind conveys the idea of approximation of the colour the speaker has in mind to the norm represented by the colour reddish-brown. Kind (of) expresses “an approach to the type” whereby the property “is, or may be, included in the class in question, though not possessing its full characteristics” (OED, 8: 437). This notion of approximation is the one that is mapped onto the use of kind (of) as a grammaticalized form (i.e. as a pragmatic marker; see Section 3.1), and possibly first as a degree modifier (see Section 3.2.6 for a discussion of the chronological rise of the intensifier and the pragmatic marker functions). 2.2. Kind of/kinda as an intensifier 2.2.1. Type and frequency In the corpora investigated, the nominal uses of kind of/kinda are by far outnumbered by uses as intensifiers of adjectives, adverbs or verbs (cf. Section 2.3.6 for a quantitative analysis). The intensifier function may represent the first step in the grammaticalization of the form from the nominal stage to the pragmatic level.1 In the first part of CSAE investigated (which I will now simply refer to as CSAE), over one third of the total occurrences of kind of/kinda are examples of the intensifying function whereas the nominal uses are scarce (4 out of 79 occurrences) or receive an ambiguous interpretation. Intensifiers are divided into different categories. Three subtypes come into play here, namely boosters, compromisers, and diminishers (Quirk et al. 1972: 439). Occurrences of kind of/kinda as a compromiser have the highest frequency rate in the three corpora (COLT, CSAE, MICASE) in which a quantitative analysis was carried out (see Section 2.3.6) while the 1. For the sake of simplicity, I refer to the three main stages in the grammaticalisation of kind of/kinda as (1) ‘the nominal use/propositional level’, (2) the intensifier use, and (3) the pragmatic marker use. This, of course, should not lead to the misinterpretation that only stage (3) sees the rise of (inter)subjective meaning. I am only using the phrase ‘pragmatic level’ here to refer to the ultimate phase in the grammaticalisation of kind of/kinda which comprises stages (3–5) of the cline represented in Figure 3 (Section 3.1).

The grammaticalisation of kind of/kinda

319

other two types of intensifiers are more unusual, their interpretation being tightly dependent on the co-text. I thus start the analysis of kind of/kinda as an intensifier with the compromiser function. 2.2.2. Compromiser In (7–10), the speaker indicates by means of the compromiser kind of/kinda that the notion conveyed by the verb or adjective pre-modified by the intensifier applies to some degree, but not fully: (7) (8) (9) (10)

I kind of like that sort of colour. (COLT) “Are you tired?” “Not really. I mean kind of but …” (MICASE) “Are they scary?” “Kind of.” (COLT) You know them hopper things you bounce on, the round balloon thing? She’s got one, it’s flat, it’s not flat but it’s kind of deflated. (BNC)

In (8), the quality expressed by tired is partially denied by means of not really, but the speaker then specifies that she is nevertheless tired to some moderate degree. Similarly, in (10), the speaker first describes the balloon as flat before she negates the integrity of this quality and finally moderates the extent to which the balloon is “deflated”. On a gradient representing the quality flat, the notion evoked by the speaker would be located half-way. In (6) above, the presence of the adjective in front of the head noun colour may give rise to a similar interpretation. We may suppose that kind of has scope over the adjective reddish-brown only.2 It would thus combine semantically quite well with the suffix -ish denoting approximation too. Just as red-brown does not exactly match what the speaker has in mind, neither is reddish-brown the most accurate term for the colour in question.

2. The syntactic parsing would then be the following: [NP a [Nƍ [AP [DegMod kind of] a reddish-brown]] colour] (adapted from Denison 2002; Tabor 1993: 6–7). The pattern may be contextually interpretable as such, but there is evidence against a similar analysis of pattern [a kind of + ADJ + N] in historical data (cf. Denison 2002). See Section 3.2.1 (examples 46–47) for further detail.

320

Hélène Margerie

2.2.3. Pragmatic meaning in the use of the compromiser In most (if not all) cases, the compromiser conveys (inter)subjective meaning, just as rather usually does (Bäcklund 1973: 69; Paradis 1997: 6, 11; Margerie 2008), which kind of indeed immediately precedes in (11): (11)

I kind of rather felt sick and went away. (COLT)

It is then not just a marker of degree; some extra (inter)subjective information is also conveyed, the meaning of which is inferred in context, as in (12): (12)

I was kind of … crude … but I was like “so what is your father doing right now? Is the man dead yet?” (CSAE)

Kind of has a pragmatic overtone in that it implies that the quality expressed by crude is unsuitable or bad. The speaker is aware that her question about her schoolmate’s father being dead is rude. She feels embarrassed and introduces the topic tentatively, as shown by the pause after kind of. Kind of here points to the awkwardness of the speaker and her awareness that the term crude suggests she behaved rudely and is therefore hard to mention because it is face-threatening for the speaker herself (Brown and Levinson 1987). (12) is thus an example of kind of as an index of intersubjectivity. (13) is another example of the intersubjective meaning of the intensifier kind of/kinda: (13)

[About a nursery] A: It’s starting to look a little bit yellow. B: I know but it’s what’s natural, it’s the real grass and it’s … A: Yeah but I mean it looks kind of sort of off. (COLT)

Speaker A starts off quite cautiously by describing speaker B’s lawn as being a little bit yellow. A little bit is a diminisher which speaker A uses as part of a politeness strategy vis-à-vis the addressee. As speaker B does not seem to agree fully with speaker A, the latter insists, but still remains extremely cautious, as indicated by the reduplication of downtoning devices,

The grammaticalisation of kind of/kinda

321

e.g. kind of and sort of (cf. Traugott this volume: 37 for the intersubjective, and more precisely intersubjectified,3 meaning of sort of as a hedge). The co-text is also quite significant in the interpretation of kind of as both a compromiser and a pragmatic marker in (14): (14)

A: I think that’s so funny. It’s a little cartoon, it’s just cheap and … like Bart Simpson kind of. B: Oh. A: Have you seen it? B: Yeah, I think I did see it. A: It’s kind of stupid but it’s kind of funny if you know what I mean cos the baby don’t talk and as soon as the Mum and Dad’s out of the room it’s let’s go and find the chocolate. It’s so funny. (BNC)

It is quite evident in the last sentence that speaker A has some reservation in telling the co-speaker that she is fond of the cartoon. Indeed she first describes the cartoon as “so funny”. It is very likely that she fears the cospeaker’s opinion on this subject because she then qualifies her own judgment by saying that the cartoon is “kind of funny”. Having just said that the cartoon is “kind of stupid”, it might be too face-threatening not to qualify the previous statement “it’s so funny”. For that reason, the high degree modifier so is given up in favour of the compromiser, although it is taken up again at the end of the sentence (“it’s so funny”) because the speaker obviously gets away with the description of the cartoon and expresses her true opinion about it. In (15), kind of is very close to the meaning of rather in that it probably goes beyond the meaning of the compromiser and comes close to a booster meaning because of the pragmatic interpretation conveyed by the expression “kind of funny”: (15)

I don’t understand why all these guys stop going out with you, and shortly afterwards get fired from Joe’s , kind of funny that ha? (COLT)

3. The search for the collocation was conducted through MICASE, COLT and CSAE. MICASE returned the greatest number of hits, probably due to its bigger size (1,848,364 words). In this corpus I found 27 occurrences of the collocation involving really and kinda, in the form of really kind of/kinda or kind of/kinda really.

322

Hélène Margerie

It is to be noted that funny counts as one of the most frequent adjectives modified by kind of, which almost turns the sequence kind of funny into a set phrase. Here are a few examples of this frequent collocation: (16) (17) (18)

I remember, he came to Bahia after one of his shows one night. ’s kinda funny. (CSAE) I remember, like I went there with this person, it’s kind of funny. This person did not want to dance. (CSAE) It’s kinda funny when you laugh so hard you cry. (MICASE)

In (15), the speaker is being ironic. She most certainly thinks that there is indeed a tight link between the two events she is mentioning. What she really means, therefore, is not that the link is just rather funny – in the sense of being weird to a moderate degree – but that it is very funny. Again, it is the ironic co-text which makes the interpretation of kind of as a booster more appropriate here. The next section discusses this function. 2.2.4. Booster In (19), the emotional co-text suggests an interpretation of the intensifier as a booster carrying the quality wicked to a high degree: (19)

Oh I like that song. That’s wicked! … There’s gonna be DJ and all that. With all hard core music. It’s safe. Kinda wicked! Well I’m buying my ticket today. (COLT)

Here the meaning is no longer that of the compromiser. The speaker is very enthusiastic about the upcoming event. Such examples are thin on the ground, to say the least. The corpora investigated yielded only one occurrence of kinda – example (19) – illustrating the booster function. However its collocation with really or other boosters points to a possible evolution of kind of/kinda towards a booster in its own right (Margerie 2007: 293–298). Here are a few examples of this quite frequent collocation:4

4. We may assume that all uses of kind of/kinda as an intensifier involve (inter)subjective meaning. Considering that such meaning does not simply arise in “relevant contexts” (Traugott this volume: 35) but is rather now coded in the

The grammaticalisation of kind of/kinda

(20)

(21)

(22)

323

You have to take your, your own educational goal very seriously, and explore check out different things and kinda go, that was an experience I can pass on next time I don’t wanna do more of that or you know I like sociology I’ll take more of that or, I liked math I even find it you know I found it really kind of interesting I like the group work I wanna go on in that. (MICASE; emphasis mine) A: The group was small and I saw myself fitting very well there. I could contribute, so it’s, been fine, so far. I guess you don’t want all those answers, huh? B: No, I am, I am interested in those, you’ll see. You’ve answered some questions that I haven’t asked yet. But, that’s great. So, the fit is really kind of important. Not just the topic, but, um, that it, that it work, as a group, that you feel good in it, feel comfortable. Yeah, yeah, right ... I think so too. People, people often sort of propose, group work, w– without really considering, how hard group dynamics can be. You know how crucial they are. (MICASE; emphasis mine) I mean isn’t this a gorgeous thing? I’m really kinda pleased how well these, show up. (MICASE; emphasis mine)

The co-text (cf. emphasis) indicates that really has left its mark on the semantics and pragmatics of kind of/kinda which it collocates with, so that the latter may be taking on yet another function within the category of degree modifiers, i.e. that of booster. Admittedly, this new function needs to be used on a much wider scale before it becomes entrenched in the language and we definitely associate it with one of the uses of kind of/kinda as an intensifier. 2.2.5. Diminisher In (23–25), kind of is best interpreted as a diminisher, i.e. a low degree intensifier: (23)

You see this, you see this area, okay, no you see this uh, section right (there?) where it kinda curves around a little bit? alright, that’s uh that’s the hippocampus. (MICASE)

use of the form as an intensifier, this is an instance of subjectified meaning as described by Traugott (this volume: 35, 37).

324 (24)

(25)

Hélène Margerie

Well it sort of flattens out, Sarah says it doesn’t flatten out I don’t know. It kinda flattens out a little bit, but it doe– certainly doesn’t go back up again right? alright so, if you’re looking at this curve and you’re trying to fit it then what what you’re gonna try to fit then, would be the, the portion … before it starts coming back up because it certainly hasn’t come back up again you know? (MICASE) It’s kinda squiggly, it’s not straight but it’s analogous to something straight, it’s very close to being straight. (MICASE)

In (23–24) a little bit, which is a diminisher itself, is a contextual clue for the interpretation of kinda as a diminisher too. It could be argued that these instances are rather examples of the approximator use (cf. Section 2.3.1) or the hedging function (Section 2.3.2). But it is also likely that when combined with a pragmatic downtoning effect – as it might actually be in (24) – the degree marker, which is most usually interpreted as a compromiser when it is an intensifier, contextually develops a diminisher function. In (25), kinda cannot be interpreted as a compromiser: the speaker’s explanation makes it clear that the property expressed by “squiggly” is not applied to a moderate degree to the object in question, otherwise the beginning of the sentence would be very much at odds with the speaker’s last remark making it clear that the object is “very close to being straight”. 2.2.6. Summary To sum up, when kind of/kinda is used as an intensifier, it has propositional meaning in the sense that it partakes in the semantic interpretation of the sentence. Removing kind of in most, if not all, examples leads to a different meaning, and indeed I have shown that its absence is quite odd due to the immediate co-text. Yet its propositional function is sometimes, if not often, coupled with an interpersonal function whereby subjective and intersubjective information is imparted. As an intensifier, kind of/kinda therefore bridges the gap between two functions which are generally clearly set apart, the propositional and the expressive functions of language (Traugott 1989). The next section illustrates its purely expressive role.

The grammaticalisation of kind of/kinda

325

2.3. Kind of/kinda as a pragmatic particle I hereafter distinguish several subtypes of pragmatic markers, which all convey the general notion of approximation derived from the nominal and intensifier uses. 2.3.1. Approximator Examples (26–27) illustrate the use of kind of/kinda as an “approximator” (Prince, Frader, and Bosk 1982: 85) or “adjuster word” (Austin 1962: 74): (26) (27)

People who are organizing foreign policy they have to, kinda work on two levels, one with the the other country and one like within, with the force within their, their own country. (MICASE) Oh I was in a really good mood; I woke up this morning in a really good mood, so I kind of danced into work. (BNC)

An approximator is a device “by whose aid, in spite of the limited scope of our vocabulary we can always avoid being left speechless” (Austin 1962: 74). Kinda invites the hearer to infer that the expression “work on two levels” in (26) is to be interpreted loosely (Andersen 2000: 48). The speaker implies that she uses a figurative expression to convey the meaning intended, so that the phrase should not be interpreted literally. In the following conversation, kinda has a very specific meaning as an approximator: (28)

Rebecca: When he came through here, did he look at you at all? Rickie: M [hm]. Rebecca: [Ini]tially? Rickie: Yeah. Rebecca: What did he do? Rickie: Just looked. He [had a] … Rebecca: [Did he] stop walking? Rickie: No, just kinda looked … and then looked, and then … walked. Rebecca: Ok. Rickie: And then he kept walking.

326

Hélène Margerie

Rebecca: When he looked at you the first time, was he sort of … looking you up and down, or … did he just look at your seat? What … Rickie: Just like … you know, look. (CSAE) Rebecca obviously tries to redefine the meaning of the verb look more precisely, hence her several questions to Rickie. Kinda is an adjuster word in the sense that it suggests that the word look may not be exactly the best term to describe what happened to Rickie on the train. More precisely, kinda here evokes a representation of the notion look on a gradient. In Rickie’s reply (in bold), the two occurrences of the verb look seem to have two slightly different meanings. The pre-modification of the first occurrence by means of kinda alters the meaning of the verb: now it is semantically closer to steal a glance at someone than to look in the sense of getting a full view of someone. When saying that she “just kinda looked”, Rickie suggests that her perception was incomplete. The full sense of look, on the other hand, is conveyed through the second occurrence of the verb, e.g. “... then looked”. What kinda seems to do is set up a gradient for the notion expressed by the verb look. It here locates the notion the speaker has in mind at some point on the gradient which is not the upper end of it corresponding to the full expression of the verb look. 2.3.2. Hedge The use of kind of/kinda as a hedge, e.g. a device meant to “make things fuzzier or less fuzzy” (Lakoff 1972: 195), is extremely frequent. Its use as an approximator could be included in this category as well, although it has an epistemic modal (evidential) function whereas the hedging device is here considered in relation to its affective (intersubjective) function. In the two corpora of American English investigated, namely MICASE and CSAE, the hedging function always outnumbers the other uses (cf. Figures 1, 2 in Section 2.3.5). The situation is different in COLT because it represents British English where kind of/kinda is not very frequent as opposed to its synonym sort of. The hedging function is illustrated in (29–31):

The grammaticalisation of kind of/kinda

(29)

(30)

(31)

327

A: Right good, alright so what do you– do you have specific questions about this for me like are there parts that you think are, weak and parts you think are particularly strong? B: I was just kinda hoping you’d read over and say this has to be changed or you know whatever. (MICASE) A: Now, that’s what printer you should get. Have you got a printer? B: Yes. A: Oh, keep it still. B: But this is kind of a, that my dad’s use. (CSAE) I like her and everything but, holy shit, I really need I do need some time to go to school and like be by myself. Do you know what I mean? It’s kind of fucking important. (COLT)

“Hedging particles are an important resource for the realisation of politeness strategies” (Aijmer 2002: 8). In (29), the hedging force of kinda is reinforced by just, and the form BE+-ING (I was hoping) also indicates carefulness on the part of the speaker who asks the hearer very politely if she might be willing to read the paper, without sounding too demanding or pushy. The difference in status between the two interlocutors might actually account for the polite reply of speaker B – the latter is a senior undergraduate whereas speaker A is a senior graduate student. In (30) the co-text makes it clear that the speaker feels awkward in telling the hearer that she cannot use the printer she just said was her own for fear that she should appear to be contradicting herself. In (31) kind of is a hedge for the use of the intensifier fucking which may sound vulgar to the addressee. 2.3.3. Marker of focus This is admittedly a most tentative subclass distinction, all the more so as (32) is the single example provided in support. It is tightly connected to other subclasses, especially the hedging function in which some would actually rather include example (32). However, the use of kind of (like) in this example might suggest a different function, i.e. focus marker: (32)

I went upstairs to the BART, and told them, and I tried to tell them what train it was, and I just … I was still kind of like shocked. (CSAE)

328

Hélène Margerie

Kind of is indeed introduced in a highly emotional co-text here. Like, following kind of, seems to function as a marker of focus (Underhill 1988: 238–240) whereby “focus” is defined as “the most significant new information in a sentence – often the point of the sentence” (Underhill 1988: 238). In (32), the speaker lays great emphasis on the adjective shocked and kind of serves to reinforce the focus device. Kind of might not be a marker of focus on its own, but there are reasons to believe that in conjunction with like, when the latter is used as a focus marker, kind of serves the speaker’s intention to highlight the content of the item modified by the phrase kind of like. Admittedly, the data provided in support of this function are much too limited to confirm whether the focus marker use constitutes a separate subclass. 2.3.4. Marker of reported speech or thought This function has hardly been mentioned as such in the literature so far. Usually, kind of/kinda combines with a marker of reported speech or thought such as go or be like (Aijmer 2002: 183), as in (33–34): (33) (34)

I went erm “excuse me can you clean something please” and she kind of went “you little shit, I’m gonna get my husband to come and break your legs.” (COLT) You’re just kinda like “that’s a weird thing. …” (MICASE).

Yet in (35) kinda functions as a marker of reported speech on its own: (35)

I just got a visual, Sharon standing in front of the class going, (SCREAM), while these little kids kinda “Señorita Flynn? Hee hee hee hee hee.” (CSAE)

In view of the uniqueness of example (35), the use of kinda as a marker of reported speech might be considered exceptional, probably resulting from the (involuntary?) deletion of like as a marker of reported speech itself (these little kids kinda like “Señorita Flynn? Hee hee hee hee hee”; cf. Blyth, Recktenwald, and Wang 1990; Romaine and Lange 1991). Still the occurrence of kinda in this example is worth mentioning as it may lead to similar future uses.

The grammaticalisation of kind of/kinda

329

2.3.5. Filler/marker of pause/reflection/repair/hesitation device (36) (37)

What I did, kind of, in regard to that, was … and I don’t have this date finalized but, s– … I am trying to go maybe the second week … of … January. (CSAE) A: Where’s that red pen gone? B: I don’t know you had it. C: Pick it up. B: Who done that? A: No one, it just fell off. B: It’s supposed to be it’s supposed erm kind of . (CSAE)

In (36), kind of marks a pause in the speaker’s discourse as she is trying to finish her sentence or to find the right words. The hesitation also transpires through the digression “and I don’t have this date finalized”. In (37), the repetition (cf. “it’s supposed”) and the marker of hesitation erm complement the use of kind of as a marker of reflection. All the occurrences of kind of/kinda as a purely pragmatic marker illustrate the subjective and intersubjective component of language. They convey no propositional meaning but mark procedural information: they are pragmatic indications for the hearer on how to process the sentence. The major idea conveyed is that of approximation: when using kind of/kinda as a pragmatic device, the speaker suggests to the hearer that the notion discussed, the speech or thought reported, or the words used are just approximate and may not represent the very idea the speaker has in mind. The pragmatic markers convey “lack of precision, which is one of the most important features of the vocabulary of informal conversation” (Crystal and Davy 1975: 111). 2.3.6. Quantitative distribution of the uses of kind of/kinda The distribution of the uses of kind of/kinda in the corpora investigated is represented in Figures 1 and 2. The few indeterminate uses have been excluded. Figure 1 represents the three main uses, e.g. “nominal”, “intensifier” and “pragmatic marker”. Figure 2 distinguishes between the hedging function and other pragmatic functions within the general category “prag-

330

Hélène Margerie

matic marker” of Figure 1. The category “other pragmatic meaning” subsumes both the approximator use and the more isolated uses presented in Sections 2.3.3 to 2.3.5. The hedging function has been set apart in order to highlight its predominance. While the “pragmatic meaning” category (Figure 1) is syntactically well distinct from the nominal and the intensifier categories, I have suggested at the beginning that the propositional and the pragmatic meanings sometimes combined, leading to a blurring of the semantic-pragmatic boundaries between the different uses that were distinguished, especially between the intensifier and the hedge uses. However, let me say again that when I refer to the “pragmatic use” of kind of/kinda in Figure 1, I am not referring to the nominal or intensifier uses. Admittedly, the latter occasionally show overlap with the category of hedges (cf. Paradis 1997: 6, 11; Margerie 2008) but they are still syntactically and pragmatically distinguishable from the purely pragmatic uses subsumed under the “hedge” and “other pragmatic meaning” rubrics represented in Figure 2. Nominal Intensifier Pragmatic meaning COLT

CSAE

MICASE

Figure 1. The distribution of the uses of kind of/kinda

Hedge Other pragmatic meaning COLT

CSAE

MICASE

Figure 2. Proportion of hedges in the pragmatic meaning category

The grammaticalisation of kind of/kinda

331

I will now show that the idea of approximation found in the intensifier and the pragmatic uses is inherited from the lexical use of [DET + kind + (of + N)] and that subjectivity is felt right from the start of the grammaticalization path of kind (of), in the original use as a noun in specific syntactic contexts. 3. Sketching a grammaticalization path for kind (of) Denison (2002, 2005) offers a constructional analysis of the historical development of kind (of). The present analysis draws on his arguments, highlighting the interaction of syntactic and semantic factors, but it mostly focuses on specific patterns involving nominal kind which already suggest a progressive transfer of the original semantic notion of approximation onto later uses. 3.1. Overview of the cline Figure 3 below represents a hypothetical cline, which is fairly similar to Denison’s speculative reconstruction (2002). The difference is that I have grouped together (at stage 2) some of the constructions that are analysed separately in Denison (2002). The aim is to show that, though distinct from a syntactic viewpoint, these constructions share a semantic value which makes them highly likely starting-points for the adverbial construction under which the intensifier and the pragmatic functions are subsumed. Denison envisages all probable scenarios for the historical development of kind of. He therefore examines different pathways, from binominal to qualifying, from qualifying to adverbial, from binominal to complex determiner, or from complex determiner to qualifying/adverbial, to name the main scenarios (see Section 3.2 for a definition of these terms). In the cline represented in Figure 3, I have included the complex determiner construction [these/those kindsing of + Npl] (see Section 3.2.3) under the same rubric as what Denison classifies as instances of the qualifying construction. This is not to say that the distinction between these two constructions is not significant. But again what I wish to shed light on is the clustering of, essentially, semantic-pragmatic (but also, to a certain extent, syntactic) properties uniting these two constructions, and their distinction, in that respect, from the adverbial construction. This is why the speculative repre-

332

Hélène Margerie

sentation in Figure 3 focuses on functional and/or semantic, not syntactic headings as in Denison (2002). The two historical pathways are supposed to complement each other insofar as Figure 3 is to be seen as an elaboration of some of the constructions delineated in Denison (2002). (1) category membership Ex. the kind of college (2) approximation Ex. a kind of reddish-brown colour

(3) (possible) intermediate stage Ex. it was kinda kind of a surprise

DET + kind + (of + NP)

A kind of N Some kind of N These kindsing(ular) of Npl(ural) A kind of a N/ADJ + N

Kind of + (a + Nsg) Kind of + Nsg/pl

(4) degree Ex. it’s not flat but it’s kind of deflated

Kind of + ADJ/ADV/V

(5) pragmatic meaning Ex. it’s kind of fucking important

Kind of + ADJ/ADV/V

Figure 3. A speculative historical development

I will concentrate on stages 2 and 3 showing kind in specific syntactic and semantic contexts. These two stages illustrate the move of kind away from its nominal status and the gradual entrenchment of the notion of approximation which is most central in the pragmatic uses of kind of/kinda. I will more precisely focus on the type of determiners involved in the construction as a most helpful indication for the semantic-pragmatic interpretation of the form in later stages (i.e. stages 4 and 5). The speculative cline in Figure 3 is supported in part by the historical data available. The OED (1989, 8: 437) does not make a distinction between the last three stages represented. These stages show kind of in its “adverbial function”, which according to the OED appeared in the late nineteenth–early twentieth centuries for kinda.

The grammaticalisation of kind of/kinda

333

Stage 2, which is the main focus of the present analysis, is illustrated through historical data and further examples are provided from present-day English. But the examples presented to illustrate stage 3 are drawn from MICASE, a corpus of present-day English, because of a lack of examples of the adverbial construction in historical corpora – Denison (2002) does not provide historical data for the construction either; his examples are extracted from present-day English corpora. I cannot therefore claim for sure that the form historically first modified a noun phrase (as in stage 3) rather than a verb phrase or an adjectival phrase (as in stages 4 and 5). Note, however, that Denison (2002) envisages a similar extension from noun phrase to verb phrase or adjectival phrase modification. The lack of historical data for stage 3 confirms for the moment the hypothetical position of constructions at stage 3 on the cline of grammaticalization. It is not a real problem for the present analysis insofar as it is stage 2 which is given a crucial role in the further grammaticalization of kind of into an adverbial, stage 3 being a potential transitional step towards stages 4 and 5. 3.2. Kind of in specific contexts In addition to the original uses of nominal kind (of) examined previously (cf. examples 1–5), there are other examples resembling the pattern [DET + kindnom(inal) + (of + NP)] but they actually differ from this string on a syntactic and semantic-pragmatic level, which explains why they are singled out as instances of a distinct construction. 3.2.1. A kind of N The pattern [a kind + of + N] can illustrate, in some contexts, what Denison (2002, 2005) calls “the qualifying construction”,5 which he distinguishes from the binominal construction illustrated in (1–5) at the very beginning of this paper. (38) is an example of the qualifying construction:

5. In other contexts, it can be an instance of the binominal construction, as in example (2), repeated here for the sake of convenience: Again if by definition an elegy is a kind of poem, then it can’t– then poetry can’t be a representation of the elegy.

334 (38)

Hélène Margerie

When thanks is not forthcoming, we feel a kind of emptiness. (borrowed from Denison 2005)

Denison argues that kind is not a noun in this construction, nor is of a preposition. Whereas kind is the head noun in the binominal construction, here kind of has a qualifying function. (38) does not imply that there are different kinds of emptiness. Rather, it means that the feeling the speaker is referring to approximates the idea of emptiness. The speaker is not able to determine what is felt precisely, although it closely resembles what is called emptiness. This is the meaning conveyed in (39) from the Corpus of Early English Correspondences Sampler (now CEECS): (39)

For newes, I wrote you of late that Shenkes had taken a town and castle in Westfalia called Werl. Synce that, the enymyes of that countrey gathered together, both the gentlemen and ablest men, and offred a kind of siege of the towne.

Surprisingly, Denison (2002) dates the qualifying construction back to the late sixteenth century – but provides no example of it at the time. Yet (39) points to an earlier development, in the early fifteenth century. (40–41) are examples of the construction in the late sixteenth century. In (40), “a kinde of tender” is a rephrasing of the preceding segment “as ’twere a tender”. Kind of combines quite well with this phrase in order to suggest some kind of distance with the term it modifies: (40)

Come Coz, come Coz, we stay for you: a word with you Coz: marry this, Coz: there is as ’twere a tender, a kinde of tender, made a farre-off by Sir Hugh here: doe you vnderstand me? (Shakespeare, The Merry Wives of Windsor; from Gutenberg)

A similar example is found in (41): (41)

Most barbarous intimation! Yet a kind of insinuation, as it were, in via, in way, of explication. (Shakespeare, Love’s Labour Lost; from Gutenberg)

In examples (42–45) from contemporary spoken English, we find several indications that the noun following kind of may not correspond exactly to

The grammaticalisation of kind of/kinda

335

what the speaker really wishes to express. It is again an approximation of her thought: (42) (43) (44) (45)

They in a sense become a kind of, uh, ornament if you will. (MICASE) The architectural work is based on a kind of um, uh m– mass production. (MICASE) [He sees him] as a kind of um, predecessor to all urban planners. (MICASE) You know, a kind of, you know, I mean … (MICASE)

The pauses as well as the markers of hesitation, “uh” or “um”, or of discourse planning (Schiffrin 1987), “you know”, “I mean”, point to the speaker’s attitude to what is being said, especially here to her quest for the right word, or to her uncertainty, as in (43). The pre-modification of the head noun by means of an adjective in (46– 47) is further indication of the particular status of kind of: (46)

(47)

Gon.: I do well beleeue your Highnesse, and did it to minister occasion to these Gentlemen, who are of such sensible and nimble Lungs, that they alwayes vse to laugh at nothing. Ant.: ’Twas you we laugh’d at. Gon.: Who, in this kind of merry fooling am nothing to you: so you may continue, and laugh at nothing still. (Shakespeare, The Tempest; from Gutenberg) Launcelot: There is but one hope in it that can do you any good; and that is but a kind of bastard hope neither. Jessica: And what hope is that, I pray thee? Launcelot: Marry, you may partly hope that your father got you not, that you are not the Jew’s daughter. Jessica: That were a kind of bastard hope, indeed. (Shakespeare, A Midsummer Night’s Dream; from Gutenberg)

The string [ADJ + N] appears as an unusual expression which may have been made up for the purpose of the conversation. In (46) kind of signals the newness of this expression and invites the hearer to infer an imagistic meaning. In (47), the speaker sets up distance because of the use of a “vulgar” expression.

336

Hélène Margerie

3.2.2. Some kind of/kinda N This pattern is quite similar in that respect. I believe that the marker some reinforces the idea of approximation or uncertainty, indicating that the referent of the head noun is qualitatively vague. Not all instances of this pattern show a non-nominal use of kind. But my purpose is to mention some particular uses in which kind can hardly be labelled a noun. (48) is an example of the use of kind as a head noun determined by some: (48)

A: So it’s some kinda dehy– dehydrogenase isn’t it? B: What is it? A: Some kinda dehydrogenase? B: Right, what kinda dehydrogenase? (MICASE)

Each speaker is trying to figure out what exactly the name of the gas is. When A uses some in her second utterance, remaining vague about the kind of dehydrogenase she is talking about, then B asks for more precision. B’s second utterance suggests that she uses kinda as a head noun and may interpret it similarly in A’s utterances. Indeed her question is clearly meant to ask for more precision about the type of gas. At the same time, the use of some suggests that speaker A cannot qualitatively determine the referents of the nouns very precisely, and kinda also leaves them unspecified. Example (49), on the other hand, shows a different type of use. Kind of reflects the speaker’s attempt to find the appropriate word or phrasing: (49)

You’re moving it into some kind of, uh, uh sort of, purely d– uh, p– uh realm of purely domestic houses … (MICASE)

Again, the markers of hesitation (“uh”), the near repetition of kind of (“sort of”) and the pauses all point in this direction. (50) shows that back in the early fifteenth century, kind of already had an ambiguous status when used with the determiner some: (50)

But yet herwith she also is very carefull how those countrees may be governed without harm to the public cause, and how hir own army, consistyng of hir people, might also be ruled and directed; of both which, though hir majesty hath had some kind of speches, yet

The grammaticalisation of kind of/kinda

337

she myndeth not to mak any resolution but by your lordships advise. (CEECS) We might consider the following analysis: some determines the head noun speches and kind of is a marker signalling approximation of the speaker’s thought. This would be an instance of Denison’s “qualifying construction”. Kind would not be a head noun, nor would of be a preposition. The sequence kind of would have a qualifying function which would then make it optional. The problem is that if we remove kind of in (50), the sentence becomes very awkward: (50ƍ)

… though hir majesty hath had some speches.

It may be possible to consider that kind of is a complex determiner – in Denison’s terms – for speches (but see Section 3.2.3). Considering the semantic deviation of kind and its indeterminate syntactic functioning in (50), we might envisage that such an ambiguous pattern contributed to some extent to the reinterpretation of kind of and its syntactic reanalysis. 3.2.3. These/those kindsing of Npl Another pattern in which kind illustrates a move towards greater grammaticalization is the occurrence of the plural determiner these or those with kind in the singular and the head noun in the plural. The pattern is attested as from the late fourteenth century (OED 1989, 8: 437). Kind shows definite signs of distance from the nominal source: the determiner these/those grammatically requires plural marking on the noun but in this particular pattern the noun bears singular marking, suggesting that it no longer shares all the characteristics of a typical noun and is going down the cline of grammaticalization – an instance of decategorialization (Hopper 1991) in the framework of grammaticalization. Example (51) dates back to the early fifteenth century; (52–53) are from contemporary English.

338 (51) (52) (53)

Hélène Margerie

And, whilest I am writing thus, I praie your lordship thinke howe such service might be done by your maritime forces and flie-boates theare, that are fittest to impeche thos kind of havens. (CEECS) You make these kind of mistakes because you get away with them in spoken speech. (MICASE) I will have a, better model of the actual, statistics, sto– for, um, the probability rate functions, for those kind of, um situations. (MICASE)

In Denison (2005), this is an instance of what he calls “the post-determiner” (or “complex determiner”) construction while adding that it is of uncertain status. In Traugott’s view (this volume: 42) kind of is a “Degree Modifier” in examples such as (51–53). The problem with this label is that it is impossible to substitute kind of for another degree modifier like rather, for instance, in the preceding examples. The syntactic role of kind of in the string [these/those kindsing of Npl] is quite complex, and though it is certainly of great importance for understanding the later uses, it is not essential for the present analysis which mainly addresses the semantic and pragmatic aspects of the construction. I will therefore leave this issue aside. Suffice to notice the gradual denominalization of kind which leads it away from the original binominal construction. I think that in the above examples, kind of gives some information about the appropriateness of the words or phrase used and indicates an epistemic or evaluative distance between the discourse and the message actually intended. That kind of imparts subjective information of this kind is quite obvious in (53) where the several pauses in the discourse suggest that the speaker is trying hard to find the right terms. In other words, it is likely that, from a purely syntactic point of view, the occurrence of kind of in this specific pattern highly influenced the grammaticalization of the form into an intensifier and then a pragmatic particle – because of the denominalization of kind. Now in semantic terms, I suppose that both the “post-determiner” construction and the qualifying construction (or its later extensions; cf. Section 3.2.4) may have provided a starting point for the development of the form into stages 3 to 5 – which, again, is the reason why they are presented at the same stage in Figure 3. The next pattern examined may have developed on the basis of the qualifying construction.

The grammaticalisation of kind of/kinda

339

3.2.4. A kind of a N/A kind of a ADJ N In these two closely related patterns, illustrated in (54–55), a kind of seems to pre-modify the NP [a + N] or [a + ADJ + N]: (54) (55)

He has acknowledged that as, a kind of a school. (MICASE) It’s a kind of a big interdisciplinary thing. (MICASE)

In (54), for instance, the speaker uses a kind of to suggest that the meaning of school is to be interpreted loosely. What she wishes to refer to is an approximate idea of a school. We might suppose that in utterances containing the qualifying construction, speakers reinterpreted the sequence [a + kind + of] as a marker indicating the looseness of discourse, so that there was no obvious determiner in the noun phrase. The addition of the determiner A just in front of the head noun – notice that A is the only determiner possible in that position – may then have been thought to be necessary, or at least quite helpful, in the syntactic parsing of the sentence. A kind of, therefore, has almost achieved the status of a set phrase in this pattern. The OED does not mention it. We cannot therefore date its first uses exactly. While there are many occurrences in present-day English, such as in (54–55), I found one single occurrence in Shakespeare’s work as compiled by Project Gutenberg: (56)

I am but a foole, looke you, and yet I haue the wit to thinke my Master is a kinde of a knaue. (Shakespeare, The Two Gentlemen of Verona; from Gutenberg)

We may surmise that the pattern gained ground later, and indeed there are numerous occurrences in, for instance, Daniel Defoe’s Robinson Crusoe (dating back to the early eighteenth century), among which are (57–60): (57) (58) (59) (60)

I took it to be a kind of a hawk. They might not see what kind of a governor they had. I made a kind of a hut for that night’s lodging. To this purpose, getting a kind of a letter of naturalization, I purchased as much land …

340

Hélène Margerie

The pauses in examples (61–63), from present-day English, confirm the speaker’s search for the phrase that will best approximate the idea the speaker has in mind: (61) (62) (63)

Kun ku is more of a kind of a, chamber drama. (MICASE) [He does all that] with a– kind of a, spirit of adventure, I mean, he’s, he … (MICASE) Uh, uh, it doesn’t have a kind of a, a substance to it. (MICASE)

If the hypothesis is correct that this construction arose later than the pattern [a kind of N], for instance, illustrative of the qualifying construction, then this is further proof that a kind of has become a set phrase by this stage, modifying the following NP, which leads us to the next construction, i.e. [kind of a Nsing/Npl]. 3.2.5. Kind of [a Nsing]/[Npl] The next step in the grammaticalization seems to be the disappearance of the determiner a before kind of, creating greater distance from the original nominal pattern and leading to increased syntactic bonding within the construction, as in (64): (64)

It was kinda kind of a surprise. (MICASE)

A higher degree of grammaticalization is reached once the head noun itself is not determined by a as in (65), and can then show plural marking, as in (66): (65)

(66)

The reason they were showing us this at college, was just … to get us, to know if our trimmer and shoer were_s … doing it right, and– plus the knowledge of knowing the bones and ligaments which we … we had it in another class too. We– that was kind of a double thing that, we had in … in another class, so it was kinda review for us. Well it was a review for some people, depend on what time of the year you took it. You know. If it was a review or not. (CSAE; emphasis mine) A: What, curled like curls? Tongs? B: Ringlets.

The grammaticalisation of kind of/kinda

341

A: Oh, like yours or ringlets? B: Ringlets, like that type of, not ringlets but tight kind of ringlets. (COLT) The co-text in (65) (cf. emphasis) makes it clear that the speaker qualifies the meaning of the word review in “it was kinda review for us”. The speaker’s hesitation as to the appropriateness of the word explains the use of kinda here, at least from a functional point of view. What remains unclear is its syntactic status. Just as Denison (2002) posits the status of post-determiner illustrated in the pattern [these/those kindsing of Npl], we might as well consider the possibility that kinda functions as determiner here. But this would be a much unexpected use. We may rather consider that kinda is here to qualify the meaning of the noun and that the intended utterance was It was kind of a review, with kinda amalgamating the determiner a. I will leave aside the issue of the syntactic status of kinda in (65). What is to be noted, though, is that the pattern resembles the intensifier and the pragmatic marker very closely. In (66), kind of does not have the plural marking expected if kind were a noun, because of the plural marking on ringlets. It here suggests that the type of hair the speaker is thinking of resembles ringlets but cannot be called ringlets. The notion of approximation is again essential. As it is no longer felt to be a typical noun, kind grammaticalizes further into a marker of approximation. In the preceding examples, kind of still premodifies nouns, although it closely resembles the function associated with stage 4 in Figure 3. This is why I regard stage 3 as a possible transitional step in the grammaticalization of the form which would then give rise to its use with phrases other than nominal ones, leading kind (of) even further away from its nominal origins. 3.2.6. Stages 4 and 5 of the cline The last two stages distinguish between the intensifier use and the different pragmatic uses (approximator, hedge, etc.). The problem is that the lack of hindsight makes it difficult to ascertain whether the intensifier use historically preceded the pragmatic use or vice versa. Traugott (this volume: 42) notes that the three constructions she focuses on, e.g. a piece of, a bit of, and a shred of, “are part of a larger set of Partitive > Quantifier > Degree Modifier changes that occurred mainly in the eighteenth and nineteenth

342

Hélène Margerie

centuries” which includes the markers kind of and sort of. She therefore derives the intensifier function directly from the nominal uses of the forms in question. Moreover, on the basis of general principles of grammaticalization and (inter)subjectification, we may consider that the more subjective meanings develop from the less subjective ones (Traugott 1989). In that case, we may argue that the intensifier uses arose earlier than, say, the hedging uses because the latter have a purely intersubjective meaning devoid of any expression of degree. Syntactically speaking, we might think that the two patterns illustrated at stages 4 and 5 are pretty much alike. However kind of/kinda at stage 4 still modifies the following item directly, be it an adjective, an adverb or a verb. The pragmatic particle, on the other hand, has potential wider scope in that it can be seen to have scope over the whole utterance, or to have an external role, considering that it expresses the speaker’s attitude to what is being said. As shown by Traugott (1995), grammaticalization does not systematically exemplify scope restriction (as argued in Lehmann [1982] 1995); it may rather involve scope extension, as in the case of discourse markers (Traugott 1995). Stage 5 of Figure 3 would then be an example of scope extension and pragmatic strengthening, supporting the claim that the uses of kind of/kinda at this stage developed later than the uses exemplified at stage 4. The indeterminacy regarding the historical development of stages 4 and 5 explains why these two stages are not further detailed. The point of this whole section was essentially to show the crucial part that stage 2 (and possibly stage 3) has (have) played in the development of the semanticpragmatic meaning of kind of/kinda at stages 4 and 5. 4. Conclusion The grammaticalization path delineated in Figure 3 shows that kind (of) has evolved along a continuum leading it towards the more (inter)subjective pole of language. It is therefore an instance of subjectification and intersubjectification as defined by Traugott (2003, this volume). The different determiners used in the construction played a key role in this evolution. They have guided the interpretation of the form in terms of pragmatic meaning, even when the uses were primarily propositional.

The grammaticalisation of kind of/kinda

343

A full account of the grammaticalization cline under investigation would require an analysis of the different uses as a pragmatic particle in terms of their development through time. Again, the recent development of these functions impedes such a reconstruction. However, in a study conducted by Fleischman and Yaguello (2004) about the pragmatic marker like and its French equivalent genre, the authors claim that “both the FOCUS and the HEDGE functions lead naturally to the QUOTATIVE function [marker of reported speech function], the most recent of the pragmatic functions to emerge in the various languages that display parallel markers” (141). This seems to tie in with the data presented in this paper, as only one occurrence of the quotative function was found, suggesting that this function developed later than the other pragmatic functions. A more thorough analysis of the historical development of the various functions as pragmatic markers will be made possible if the new functions that kind of/kinda occasionally exemplifies develop further in future time. Sources of data BNC CEECS COLT CSAE MICASE Gutenberg

British National Corpus Corpus of Early English Correspondences Sampler Corpus of London Teenage Language Corpus of Spoken American English, Part 1 Michigan Corpus of Academic Spoken English http://www.gutenberg.org

References Aijmer, Karin 1984 Sort of and kind of in English conversation. Studia Linguistica 38: 118–128. Aijmer, Karin 2002 English Discourse Particles: Evidence from a Corpus. Amsterdam/Philadelphia: Benjamins. Andersen, Gisle 2000 Pragmatic Markers and Socio-Linguistic Variation: A RelevanceTheoretical Approach to the Language of Adolescents. Amsterdam/Philadelphia: Benjamins.

344

Hélène Margerie

Austin, John L. 1962 How to Do Things with Words. Cambridge: Harvard University Press. Bäcklund, Ulf 1973 The Collocation of Adverbs of Degree in English. Acta Universitatis Upsaliensis. Uppsala. Blyth, Carl Jr., Sigfrid Recktenwald and Jenny Wang 1990 I’m like, ‘Say what?!’: A new quotative in American oral narrative. American Speech 65: 215–227. Brinton, Laurel J. 1996 Pragmatic Markers in English: Grammaticalization and Discourse Functions. Berlin/New York: Mouton de Gruyter. Brown, Penelope and Stephen C. Levinson 1987 Politeness: Some Universals in Language Usage. Cambridge: Cambridge University Press. Crystal, David and Derek Davy 1975 Advanced Conversational English. London: Longman. Denison, David 2002 History of the sort of construction family. Paper presented at ICCG2, Helsinki, 6–8 September 2002. Denison, David 2005 The grammaticalisation of sort of, kind of and type of in English. Paper presented at NRG 3, Santiago de Compostela, 17–20 July 2005. Fleischman, Suzanne and Marina Yaguello 2004 Discourse markers across languages: Evidence from English and French. In Discourse across Languages and Cultures, Carol Lynn Moder and Alda Martinovic-Zic (eds.), 129–148. Amsterdam/Philadelphia: Benjamins. Hopper, Paul 1991 On some principles of grammaticalization. In Approaches to Grammaticalization, vol. 1, Elizabeth Closs Traugott and Bernd Heine (eds.), 17–35. Amsterdam/Philadelphia: Benjamins. Lakoff, George 1972 Hedges: A study in meaning criteria and the logic of fuzzy concepts. In Papers from the Eighth Regional Meeting, Chicago Linguistic Society, Peranteau Paul M., Judith N. Levi and Gloria C. Phares (eds.), 183–227. Chicago: Chicago Linguistic Society. Lehmann, Christian 1995 Thoughts on Grammaticalization. Munich/Newcastle: LINCOM EUROPA. Original publication in Arbeiten der Kölner UniversalienProjektes 48, University of Cologne, 1982. Longman Dictionary of Contemporary English. 2nd ed. London: Longman.

The grammaticalisation of kind of/kinda

345

Margerie, Hélène 2007 From downgrading to (over)intensifying: A pragmatic study in English and French. In Explorations in Pragmatics: Linguistic, Cognitive and Intercultural Aspects, Ivan Kecskes and Laurence Horn (eds.), 287–311. Berlin/New York: Mouton de Gruyter. Margerie, Hélène 2008 A historical and collexeme analysis of the development of the compromiser fairly. Journal of Historical Pragmatics 9: 288–314. Paradis, Carita 1997 Degree Modifiers of Adjectives in Spoken British English. Lund: Lund University Press. Prince, Ellen F., Joel Frader and Charles Bosk 1982 On hedging in physician-physician discourse. In Linguistics and the Professions: Proceedings of the Second Annual Delaware Symposium on Language Studies, Robert J. di Pietro (ed.), 83–97. Norwood, N.J.: Ablex. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech and Jan Svartvik 1972 A Grammar of Contemporary English. London: Longman. Romaine, Suzanne and Deborah Lange 1991 The use of like as a marker of reported speech and thought: A case of grammaticalization in progress. American Speech 66: 227–279. Schiffrin, Deborah 1987 Discourse Markers.Cambridge: Cambridge University Press. Simpson, John and Edmund Weiner 1989 Oxford English Dictionary. 2nd ed. Oxford: Oxford University Press. Traugott, Elizabeth Closs 1989 On the rise of epistemic meanings in English: An example of subjectification in semantic change. Language 65: 31–55. Traugott, Elizabeth Closs 1995 The role of the development of discourse markers in a theory of grammaticalization. Paper presented at ICHL XII, Manchester, 1995. Traugott, Elizabeth Closs 2003 From subjectification to intersubjectification. In Motives for Language Change, Raymond Hickey (ed.), 124–139. Cambridge: Cambridge University Press. Traugott, Elizabeth Closs This vol. (Inter)subjectivity and (inter)subjectification: A reassessment. In this volume, 29–71. Traugot, Elizabeth Closs and Ekkehard König 1991 The semantics-pragmatics of grammaticalization revisited. In Approaches to Grammaticalization, vol. 1. Elizabeth Closs Traugott

346

Hélène Margerie

and Bernd Heine (eds.), 189–218. Amsterdam/Philadelphia: Benjamins. Underhill, Robert 1988 Like is, like, focus. American Speech 63: 234–246.

Part 4 (Inter)subjectification and grammaticalization involving modals

Better as a verb David Denison and Alison Cort Abstract We trace the present-day pattern seen in You better watch out! from its Old English origins, where the adjective better occurs in construction with BE, and from Middle English also with HAVE. The intersecting history of a number of different patterns is shown to lead to the modern layering of the forms had better, ’d better and plain better, which are similar to modal verbs in many but not all respects. The verbal uses are subsumed in the paper under the label BETTER. On the way the complementation of the BETTER construction becomes more tightly constrained: explicit comparison is lost, as are clause types other than plain infinitivals. Conversely, the subject slot becomes less constrained. However, a strong preference for first and second person subjects is evident. Interaction with proverbs is found to play a part in the licensing of clipped sentences without expressed subject. Once the form history has been established, the semantic history is considered and related to changes of form. We explore deontic and epistemic semantics, including mutual compatibility in a single example, as well as counterfactual use. Somewhat orthogonal to the development of epistemic meaning is an increase of intersubjectivity that is associated with the rise of the current meaning of advisability.

1. Introduction1 The lexical item had better ~ ’d better ~ better has been on the fringes of the modal verb system for a long time. For brevity we will refer to the verb in its various forms as BETTER.2 Its history has not been widely discussed, nor has great attention been paid to its PDE behaviour (though note here 1. The corpus data and much of the work for the oral paper in 2005 were contributed by AC. The present revision is mostly DD’s work, hence the fluctuation between authorial we and I. I am grateful to Martina Faller and John Payne for helpful advice on modality and logic, though I don’t expect to have satisfied either of them, and to two anonymous referees for their detailed comments, some of which at least I have been able to take account of. 2. We adopt the notational convention of small caps for verb lexemes.

350

David Denison and Alison Cort

Mitchell 2003, which we have found valuable). However, both its form history and its semantics are interesting and intricate, and the latter in particular call into question some familiar assumptions about modality and about semantic and pragmatic change. We will not give much space to the question of whether – or to what extent – BETTER belongs to the category Modal, although we will frequently look at its behaviour, distributionally and semantically, in terms that would be appropriate for a modal verb. Our main goal is to throw light on its semantic history in Section 3, then to trace the growth of interpersonal usage in Section 4. A necessary preliminary in Section 2 is to trace the morphosyntactic history – in part to justify classing BETTER as a verb at all. A brief afterword forms Section 5. 2. Form history 2.1. Present-day English grouping In PDE BETTER seems to belong to a group of phrasal items of similar shape. Consider: (1)

a. b. c.

I’d better get a takeaway. I’d rather get a takeaway. I’d sooner get a takeaway.

These three items are lumped together by Quirk et al. among the ‘modal idioms’ (1985: 137, 141–143), and by OED s.v. have v. 22a. All three have as their (original) lexical core an item in the comparative which is adjectival or adverbial or sometimes indeterminately either, used with a light or auxiliary verb and complemented, in the constructions which are of interest to us, by a clausal element. They look similar in form and show some – but only some – similarities of meaning as well. As for the typical modal semantics of obligation, volition and necessity, BETTER would appear to approach an obligation or necessity sense, though its actual meaning is closer to advisability (see Section 3.4 below). With had/would rather and had/would sooner there is more of a sense of preference, which is related to volition. However, OED gives ‘preference or comparative desirability’ as the general sense of all these idioms and more.

Better as a verb

351

2.2. Basic construction types To what extent can a common history be attached to the set of lexemes and constructions exemplified in (1a–c)? Part of their interest lies in the trading relationship between A/Adv and the verbally conjugated item. With BETTER, the verb at the outset was BE, whereas in the earliest (and still surviving) pattern for the other two, it was the originally volitional would: (2)

NPsubj would A/Adv + clause

Subsequently with those two, had and occasionally shall or should became alternatives to would as the verbal element. For a detailed account, see Hall (1881).3 BETTER clearly has a different beginning. The items which do resemble BETTER to begin with are patterns involving lever, the comparative of lief ‘agreeable’, no longer in use. In medieval times the lever constructions were far commoner than BETTER and can therefore be used with caution to fill out details of the early history of BETTER. The history of BETTER (and lever) in OED and elsewhere involves the following patterns: (3) (4) (5) (6)

(h)it is A + clause NPobl is A + clause NPsubj be A + clause NPsubj have A/Adv + clause

Pattern (4) is the oldest involving a referential NP as argument, while (6) is the ModE one, but what is the status of (5)? An example of (5) (with trivial word order difference) is: (7)

a1450(?c1421) Lydg. ST (Arun 119) 2024: He better war to ha ben in pees. (MED s.v. betere adj. 1b[b])

OED changes its mind as to the history of this phase, stating s.v. better a. and adv. A.4b that the dative pronoun of (4) ‘was subsequently changed into the nominative’ of (5), then ‘[f]inally this was given up for the current’ (6). An alternative account appears s.v. have v. B.22c: that (5) only arose 3. Much of Hall’s discussion is framed in the form of orotund footnotes which could almost have been the model for Flann O’Brien’s preposterous de Selby commentaries.

352

David Denison and Alison Cort

out of the confusion between (4) and (6), along with another and clearly blended form, (9). The dates of appearance of (5) and (6) are too close to decide the issue, as we will see below. We have expressed the patterns (3–6) in formulaic terms. NP has human reference and is often pronominal,4 so that even after the loss of OE case, an oblique-case NPobl can usually be discriminated from a subjective NPsubj. Pattern (3) as formulated lacks NP altogether; we have not come across examples with an oblique NP, though from ME onwards such an NP can occur in a for-phrase. A/Adv can be lief, leofre/lever, levest, bet, better, best, selest, well;5 comparative forms are the most common, superlative less so, while the positive forms are typically used with as. Clause can be a thatclause, contact clause, to-infinitive or plain infinitive, though with the advent of forms with HAVE in (6), only infinitival clauses are used. One piece of evidence which supports the first of OED’s accounts – of a sequential development (4) > (5) > (6) – is a stage (4ƍ): (4ƍ)

NPsg

is A + clause

The NP is singular and of indeterminate case, while the verb inflection is 3 sg, so that it is unclear whether the NP is subject or oblique. Pattern (4ƍ) can be regarded as intermediate between (4) and (5). We have noted some five probable examples for BETTER, including (8), and pattern (4ƍ) certainly occurs for lever too: (8)

a1325 SLeg.(Corp-C 145) 131/91: A mon were betere for is sunne be[o] sori and vnssriue þanne issriue wiþoute sorinesse. (MED s.v. unshriven [ppl.])

4. Although there are examples in OE where NP is nominal (e.g. þam menn, þam hæþenan cilde, ælcum men, Ðæm lytegan, þam men, Ðam yrmingan, ælcum cristenum men), they are greatly outnumbered by pronominal NPs, a preponderance which we can assume continued into ME. As we shall see in Section 2.5.2, the successor constructions in ModE also show a preference for pronominal subjects. 5. Examples of the latter, somewhat rare, lexemes: (i) For þan us is selest ær ðam deaþe, þæt we onginnan þisse forgifenan tide brucan (HomM 11 [ScraggVerc 14] B3.5.11) therefore us is best before the death that we begin this given time to-enjoy (ii) 1844 MOZLEY Ess. (1878) II. 27 You must give way; and you had as well do so voluntarily. (OED)

Better as a verb

353

There are several further complications. One is an apparently blended construction: (9)

NPobl

had A + clause

This appears briefly in late ME, completing the analogical proportion (10)

NPobl be : NPsubj be :: ? : NPsubj have

See Visser (1963–73: i 33–34), who refers to van der Gaaf (1904). In fact, on the rather modest evidence that Visser gives,6 (9) appears at much the same date as (6), so it is unclear whether the development was (5) > (6), or (4) > (9) > (6). The dating of lever constructions tends to support the former, but rather brings in a blended (9) construction well before (6). Another problem is the use of such patterns as me/I had been better + clause. Visser lists them (plus some irrelevant examples) separately from those with tensed BE (1963–73: i 34). Conceivably the HAVE form in such patterns could be regarded as the principal verbal item and hence parallel to HAVE in (9) or (6), respectively. The impression is reinforced by examples like (11): this looks very much like pattern (6) with a superfluous been. (11)

1604 SHAKES. Oth. III. iii. 363 Thou had’st bin better haue bin borne a Dog Then answer my wak’d wrath. (OED, Visser)

Logically, however, had’st in (11) is a mere auxiliary of the perfect, with as the principal verb, as in (4) or (5), and we shall provisionally assume this latter, simpler analysis. The had been better forms with a clausal complement postdate the patterns with non-perfect BE. A referee adds another ingredient: the sporadic use of better as a sentence adverb with other verbs, as in BE

(12)

c1340 Cursor M. (Laud MS.) 9815 His hert ought bettyr breke in iije Then fro his byddynges to fle. (OED)

Whether such examples facilitated a switch from adjective use with BE to more adverb-like use with HAVE, or whether they merely show analogical

6. Visser only gives two examples of (9) with better, marked “14th c.” and “End 14th c.”, respectively, but the first comes from a manuscript dated a1425.

354

David Denison and Alison Cort

influence from rather, (s)he leaves open. It can be seen that the range of forms in the early period is something of a tangle. The later history of BETTER brings more subtle changes. From the start the verb HAVE is nearly always in the form had (or hadst), which OED calls a past subjunctive, and it is not clear how often other forms of HAVE are used; OED’s and MED’s non-past examples all involve lever or rather, but cf. perhaps (13)

1561 T. HOBY tr. Castiglione’s Courtyer IV. Uvijb, You haue better declared the vnluckinesse of yonge men, then the happynesse of olde menn. (OED)

In time, anyway, the inflected verb becomes confined to past tense had. In turn the form had is increasingly often reduced to ’d and finally to zero. In the layering which is typical of a grammaticalization process, all three forms continue to co-exist. To judge from the evidence of OED, MED and Visser (1963–73), date ranges are approximately as shown in Table 1, though the dates quoted are no more than indicative. It is clear that in some of its early history BETTER patterns with lever, while later it partly resembles rather. Table 1. Date ranges for patterns of better, lever and rather pattern (3) (4) (4ƍ) (5) (2) (9) (6)

lever never found OE to 1614 a1340 to c1380 c1320 to c1450 c1450 c1390 to mid-15th c c1300 to 1750

better OE to PDE OE to 1470–85 c1250 to c1333–52 [??] 1303 to 1647 [irrelevant?] a1425 c1410 to PDE

rather never found never found never found never found 1280–90 to PDE a1325(c1280) c1450 to PDE

2.3. Clausal complementation Meanwhile, complementation is always non-finite with the had better patterns. The complementing clause is nearly always headed by a plain infinitive. Exceptions involve counterfactuals, which have been possible intermittently in ModE and which may permit to-infinitives beside the more common plain infinitives. The original adjectival it BE better construction of (3) has always permitted various types of complement clause.

Better as a verb

(14) (15) (16)

355

1542 Brinklow’s Complaint 64, 14 Ye had bene better to have suffred death (van der Gaaf, Visser) 1723 CHAMBERS tr. Le Clerc’s Treat. Archit. I. 79 It had been better..to have fluted the upper part. (OED s.v. flute v. 3a) 1866 Evening Star 19 Mar., He had better have been a dead man than have emitted from his mouth..such a rum-bred pestilence of breath. (OED)

Both adjectival better and counterfactuality are simultaneously exemplified in (14) and (15). 2.4. Comparison Example (16) illustrates another salient development: that it has become ungrammatical with BETTER to include the standard of comparison in a than- or as-phrase (see also Mitchell 2003: 140–141). This change applies to the BETTER construction generally, not just in counterfactual use. Thus the comparison is included in the seventeenth-century (17) and nineteenthcentury (16) and (18) but has become impossible in the PDE (19): (17) (18) (19)

1613 SHAKES. Hen. VIII, V. iii. 132 He had better starue Then but once thinke his place becomes thee not. (OED) [1961] I insisted on takeing the field and prevailed – thinking that I had better die by rebel bullets than [by] Union quackery. (Brown 7 F18 1580) a. *We had better get a takeaway than start cooking now. b. *We’d better get a takeaway than start cooking now. c. *We better get a takeaway than start cooking now.

2.5. Subject 2.5.1. Animacy of subject From the eighteenth century we start to get examples where the NP subject, previously always human in reference, is inanimate or a dummy NP:

7. Example (18) appears to be a quotation from a Civil War letter.

356 (20) (21)

David Denison and Alison Cort

1712 ADDISON Spect. No. 287 (page)3 There had better be none at all. (OED s.v. have v. B.22a) We’ve come a long way and spent a lot of money. It had better be good. Don’t worry about the Acropolis. It is awe-inspiring. (1961 Brown G51:35)

It is of course typical of auxiliaries not to impose any subject selection. 2.5.2. Constraints on expressed subject There is no absolute restriction on the kinds of NP found with BETTER either. A range is illustrated in (22–25): (22) (23) (24) (25)

However, during the summer Lord King warned that Boeing had “better get going” if it was to beat Airbus to the order. (BNC A9D 38) What’s so terrible about that? As a matter of fact, Grandpa better fork over some dough and pretty fast or I’m going to make his little Sarah into a shiksa. (ARCHER 1964gelb.d8a) “When Margaret Thatcher gets in you and your bloody unions better watch out.” (BNC AOU325) Oh, there had better be two of you. (BNC AT7 2570)

However, one kind of subject is very strongly favoured with BETTER: pronoun subjects. BETTER has a markedly higher level of subject pronouns than the semantically similar modals should and ought to, for instance. In Cort’s (2006) dataset, while should has a pronominal subject just over 40% of the time and ought to roughly 50%, with BETTER nearly 82% of examples have pronouns as subject; see Table 2. Particularly prominent are first and second person pronouns, which between them account for more than 70% of all instances. The predominance of first and second person is actually even higher than that, since – as we shall shortly see in Section 2.5.3 – a further 13% of BETTER examples lack an expressed subject, and such examples are nearly all implicitly first or second person. The subject NPs illustrated in (22–25), fully grammatical though they are, are actually rather exceptional statistically.

Better as a verb

357

Table 2. Pronominal subjects with better, should and ought to in six corpora BETTER

I you he she it we they total PRO sub total ex

I you he she it we they total PRO sub total ex

ARCHER 19 20% 44 46.32% 4 4.21% 3 3.16% 0 0% 12 12.63% 3 3.16% 85 89.47% 95

LOB 11 22% 14 28% 1 2% 1 2% 0 0% 10 20% 0 0% 37 74% 50

ARCHER 39 10.74% 25 6.89% 26 7.16% 8 2.20% 19 5.23% 25 6.89% 22 6.06% 164 45.18% 363

LOB 13 12.62% 8 7.77% 3 2.91% 6 5.83% 4 3.88% 7 6.80% 9 8.74% 50 48.54% 103

LLC 21 52.5% 10 25% 1 2.5% 1 2.5% 0 0% 3 7.5% 0 0% 36 90% 40

Brown 6 15% 15 37.5% 5 12.5% 2 5% 1 2.5% 2 5% 1 2.5% 32 80% 40

FLOB 6 16.22% 13 35.14% 1 2.7% 0 0% 0 0% 7 18.92% 1 2.7% 28 75.68% 37

Frown 5 13.89% 14 38.89% 5 13.89% 0 0% 0 0% 2 5.56% 0 0% 26 72.22% 36

total 68 22.82% 110 36.91% 17 5.70% 7 2.35% 1 0.34% 36 12.08% 5 1.6% 244 81.88% 298

ought LLC Brown 28 11 17.18% 17.19% 20 5 12.27% 7. 35% 10 7 6.13% 10.29% 2 3 1.27% 4.41% 9 2 5.52% 2.94% 28 3 17.18% 4.41% 12 3 7.36% 4.41% 109 34 66.87% 50% 163 68

FLOB 5 8.62% 2 3.45% 3 5.17% 3 5.17% 5 8.62% 5 8.62% 2 3.45% 25 43.10% 58

Frown 2 4% 2 4% 3 6% 1 2% 3 6% 6 12% 1 2% 18 36% 50

total 98 12.17% 62 7.70% 52 6.46% 23 2.86% 42 5.22% 74 9.19% 49 6.09% 400 49.69% 805

358

David Denison and Alison Cort

I you he she it we they total PRO sub total ex

ARCHER 483 21.32% 111 4.90% 138 6.93% 47 2.08% 108 4.77% 136 6.00% 78 3.44% 1101 48.61% 2265

LOB 107 8.22% 36 2.77% 70 5.38% 27 2.08% 92 7.07% 79 6.07% 49 3.77% 460 35.36% 1301

should LLC Brown 165 34 27% 3.83% 42 30 6.87% 3. 38% 18 44 2.95% 4.96% 7 5 1.15% 0.56% 44 71 7.20% 8% 64 61 10.47% 6.88% 36 30 5.89% 3.38% 376 275 61.54% 31% 611 887

FLOB 63 5.65% 37 3.32% 51 4.57% 15 1.35% 81 7.26% 58 5.20% 94 8.43% 399 35.78% 1115

Frown total 36 888 4.76% 12.80% 31 287 4.1% 4.14% 28 349 3.70% 5.03% 13 114 1.72% 1.64% 41 437 5.42% 6.30% 43 441 5.68% 6.36% 28 315 3.7% 4.54% 220 2831 29.06% 40.82% 757 6936

2.5.3. No expressed subject In a high proportion of instances BETTER occurs in clipped constructions without any expressed subject. (Unstressed had is never retained when the subject is omitted, and in the absence of a phonological host, contracted ’d is obviously impossible; all such examples therefore involve the zero-form better.) (26)

Where should I go? To my room? Better stay in the hotel lobby, where the walls looked good and thick. (Brown F24 0090)

Considered as a kind of modal verb, BETTER is unusual in this respect. Although other modals do sometimes undergo subject clipping of this type, as in examples (27–30), it is highly infrequent in the corpus data collected by Cort (2006); see Table 3. (27)

Leiter turned at the door. ‘Take it easy, James. Be back in an hour and we’ll go and get ourselves a good dinner. I’ll find out where they’ve taken Tingaling and we’ll mail the dough to him there.

Better as a verb

(28) (29) (30)

359

Might cheer him up a bit, the poor little bastard. (ARCHER 1956flem.f8b) 7 P.M. Sits with its head down, engaged in picking at imaginary objects in front of it. Can find its way in and out of its cage when roused to action. (ARCHER 1873ferr.s6b) Mary and I stood by here for the call on December 30th and 31st but it didn’t come through. Must have been caught up in the New Years telephone overcrowding. (ARCHER 1961hmwy.x8a) My period – right on time. Couldn’t be more regular. (ARCHER 1978ryan.j8a)

Table 3. Clipping with modal verbs ARCHER Brown FLOB Frown clips total clips total clips total clips total Can 6 3084 0 1762 0 1757 0 1725 0.19% Could 7 2765 0 1776 2 1569 3 1471 0.25% 0.13% 0.2% May 1 2747 0 1301 0 1100 0 883 0.04% Might 4 1301 0 663 4 641 1 638 0.31% 0.6% 0.16% Must 5 2242 2 1013 0 803 1 662 0.22% 0.2% 0.15% Ought to 1 363 0 68 0 58 0 51 0.28% Shall 3 2109 0 267 0 200 0 149 0.14% Should 3 2265 0 887 1 1115 1 756 0.13% 0.09% 0.13% Will 4 – 0 2344 0 2215 1 1793 0.06% Would 3 4222 0 2842 0 2293 0 2410 0.07% total 37 – 2 12406 7 11751 7 10389

LOB clips total 0 2147 1 1604 0.06% 1 1338 0.07% 6 779 0.77% 3 1096 0.27% 1 103 0.97% 0 351 2 0.15% 1 0.04% 1 0.04% 16

1301 2347 2774 13840

The relative rarity of clipped structures with these modals, never rising above a fraction of one per cent of occurrences, contrasts quite dramatically with BETTER, where clipped constructions account for a significant propor-

360

David Denison and Alison Cort

tion of the data – between 10 and 20% in all but one of the corpora; see Table 4. Table 4. Clipping with BETTER BETTER

ARCHER Brown FLOB FROWN LLC LOB total

clipped structures 10 (10.5%) 9 (22.5%) 7 (18.9%) 1 (2.8%) 4 (10%) 9 (18%) 40 (13.4%)

total data 95 40 37 36 40 50 298

2.5.4. Proverbs The unusually high level of clipped structures with modal BETTER calls to mind some familiar proverbial patterns that begin with better: (31) (32) (33)

Better to have loved and lost than never to have loved at all. Better sit still than rise and fall. Better the devil you know than the devil you don’t.

This type of proverb, with or without initial clipping, derives from (3) in form, with explicit comparison retained. Although there are many possible variations on the surface form available, the underlying construction can be expressed as (34)

(It is) better (to) XP1 (than XP2)

The more specific constructions which occur are as follows (most examples from Dykes 1713, Kelly 1721, Ludovici 1791, MacIntosh 1785, Pinker and Jackendoff 2004, Ray 1742, Speake 2003 = “ODP”): – (It) is better (to) VP1 than (to) VP2 ± ellipsis of verb (35)

c 1390 Gower Confessio Amantis v. 7725 Betre is to yive than to take. (ODP)

Better as a verb

(36) (37) (38) (39)

361

It’s better to be stung by a nettle than prick’d by a rose (Ray 1742: 20) It’s better to be happy than wise (Ray 1742: 116) It is better to knit than blossom (Ray 1742: 127) 1837 S. Lover Rory O’More II. xxi. ‘Jist countin’ them,—is there any harm in that?’ said the tinker: ‘it’s betther be sure than sorry’. (ODP)

– Better to VP1 than (to) VP2 (40) (41) (42)

a 1628 in M. L. Anderson Proverbs in Scots (1957) no. 320 Better to wow [woo] over middin, nor [than] over mure. (ODP) 1911 G. B. Shaw Getting Married 116 St Paul’s reluctant sanction of marriage;..his contemptuous ‘better to marry than to burn’ … (ODP) Better to die a beggar than live a beggar (Ray 1742: 2)

– Better VP1 than VP2 (43) (44)

Better have it than hear of it (Ray 1742: 57) Better hold by a hair as draw by a Tedder (Kelly 1721: 55)

– With ellipsis of duplicate verbal material in VP2 comparison (45) (46)

1546 J. Heywood Dialogue of Proverbs I. xi. D2V Sonne, better be envied then pitied, folke sey. (ODP) Better wait on the Cooks as the Leaches (Kelly 1721: 57)

– Better NP1 than NP2 (47)

1857 Trollope Barchester Towers II. vii. ‘Better the d—you know than the d—you don’t know,’ is an old saying..but the bishop had not yet realised the truth of it.

– Better BE NP1 than NP2 (48)

1678 J. Ray English Proverbs 204 Better are small fish then an empty dish. (ODP)

362

David Denison and Alison Cort

2.5.5. Influence of proverbial patterns on BETTER Crucial for our purposes is the pattern illustrated by (43–44), which could also be formulated as a variant of (49)

Better + Vinf + X

Notice that (49) can arise in two ways: by ellipsis of it is in proverbs, and by ellipsis of NP had before modal BETTER. Which came first? As we have seen, clipped proverbs are already attested in the seventeenth century, and in the specific form of (49) by the early eighteenth century; see (40) and (42–48) above. As for clipped modal BETTER, OED has one example from the late eighteenth century and several from the nineteenth century, while ARCHER has this type from the mid-/late nineteenth century only and Visser only from 1922 (1963–73: iii-2 1828): (50) (51) (52) (53)

1794 MATHIAS Purs. Lit. (1798) 73 Better preach With silky voice, and sacred flow’rs of speech. (OED) Better send for her sister. (ARCHER 1867robe.d6b) Wheedle her a bit. Better wheedle her. Safer. (ARCHER 1944bagn.d7b) ALICE: … Mabel Cantwell and I face the press together. Can I get out of it? RUSSELL: No. Better not. (ARCHER 1960vidl.d8a)

Initial clipping happened first in proverbs, therefore. It seems plausible that the pattern of sentence-initial better in proverbs may have helped to license the superficially identical structure for modal BETTER, contributing to the high frequency of ellipsis with BETTER as compared with other modals. Blended structures which seem to combine elements of both can be found on the internet: (54)

(55)

... without believing, the only reason I could think of is if you say “I don’t really believe, but just in case it IS true, I had better be safe.” But isn’t this ... www.authorsontheweb.com/ubb/Forum1/HTML/000258.html ... Upon conferring with Oneonta fire and police departments, Piscitelli said, “we’d better be safe than sorry,” since a cause couldn’t be determined. ...

Better as a verb

(56)

363

www.thedailystar.com/news/stories/2005/01/27/odor3.html – 4 Feb 2005 ... I did not intend to frighten the students that would like to participate in this program, but warn them since we all know that “You better be safe than sorry ...” www.spainexchange.com/student_info/erasmus_tips.php – 4 Feb 2005

This at least lends support to the idea of interaction between proverb structures and the general history of BETTER. 2.6. Counterfactuals There is one other formal property to be discussed, one of the standard tests for modalhood. Can BETTER be used as first verb in the apodosis of a past unreal conditional? Huddleston (1980, table) and Mitchell (2003: 134) say not. To us – and apparently also to Huddleston and Pullum (2002: 113) – the judgement is a little less certain. Perhaps example (57a) is not completely out, though it is certainly marginal at best; (57b) is worse: (57)

a. b.

?*If Blair hadn’t won the 1997 election, he had better have left politics. *If Blair hadn’t won the 1997 election, he(‘d) better have left politics.

To express such ideas in bullet-proof standard English, awkward circumlocutions must be used instead of BETTER, with BE8 rather than HAVE as the last verb before non-verbal better: (58)

a. b. c.

… he would have been better (off) leaving politics. … he would have been better to leave politics. … it would have been better for him to leave politics.

We will return in Section 3.7 below to the PDE unsuitability of BETTER for counterfactuals.

8. Or perhaps DO: (i) … he would have done better to have left politics.

364

David Denison and Alison Cort

2.7. Summary of form history Putting together the history discussed so far, the main stages are as follows: (59) (60) (61) (62) (63) (64) (65) (66) (67)

(h)it is better + clause (+ than) NPobl is better + clause (+ than) NPsubj BE better + clause (+ than) NPhuman HAVE better + (to) + Vinf (+ than) a. NPhuman had better + (to) + have Vpa.ptcp (+ than) b. NPhuman had better + Vinf a. NPhuman ’d better + (to) + have Vpa.ptcp (+ than) b. NPhuman ’d better + Vinf better + Vinf NPhuman better + Vinf NPhuman/inanimate better + Vinf

3. Semantics and pragmatics 3.1. Possible correlations between syntactic change and semantics We can suggest some probable correlations between morphosyntax and semantics whereby transitions between the main stages in the form history correspond to differences of meaning (Table 5). In OE and as long as oblique human NPs are allowed in ME, we can say that the human NP and (h)it are in complementary distribution. The oblique NP is an optional element, and when it does occur, it has a meaning standardly encoded as dative, namely beneficiary. The meaning of the predicate BE better is essentially evaluative. However, we have not found a convincing line of argument to associate a particular semantics with either the NP or the predicate in the NPsubject pattern, (61) – merely that the NP is likely to be an argument of the lower verb and optionally of BETTER as well. Once the BETTER construction comes on the scene, always with a subject NP of human reference – (63) and later types – the meaning can become deontic. Finally, when non-human subjects become available, the meaning can become epistemic. As will be discussed below, the availability of first deontic and then epistemic meanings does not entail that such meanings are necessarily present.

Better as a verb

365

Table 5. Changes of form and change of meaning stages (60) ~ (61), (62)

earlier stage form meaning oblique beneficiary, human NP evaluative modal

(62) ~ (63b)

toinfinitive

separate tense domain

(63a) ~ (63b)

better to have V

(62) ~ (63b)

better X than Y

(63), (64), (66) ~ (67)

animate NP subject

evaluative, implicit advice comparison, evaluation, advice deontic

later stage form meaning subject NP argument of lower verb and sometimes of BETTER too plain temporal coninfinitive currence of modal and lower verb; immediacy better V explicit advice better X

(stronger) deontic

inanimate or dummy subject

not argument of BETTER; clause may be epistemic

3.2. Immediacy and specificity Dictionary evidence suggests that infinitives with BETTER have always been predominantly plain rather than to-infinitives. However, there have certainly been examples with to. Recent scholars (Duffley 1992; Fischer 1993, 2000, among others) have argued that to-infinitives in general encode a tense domain separate from that of the higher verb, whereas plain infinitives do not. It is possible, therefore, that the now-categorial selection of plain infinitives is associated with a temporal coincidence of BETTER and the lower verb, which may help to explain the sense of immediacy of the advice that we detect in BETTER, as well as the specificity which has been observed, typically ‘a recommendation for a specific occasion’ (Westney 1995: 182); Mitchell too shows how BETTER is appropriate for advice on a particular occasion rather than in general or habitual situations (2003: 141).

366

David Denison and Alison Cort

3.3. Person We must try to account for the extraordinary predominance with BETTER of first and second person subjects, explicit and implicit (Section 2.5.2). First and second person pronouns are top of the animacy hierarchy (Silverstein 1976; Traugott and Dasher 2002: 69), but why should NPs lower down the scale appear so infrequently? Keith Mitchell claims that BETTER is used not only to impart advice favoured by a speaker but to ‘imply that the speaker is also deciding that the advice should be acted upon’ (2003: 143). In connection with the association of BETTER with specific reference, Westney mentions a (consequent?) expectation of fulfilment (1995: 182). If we put these semantic-pragmatic traits together – speaker decision, specific occasion, expectation of fulfilment – then it does not seem surprising that BETTER should be associated overwhelmingly with the actual participants in the speech situation, namely first and second person: they represent the most likely effective scope of a speaker’s advice/decision on some immediate event. (It is interesting that examples with third person subjects often do not refer to a specific occasion; see those cited in Section 3.4.2 below.) The properties discussed in this and the previous section range over semantics, pragmatics and morphosyntax. Some represent statistical tendencies, others have become categorical. Rather than trying to tease out a neat causal chain in which property X leads to Y leads to Z, it seems better to regard this as an instance of pattern-strengthening in which the whole constellation of mutually reinforcing properties gradually coheres around the constructional idiom involving BETTER. 3.4. Deontic meaning 3.4.1. Advisability BETTER is often used to express the weakly deontic function of advisability, a concept discussed by Traugott and Dasher: “It includes the sense that action sought of the subject is not only normatively wished for but is also beneficial to the subject …” (2002: 106). Although their example involves modal OUGHT, it seems fully applicable to BETTER, except that historically the order should be reversed for BETTER: the action sought is beneficial to the subject and is also wished for by the speaker. The latter starts off as a

Better as a verb

367

generalized invited inference from the former, one that is particularly salient in the frequent early use of better constructions in conditional or even counterfactual contexts: why mention that you think it better for the subject that they should do/should have done something unless you wish them to do it/wish they had done it? Over time the element of speaker’s wish becomes semanticized. 3.4.2. Third-person subjects The subjects of deontic advice expressed using BETTER are mostly first or second person, therefore personal by definition, but third person subjects are found too. A third person subject does not in itself rule out deontic meaning. To see this, consider a classic case of unequivocal deontic use, MUST with a second person subject, as in (68): (68)

a. b. c.

You must repay the debt within six months. The debt must be repaid within six months. The borrower must repay the debt within six months.

The meaning remains essentially the same, hence deontic, even if the sentence is passivized or given a human third person subject. Now BETTER, like MUST, is voice-neutral, so the same relationship holds between the a and b sentences in (69): (69)

a. b. c.

You had better repay the debt within six months. The debt had better be repaid within six months. The borrower had better repay the debt within six months.

And both (69b) and (69c) remain deontic. Let us consider a selection of inanimate third person subjects: (70)

(71) (72)

1885 O. HEAVISIDE in Electrician 4 Sept. 311/2 We pass to electric displacement, the analogue of magnetic induction (noting by the way that it had better not be called the electric induction..but be called the displacement). 1905 H. J. SPOONER Motors & Motoring 19 Grinding in valves is an operation that had better be left to the trained mechanic. He slowly imagines the character, moulds her into shape, and then – probably the last thing of all – pops a pair of glass eyes into

368

David Denison and Alison Cort

those empty sockets. Eyes? Oh yes, she’d better have eyes, he reflects, with a weary courtesy. (BNC G1A 574) In examples (70) and (71), both somewhat archaic, the writer gives advice, enjoins someone unspecified (the reader, people in general) to act in a certain way. In (71) we have free indirect speech, and the recipient of the advice is the speaker/thinker himself: this is speaker decision. All three are deontic. We are arguing, therefore, that deontic meaning is possible even if a human recipient of advice or direction is not actually expressed, so long as one can be inferred. 3.4.3. Comparison Up to late ModE, BETTER in the form had better could be used, albeit rarely, with a than-phrase of comparison: (73)

= (17) 1613 SHAKES. Hen. VIII, V. iii. 132 He had better starue Then but once thinke his place becomes thee not. (OED)

As we have seen, this possibility no longer exists. Mitchell regards this as confirmation that the meaning of BETTER now extends beyond the pure giving of advice to directing the behaviour of others or announcing decisions about one’s own. “In other words, here the deciding function overrides the comparative advantage element in the meaning …” (2003: 143). Directing others is more strongly deontic than advising them. 3.4.4. Semantics of proverbs The proverbial patterns of Section 2.5.4 involve advice-giving, like BETTER. Unlike BETTER, the proverbs offer general (generic?) advice (i.e. what is commonly/historically held to be true) and need not reflect the opinion of the speaker. They do not convey any element of decision, any real expectation that the advice will be acted upon, any sense of immediacy. They are not, or barely, deontic. The interpersonal element is low.

Better as a verb

369

3.5. Epistemic meaning Scholars routinely distinguish deontic and epistemic modality, as in (74) (75)

You must hurry up. John must be stuck in traffic.

respectively. Here is Coates’s definition of epistemic modality: It is concerned with the speaker’s assumptions or assessment of possibilities and, in most cases, it indicates the speaker’s confidence (or lack of confidence) in the truth of the proposition expressed. (1983: 18)

Westney denies flatly that BETTER can be used epistemically (1995: 181 and 183 note 12), while Mitchell asserts that it can (2003: 145–146). The disagreement is instructive. The examples Mitchell chooses are decontextualized snippets from the BNC with third person subjects. Mitchell’s alleged epistemic examples of BETTER (2003: 145, his examples 73–75) are as follows. We have expanded the context from the full BNC, apparently unavailable to Mitchell. (76)

(77)

(78)

[“Kurt here. I have urgent information. There have been serious developments. Can we meet? …” “I’ll meet you in the lobby of the Frankfurter Hof half an hour from now.] It had better be important. [” He slammed down the phone before Meyer could reply …] (BNC ARK2617) [I had hoped she would be able to get rid of Dennis quickly, but it was almost 4 o’clock before the red BMW finally appeared and roared away in the direction of the ring road. By that time I was chilled to the bone, exhausted from the relentless battering of the traffic, sullen and depressed.] This had better be good, I thought grimly as I crossed the road and walked up the cul-de-sac to the Parsonage. [This had better be bloody good.] (BNC BMR408) [“It was here. I haven’t made it up.] I swear the bodies are somewhere near here.” “They had better be,” Goreng said. [“Or there will be new ones.”] (BNC H9N288)

All three come from fiction (imaginative prose). Mitchell correctly guesses the context of (76) and asserts that the speaker expresses the hope that it is an important matter (and an alternative utterance in this context would be I hope it’s important).

370

David Denison and Alison Cort

He claims that of the three, (76) and (77) at least ‘are resistant to a deontic interpretation’ (2003: 145); (77) and (78) are not otherwise discussed. He goes on to cite Givón (1994: 280) and Palmer (2001: 134) in support of the claim that hope is an epistemic illocutionary act, an instance of epistemic volition. We have some difficulty in accepting all of these claims. Would I hope it’s important really have served as an alternative to It had better be important, and can the claimed absence of deontic meaning be sustained? Epistemic volition is a possible characterization of that part of the meaning of BETTER concerned with the truth of the proposition (‘I wish it were true / I wish it to be true’),9 but it fails to capture that part of the meaning in which the speaker not so much comments on the truth of a proposition as actually tries to influence events (if only counterfactually) by impressing a course of action on a participant or imposing an obligation. One interpretation is that all three of (76–78) incorporate simultaneously an epistemic and a deontic element. (We assume here that the kinds of modality available in English range over those discussed by Palmer 1979.) The epistemic component is certainly akin to hope, while the deontic element involves the imposition of an obligation by the speaker/thinker. The meaning of (76) is difficult to paraphrase precisely, but it does seem to be rather more than merely ‘I hope it’s important’. We detect in addition two linked meanings: that someone (here the addressee) is in some way responsible for the situation, and that that person should endeavour – or should have endeavoured – to produce a favourable outcome (plus perhaps the suggestion that they will suffer adverse consequences otherwise). With hope there are no such additional meanings. Much the same analysis holds for (78). As for (77), the emphatic repetition (not given by Mitchell) makes the writer seem more vehement: someone is responsible, probably the housewife who had previously made sexual advances, or perhaps the writer himself, or both. It is generally taken for granted in the literature that deontic and epistemic modalities are mutually exclusive, which may explain both Westney’s position (if he thinks deontic meaning is always present with BETTER) and Mitchell’s (who argues from an apparent absence of deontic meaning in certain examples). Some scholars do acknowledge the existence of modal verb examples which are either equivocal or ambiguous between these 9. However, it is not necessarily legitimate to conflate desire for an outcome with belief about a proposition; they are kept distinct in Ginzburg & Sag (2000), for instance. On the other hand, I do not accept an anonymous referee’s claim that hope concerns the desirability of a situation and so (if modal at all) must be deontic: hope has no performative element and does not convey an obligation.

Better as a verb

371

two kinds of modality – Coates (1983) is a notable case in point – but I am not aware of analyses that attribute both deontic and epistemic modality to the same sentence without either ambiguity or vagueness. I see no reason why they should not coexist. Some scholars operate with a much wider armoury of modalities: bouletic, doxastic, and so on. Again, I see no particular advantage in this case in assigning modalities to a more fine-grained classification. Better insight may be gained by invoking the concept of intersubjectivity; see Section 4. Keith Mitchell has drawn my attention to a more recent discussion of modality of his, where the following tabulation occurs: Table 6. Two kinds of modality, adapted from Mitchell (2009) DEONTIC MODALITY

EPISTEMIC MODALITY

speaker’s decisions about the occurrence of situations (potential ‘acts’)

speaker’s conclusions about the truth of propositions (potential ‘facts’)

function: directive/conative You can take my car

function: speculative/verdictive They may be at the pub

It seems to me that the BETTER clause of (76) is both directive and speculative. The speaker is doing two things at once: imposing a retrospective obligation on a speech participant, and at the same time making a judgement about the truth of the proposition. 3.6. ‘Threat’ or other additional meanings Does BETTER encode a threat of adverse consequences if the speaker’s recommendation is not followed by the second or third person addressee? The possibility has been canvassed in the literature, and the idea was apparently salient for one group of ordinary speakers whose invented examples of sentences with BETTER tended to include a threat element.10 Examples in10. These were native and non-native speakers, linguistically sophisticated, who were questioned by AC as to how verbal BETTER was used. One interpretation of this (suggested by Susan Fitzmaurice, p.c.) is that a threat context was the clearest way of signalling their intuitive sense of an intersubjective meaning (on which see Section 4).

372

David Denison and Alison Cort

volving an explicit or implicit threat do indeed occur, as in the corpus examples (79–82): (79) (80) (81)

(82)

I saved out that lunch money and they took it. And one of them push him down and tell him he better get some protection for himself. (Frown L12:28) You’d better not show your face back here, unless you want me to work on it with a razor. (LOB L10:112) “You see, Mr. Tisdal, you had better behave peaceably,” said Garrett. “There is nothing to be gained by violence. We are protected, and you in every way in our power – … and I tell you fairly, that, except with my permission, you shall not leave this room alive.” (ARCHER 1847lefa.f5b) As a matter of fact, Grandpa better fork over some dough and pretty fast or I’m going to make his little Sarah into a shiksa. (ARCHER 1964gelb.d8a)

But they are not frequent, amounting to just a handful in the smaller corpora and the BNC sample. (There is an analogy here with the deontic permission sense of MAY, which is perceived as more prototypical than epistemic uses despite its comparatively low frequency of occurrence.) More important, any sense of threat with BETTER is either encoded elsewhere in the linguistic context or is merely a pragmatic implication in a situational context where menace is to be expected; see here Westney (1995: 183–184) and Mitchell (2003: 146). Threat is not inherently part of the semantics of BETTER. Examples like (82) are reminiscent of the so-called pseudo-conditional, in which an imperative clause X is coordinated with a declarative clause Y, roughly equivalent to protasis and apodosis of a conditional sentence, and with Y usually representing an adverse consequence for the addressee: (83) (84)

Hand over the money, or I’ll shoot. = X or Y ‘If you don’t X, then Y’ Scream and you’ll be sorry. = X and Y ‘If you X, then Y’

To a certain extent the BETTER construction is equivalent to a hedged version of such an imperative clause, but it is far less grammaticalized in this role: with BETTER the consequence need not be stated, and there is no equivalent to (84) (where the imperative actually conveys ‘don’t scream’,

Better as a verb

373

the converse of its literal meaning outside the pseudo-conditional construction). If threat is not part of the semantics of BETTER, are there nevertheless additional shades of meaning in the pragmatics? If so, they must be tested against definitions of various kinds of implicature. Consider the clause it had better be important of (76). Suppose this conveys additional meanings something like the following (the first two are perhaps more convincing): (85)

a. b. c.

You should have made sure that the information is important. I hold you responsible. If the information turns out not to be important, you will suffer adverse consequences.

Are such meanings defeasible or detachable? (Huang 2007: 32–35; Levinson 1983: 114ff.) I believe that (85c) and probably (85b) are defeasible, while (85a) is not and so cannot be an implicature: (86)

a. b. c.

!It had better be important, but you were under no obligation to make sure. It had better be important, but I guess you can’t know whether it is. It had better be important, but you won’t suffer if you made an honest mistake.

The continuation shown in (86a) does not seem felicitous. It is harder to test whether the meanings of (85) are detachable – not specifically linked to the linguistic form of (76) – because it is so hard to find a good synonym for BETTER. Consider these possible paraphrases: (87)

a. b.

It would be advantageous for it to be important. I hope that it is important.

If either of (87a–b) is close to synonymy with the BETTER clause of (76), then in my judgement it carries no additional implicatures or entailments, but the additional meanings (85) are lost, from which it apparently follows that (85b–c) cannot be conversational implicatures. But not only is the test of detachability problematic (Levinson 1983: 119–120), the supposed synonymy of (87) with (76) is far from clear. It is tempting to see (85b–c) instead as conventional implicatures of BETTER, which would sit better with meanings tied to the particular idiom, but defeasibility would argue against

374

David Denison and Alison Cort

this (Levinson 1983: 128). In Traugott and Dasher’s terms (2002), these might be invited inferences, not yet generalized. In the absence from the context of a suitable ‘fall-guy’, someone to blame, the apparent implicatures or invited inferences and hence the deontic element disappear. Consider these invented pairs (cf. the genuine 21 above): (88)

a. b.

(89)

a. b.

(You’ve talked me into an expensive holiday.) I hope the weather is good. (You’ve talked me into an expensive holiday.) The weather had better be good. (The annual parade is in September.) I hope the weather is good. (The annual parade is in September.) The weather had better be good.

Examples (88a–b) are not synonymous: in (88b) but not (88a), ‘you’ will be (partly) to blame if my money is wasted. The context of (89), on the other hand, allows (89a–b) to be closer to synonymy, except perhaps that (89b) expresses a stronger hope, almost a jokey requirement (of whom? – the gods?) that the weather be good. Thus (88a) and (89a–b) are not deontic and are best seen as epistemic.11 The following, historical example may belong here too; it repeats (20) with a little more context: (90)

1712 ADDISON Spect. No. 287 (page)3 if [liberty] only spread among particular branches, there had better be none at all, since such liberty only aggravated the misfortunes of those who are deprived of it, by setting before them a disagreeable subject of comparison.

The natural interpretation is of the writer’s preference for what the state of affairs should be, a hope. Epistemic uses of BETTER come later than deontic ones (indeed example 90 is unusually early). The late development of epistemic usage is of course a classic symptom of grammaticalization among modals. A non-human 11. An anonymous referee tentatively appeals to van der Auwera and Plungian’s classification of the modality of possibility and necessity (1998): a weakening of deontic meaning in BETTER could then be seen as a shift from participantinternal to participant-external modality.

Better as a verb

375

subject is probably a necessary but certainly not a sufficient condition for epistemic meaning. 3.7. Counterfactual semantics As we saw in Section 2.6, BETTER has been used in counterfactual contexts to comment on what might have been. Here are two more examples: (91) (92)

1814 “Poor dear Fanny,” cried her cousin, “how ill you have been used by them! You had better have staid with us.” (Austen, Mansfield Park 86, cited by Jespersen 1909–1949: IV.146) 1813 It had better have happened to you, Lizzy; you would have laughed yourself out of it sooner. (Austen, P&P xxv)

Evidently if there is an advice element in (91), it cannot be direct advice as to immediate future action: at best it is implicit advice on future behaviour in similar circumstances, but the main effect is of evaluation of a past action and perhaps of implicit speaker wish. There is no advice element at all in (92). Such irrealis functions are regarded by Mitchell as positively ungrammatical in contemporary usage (2003: 143). Nonetheless they do still occur, if infrequently; there are two examples in written texts in the corpus data, (93) and (94): (93)

(94)

He declares that he has come to prevent Charity from getting into trouble, or to help her evoke a marriage proposal from Lucius, but he concludes the episode by saying in front of Lucius that Charity is a promiscuous “woman of the town” just like her mother. “I went to save her from the kind of life her mother was leading – but I’d better have left her in the kennel she came from” (203–204). (Frown G43:10) “Well, I am now. Simon beat Pippa for these.” “No wonder.” “He’d better have beaten me.” Sara swung round, and Thomas was half touched by the horror in her. (FRH 1761 BNC)

Perhaps the most natural-seeming paraphrase for this expression of unreal past in standard English is the rather lengthier ‘It would have been better if X had happened [rather than Y]’ (cf. Section 2.6), although this loses something of the speaker investment.

376

David Denison and Alison Cort

What kind of modality is this? In meaning, (91) is not quite the same as (92–94). All four express a present-time speaker comment on the past, but only (91) contains a small element of direct (albeit belated) advice from speaker to hearer and can be analysed as – just about – deontic. In (92–94), instead of advising a second person subject of BETTER, the speaker expresses a preference towards the proposition. And this looks very much like epistemic use, defined as follows by Mitchell: [t]he speaker expresses a volitional attitude (a desire, a preference) towards a proposition that is potentially true: “I would like/prefer it to be the case that p”. Hoping is a type of epistemic volition: a wish that a proposition whose truth is unknown turns out to be true. (2003: 145)

In these counterfactual cases, however, what is being expressed is a volitional attitude with a converse formulation: a wish that a proposition whose untruth is known might turn out to be true. Notice, though, that these particular epistemic uses have almost disappeared from current English.12 Note too that BETTER historically involves or derives from a past tense form, yet its gravitation away from counterfactuals and unreal conditionals towards present and future open conditionals suggests that it fits a generalization about obligation modals (apart from HAVE to) losing their productive morphological past-nonpast pairing and developing a “hypothetical present” with past form (Traugott and Dasher 2002: 108, 132–137). Traugott and Dasher take the appearance of the have-perfect after a modal (must, in their discussion) as unambiguous demonstration that the modal “was no longer understood as marked for past tense” (2002: 137). If the inference is safe, then the same should hold of BETTER from the fifteenth or sixteenth century iff it is a modal: (95) (96) (97) (98)

[c1435 Torr. Portugal 1186 Better he had to have be away.] (OED s.v. better, a., (n.), and adv. A.4b) 1537 Thersytes, Four O. Plays (1848) 69 They had better haue set me an errande at Rome. (OED s.v. better, a., (n.), and adv. A.4b) c1620 T. GOFFE Careless Shepherdess III. i, You had better have been hang’d at first, as I wo’d had you. (OED s.v. will v.1 A.8) 1621 SANDERSON Serm. I. 214 Vzza had better have ventured the falling, than the fingering of the ark. (OED s.v. falling vbl.n. 1)

12. Alternatively, as an anonymous referee argues, if what is expressed is advisability, then such examples are not epistemic at all but merely unfulfilled past time deontics.

Better as a verb

(99)

377

1684 H. MORE Answ. 214 The Remarker had better have let things alone. (OED)

However, the superfluous use of infinitive have to express or reinforce unrealized action is well known (Denison 1998: 140), and many examples follow (non-modal) verbs which certainly retain a clear past-nonpast distinction: (100)

1595 V. SAVIOLO Practise I. i. sig. **2, The Padouan replyed, that he meant to have cut of the Spaniards heade firste. (OED s.v. Paduan 2)

Given the uncertain status of BETTER as a modal, especially in the early ModE period, we are unwilling to use this test. Nevertheless, BETTER can express a hypothetical present with its past form. 3.8. Discourse markers Since Traugott and Dasher raise the question of a further semantic change from epistemic into speech act modalities, we have looked briefly at the evidence of BETTER. They suggest that “evidence in the domain of modal auxiliaries has remained elusive” (2002: 77–78), and cf. also their (2002: 115–116) uncertainties about timing of development of speech act modalities. In late ModE the combination of BETTER with a verb of speaking, thinking or motion can act as a discourse marker. Such discourse markers often serve to indicate (i) a (brief) divergence from the main topic of discussion, (ii) an outright change of topic, (iii) the introduction of a topic, (iv) the closure of a topic/discussion: (101) (102) (103)

1897 O. WILDE Let. ?7 June (1962) 601, I had better say candidly that he is not ‘a beautiful boy’... He is simply a manly simple fellow, … (OED s.v. mulierast n.) “^I think we’d better go on to the next question fr=om# (name)” (LLC 5:4 Heading) 1992 The most important example and one easily overlooked is the symbiotic relationship of the stony corals which act as hosts to microscopic algae living within their tissues. Perhaps I better explain

378

(104)

(105) (106)

David Denison and Alison Cort

that symbiosis is popularly defined as a relationship between two differing life forms for their mutual benefit. (BNC C95:2088) Since I mention native speakers’ feelings in this connection, and since I am elsewhere rather sceptical about appeals to native speakers’ feelings, I had better explain that in this case my evidence comes from the native speakers of English I have taught in practical classes on transcription over many years. (BNC K93 175) “Is there something wrong?” Grace asked. “I figure you’d better know someone’s been following us, Grace. They may be here.” (FLOB N14:31) I’d better introduce myself … (Westney 1995: 183, example 5.75)

But it is the combination of BETTER and predicate which carries the discourse function: the function of BETTER alone is almost purely one of deciding, with perhaps sufficient connotation of ‘advice to self’ to be regarded as deontic. It is very doubtful, therefore, that BETTER by itself can be considered a discourse marker. 4. Subjectivity and intersubjectivity Evaluation is inherent in the meaning both of the adjective/adverb better and of the verbal BETTER, hence an element of subjectivity is necessarily involved in their use. Many early examples of the non-verbal better involve third person NPs and express the speaker/writer’s evaluation of a situation and no more. (There is, however, no epistemic element, since the evaluation is part of the propositional content; cf. here also Westney 1995: 183 n.12.) Although pronominal NPs have always been predominant, it is not clear whether first and second person had such an overwhelming preponderance as they now do: we have not undertaken a corpus count of earlier material, and in any event, first and second person pronouns might well be under-represented in the text types available. Counterfactual uses, which were once common, are perhaps somewhat more strongly subjective, since they involve both the projection of an imaginary outcome by the speaker/writer and its evaluation, while proverbial uses are weaker. In Section 3.6 we discussed a group of examples which seemed to convey at one and the same time both deontic and epistemic semantics. An alternative account would bring in intersubjectivity. We repeat (88) and (89) below:

Better as a verb

(88)

a. b.

(89)

a. b.

379

(You’ve talked me into an expensive holiday.) I hope the weather is good. (You’ve talked me into an expensive holiday.) The weather had better be good. (The annual parade is in September.) I hope the weather is good. (The annual parade is in September.) The weather had better be good.

Where (88a) and (89a–b) are epistemic and subjective, (88b) is intersubjective and epistemic/deontic: the speaker/writer projects the hope onto the addressee as well (‘you too should hope for a favourable outcome’). It looks as if a cluster of properties is lost at much the same time in late ModE: complementation by a to-infinitive, inclusion of a comparison, and the possibility of counterfactual meaning (see Sections 2.3, 2.4 and 2.6 above). The meaning of advisability comes to the fore and with it an intersubjective element: the speaker/writer doesn’t just express an opinion but attempts to co-opt the hearer/reader into accepting the advice. (Note that the ‘hearer’ and speaker may be the same individual, since first person use is semantically reflexive). The greater the degree of speaker decision, the stronger the interpersonal element. Likewise, any invited inferences involving a second person subject, whether of being held responsible or even of being at risk of adverse consequences for non-compliance, suggest a great degree of intersubjectivity. The preceding paragraphs rather suggest a steady diachronic progression from subjectivity to intersubjectivity, thus in Traugott’s terms (this volume), a process of intersubjectification. However, there is another line of development which need not involve interpersonal use and therefore need not contribute to intersubjectification. This is the diachronic development of modality from deontic to epistemic, especially in connection with inanimate and dummy NP subjects. After all, a natural interpretation of the syntax-semantics interface is to assume that deontic modality requires a twoplace verb, where the person on whom the obligation falls is one argument (and a proposition the other). Intersubjectivity arises because of the involvement of the speaker not as a syntactic argument but pragmatically. An epistemic modal is a one-place verb, taking only a proposition as argument and involving the speaker’s perspective pragmatically: the perspective of the hearer does not have to be addressed by the speaker, hence intersubjectivity is low or absent altogether.

380

David Denison and Alison Cort

5. Afterword In the course of a history that lasts at least 1200 years, the BETTER constructions have exhibited various attributes of grammaticalization, and in particular a growth of interpersonal usage that is reflected in their semantics and in their selective preference for first and second person subjects. There is a typical chronological priority of deontic before epistemic use with inanimate or dummy subjects, just as if BETTER was (or was becoming) a full-blown modal.13 We have not detected, however, an inevitable and unidirectional change towards pure modalhood. For one thing, BETTER is probably older than the morphosyntactic category Modal. For another, the category is a moving target – but that is another story. Secondary sources We quote Cort (2006) on the sources used: The corpora used were: the British National Corpus (BNC), A Representative Corpus of Historical English Registers (ARCHER),14 the London–Lund Corpus (LLC), the 1961 … Brown Corpus and the LOB Corpus …, and their early 1990s Freiburg counterparts FROWN and FLOB. This selection of corpora encompasses a range of Englishes: spoken and written, British and American, historical and contemporary.

13. On this question the formal property of being a ‘NICE’ verb might be thought relevant, as in internet data such as the following: (i) its guna b so fun we better b in tha same team bettnt we! (ii) lol well you better start staying in then betternt you!!!!!! lol Example (i) comes from Google’s cache of as retrieved on 10 June 2007 19:34:51 GMT, search dated 13 August 2007, while (ii) is from , sampled 13 August 2007. But bettern’t forms have never been particularly frequent, and even an apparently central modal verb like MAY has lost the form mayn’t. 14. The examples from ARCHER and the word counts were gathered from an early version of the corpus by AC; citations have been referred by DD to the filenames used in the current release of ARCHER, version 3.1.

Better as a verb

381

Table 7. Corpora cited corpus ARCHER BNC BROWN FLOB FROWN LLC LOB

AmE

BrE

spoken

9 9 9 – 9 – –

9 9 – 9 – 9 9

9 9 – – – 9 –

written 9 9 9 9 9 – 9

timespan

words (millions)

1650–1991 1960–1993 1961 1991 1992 1959–1975 1961

1.7 100 1 1 1 0.5 1

A number of citations come from OED, and other sources for occasional citations are indicated after the quotation concerned. References Coates, Jennifer 1983 The Semantics of the Modal Auxiliaries. London: Croom Helm. Cort, Alison 2006 Recent and current change in the modal verb. Submitted (but never corrected and revised) PhD dissertation, University of Manchester. Denison, David 1998 Syntax. In The Cambridge history of the English language, vol. 4, 1776–1997, Suzanne Romaine (ed.), 92–329. Cambridge: Cambridge University Press. Duffley, Patrick J. 1992 The English Infinitive. London and New York: Longman. Dykes, Oswald 1713 English Proverbs, with Moral Reflexions; in imitation of Sir Roger L’Estrange’s Æsop. ... The third edition. To which is added, the Union-proverb, ... 3rd ed. London: G. Sawbridge. Fischer, Olga 1993 The distinction between to and bare infinitival complements in late Middle English, the expression of transitivity and changes in the English system of case. Ms. Fischer, Olga 2000 Grammaticalisation: Unidirectional, non-reversable? The case of to before the infinitive in English. In Pathways of Change: Grammati-

382

David Denison and Alison Cort

calization in English, Olga Fischer, Anette Rosenbach and Dieter Stein (eds.), 149–169. Amsterdam/Philadelphia: Benjamins. Ginzburg, Jonathan and Ivan A. Sag 2000 Interrogative Investigations: The Form, Meaning, and Use of English Interrogatives. Stanford: CSLI Publications. Givón, Talmy 1994 Irrealis and the subjunctive. Studies in Language 18: 265–337. Hall, Fitzedward 1881 On the origin of “had rather go” and analogous or apparently analogous locutions. American Journal of Philology 2: 281–322. Huang, Yan 2007 Pragmatics. Oxford: Oxford University Press. Huddleston, Rodney 1980 Criteria for auxiliaries and modals. In Studies in English Linguistics: For Randolph Quirk, Sidney Greenbaum, Geoffrey Leech and Jan Svartvik (eds.), 65–78. London and New York: Longman. Huddleston, Rodney and Geoffrey K. Pullum 2002 The Cambridge Grammar of the English Language. Cambridge: Cambridge University Press. Jespersen, Otto 1909–49 A Modern English Grammar on Historical Principles, 7 vols. Heidelberg: Carl Winters Universitätsbuchhandlung. Repr. in George Allen and Unwin, London, 1961. Kelly, James 1721 A Complete Collection of Scotish Proverbs explained and made intelligible to the English reader. London: W. and J. Innys and J. Osborn. Levinson, Stephen C. 1983 Pragmatics. Cambridge: Cambridge University Press. Ludovici, C[hristian] (ed.) 1791 A Dictionary English, German and French. Containing not only the English words ... but also their proper accent, phrases, figurative speeches, ... 4th edn. Leipzig: John Frederick Gleditsch. MacIntosh, Donald 1785 A Collection of Gaelic Proverbs, and Familiar Phrases; accompanied with an English translation, intended to facilitate the study of the language; ... Edinburgh: Donaldson, Creech, Elliot, and Sibbald. Mitchell, Keith 2003 Had better and might as well: On the margins of modality? In Modality in Contemporary English, Roberta Facchinetti, Manfred Krug and Frank Palmer (eds.), 129–149. Berlin/New York: Mouton de Gruyter.

Better as a verb

383

Mitchell, Keith 2009 Semantic ascent, deixis, intersubjectivity and modality. In Modality in English: Theory and Description, Raphael Salkie, Pierre Busuttil, and Johan van der Auwera (eds.), 55–78. Berlin/New York: Mouton de Gruyter. Palmer, Frank R. 1979 Modality and the English Modals. 1st ed. London: Longman. Palmer, Frank R. 2001 Mood and Modality. 2nd ed. Cambridge: Cambridge University Press. Pinker, Steven and Ray Jackendoff 2004 The faculty of language: What’s special about it? Cognition 95: 201– 236. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech and Jan Svartvik 1985 A Comprehensive Grammar of the English Language. London and New York: Longman. Ray, John 1742 A Compleat Collection of English Proverbs; also the most celebrated proverbs of the Scotch, Italian, French, Spanish, and other languages ... 3rd ed. London: H. Slater, F. Noble, W. and T. Payne, T. Wright, and J. Duncan. Silverstein, Michael 1976 Hierarchy of features and ergativity. In Grammatical Categories in Australian Languages, Robert M. W. Dixon (ed.), 112–171. NJ: Humanities Press. Speake, Jennifer (ed.) 2003 The Oxford Dictionary of Proverbs. Oxford University Press. Traugott, Elizabeth Closs and Richard B. Dasher 2002 Regularity in Semantic Change. Cambridge: Cambridge University Press. van der Auwera, Johan and Vladimir A. Plungian 1998 Modality’s semantic map. Linguistic Typology 2: 79–124. van der Gaaf, W[illem] 1904 The Transition from the Impersonal to the Personal Construction in Middle English. Heidelberg: Carl Winter’s Universitätsbuchhandlung. Visser, Frederic Th. 1963–73 An Historical Syntax of the English Language. 4 vols. Leiden: E. J. Brill. Westney, Paul 1995 Modals and Periphrastics in English: An Investigation into the Semantic Correspondence between Certain English Modal Verbs and their Periphrastic Equivalents. Tübingen: Niemeyer.

(Inter)subjectification in the domain of modality and mood – Concepts and cross-linguistic realities Heiko Narrog Abstract This paper consists of three parts. The first (Sections 1 and 2) reviews the concepts of subjectification and (inter)subjectification as espoused by Traugott and Langacker and their application to modality. In the second part (Section 3), the author presents his own take on the issue. Here, the most common cross-linguistic changes in modal meaning are discussed with respect to (inter)subjectification and are illustrated with examples from English, German and Japanese language history. In the third part (Section 4), it is argued that (inter)subjectification and meaning change in general do not take place in a social and cultural vacuum, and that therefore (inter)subjectification cannot be a purely conceptual issue. In particular, the question is addressed why ‘strong obligation’ – ‘certainty’ (e.g. English must) polyfunctionality is cross-linguistically much less common than one might expect on the basis of data from the Germanic languages. Cross-linguistic data are presented that show the distribution of obligation markers, their type, and their etymologies, which vary widely across languages. It is suggested that grammaticalization and semantic change of obligation markers may depend on how speakers in a specific language/culture pragmatically deal with the expression of obligation.

1. Preliminaries While research on subjectification is diverse, there are currently two clearly dominant conceptions of subjectification in linguistics, one by Langacker (1990, 1998, 1999), and another one by Traugott (e.g. 1989, 1995, 2003). Langacker initially characterized subjectification as “the realignment of some relationship from the objective to the subjective axis” (Langacker 1990: 17), but later re-defined it as “a gradual process of progressive attenuation,” in which “an objective relationship fades away, leaving behind a subjective relationship that was originally immanent in it” (Langacker 1998: 75f). Langacker’s definition aims at being maximally precise within a specific model of grammatical description, viz. Cognitive Grammar. The

386

Heiko Narrog

focus is on the event and the role of event participants, and the comparison of related constructions in synchrony constitutes the material basis for analysis. Traugott’s approach, in contrast, is based on pragmatics and the historical analysis of semantic change in context. According to her, subjectification is “the mechanism whereby meanings come over time to encode or externalize SP[speaker]/W[writer]’s perspectives and attitudes as constrained by the communicative world of the speech event, rather than by socalled ‘real-world’ characteristics of the event or situation referred to” (Traugott 2003: 126). Put more simply, subjectification involves “meanings [being] recruited by the speaker to encode and regulate attitudes and beliefs” (Traugott this volume: 35). Importantly, a dimension of intersubjectification has been added, i.e. the process whereby “meanings come over time to encode or externalise implicatures regarding SP[speaker]/ W[writer]’s attention to the ‘self’ of AD[addressee]/R[reader] in both an epistemic and a social sense” (Traugott 2003: 130). Intersubjectification presupposes subjectification. In two more recent proposals, I have put forward a model that combines subjectification and intersubjectification in the dimension of speakerorientation (Narrog 2005b, 2007). De Smet and Verstraete (2006) have also contributed to Traugott’s line of thought about subjectivity and subjectification, by distinguishing between ‘ideational’ subjectivity and ‘interpersonal’ subjectivity and making a new distinction between pragmatic and semantic subjectivity. Subjectification and intersubjectification are semantic phenomena grounded in pragmatics and discourse. They are not necessarily accompanied by structural change (cf. Traugott and Dasher 2002: 87–90), but it is plausible that particularly subjectification, in relation to the motivating factor of expressivity (cf. Hopper and Traugott 2003: 72–74), can play a role in structural change. According to Traugott, this is especially true in the case of the initial stages of grammaticalization (Traugott 1995: 47, this volume). Assuming semantic change may precede and perhaps even motivate structural change (cf. Hopper and Traugott 2003: 100; Traugott and Dasher 2002: 283), it seems reasonable to expect there to be cases in which subjectification is reflected in the structural properties of grammaticalization (cf. Heine and Reh 1984: ch. 1 for a list of properties). I have pointed out (Narrog 2005b), however, that while increased speaker-orientation may be reflected in the grammaticalization of an individual expression, the more grammaticalized item in a specific semantic field is not necessarily more subjectivized. A good example of grammaticalization accompanied by gra-

(Inter)subjectification in the domain of modality and mood

387

dual subjectification and intersubjectification is the Modern Japanese addressee honorific -mas-u (Narrog 1997; Traugott and Dasher 2002: 241). It started out as a morphologically complex main verb *mawi-ir-as-uru ‘let come (humilitive)’ in Old Japanese, came to be used as a suffix verb indicating humility without lexical content, and finally became the modern honorific addressee marker. Semantically, it first underwent subjectification in the sense of Traugott, indicating the speaker’s evaluation of his or her social standing in comparison to some other person, and later intersubjectification, expressing politeness towards the addressee. This was accompanied by morphological and phonological changes, namely, loss of morphosyntactic independence, and morphological and phonological condensation (*mawi-ir-as-uru > mawir-as-uru > mair-as-uru > maras-uru > mas-uru > mas-u). The focus of this paper is more on semantic change than on the structural changes that may or may not accompany it. The topic is historical change in the area of modality and mood, and the question is addressed what type of change can be observed cross-linguistically, especially against the background of existing hypotheses on (inter)subjectification and speakerorientation. Section 2 gives an overview of Traugott’s and Langacker’s proposals concerning semantic change and grammaticalization in the area of modality, and Section 3 sets out my own model, backed up by some cross-linguistic data. Section 4 broaches a specific topic in some detail, namely the question of why deontic-epistemic polysemy of modal markers is cross-linguistically not as common as the situation in many IndoEuropean languages might suggest. 2. Subjectification of modal markers – Traugott and Langacker Traugott and Langacker, the leading proponents of subjectification as a mechanism in meaning change, have treated modal verbs as prominent examples of subjectification and grammaticalization. Their research counts as a benchmark in this area, and any discussion of the topic necessarily has to start out with a critical assessment of their hypotheses. In this section, their accounts of the subjectification of modal markers, particularly the English modals, are briefly reviewed, first Langacker’s and then Traugott’s. Langacker’s (1990, 1991, 1998, 1999, 2003) proposals on the subjectification of modal verbs make crucial reference to the force-dynamic model of modal meanings espoused by Talmy (1988), Sweetser (1982, 1990) and

388

Heiko Narrog

Johnson (1987), which he integrates into his own cognitive theory. The development of the English modals according to Langacker can be summarized as follows.1 The modals in English historically derive from main verbs with a purely objective meaning,2 but with force dynamic notions that designate some kind of potency.3 The action described by the verb complement remains potential rather than actual (Langacker 1990: 25, 1998: 82, 1999: 163). Their crucial development starts from here. First, there is a shift from physical to social force when verbs originally denoting some potency of the subject come to have deontic meanings (obligation, permission). Second, “the source of potency is no longer identified with the subject, but is implicit and subjectively construed” (Langacker 1999: 163). It can be the speaker or some individual from the context, but also some general authority. That is, the source of potency becomes diffuse. At the same time the target of the potency (the agent) also becomes diffuse. Modals with an epistemic meaning, as a further development from root modals, are maximally diffuse with respect to the source and the target of potency. Furthermore the potency itself is reconceptualized as the “evolutionary momentum of reality itself, as assessed by the speaker” (Langacker 1998: 85, 1999: 164). According to Langacker, this is a stage which has also been reached by German modal verbs (Langacker 1990: 27), but there is a third stage of development reached only by the English modals in their current status as grounding predications. At this stage, the conception of a directed potency loses its profiled status and the modal only designates a process deemed necessary (or possible). Root and epistemic meanings coexist at both the second and the third stage (Langacker 1990: 27f, 1998: 83f). Furthermore, “extreme subjectification” takes place when the epistemic sense comes to be used with a present rather than a future orientation. In the present-oriented use, the force-dynamic value of the modals does not pertain anymore to the evolution of reality per se, but instead to the evolution 1. Note that in his writings of 1990 and 1991, Langacker refers to three stages of development while in his later writings he only compares the original main verb use and the modern use and states the developments that presumably took place in between. 2. “To put the evolution of English modals into proper perspective, let us start at the period (Old English and beyond) at which they were main verbs whose essential content was purely objective” (Langacker 1990: 25). 3. Langacker (1990: 333) defines ‘potency’ as follows: “a physical or mental force that, once unleashed, tends to cause its [= the process’s] occurrence.”

(Inter)subjectification in the domain of modality and mood

389

of the knowledge of reality (Langacker 1998: 86, 2003). Langacker does not explicitly ascribe extreme subjectification either to the second or the third stage, but since the German modal verbs also have these presentoriented epistemic uses, we must assume that it belongs to both stages. Overall, for Langacker, the development of the English modal verbs is an example of the attenuation of subject control and the development of grounding predications. The changes posited by Langacker can be summarized as follows: Table 1. Subjectification in the history of the English modals according to Langacker Stage

Semantic change

I > II

domain of force shifts from physical to social diffusion of source and target of potency maximal diffusion of source and target of potency potency comes to pertain to the evolution of the [speaker’s] knowledge of reality conception of directed potency loses profiled status

I > II II II, III II > III

Corresponding change in the history of English modals change from main verb to modal verb change from main verb to modal verb change from deontic to epistemic meaning development of presentoriented epistemic meanings modals become grounding predications

Langacker’s concept is novel in relating facts in the development of English modals to a cognitive theory of language description. The suggestion that the attenuation of the role of the grammatical subject is an important aspect of the semantic development of English modal verbs is intuitively plausible, even if he does not support it with historical data. In the past ten or so years, a number of scholars that associate themselves with the framework of Cognitive Grammar have, to various degrees, taken up Langacker’s ideas and have suggested several modifications. Three of these modifications will be mentioned here. First, a point on which many authors agree is that the status of grounding predication should not be as tightly associated with linguistic form (particularly presence or absence of tense inflection) as Langacker suggests. Modal verbs used epistemically in German, Dutch and Spanish, for example, seem to share their conceptual structure with English epistemic modal verbs but not their morphosyntactic properties (cf. Cornillie 2005, 2006; Mortelmans 2002; Nuyts 2002). Nuyts (2002) even goes so far as to propose that the status of groun-

390

Heiko Narrog

ding predications should be considered as something purely conceptual rather than structural. Second, it is suggested that the differences between deontic and epistemic uses of modal verbs are larger both conceptually and grammatically than Langacker posits in his model (e.g. Pelyvás 2001a, 2006; Cornillie 2005). Thirdly, Pelyvás (2001b, 2006) suggests that cognitive predicates have a conceptually similar or equal status to modal verbs, which is in consonance with Nuyts’s position. The most detailed account of change in modal meaning in terms of Traugott’s concept of subjectification can be found in Traugott and Dasher (2002: ch. 4), where the authors analyse the history of must and ought to (see earlier work on the topic in Traugott 1989; Nordlinger and Traugott 1997. Traugott’s account of must is strongly influenced by Goossens 2000, and others). Unlike Langacker, who mainly extrapolates an idealized sequence of diachronic changes from observations on synchronic structure, Traugott’s approach is historically informed. Traugott also posits three stages, which are quite different from those of Langacker, as they point to concrete historical semantic shifts. At stage I, must has the meaning of ability and permission, at stage II (starting with late OE) it gains a deontic meaning of obligation, and at stage III (mid-Middle English), it develops epistemic meaning. Ought to denotes possession at stage I, that is, a clearly lexical meaning. At stage II (late OE), it develops deontic meaning, and at stage III (early Modern English) epistemic meaning. Traugott and Dasher (2002) note the following subjectification processes: concerning must, at stage II they observe the emergence of more subjective deontic uses, when “no social, religious, or other normative force is specified or implied” (Traugott and Dasher 2002: 125). That is, similar to Langacker, it is the nature of the modal force that is taken as a measure for subjectification at this stage. Furthermore, the deontic meaning is seen as subjective if it is the speaker who obliges someone, in which case the deontic modal “has been preempted to the subjective self of SP/W” (Traugott and Dasher 2002: 125). This second indicator of subjectification seems to be lacking in Langacker’s theory and constitutes a difference between Langacker and Traugott.4 Traugott and Dasher also observe subjectification at stage III, when examples of the epistemic sense start not to “invoke general opinion …, nor are they contextualized in logical reasoning …, but simply

4. However, one of the anonymous reviewers has pointed out to me that this notion of subjectification might be included in Langacker’s concept of subjectification as grounding.

(Inter)subjectification in the domain of modality and mood

391

express the speaker’s confidence in the inference being made at the moment” (Traugott and Dasher 2002: 129). With respect to ought to, partially parallel subjectification processes can be observed: initial uses expressing obligation at stage II are weakly subjective “in that the speaker is merely reporting an obligation which is imposed by a conventionalized, accepted set of rules and expectations (such as the Church or the Law)” (Traugott and Dasher 2002: 141). Later on in this stage, examples are found “that suggest that the speaker is identifying with these values, and is presenting him- or herself rather than society as the force behind the obligation” (Traugott and Dasher 2002: 141). The epistemic sense, however, is apparently not as far subjectified as in the case of must, as “more strongly subjective examples grounded in the speaker’s personal expectations rather than those of a wider group are rare” (Traugott and Dasher 2002: 142). The development of epistemic meaning, then, is conceived as a, or even the, typical instance of subjectification in modal meaning, since the development deontic > epistemic, unlike the meaning shifts detailed above, has often been cited as part of the tendency to recruit meanings into increasingly subjective semantics (Traugott and Dasher 2002: 149, 281; Hopper and Traugott [1993] 2003: 92). According to Traugott (1989: 37), “by Tendency III [Meanings becoming increasingly based in the speaker’s subjective belief state/attitude toward the proposition, H.N.] they [the modals, H.N.] acquired epistemic meanings that are primarily focused on the internal world of the speaker’s belief and knowledge states.” In summary, Traugott also sees the type of force and the source of the force expressed by the modal as a factor in semantic change in modals. However, as she states, “there are reasons to doubt that the forces and barriers metaphor is a key to the semantic development of either deontics or epistemics. … [T]he lexical sources of the modals show little evidence of the semantics of forces and barriers.” (Traugott and Dasher 2002: 111). Although not mentioned specifically in relation to subjectification, contexts for change from deontic to epistemic in her view include impersonal constructions and “wide scope generalized deontic necessity” (Traugott and Dasher 2002: 127, 130). These factors clearly imply the attenuation of the role of the grammatical subject, and also a diffusion of the source and target of the modal force. Here we find apparent similarities with Langacker. However, in contrast with Langacker, she includes pragmatic factors concerning the internal development of deontic and epistemic modalities that show how far the SP/W has preempted a modal expression to the subjective

392

Heiko Narrog

self. As noted above, it is rather these pragmatic factors which are central to her concept of subjectification while the attenuation of the role of the subject is merely a factor in the concrete case of the subjectification of the modals. The approach in Traugott’s theory of subjectification sets itself apart from Langacker’s theory by the following three points. First, it is solidly based on historical facts and language data. Second, it refrains from equating the degree of subjectification with structural properties such as person – number marking. Third, it takes into account pragmatic factors of the context. In my view, an empirical foundation is indispensable for making a theory valid. However, it has also been suggested that Langacker’s purely conceptual perspective on subjectification and Traugott’s empirically based approach complement each other (cf. Evans and Green 2006: 733). 3. Proposed approach (Narrog 2005a, 2005b, 2007) Building on Traugott’s concept, I (Narrog 2005b, 2007) have claimed that an adjustment in perspective is called for in order to account for a wider range of data in the area of modality and mood. The study of modality is a vast field, with many competing views being held. The basic position assumed here is that of Narrog (2005a, 2005b), which considers modality in terms of lack of factivity, or, from a different perspective, as the relativization of the validity of a proposition with respect to a certain background. It thus follows scholars such as Kiefer (1987, 1997), Dietrich (1992), and, only partially, Lyons (1977) and Palmer (2001). It is in distinct contrast to definitions of modality which are characterized in terms of speakers’ attitudes or subjectivity. In this view, subjectivity and speakers’ attitudes are expressed pervasively across choices of lexical items and grammatical categories in natural discourse, and cannot be limited to a specific grammatical category.5 It is not reasonable to define

5. Rumsey (2003: 184; in the context of denying attributing subjectivity chiefly to one category, namely person, as Benveniste did): “[It] is a mistake to see [subjectivity] as grounded in any one area of grammar, be it the category of person, modality, spatial deixis or whatever. It does seem to be the case, however, that there is a certain set of functionally interrelated grammatical categories which play a leading role in this respect. These are the ones which are indexically grounded in the speech situation, presupposing the roles of speaker

(Inter)subjectification in the domain of modality and mood

393

(that is delimit in contrast to others) a grammatical category in terms of speakers’ attitudes and subjectivity, because the means of expression of the speakers’ attitudes are far too varied to be subsumed under one category label. 3.1. The model In Narrog (2005b), a model of modality was proposed along two dimensions, the first of which is volitivity, which distinguishes volitive from nonvolitive modality. This distinction is based on the insight by Jespersen ([1924] 1992: 313ff), further developed in Heine (1995a), claiming an “element of will” as the most fundamental distinguishing element between different kinds of mood. Consider, for example, sentence (1): (1)

They must be afraid of us.

The modal must in this sentence can basically be interpreted in two ways. Either it denotes the imposition of the speaker’s volition6 (‘I wish/ demand/order that the state-of-affairs [they are afraid of us] is brought about’) (= deontic interpretation) or it denotes an assessment of a present state-of-affairs (‘I conclude that they are afraid of us’) (= epistemic interpretation). In this way, the basic idea is that in deontic uses of modals an element of will is present while it is absent in the case of epistemic interpretation. Modalities where (at least typically) an element of will is present are labeled ‘volitive’, while those where it is absent are labeled ‘nonvolitive’. If this distinction is extended beyond deontic and epistemic, boulomaic modality (wish, want) will be classified as ‘volitive’, and evidentiality and dynamic modality (can, be able to) as ‘non-volitive’. The status of boulomaic modality and evidentiality with respect to this categorization should be entirely uncontroversial. Dynamic modality, denoting physical and situational possibility, is also independent of the subject’s or the speaker’s volition. Consider the following example: and addressee, and shifting in their reference and predicational value according to who is doing the talking and who is the addressee.” 6. In context, it could also be someone else’s volition, if, for example, the speaker relates They must be afraid of us as the will or command of a third person higher authority.

394 (2)

Heiko Narrog

Todd can speak some Chinese (but only if he’s drunk).

Can denotes some ability of the subject in this sentence. If the clause in brackets is added, it acquires the meaning of situational possibility. In any case, the ability to speak Chinese is independent of the subject’s or the speaker’s volition. In the case of acquired/learnt ability, the process of acquisition itself certainly involves volitional activity, but once the ability is acquired, it is like innate ability (Birds can fly) or situational possibility, a disposition beyond volition. However, if can is used as a permission (You can leave now), we deal with deontic possibility, and therefore a ‘volitive’ category. The second dimension is a scale of increasing speaker-orientation. The term speaker-orientation refers to the speaker her- or himself and the speech situation, including the hearer, and thus subsumes both subjectivity and intersubjectivity in Traugott’s sense. “[A]s speaker-orientation means orientation towards the speaker and the speech situation, it potentially also includes an increasing orientation towards the addressee as part of the speech situation” (Narrog 2005b: 692). Now, crucially, it is claimed that diachronically, modal meanings always shift in the direction of increased speaker-orientation. The increase in speaker-orientation is claimed to be essentially independent of the dimension of volitivity. The model proposed here is represented in Figure 1: illocutionary force modulation speaker-oriented mood

modality

event-oriented volitive

Figure 1. A semantic map of modality and mood

non-volitive

(Inter)subjectification in the domain of modality and mood

395

The vertical dimension is from event-oriented (down) to speaker-oriented (up). It defines the principal direction of semantic change (indicated by the arrows), namely towards more speaker-orientation, and in grammatical terms, from modality to mood and illocutionary force. The horizontal direction has a volitive and a non-volitive side. That is, in principle, change within the range of volitive (e.g. deontic) meanings and non-volitive (e.g. epistemic) meanings is equally possible as change from volitive to nonvolitive and non-volitive to volitive, as long as these changes involve an increase (or at least no decrease) in speaker-orientation. Although change across the horizontal dimension from deontic (volitive) to epistemic (nonvolitive) occurs, this type of change is not defining for the overall directionality of change. Accordingly, it is suggested that the prevailing focus on the shift from deontic to epistemic meaning as the representative change in modal meaning might be misplaced (cf. Narrog 2005b: 696–699). This focus is based on salient historical developments in English (and a number of related languages) but may not be justified in a broader cross-linguistic perspective. Also, unlike the model of van der Auwera and Plungian (1998), it is not assumed that participant-external modality is in general more advanced than participant-internal modality, nor that epistemic modality is in general more advanced than deontic modality. Furthermore, future/prediction meanings are viewed as part of the area of modality and mood. One characteristic of this model is that it attempts to integrate mood and illocutionary force in the same semantic map as modality proper. ‘Mood’ is understood here not as a morphological category (although prototypically it coincides with formal expression through mood markers) but as a semantic category comprising sentence-final clausal moods such as imperatives and hortatives, which carry illocutionary force, i. e. Bybee et al.’s (1994) speaker-oriented modality, and subordinating moods, which indicate the position of a clause in discourse. This category certainly needs further differentiation but the internal structure is intentionally left open, as we are solely concerned with its function as the target of semantic change out of modality proper, and it is further assumed that we do not know enough about this area yet. The same holds for illocutionary force modulation, a category which can be exemplified by the final particles of Japanese or the modal particles of German. This is in my view also a “target” category for markers of modality proper in the case of semantic change. As modal categories do not have to go through a phase of ‘mood’ marking in order to

396

Heiko Narrog

become illocutionary force modulators, this category is depicted as horizontally encompassing mood. Evidence that supports such a model is adduced from diachronic data from English and Japanese, and, notably, from the cross-linguistic study on modality grams by Bybee et al. (1994). The data from Bybee and research associates show that the shift from deontic to epistemic meaning does exist outside of the well-known Indo-European languages as well, but is mainly confined to the Eurasia language area.7 Overall, other changes are more frequent. They include change within the volitive area (from obligation to imperative), within the non-volitive area (from ability/root possibility to epistemic possibility), and from non-volitive to volitive (e.g. from future/prediction to imperative and from root possibility to permission). Table 2 below shows the eight most frequent changes, and the number of “grams” (markers) and languages in which they occurred.8 Table 2. Changes of modal meanings in Bybee et al. (1994, ch. 6) source meaning

new meaning

grams (languages)

Bybee et al. (1994)

1 future/prediction

imperative

13

p. 210–211

2 root possibility

permission9

9

p. 193

3 root/epistemic possibility

admonitive

5

p. 211

4 obligation

imperative

4

p. 211

5 ability/root possibility

epistemic possibility

4

p. 195, 204

6 strong obligation

certainty

3

p. 195, 203

7 weak obligation

probability

2

p. 195, 203

8 prediction/future

probability

2

p. 202

7. In a related observation, van der Auwera et al. (2005: 247, 256) suggest that modal polyfunctionality (i.e. the same modal having situational and epistemic meanings) is typical for “Standard Average European” as opposed to other language areas. 8. In the cases discussed here, one gram per language is concerned. 9. Bybee and associates suggest that circumstantial (root) possibility meaning might entail permission (Bybee et al. 1994: 193).

(Inter)subjectification in the domain of modality and mood

397

In the remainder of this section, I will elaborate on the types of diachronic change that usually receive less attention than those from deontic to epistemic, but are by no means less important, specifically the change from non-volitive (incl. epistemic) to volitive (incl. deontic) and changes within the confines of volitive (particularly deontic) and non-volitive (particularly epistemic) meanings. I will particularly focus here on the relationship between these changes of modal meanings and speaker-orientation (i.e. (inter)subjectification). Also, I will put some emphasis on the change out of modality proper into the area of mood, which is an integral part of my model but has received relatively little attention in research focused on English, where this kind of change is not historically documented. 3.2. From non-volitive to volitive, including from epistemic to deontic The cross-linguistically most common changes in Table 2 above, that is changes #1 through #3, go in a rather unexpected direction, namely from non-volitive to volitive, including from epistemic to deontic (“speakeroriented” in Bybee et al.’s 1994 terminology). Change #2 from root possibility to permission is well-known in English and presumably uncontroversial. I will therefore concentrate on changes #1 and #3. However, the discussion will be kept short here since the issue has already been discussed in more detail in Narrog (2005b: Section 3.5). The acquisition of imperative functions by future/prediction markers was documented in as many as 13 out of the 76 languages of Bybee et al.’s (1994: 273) sample. Bybee et al. (1994: 211) explain this change as follows: “In a situation in which the speaker has authority over the addressee, a prediction about the addressee can be interpreted as a command”. While English future markers have arguably not semanticized an imperative function, an imperative-like reading of will is available in context, and has frequently been mentioned in the literature. Visser (1969: 1695) gives the following modern example: (3)

You will tell your father, Sam…I shall be most willing and ready to lend him any aid in my power (Dickens, The Pickwick Papers, 1837).

Visser notes the first related examples of use in the 16th century. He comments, that “[i]n spite of the variety of connotations inherent in the you will

398

Heiko Narrog

request, it is always understood that the persons addressed will perform the action denoted by the main verb. The relation to the future is consequently still there” (Visser 1969: 1696). In other words, a command with will pragmatically presupposes that the addressee is at the same time the performer of the action. This use has not emancipated itself from the future sense because it can still be derived from it in context. The appropriation of root or epistemic possibility markers for an admonition (change #3 in Table 2) has no apparent counterpart in English, German, or Japanese, but it is basically parallel to the one of future/prediction marking for imperatives. The future/possibility marking of the indefinite future in Chepang which is used as a prohibitive (Bybee et al. 1994: 212; Caughley 1982: 102) is a good example. The starting point is an unpleasant inference such as ‘You may be sick’. Note that the subject has no control over the state expressed by the adjective. According to Caughley, such predictions were gradually extended to utterances where the addressee does have control over the situation as ‘You may fall from that rock’ (= ‘Don’t fall from that rock!’), and finally ‘Don’t you two ever go out’ (Caughley 1982: 102). Note that with the last example, a reading ‘You may go out sometime’ makes no sense. Thus, the new prohibition sense has become semanticized.10 Changes #1 and #3 are clear cases of increased speaker-orientation. While the future/prediction or possibility senses still describe the outside world (i.e. future-projected reality), the imperative and admonitive/prohibition senses directly mark speech acts. With respect to the model proposed in 3.1, they are located in the area of mood, at the same time frequently carrying illocutionary force. They are thus clearly anchored in the speech situation and the interaction between speaker and hearer, more so than expressions of possibility or predictions. In a theory that posits a directionality of change from deontic to epistemic, the change from epistemic to deontic could be considered as counter-directional. In the model of change towards more speaker-orientation proposed here, it is not. Similar to the well-known changes from deontic to epistemic, it is a change towards more speaker-orientation. Changes #1 and #3 lead from the area of modality to the more grammaticalized and more subjectified area of mood. In contrast, a change from probability to (eventoriented) obligation, for instance, would be a counterexample, but such a change has, to my knowledge, not been documented. 10. Pakendorf and Schalley (2007) offer a more detailed study of this phenomenon in a number of languages.

(Inter)subjectification in the domain of modality and mood

399

The closest thing to a (counter-directional) shift from epistemic to deontic within the area of modality (and not mood) that I have noted so far is the use of a future marker as obligation in Lakota and Dakota, as noted by Buechel (1939) and Boas and Deloria (1941). Boas and Deloria (1941: 105) state that “the future expresses also the obligatory must”. Buechel (1939: 297) writes that “Lakota has no way of expressing absolute necessity. There is no word corresponding to our ‘must’. The idea must be expressed indirectly. In some instances, the future tense will do it, … [s]ometimes the imperative may be used”. However, the example given by Buechel (1939), ‘Thou shalt not (must not) steal’, has a second person referent, and therefore it seems unlikely that the Lakota and Dakota future semanticized in a real obligation sense. Instead, it seems to be on the way to developing an imperative sense fixated on the second person, as would be predicted. In fact, Buechel (1939: 276) notes that “[t]he future tense is often used instead of the imperative”. 3.3. Change and subjectification within the area of volitive (deontic) meanings As indicated above, Traugott and her research associates have a place for subjectification not only in the change from deontic and epistemic, but also within deontic meaning. This type of change is, however, less highlighted or thought to be less central than the change from deontic to epistemic. In the history of *motan/must, they observe subjectification when uses emerge where “no social, religious, or other normative force is specified or implied” (Traugott and Dasher 2002: 125). In the history of agan/ought to, subjectification has taken place when a use arises in which “the speaker is … presenting him- or herself rather than society as the force behind the obligation” (Traugott and Dasher 2002: 141). Thus, in the following earlier example of *motan/must (4), an external necessity is superimposed on the subject, while in (5), the meaning is participant-internal and motivated by the volition of the speaker.11 (4)

Ær ic moste in ðeossum before I had.to in this

atolan æðele black territory

gebidan. stay-INF

11. The following abbreviations are used in this paper: ACC – accusative; ADV – adverbial; BOU – boulomaic; CON – conditional; GER – gerund; INF – infinitive; NEG – negation; NPS – non-past; VRB – verbalization.

400

(5)

Heiko Narrog

‘Before/earlier I had to stay in this place’ (Fallen Angels [ca. 1000], quoted from Traugott and Dasher 2002: 124) I moste han of the peyrs that I se, Or I moot dye. ‘I must have some of the pears that I see, or I will die’ (Chaucer [14th century], quoted from Traugott and Dasher 2002: 125)

A similar development can be found with the cognate of *motan in the history of German. Like OE *motan (cf. Goossens 1987a; Van Herreweghe 2001), OHG *muozan had a meaning mainly associated with permission. The eventually dominating sense of obligation was only secondarily available. Fritz (1997) offers a highly original and persuasive account of how these two meanings relate to each other. Note that in the history of English as well, the apparent shift from permission to obligation of *motan/must has not yet found a final solution, and Fritz’s explanation is suggestive for the English cognate as well. Fritz (1997: 90) assumes that *muozan in OHG signifies very generally that an event is determined by the external situation. This means that the basic meaning was close to ‘(due to circumstances) someone is in a position to do something’. Being in a position to do something can include both not being hindered and being necessitated. While early in the development the circumstances were such that the meaning of ‘not being hindered’ dominated, they shifted to ‘being necessitated’ in late OHG and Middle High German (MHG). This can be viewed as a kind of pragmatic strengthening. Crucially for the issue of subjectification of deontic meaning, the early obligation examples of *muozan, as in OE, involve necessitation through external circumstances, or through the will of higher religious forces, in the sense of ‘being destined to do something’, rather than personal volition. Cf. example (6): (6)

… sie muosen aber dienon unholden herron … they had.to again serve evil masters ‘…[due to God’s wrath] they had to [= came into the situation to] serve evil masters again.’ (Notker [9th century], quoted from Fritz 1997: 92)

Gamon (1994), although less explicit than Fritz, presents a similar view of the meaning of *muozan in OHG. He states that OHG uses “do not seem to have to do with social obligation or permission; rather external circumstances – not the speaker or some other locus of social force – result in a sort of inevitability of occurrence” (Gamon 1994: 159), and, “only some of

(Inter)subjectification in the domain of modality and mood

401

the early (late OHG) attestations of ‘compulsion’ müssen are specifically deontic, having rather to do with some sort of inevitability or necessity imposed by external circumstances” (Gamon 1994: 160). Note that this account is consistent with the concept of subjectification by Traugott (1993, 2003, 2007), and the concept of speaker-orientation proposed by Narrog (2005b) in suggesting that the deontic meaning originally described external circumstances before being appropriated to express the speaker’s volition and opinions. Note, however, that it essentially contradicts approaches that posit that internal meanings precede external ones (e.g. van der Auwera and Plungian 1998). Also, pace Langacker (1990, 1998, 1999), as in OE and OHG the force is external, and often vaguely attributable to god, heaven, or destiny in general, it becomes questionable that the diffusion of the modal force represents a later stage in development.12 The above examples illustrated subjectification in the early stages of grammaticalization of modal markers. In the remainder of this section, I will adduce examples that show that subjectification within volitive meanings can also occur at later stages, and go even much further. In the examples that I refer to below, subjectification operates on a larger scale, crossing from modality into mood, and in an example from Japanese, it is even connected to grammaticalization. In the cross-linguistic data taken from Bybee et al. 1994 (see Table 2), change #4 from obligation to imperative corresponds to change towards more speaker-orientation, which is confined within the area of volitive (in this case, deontic) modality. Unfortunately, Bybee et al. do not give the names of the languages in which this change occurred, and it is impossible to reconstruct them from the context. However, the pragmatic reasoning behind this change can be comprehended easily from their description: “… we would claim that use of an obligation marker in second person can easily be reinterpreted as an imperative. For instance, sentence (87) [represented here as (7)] can be interpreted as the speaker issuing a directive. (7)

You must call your mother.” (Bybee et al. 1994: 211)

12. Interestingly, Bybee et al. also note that the concept of destiny and inevitability is an important notion related to ‘obligation’. They suggest that English should originally denoted destiny and inevitability, and developed a sense of personal obligation in Modern English (Bybee et al. 1994: 186f).

402

Heiko Narrog

Note that the change from obligation to imperative involves a clear increase in speaker-orientation. The meaning of the construction becomes grounded in the speech situation itself. In my model (see Figure 1 above), it is a change “upwards” from modality into mood. As the imperative is typically directed towards the hearer, the change from obligation to imperative is also a case of intersubjectification in Traugott’s model. A similar change, clearly associated with grammaticalization, occurred in Japanese, with the polyfunctional modal marker besi. It was discussed in some detail in Narrog (2002; 2005b: Section 3.4) and Narrog (2006: Section 3.2) and will be repeated here only briefly. The suffix verb besi had a basic deontic and a basic epistemic sense in Old Japanese. The deontic sense was should-like and can be described as ‘appropriateness’. It was equally applicable to all kinds of human subjects, examples with 3rd person subjects being by far the most common in Old Japanese texts. However, later it came to specialize on 1st person contexts and became a hortative mood. At the same time it lost its inflections and became a particle on the periphery of verb phrases. It can be hypothesized, then, that if volitive markers grammaticalize and shift into the area of mood marking (imperatives, hortatives, desideratives etc.), they typically become focused on the speech act participants as sources and targets of the modal force. This is a paradigm example of increased speaker-orientation or (inter)subjectification. In contrast, this type of development seems to contradict a concept of subjectification which relies on the diffusion of the locus of potency and the attenuation of the grammatical subject. 3.4. Change and subjectification within the area of non-volitive meanings Traugott and Dasher notice subjectification at stage III of the development of must, when examples of the epistemic sense start not to “invoke general opinion …, nor are they contextualized in logical reasoning …, but simply express the speaker’s confidence in the inference being made at the moment” (Traugott and Dasher 2002: 129). This is an unproblematic example of subjectification proceeding within the area of non-volitive, epistemic meaning. This type of subjectification also fits well into the model of change towards more speaker-orientation suggested here. Among the cross-linguistic changes listed in Table 2, change #5, from ability/root possibility to epistemic possibility, and change #8, from fu-

(Inter)subjectification in the domain of modality and mood

403

ture/prediction to probability, represent such changes towards more speaker-orientation. In the case of change #5, which is traditionally labeled change from agent-oriented to epistemic modality, this should be entirely uncontroversial. In the case of change #8, the labeling is somewhat obscure.13 These are cases of what Langacker (2003), with reference to the development of the English modals, has called “extreme subjectification”, that is, future-oriented epistemic modals acquiring present-oriented meaning. The claim made here should in fact be relatively uncontroversial. Numerous researchers have pointed to the fact that future markers which also have a modal epistemic meaning of prediction can later develop a full epistemic meaning, which is not bound to prediction about the future (cf. Bybee et al. 1991: 24–29; Heine 1995b: 126f). With respect to English will, Aijmer (1985: 17) comments on this development as follows: “a sentence with a future form is similar to a sentence with an epistemic modal auxiliary. The modal meaning associated with will at this stage of the change [future > modality] is to express a high degree of certainty.” Originally, will was a verb of volition, but according to Aijmer (1985), and Visser (1969: 1692), a pure (i.e. non-volitive) future/prediction reading already had developed in OE. (8)

Hit wile þincan ungeleaffulic it will seem unbelievable ‘It will seem unbelievable’ (Ælfric [10th ct] quoted from Visser 1969: 1692)

Note that although this use already appears to have a modal coloring, it is still future- and not present-oriented. According to Visser (1969: 1700f), uses that express “a supposition, a deduction, or an inference with the notion of futurity obscured or lost” can be found no earlier than the 15th century.14 This present-oriented use can be illustrated by the following example:

13. Bybee et al. (1994: 202) cite “grams” in Basque, Haka, and Spanish (the synthetic future) as concrete examples. 14. Aijmer (1985: 17) suggests that such uses occurred in the 14th century but provides no examples earlier than the 15th century.

404 (9)

Heiko Narrog

Hwæt wile þis wesan? What will this be ‘What will this be?’ (15th ct text quoted from Visser 1969: 1701)

For German werden, a similar development can be assumed. Werden took over the future function mainly from soln/sollen in Late MHG and Early New High German. The question whether werden in its beginnings was primarily modal or primarily temporal has been controversial in German linguistics. Shigeto (2004: 77) takes the middle position that werden even if used primarily temporally always also had a modal connotation in which the speaker expressed her or his expectation and inference. Crucially, examples with a proposition with present-time reference come relatively late. The following example is the first cited in the Deutsches Wörterbuch (DWb), and is dated in the 16th century (the first example for future use in the DWb is from the 13th century). The epistemic sense is supported by an adverb. (10)

Jonas: ... ey, es ist Spalatinus, der wirdt freilich auch (name) oh it is (name) he will certainly also etwas davon wissen ... something thereof know ‘Jonas:…Oh, it is Spalatinus. He will certainly also know something thereof.’ (Cochläus 1530; as quoted in DWb, vol. 29, col. 255)

As mentioned above, these are examples of Bybee et al.’s (1994) change #8. Bybee et al. (1994) exemplify it with the Spanish synthetic future. This is an even more advanced example than English will and German werden, since, according to Bybee et al. (1994: 224, 227), the Spanish synthetic future is currently losing its future function and is in the process of specializing on present probability. 4. Rethinking the roles of obligation (deontic necessity) and ability/ situational possibility in semantic change The polyfunctionality between strong obligation and strong epistemic necessity, as epitomized by the English modal must, is the most salient and, so-to-speak, spectacular pairing of deontic and epistemic meaning in one

(Inter)subjectification in the domain of modality and mood

405

marker in English and in those languages which share the same phenomenon. Thus, it is not surprising that this polyfunctionality plays a privileged role in Sweetser’s influential explanation of modal polyfunctionality (e.g. Sweetser 1990: 49, 57), and related accounts (e.g. Johnson 1987; Radden and Dirven 2007). However, as pointed out in Section 2, this type of polysemy and change does not appear to be particularly common crosslinguistically, at least not outside the Eurasian language area. In fact, only 3 out of 42 changes listed by Bybee et al. (1994) belonged to this type.15 Furthermore, it appears that overall ability/situational possibility (with 18 changes) was a more common source concept than obligation (which features as source in 9 changes). In this section the notions of obligation and ability/situational possibility will be scrutinized further from a typological and historical perspective. 4.1. How common are ‘obligation’ and ‘ability’/‘situational possibility’ cross-linguistically? For a speaker of English or a similar Indo-European language, the grammaticalized concept of strong obligation as expressed by must is a matter of course, and intuitively may appear to be universal. However, linguistically speaking, this is not the case. This fact was already indicated in the quote from Buechel (1939: 297) about Lakota that this language “has no way of expressing absolute necessity. There is no word corresponding to our ‘must’. The idea must be expressed indirectly.” Another language to which the same observation applies, but in a different manner, is pre-modern Japanese, which had weak obligation markers in besi, and later, in Middle Japanese, in -(y)oozu, but no strong must-type obligation marker. It only emerged in Modern Japanese with the grammaticalization of the doublenegated must-type conditional construction -(a)nakereba naranai, which is, however, strictly speaking a marker of necessity and not of obligation, that is, the obligation meaning arises secondarily to its basic necessity meaning. When we speak about the absence of marking for obligation or ability/ situational possibility we mean the absence of a linguistic construction dedicated to express such a concept. We do not claim that the speakers of such a language lack the concept or are completely unable to express it. As Buechel remarked for Lakota, indirect means of marking may be available 15. Bybee et al. (1994: 195) themselves write that “there is not a large number of grams in our database that express both agent-oriented and epistemic senses”.

406

Heiko Narrog

to speakers. In any case, to get a better idea how wide-spread various eventoriented modalities are, we investigated two data sets. The first one is extracted from a sample of 200 languages based on Ruhlen’s (1987) language classification and Rijkhoff and Bakker’s (1998) sampling method (cf. Narrog to appear). Information was sought on whether each language has at least one marker or construction dedicated to either obligation/root necessity or ability/situational possibility. These pairs of categories are seen here as natural clusters which are labeled NEC (necessitive) and POT (potential) respectively. Table 3 shows the number of languages in the sample which had grammaticalized one of these categories or both (or neither), split by area: Table 3. Obligation and ability/situational possibility in a 200 languages sample16 both NEC and POT

only NEC

only POT

neither

total

Africa

27

3

9

7

46

Americas

32

4

7

4

47

Australia

1

2

1

12

16

Eurasia

15

0

0

0

15

Pacific (Austronesian and Papuan)

27

1

17

11

56

South and Southeast Asia

19

0

1

0

20

total

121

10

35

34

200

The data suggest that many (34) of all languages do not have any constructions dedicated to the expression of either obligation or ability/situational possibility. Furthermore, only 131 (121+10), that is somewhat less than two-thirds of all languages, had a grammatical construction for obligation/ root necessity, which is less than the 157 (121+35) which had one for ability/situational possibility. There are also distinct areal biases, with Eurasia being the area where these modal categories are most regularly present, and the Pacific area including Australia being the area where they are most regularly absent from grammar.

16. The appendix (p. 421–422) lists the 200 languages selected.

(Inter)subjectification in the domain of modality and mood

407

Since I have no direct knowledge of most of the languages in the sample, and descriptions vary in depth and quality, the data contains a certain margin of error. However, as most of the authors of the grammars which were consulted are native speakers of a European language, it must be assumed that they are sensitive to the existence of modal constructions if they existed in the target language, and did not simply overlook them. In fact, some authors, like Buechel above, explicitly refer to the absence of certain types of modality. Thus, Everett and Kern (1997: 327), for example, in their description of Wari’, which is based on the Lingua Questionnaire and therefore requires paragraphs on “debitive mood” and “potential mood” (i.e. obligation and ability and/or situational possibility) state that “there are no forms that mark the potential mood exclusively”, and “debitive mood is expressed by irrealis”. Myhill and Smith (1995: 240, 271), in a crosslinguistic study of obligation expressions, found that Biblical Hebrew had no such expression while Hopi, for example, has only one, which, however, is used but infrequently. In fact, it is not uncommon for languages to only have a general irrealis or future mood, which in context assumes an obligation or ability meaning, and not to have constructions dedicated to these categories. This recalls the situation of older Indo-European, for which many scholars assume (e.g. Plank 1984) that mood forms originally fulfilled the functions that modern modal auxiliaries fulfil, but were semantically more general. The next question I am concerned with here is how common the strong must-type of obligation is. This is shown in Table 4. Table 4. Types of deontic necessity cross-linguistically deontic type strong weak neutral indeterminable

number of languages 60 62 22 32

60 out of 200 languages had a must-type of marker, 62 languages a shouldtype, 22 a type that was described as “neutral” or belonging to neither type, and 32 were indeterminable.17 Note that Myhill and Smith have already pointed out that the categories of ‘strong’ vs. ‘weak’ obligation as observed in English must vs. should are not universal types of obligation available in 17. Note that 44 languages had markers of more than one type. Therefore, the total is larger than the 131 in Table 3.

408

Heiko Narrog

any language (Myhill and Smith 1995: 242). In any case, the number of languages that have a must-type of obligation are only a minority, even if some of the “indeterminable” markers would turn out to be of the musttype. It is furthermore of interest that in some minority languages the musttype marker is borrowed from a neighboring major language. This was explicitly noted in the grammar of six languages (Dongolese Nubian, Maba: borrowed from Arabic; Itelmen: borrowed from Russian; Biak: borrowed from Indonesian; Semelai, Tetun Fehan: borrowed from Malay). 4.2. Why are there relatively few languages with must-type deontic markers, and why is the deontic necessity – epistemic necessity polyfunctionality not more common? The question posed here can be addressed either with language-internal arguments or with language-external arguments. The former type of explanation might centrally refer to the possibility of expressing strong obligation indirectly, for example with a weak obligation marker or a general irrealis marker that has strengthening elements in the linguistic or nonlinguistic context. Nevertheless, the question remains why certain languages offer their speakers the choice to directly express an obligation, and more specifically a strong obligation, while others don’t. I suggest that in order to answer this question we might consider turning to the fact that the specific vocabulary and structure of languages are the diachronic product of the activities of their users in specific cultures and societies. This point leads us to the fact that obligation is quite different from epistemic modality, and also ability and possibility, in being socially defined, namely by morals, laws or shared values. Obligations are thus clearly sensitive to society and culture and not merely to cognition or linguistic structure. Depending on specific societies and cultures, the nature of obligations, that is the questions of who is entitled to oblige others to certain actions or behaviors, who has moral authority, who is responsible for carrying out obligations (e.g., individuals and/or groups), what kind of obligations exist (e.g. religious, legal, moral), or if obligations systematically exist at all in a culture, may vary widely. There is the danger that in the discussion of ‘obligation’ or ‘deontic necessity’, we unconsciously conceptualize and analyse them in terms of a specific language and culture, which we assume by default, but probably incorrectly, to be universal.

(Inter)subjectification in the domain of modality and mood

409

In the context of language, markers of obligation are closely related to imperatives. The most salient semantic-pragmatic difference is that, unlike markers of obligation, imperatives are in principle always used performatively. Furthermore, imperatives are commonly used when the speaker has full authority over the hearer. In contrast, obligation markers are associated mainly with situations that diverge from this ‘unmarked’ scenario, namely either describing and reporting obligations, or, if used performatively, without the speaker having straightforward authority over the hearer (cf. Palmer 1986: 29–30; Myhill and Smith 1995: 275, 283). Imperatives are believed to be practically universal to all languages (cf. König and Siemund 2007: 303), but obligation markers and constructions are absent in a large number of languages. This may be due to the fact that they are associated with marked and socially problematic scenarios. Obligations potentially put human relationships at risk, since non-compliance is associated with all kinds of sanctions. Talking directly about obligations may be face-threatening or even menacing, and consequently, if the speaker is not in a position of full authority, puts the speaker her- or himself at risk. This is presumably even much more the case in small-scale, closely-knit, group-oriented societies, where the maintenance of human relationships is vital to the functioning of the group, than in large-scale, individualized cultures. This leads directly to the reason for the relative cross-linguistic paucity of strong obligation markers and constructions that I would like to suggest in this section: in many cultures, direct reference to obligations is habitually avoided for social reasons, especially if the speaker is not in a position of absolute authority. Obligations are then habitually referred to in an indirect manner if at all. Therefore, languages in such cultures are less likely to grammaticalize obligation markers and constructions, particularly not strong ones, since grammaticalization presupposes frequency of use. I wish to illustrate this idea with Japanese, which historically only had a weak obligation marker (besi), and no must type of constructions, but in Modern Japanese developed a strong obligation construction as well. This strong obligation construction, however, denotes primarily a general (nonepistemic) necessity, meaning literally ‘something does not “become” if something is not done’, thus implying, depending on context, the obligation to do something, as in (11), where it denotes a necessity and may consequently be an obligation.

410 (11)

Heiko Narrog

Kanozyo-o sagasi-das-ana-kereba nar-ana-i. look.for-bring.out-NEG-CON become-NEG-NPS 3S-ACC ‘I have to find her.’ (Kǀbǀ Abe: Hakootoko, 1973)

Although Modern Japanese now does have this marker that can express strong obligation, there is some evidence that nevertheless suggests that its use is more restricted than must in English, presumably corresponding to a different social fabric which is reflected in the well-known propensity for the use of negative politeness in Japanese language and culture (cf. Brown and Levinson 1987: 245–247) and for the avoidance of face-threatening acts. The first piece of evidence that I adduce to support my argument here comes from a corpus study on Japanese and Chinese modal markers and person (Chiang 2007).18 The study showed that it is extremely uncommon in Japanese to use the strong necessity construction -(a)nakereba naranai with second person subjects, that is, referring to a possible obligation or necessity of the interlocutor. As can be seen in Table 5 below, Chiang found no single instance of the construction with a second person subject. Table 5. Frequencies of person in Japanese strong necessity construction (Chiang 2007: 72)

-(a)nakereba naranai

1st person subject

2nd person subject

3rd person subject

total

52

0

63

115

Note that the non-occurrence of strong necessity -(a)nakereba naranai with second person subject is not due to any morphosyntactic constraint. This kind of usage is simply avoided for pragmatic (social) reasons. In a much larger sample we might find some examples, but they would still be scarce. However, language users have, so to speak, created a convenient loophole here. The construction is a conditional and it is possible to omit the consequent (apodosis) which indicates negative evaluation. If this happens, the antecedent is usually abbreviated to -(a)nakya/nakutya. In this case, the negative consequent (‘[it] does not become’) is left to be inferred by the hearer. In principle it is left unclear whether the consequent contains a negative evaluation or not, although the entrenchment of the strong necessity 18. Chiang used a mixed corpus of written and spoken language.

(Inter)subjectification in the domain of modality and mood

411

construction with the negatively evaluated consequent leads the hearer to infer the negative evaluation and thus the necessity/obligation. (12)

Haya-ku it-te kure-na-kyaa. say-GER give-NEG-CON fast-ADV ‘You should tell me earlier.’ (Mainichi Newspaper 6/2003)

In this way, the construction is mitigated, and apparently only in this form does it become suitable for the use with the interlocutor as subject. As Table 6 shows, it now regularly (although not particularly frequently) cooccurs with second person subject. Table 6. Frequencies of person in Japanese strong necessity construction (Chiang 2007: 72)

-(a)nakya/naktya

1st person subject

2nd person subject

3rd person subject

total

35

13

4

52

If the numbers for the full construction and for the abbreviated construction are combined, we get 52% of uses with first person, 40% with third person, and 8% with second person. The first person use of this construction appears to be overall the most common use. Recall, however, that must-type markers with first person cross-linguistically are often used not to express an actual obligation, but instead pragmatically express a justification for an action of the speaker that potentially inconveniences the hearer (cf. e.g. Myhill and Smith 1995: 249, 261), or simply something that the speaker has an urge to do. Typical and common expressions in English that belong to the latter type are I must admit/say/confess etc. (cf. Coates 1983: 35–36). As they do not refer to any real life obligation one could as well label them ‘pseudo-obligations’.19 Use with second person subject only becomes slightly better with the weak should-type marker beki, for which 12% second-person use were recorded (cf. Chiang 2007: 108). Instead of these constructions that directly express a non-epistemic necessity, when referring to an obligation that pertains to the second person, so-called ‘advice’ or ‘recommendation’ constructions that only indirectly hint at an obligation, such as hoo-ga ii ‘this 19. The matter how the obligation constructions with first person subject are actually used in Japanese was not investigated by Chiang (2007).

412

Heiko Narrog

way is better’ are preferred. (Chiang [2007: 108] has 39% of second person subjects with this construction.) The same restrictions that hold for Japanese apparently do not hold for English, at least not to the same extent. Table 7 shows counts from two corpora. First, I counted the deontic uses of must (instances from written text only) among the first 600 examples in the 5-million word extraction of Collins Cobuild Bank of English (2001). Second, I counted all instances of (clearly) deontic must in the London–Lund Corpus. Table 7. Frequencies of person with deontic must in British English corpus data20 1st person subject

2nd person subject

3rd person subject

total

Collins Cobuild (written)

60

63

218

341

London–Lund Corpus (spoken)

130

53

59

242

Total

190

116

277

583

Overall proportion

0.33

0.20

0.48

In written language there is a strong propensity towards third person subjects, presumably due to the fact that in written language the majority of texts deal with third person subjects. In contrast, in spoken language, a strong propensity can be observed to use deontic must with first person subjects. This is to a certain extent due to the use of fixed ‘pseudoobligation’ expressions with verbs of locution such as I must say or I must admit (36 out of 130 examples) in colloquial language. Examples with second person subject are not privileged in either register, but occur constantly in both spoken and written language. The fact that the number is still the lowest for all persons might be motivated in a similar way as I suggested for Japanese: the use of a strong obligation marker referring to an action of the hearer is potentially imposing and face-threatening. Nevertheless, this probably universal fact about strong obligation markers apparently does not have the same dramatic effect on its use in English as it does in Japanese. Coates (1983: 36) also mentions second-person subject use of 20. The 341 deontic examples of must in the 5-million word extraction from the Bank of English occurred in a total of 600 examples, while the 242 examples from the London–Lund Corpus are part of a total of 496 examples of must in the corpus.

(Inter)subjectification in the domain of modality and mood

413

must as one of the parameters along which its semantics varies, and Myhill and Smith (1995: 249, 253) cite specific uses of English obligation markers with second person subjects, which even include ‘reprimands’ and ‘negative effect on listener, speaker unsympathetic’, that is, face-threatening acts. Another piece of evidence pointing at cultural and linguistic differences concerning deontic expressions, as shown by comparing English and Japanese, is the general paucity of deontic performative verbs in the latter language. For example, the following verbs can at least in principle be used performatively and deontically in English with a first person subject and a second person subject in the subordinate clause which is obliged or requested to do something: order, command, insist, demand, oblige, require, request, ask, tell (cf. Vanderveken 1990: 189–196). It is possible to find translational equivalents for them in Japanese: order, command, require, ask, tell – meiziru, meirei-suru; demand – yookyuu-suru; insist – syutyoosuru; oblige – gimu-dukeru. However, they are commonly used only descriptively.21 Moreover, all of these verbs belong to Sino-Japanese vocabulary, that is, they are not part of the indigenous Japanese vocabulary. In fact, the Nihon Kokugo Daijiten, the Japanese equivalent of the OED, lists their first appearance as verbs for the Meiji period of Westernization or even later.22 All of them are generally associated with written rather than spoken language. Finally, it should be mentioned that the most common form of deontic expression in Modern Japanese, namely through conditionals, is logically based on the concept of possibility and not necessity, as has already been pointed out by Akatsuka (1992: 7). One finds the expression of necessity only through negating possibility. Overall, I would like to suggest that there are languages with a cultural and social background in which the overt expression of deontic necessity is avoided, and where for this reason deontic concepts are less likely to be grammaticalized. I have illustrated this only with Japanese. However, cross-linguistic data show that the number of such languages may be large. If the expression of deontic necessity is less grammaticalized, or relatively 21. All of them, except syutyoo-suru, are used performatively in legal documents. However, as the legal system as such was imported from European countries in the Meiji era, this specific use as well is presumably an import (which has not percolated through to other registers). 22. For meiziru, yookyuu-suru, and syutyoo-suru, the first examples are listed in a period between 1867 and 1871, for gimu-dukeru and meirei-suru in the second half of the 20th century.

414

Heiko Narrog

infrequently used even if grammaticalized, the category is also less likely to become a source of expression for other modal categories including epistemic modality. This is certainly one promising explanation for the fact that cross-linguistically a change from strong deontic necessity to strong epistemic necessity, as witnessed in the case of English must, is relatively uncommon despite the close conceptual relationship between deontic and epistemic modality. 4.3. What is this thing called obligation? – The many natures of deontic necessity I have already pointed out the possibility that the way that deontic necessity (obligation), including its role in semantic change, is conceptualized in modern linguistics is shaped by our perception of deontic necessity in modern European languages (particularly the modern Germanic languages) and cultures, which we tend to assume to be universal without giving it much thought. However, obligation is a notion which is apparently extremely sensitive to the social and cultural background, and may differ substantially from language to language. Myhill and Smith (1995: 285) in their crosslinguistic study on obligation expressions remarked in this respect that if a list of obligation functions were to be created “on the basis of … Hopi data, we could make a category like ‘instructions for performing a ritual spiritually beneficial to the community as a whole’…. However, this context would occur rarely if ever in original texts in languages such as English and Chinese…”.23 We also face the possibility that different conceptualizations of deontic necessity may lead to different diachronic pathways of change, or, more generally, that the way deontic necessity is conceptualized is reflected in the way it changes semantically. I wish to address this issue from two different viewpoints here. First, historically within the history of English; second, cross-linguistically, examining various sources of deontic necessity markers. 23. A similar type of obligation is noted by Willett (1991: 176) for Southeastern Tepehuan (Uto-Aztecan): strong obligation can only be conveyed in conjunction with negation. That is, the taboos associated with sacred rites are always obligatory for the participant, but they are usually things that he must not do. To express in a positive way what a devout person should do, the Spanish phrase tiene que ‘have to’ is borrowed.

(Inter)subjectification in the domain of modality and mood

415

Our use of obligation markers is bound to a specific social and cultural background, which in the case of modern Western societies is based on the idea of rights and obligations of the individual human being. We have the freedom to make choices that influence or even determine our future, and on the other hand we may be subject to obligations which constrain our choices and for which we bear personal responsibility. This world-view, however, is not common to all societies, and even in our own cultures, a look into older texts, particularly those up to the middle ages, reveals a rather different world view, which may be just as exotic for us as that of non-Western cultures. Few people had much choice over how to conduct their lives, and what futures to choose, and in addition to heavy social constraints, Christian doctrine and faith reigned people’s thinking to an extent that is difficult to imagine in modern society. This social background is, I suggest, reflected in the meaning and use of the deontic necessity expressions of the time. I take shall/should as an example. The original meaning of shall as a main verb, according to the OED (shall, v, BI), is ‘to owe (money etc.)’. It is the derived perfect of a now lost present stem, and the Germanic etymology, according to the DWb (Grimm and Grimm 1905: 1452) is probably something like ‘have killed someone (and therefore owe the payment of a penalty)’. Its most basic meaning in Old and Middle English when accompanying an infinitive complement was, according to the OED (shall, v, BII), the expression of ‘what is right and becoming’. This is a meaning which in Modern English is expressed either by the past subjunctive should, or by ought. Note that ‘what is right and becoming’ is not itself an obligation, but that an obligation can merely be inferred from ‘what is right and becoming’, in the manner of ‘doing P is right and therefore I have the obligation to do P’. It further denoted ‘necessity of various kinds’, and, corresponding to Modern English is to or am to “indicating what is appointed or settled to take place”, and meanings pertaining to commands, instructions, and the speaker’s determination. Visser (1969: 1581–1582) comments on the deontic sense as follows: Since here the main notion [of shall] is that of obligation it is natural to ask who or what the force is that must be considered as imposing that obligation. In pagan times the gods, and in Christian times God were consciously or unconsciously regarded as the disposers of future events, so that ‘he sceal + infinitive’ was originally used for events that were seen as predestined or providentially decreed …, and later, when the idea of divine interposition was weakened or absent, for events predetermined by fate or general necessity, and eventually for events that were sure to happen independently of anybody’s will. (Visser 1969: 1581–1582)

416

Heiko Narrog

Concerning the non-deontic interpretation, shall had (as also implied in Visser’s comment) already acquired a future sense by Old English. The Old English future sense is characterized by the OED as “announcing a future event as fated or divinely decreed”, and further, “[h]ence, shall has always been the auxiliary used, in all persons, for prophetic or oracular announcements of the future …”. The following is an early example: (13)

Sceal anra gehwylc fore Cristes cyme cwic arisan. shall one every at Christ’s coming alive arise. ‘Each one shall arise alive at the coming of Christ.’ (Cynewulf, Crist, ca. 900; glosses H.N.)

Further epistemic senses that occurred later are listed. The point that deserves particular attention here is that neither the Old English deontic sense nor the Old English future sense are necessarily the same as the Modern English senses of shall and should, and that the way future and obligation were conceptualized in historical times probably also reflected on the (diachronic) relationship between these senses. Specifically, the deontic and the non-deontic sense both share a salient element of destiny or things being predetermined by higher forces beyond the will of the subject (or even human volition in general). In the deontic sense, we find a view of predetermination particularly in the sense of “indicating what is appointed or settled to take place”. This directly leads to the future sense of “announcing a future event as fated or divinely decreed”. This element that links deontic and epistemic (future) meaning is bound to a specific cultural conceptualization of deontic necessity and future, and it is probably also diachronically the decisive linking element between the two senses (we cannot know this for sure, since this specific epistemic (future) sense was already extant in the Old English documents).24 4.4. Cross-linguistic diversity of sources of deontic markers At this point I wish to direct attention to the cross-linguistic diversity in sources for deontic necessity meanings. The most common explanation for the emergence and development of deontic modal meanings, and even their 24. Parallels to the deontic and epistemic (future) tense of should can be found in the history of the Old Japanese marker besi, which also basically denoted an inevitability or determinedness in both senses (cf. Narrog 2002).

(Inter)subjectification in the domain of modality and mood

417

extension to epistemic meanings, in English in the field of Cognitive Linguistics is associated with the concept of ‘force dynamics’ (Sweetser 1982, 1990; Talmy 1988; Langacker 1990, 1991; Radden and Dirven 2007; and others). This concept has been applied to the analysis of Modern English modals and projected into diachrony. I believe that a closer investigation is needed in order to determine whether this concept is as valid for the description of the actual history of the English modals, as it is elegant for the synchronic description.25 However, I will leave this point for a separate study and simply wish to point out that cross-linguistic sources for grammaticalized deontic necessity marking is quite diverse, and it is questionable whether the concept of ‘force dynamics’ is universally applicable to these sources. The lexical sources for modal markers of deontic necessity found in the data from the 200 languages sample introduced in 4.1., are listed in Table 8 below.26 The evaluative ‘be good’-type which is most common crosslinguistically is also the most common type in Japanese, and can be exemplified by the construction -(r)eba ii ‘it is good if’ as in (14): (14)

Yase-ta-kat-tara, kuw-ana-kereba i-i good-NPS slim.down-BOU-VRB-CON eat-NEG-CON ‘If you want to slim down, you should eat less’ (lit. ‘It would be good if you ate less, if you want to slim down’) (Takanashi 1995: 238, 242)

25. Similar skepticism is voiced by Traugott and Dasher (2002: 111) who state that “there are reasons to doubt that the forces and barriers metaphor is a key to the semantic development of either deontics or epistemics. For one, the lexical sources of the modals show little evidence of the semantics of forces and barriers.… For another, … metaphorization does not appear to be crucial to an understanding of the processes behind semantic change as revealed by the textual evidence.” 26. There is an overlap between ‘be’ and ‘have’ with two languages and ‘be’ and ‘become’ with one language because the source construction can have both meanings.

418

Heiko Narrog

Table 8. Lexical sources of obligation markers Marker/construction be good/with goodness construction with ‘be’ be fitting/be suitable; or: fit/meet want/wish be correct/right/proper get, obtain have be a duty, be enough/suffice, be a thing, attach/affix, be appropriate, become, be better, belong to, come, be prepared, be a custom, do/make, follow, fall, stick, go, owe, hold/keep, succeed, put, say, see, impinge/affect, transform, portion/lot, have way

Number of languages 12 11 11 8 6 4 4 ”2

Practically all Modern Japanese deontic modal expressions (with the exception of beki, which is historically ancient and has no specific etymology) contain either the element ‘good’ or a negatively evaluated element. That is, Modern Japanese deontic modality is fundamentally ‘valuative’ in nature. If any ‘forces’ are involved here at all, they can only be secondarily implied by the valuative judgment. Evaluation, which is only marginally represented in the English modal system, e.g. had better (see Denison and Cort this volume), may at least in some languages be more important in conceptualizing deontic modality than ‘forces’ and ‘power’. Givón, for example, has made allowances for this fact, in relabeling the category traditionally known as ‘deontic modality’ as ‘valuative’ (Givón 1995: 112–113). The cross-linguistically second most frequent types are constructions involving either an existential verb or a verb or adjective meaning something like ‘fit’ or ‘be suitable’. (15) is an example of English be to in a deontic meaning. (English is not part of the sample but it is cited here because it is more easily accessible than examples from languages which are not widely known.) (15)

He was simply to listen and report back what he heard. (Mindt 1995: 151)

According to Mindt’s statistics (1995: 150), the deontic use is minor, accounting for only 26% of the various readings of the be to-construction. More commonly, be to denotes a ‘possibility/high probability’ or a ‘certain-

(Inter)subjectification in the domain of modality and mood

419

ty/prediction’ (totalling 64%). The following is an example of the ‘prediction’ sense. (16)

This was the beginning of the awareness which was to come in the Sixties. (Mindt 1995: 151)

Interestingly, be to, as mentioned in the preceding paragraphs, partly corresponds to the meaning of shall in Old English. The feature that sets it apart from more common obligation constructions in Modern English, such as must and should, is that the obligee is not conceptualized as active, but rather as being bound or scheduled to perform or undergo some event which is predetermined. It is no coincidence that in English, the passive as well is a construction with be. Bybee et al. (1994: 185) point out that syntactically “some of the source constructions for obligation are passive-like in structure; that is, the one who is obliged is treated like the object or patient in the clause.” The passive-like conceptualization of a deontic necessity is also reflected in some other source meanings, such as ‘fall’/‘befall’ or ‘become’ (which happens to be at the centre of the Modern German passive). According to Bybee et al. (1994: 184) ‘have’-type constructions are similar to the ‘be’-constructions in that both “associate” an agent with an activity. The ‘be fitting’-type of constructions are also clearly close to the ‘be’-constructions in that they denote that an activity matches a frame that is set up by some secular or religious authorities. It appears to me that the ‘be to’ and ‘be fitting’-type describe a scheduling or predetermination of events to which an agent is passively subjected. Whether or not this is conceived as happening under the influence of force, or simply as a naturally predetermined course of events is a question that may be culturally determined. In the case of the ‘have’-source meaning, an apparent contradiction arises with a ‘force dynamics’ conceptualization because a possessor should in fact be expected to yield power and force and not be in the exact opposite position, namely as the target of an external force. None of these meanings can be straightforwardly described by the concept of ‘force dynamics’. Such a type of description will at least take some sophisticated argumentation. It should be mentioned here that in English itself the most straightforward match for a ‘force dynamic’ explanation is may, which indeed originally had the meaning of the subject yielding power (cf. OED may, v1, B), but this is a verb of deontic possibility and not of necessity.

420

Heiko Narrog

5. Conclusions In this paper, I have discussed diachronic aspects of modality from the perspective of grammaticalization and (inter)subjectification. I critically reviewed the concepts of subjectification of Langacker and (inter) subjectification of Traugott with respect to their accounts of change in modal markers (the English modals), and presented my own concept of increased speaker-orientation as the crucial dimension in cross-linguistic change of modal markers. This model extends from modality proper into mood and illocutionary force modulation, categories which are seen as natural targets for markers of modality proper in semantic change. One of the characteristics of research on modality and semantic change including subjectification has been a focus on some particularly salient changes, especially that from strong obligation to epistemic certainty, as exemplified by English must. However, this type of change is crosslinguistically not especially common. In fact, changes based on root possibility enjoy a higher frequency cross-linguistically. I related this observation to the fact that the grammaticalization of obligation markers, particularly strong obligation markers, is cross-linguistically much less frequent than one would suspect from the situation in English, which has several such markers and constructions. In addition, even if obligation markers and constructions are grammaticalized, they may still be used less frequently than in English, or be subject to various constraints, thus making further grammaticalization and semantic extension less likely. I suggested that the non-development of obligation markers in many languages may be related to the social and cultural background in which these languages are spoken. Obligations are pragmatically impositions. They are thus potentially face-threatening, and their direct expression might be avoided. Furthermore, it was suggested that the concept of obligation itself is not universal but may differ from culture to culture, leading to different diachronic pathways. These points were illustrated by cross-linguistic data concerning the distribution and etymology of obligation markers, and with pragmatic constraints on the use of obligation markers in Japanese (see also Myhill 1995, 1997; Burridge 1998 for studies on cultural influences on the grammaticalization of modal verbs in American English and Pennsylvania German, respectively). One possible conclusion from these observations is that subjectification and intersubjectification do not take place in a social vacuum or (only) in an individual’s cognition but also in speaker-hearer interaction in social contexts and in specific communities and cultures (see

(Inter)subjectification in the domain of modality and mood

421

Traugott 2007: 296–297 for a similar view). It would perhaps be going too far to claim on the basis of the relatively sketchy evidence provided in this paper that social norms and social practices ‘motivate’ semantic change, but it is certainly possible to suggest that they are part of the conditions for semantic change. This is an issue that warrants further empirical investigation in geographically and historically diverse languages (including the well-known Indo-European languages). Acknowledgements First, I wish to thank the editors for inviting me to this volume. I am also grateful for the eye-opening advice that I received from Elizabeth Traugott, Kristin Davidse, and two anonymous reviewers during the revision process. I wish I would have been able to reflect it more exhaustively and faithfully in this paper. Finally, I wish to express my gratitude for Kristin Davidse’s thorough proof-reading at the end of the process, which (I hope) made this article much more readable. Appendix: 200 languages sample The following 200 languages were selected (alphabetical order within phyla): Amharic, Arabic, Dhaasanac, Hausa, Hdi, Iraqw, Maale, Mali Tamashek/Tamasheq, Mina, Miya, Mupun, Musgu, Somali (Afro-Asiatic); Japanese, Khalka Mongolian, Turkish, Udihe (Altaic); Awa Pit, Aymara, Blackfoot, Chalcatongo Mixtec, Chimalapa, Comanche, Eastern Pomo, Halkomelem, Hixkaryana, Huallaga Quechua, Itzaj, Jarawara, Koasati, Kwaza, Lakhota, Mam, Maricopa, Misantla Totonac, Moseten, Movima, Nez Perce, Nootka, Northern Paiute, Paumari, Pawnee, Retuara, Sabanes, Shoshone, Sierra Miwok, Sochiapan, Tariana, Tepehuan, Teribe, Thompson, Timucua, Trumai, Tzutujil, Urarina, Waiwai, Wari’, Wintu, Yurok (Amerind); Arabana-Wangkangurru, Bininj Gun-wok, Diyari, Dyirbal, Gagadu, Gooniyandi, Kayardild, Mangarayi, Martuthunira, Ngiyambaa, Nunggubuyu, Nyangumarta, Wambaya, Wardaman, Yidiny, Yindjibarndi (Australian); Araki, Begak, Biak, Blue Hmong, Chamorro, Eromangan, Fijian, Hawaiian, Indonesian, Kambera, Khmer, Kwaio, Leti, Loniu, Maori, Mualang, Mwotlap, Paiwan, Palauan, Rapanui, Santali, Semelai, Sinaugoro, Taba, Tetun Fehan, Thai, Tinrin, Tukang Besi, Tuvaluan, Vietnamese,

422

Heiko Narrog

Yapese, Yogad (Austric); Abkhaz, Georgian (Caucasian); Itelmen (Chuckchi-Kamchatkan); Berbice Dutch Creole, Ndyuka, Tok Pisin (Creoles); Kangiryuarmiut (Eskimo-Aleut); Breton, Catalan, Kashmiri, Marathi, Modern Greek, Punjabi, Russian, Swedish, Zazaki (Indo-Hittite); Abui, Abun, Alamblak, Ambulas, Amele, Awtuw, Bilua, Bongu, Bukiyip, Dani, Dla, Eipo, Enga, Fore, Hatam, Hua, Imonda, Kâte, Kobon, Korowai, Lavukaleve, Mai Brat, Menya, Nabak, Sahu, Suena, Tauya, Telefol, Yimas (Indo-Pacific); Ainu (Isolates); Kung-Ekoka, Nama (Khoisan); Haida, Slave (Na-Dene); Adioukrou, Babungo, Baule, Bobangi, Degema, Ewe, Fongbe, Hone, Jalonke, Jeli, Kana, Koromfe, Krongo, Leggbo, Lele, Mundang, Nkore-Kiga, Noon, Samba Leko, Supyire, Swahili (Niger-Kordofanian); Anywa, Dongolese Nubian, Gula, Kabba, Koyra Chiini, Lango, Maba, Madi, Mbay, Ngiti (Nilo-Saharan); Burmese, Chinese, Dolakha Newar, Garo, Lahu, Meithei, Mongsen Ao, Qiang, Tshangla (Sino-Tibetan); Finnish (Uralic-Yukaghir). References Aijmer, Karin 1985 The semantic development of will. In Historical Semantics, Historical Word-Formation, Jacek Fisiak (ed.), 11–21. Berlin: Mouton. Akatsuka, Noriko 1992 Japanese modals are conditionals. In The Joy of Grammar: A Festschrift in Honor of James D. McCawley, Diane Brentari, Gary N. Larson, and Lynn A. Macleod (eds.), 1–10. Amsterdam/Philadelphia: Benjamins. Boas, Franz and Ella Deloria 1941 Dakota Grammar. New York: AMS Press. Brown, Penelope and Stephen C. Levinson 1987 Politeness: Some Universals in Language Usage. Cambridge: Cambridge University Press. Buechel, Eugene 1939 A Grammar of Lakota: The Language of the Teton Sioux Indians. Rosebud: Rosebud Educational Society. Burridge, Kate 1998 From modal auxiliary to lexical verb: The curious case of Pennsylvania German. In Historical Linguistics 1995. Selected Papers from the 12th International Conference on Historical Linguistics, Manchester, August 1995, vol. 2, Richard M. Hogg and L. van Bergen (eds.), 19–33. Amsterdam/Philadelphia: Benjamins.

(Inter)subjectification in the domain of modality and mood

423

Bybee, Joan, William Pagliuca and Revere Perkins 1991 Back to the future. In Approaches to Grammaticalization, vol. II, Elizabeth Closs Traugott and Bernd Heine (eds.), 17–58. Amsterdam/Philadelphia: Benjamins. Bybee, Joan, Revere Perkins and William Pagliuca 1994 The Evolution of Grammar: Tense, Aspect, and Modality in the Languages of the World. Chicago: Chicago University Press. Caughley, Ross C. 1982 The Syntax and Morphology of the Verb in Chepang. Canberra: Australian National University. Caughley, Ross C. 2000 Dictionary of Chepang: A Tibeto-Burman language of Nepal. Canberra: Australian National University. Chiang, Chien-Ling 2007 Nihongo no modariti hyǀgen to ninshǀ ni tsuite – Chnjgokugo to no hikaku o chnjshin ni [Japanese modal expressions and person – a comparison with Chinese]. PhD dissertation. Sendai: Tohoku University. Cornillie, Bert 2005 On modal grounding, reference points, and subjectification. The case of the Spanish epistemic modals. Annual Review of Cognitive Linguistics 3: 56–77. Cornillie, Bert 2006 Conceptual and constructual considerations on the subjectivity of English and Spanish modals. In Subjectification: Various Paths to Subjectivity, Angeliki Athanasiadou, Costas Canakis and Bert Cornillie (eds.), 177–205. Berlin/New York: Mouton de Gruyter. Croft, William 2000 Explaining Language Change: An Evolutionary Approach. London: Longman. De Smet, Hendrik and Jean-Christophe Verstraete 2006 Coming to terms with subjectivity. Cognitive Linguistics 17: 365– 392. Dietrich, Rainer 1992 Modalität im Deutschen. Opladen: Westdeutscher Verlag. Diewald, Gabriele 1999 Die Modalverben im Deutschen: Grammatikalisierung und Polyfunktionalität. Tübingen: Niemeyer. Evans, Vyvyan and Melanie Green 2006 Cognitive Linguistics: An Introduction. Edinburgh: Edinburgh University Press.

424

Heiko Narrog

Everett, Daniel Leonard and Barbara Kern 1997 Wari: The Pacaas Novos Language of Western Brazil. London: Routledge. Fritz, Gerd 1997 Historische Semantik der Modalverben. Problemskizze – Exemplarische Analysen – Forschungsüberblick. In Untersuchungen zur semantischen Entwicklungsgeschichte der Modalverben im Deutschen, Gerd Fritz and Thomas Gloning (eds.), 1–157. Tübingen: Niemeyer. Gamon, David 1994 On the development of epistemicity in the German modal verbs mögen and müssen. Folia Linguistica Historica 14: 125–176. Givón, Talmy 1995 Functionalism and Grammar. Amsterdam/Philadelphia: Benjamins. Goossens, Louis 1987a Modal tracks. The case of magan and motan. In Studies in Honour of René Derolez, Anne-Marie Simon-Vandenbergen (ed.), 216–236. Ghent: University of Ghent. Goossens, Louis 1987b Modal shifts and predication types. In Ins and Outs of the Predication, Johan van der Auwera and Louis Goossens (eds.), 21–37. Dordrecht: Foris. Goossens, Louis 2000 Patterns of meaning extension, “parallel chaining”, subjectification, and modal shifts. In Metaphor and Metonymy at the Crossroads: A Cognitive Perspective, Antonio Barcelona (ed.), 149–169. Berlin/New York: Mouton de Gruyter. Grimm, Jacob and Wilhelm Grimm 1854–1971 Deutsches Wörterbuch. [dWb] Leipzig: Hirzel. Haspelmath, Martin and Ekkehard König 1998 Concessive conditionals in the languages of Europe. In Adverbial Constructions in the Languages of Europe, Johan van der Auwera and Donall Ó Baoill (eds.), 563–640. Berlin/New York: Mouton de Gruyter. Heine, Bernd 1995a Agent-oriented vs. epistemic modality: Some observations on German modals. In Modality in Grammar and Discourse, Joan Bybee and Suzanne Fleischman (eds.), 17–53. Amsterdam/Philadelphia: Benjamins. Heine, Bernd 1995b On the German werden future. In Discourse Grammar and Typology, Werner Abraham, Talmy Givón, Sandra A. Thompson (eds.), 119–138. Amsterdam/Philadelphia: Benjamins.

(Inter)subjectification in the domain of modality and mood

425

Heine, Bernd and Mechthild Reh 1984 Grammaticalization and Reanalysis in African Languages. Hamburg: Helmut Buske. Jespersen, J. Otto 1992 Reprint. The Philosophy of Grammar. Chicago: University of Chicago Press. Original publication, 1924. Johnson, Mark 1987 The Body in the Mind: The Bodily Basis of Meaning, Imagination, and Reason. Chicago: University of Chicago Press. Kiefer, Ferenc 1987 On defining modality. Folia Linguistica XXI: 67–94. Kiefer, Ferenc 1997 Presidential address – Modality and pragmatics. Folia Linguistica XXXI: 241–253. König, Ekkehard 1992 From discourse to syntax: the case of concessive conditionals. In Who Climbs the Grammar-Tree, Rosemarie Tracy (ed.), 423–433. Tübingen: Niemeyer. König, Ekkehard and Peter Siemund 2007 Speech act distinctions in grammar. In Language Typology and Syntactic Description, vol. 1, Timothy Shopen (ed.), 276–324. Cambridge: Cambridge University Press. Langacker, Ronald W. 1990 Subjectification. Cognitive Linguistics 1: 5–38. Langacker, Ronald W. 1991 Foundations of Cognitive Grammar. Volume 2: Descriptive Application. Stanford: Stanford University Press. Langacker, Ronald W. 1998 On subjectification and grammaticalization. In Discourse and Cognition: Bridging the Gap, Jean-Pierre Koenig (ed.), 71–89. Stanford: CSLI. Langacker, Ronald W. 1999 Losing control: grammaticization, subjectification, and transparency. In Historical Semantics and Cognition, Andreas Blank and Peter Koch (eds.), 147–175. Berlin/New York: Mouton de Gruyter. Langacker, Ronald W. 2003 Extreme subjectification: English tense and modals. In Motivation in Language: Studies in Honor of Günter Radden, Hubert Cuyckens, Thomas Berg, René Dirven and Klaus-Uwe Panther (eds.), 3–26. Amsterdam/Philadelphia: Benjamins. Lyons, John 1977 Semantics. Cambridge: Cambridge University Press.

426

Heiko Narrog

Mindt, Dieter 1995 An Empirical Grammar of the English Verb: Modal Verbs. Berlin: Cornelsen. Mindt, Dieter 2000 An Empirical Grammar of the English Verb System. Berlin: Cornelsen. Mortelmans, Tanja 2002 Wieso sollte ich dich küssen, du hässlicher Mensch! A study of the German modals sollen and müssen as grounding predications in interrogatives. In Grounding: The Epistemic Footing of Deixis and Reference, Frank Brisard (ed.), 391–432. Berlin/New York: Mouton de Gruyter. Myhill, John 1995 Change and continuity in the functions of the American English modals. Linguistics 33: 157–211. Myhill, John 1997 Should and ought: the rise of individually oriented modality in American English. English Language and Linguistics 1: 3–23. Myhill, John and Laura A. Smith 1995 The discourse and interactive functions of obligation expressions. In Modality in Grammar and Discourse, Joan Bybee and Suzanne Fleischman (eds.), 239–292. Amsterdam/Philadelphia: Benjamins. Narrog, Heiko 1997 -mas.u: Die Grammatikalisierung von Höflichkeit. In Referate des 10. Deutschsprachigen Japanologentages vom 9. bis 12. Oktober 1996 in München, Ulrich Apel et al. (eds.), 688–696. München: iudicium. Narrog, Heiko 1999 Japanische Verbflexive und flektierbare Verbalsuffixe. Wiesbaden: Harrassowitz Verlag. Narrog, Heiko 2002 Polysemy and indeterminacy in modal markers – the case of Japanese beshi. Journal of East Asian Linguistics 11: 123–167. Narrog, Heiko 2005a On defining modality again. Language Sciences 27: 165–192. Narrog, Heiko 2005b Modality, mood, and change of modal meanings – a new perspective. Cognitive Linguistics 16: 677–731. Narrog, Heiko 2007 Modality and grammaticalization in Japanese. Journal of Historical Pragmatics 8: 269–294.

(Inter)subjectification in the domain of modality and mood

427

Narrog, Heiko to appear Voice and non-canonical marking in the expression of event-oriented modality – a cross-linguistic study. Linguistic Typology. Nuyts, Jan 2001 Subjectivity as an evidential dimension in epistemic modal expressions. Journal of Pragmatics 33: 383–400. Nuyts, Jan 2002 Grounding and the system of epistemic expressions in Dutch: A cognitive-functional view. In Grounding: The Epistemic Footing of Deixis and Reference, Frank Brisard (ed.), 433–466. Berlin/New York: Mouton de Gruyter. Oxford English Dictionary 1999 Oxford English Dictionary on CD-Rom. Version 2.0. Oxford University Press. Pakendorf, Brigitte and Ewa Schalley 2007 From possibility to prohibition: A rare grammaticalization path. Linguistic Typology 11: 515–540. Palmer, Frank Robert 1986 Mood and Modality. Cambridge: Cambridge University Press. Palmer, Frank Robert 2001 Mood and Modality. 2nd ed. Cambridge: Cambridge University Press. Pelyvás, Péter 2001a On the development of the category modal: a cognitive view. In Wort und (Kon)text, Piroska Kocsány and Anna Molnár (eds.), 103– 130. Frankfurt: Peter Lang. Pelyvás, Péter 2001b The development of the grounding predication: epistemic modals and cognitive predicates. In Pragmatics and the Flexibility of Word Meaning, Enikõ T. Németh and Károly Bibok (eds.), 151–174. Amsterdam: Elsevier. Pelyvás, Péter 2006 Subjectification in (expressions of) epistemic modality and the development of the grounding predication. In Subjectification: Various Paths to Subjectivity, Angeliki Athanasiadou, Costas Canakis and Bert Cornillie (eds.), 121–150. Berlin/New York: Mouton de Gruyter. Plank, Frans 1984 The modals story retold. Studies in Language 8: 305–364. Radden, Günter and René Dirven 2007 Cognitive English Grammar. Amsterdam/Philadelphia: Benjamins. Rijkhoff, Jan and Dik Bakker 1998 Language sampling. Linguistic Typology 2: 263–314.

428

Heiko Narrog

Ruhlen, Merritt 1987 A Guide to the World’s Languages. Volume 1: Classification. Stanford: Stanford University Press. Rumsey, Alan 2003 Language, desire and the ontogenesis of intersubjectivity. Language and Communication 23: 169–187. Shigeto, Minoru 2004 Modale Bedeutung des Futurs in der historischen Entwicklung. Neue Beiträge zur Germanistik 3: 74–82. Sweetser, Eve 1982 Root and epistemic modals: causality in two worlds. Proceedings of the eighth Annual Meeting of the Berkeley Linguistics Society, Monica Macaulay, Orin D. Gensler, et al. (eds.), 484–507. Berkeley, CA: Berkeley Linguistics Society. Sweetser, Eve 1990 From Etymology to Pragmatics: Metaphorical and Cultural Aspects of Semantic Structure. Cambridge: Cambridge University Press. Takanashi, Shino 1995 Suru-to ii to sureba ii to shitara ii. Jǀken setsuzoku keishiki ni yoru hyǀkateki fukugǀ hyǀgen 1 [Suru-to ii, sureba ii and shitara ii. Complex evaluative expressions through conditional forms 2]. In Nihongo Ruigi Hyǀgen no Bunpǀ (1), Tatsuo Miyajima and Yoshio Nitta (eds.), 235–243. Tǀkyǀ: Kuroshio Shuppan. Talmy, Leonard 1988 Force dynamics in language and cognition. Cognitive Science 12: 49–100. The Bank of English (5 million words). Contained in: Collins Cobuild English Dictionary for Advanced Learners. London: Harper Collins, 2001. Traugott, Elizabeth Closs 1989 On the rise of epistemic meanings in English: an example of subjectification in semantic change. Language 65: 31–55. Traugott, Elizabeth Closs 1995 Subjectification in grammaticalization. In Subjectivity and Subjectivisation: Linguistic Perspectives, Dieter Stein and Susan Wright (eds.), 31–54. Cambridge: Cambridge University Press. Traugott, Elizabeth Closs 2003 From subjectification to intersubjectification. In Motives for Language Change, Raymond Hickey (ed.), 124–139. Cambridge: Cambridge University Press. Traugott, Elizabeth Closs 2007 (Inter)subjectification and unidirectionality. Journal of Historical Pragmatics 8: 295–309.

(Inter)subjectification in the domain of modality and mood

429

Traugott, Elizabeth Closs This vol. From ideational to interpersonal. A reassessment. In this volume, 29–71. Traugott, Elizabeth Closs and Richard Dasher 2002 Regularity in Semantic Change. Cambridge: Cambridge University Press. van der Auwera, Johan and Vladimir A. Plungian 1998 Modality’s semantic map. Linguistic Typology 2: 79–124. van der Auwera, Johan, Andreas Ammann and Saskia Kindt 2005 Modal polyfunctionality and Standard Average European. In Modality. Studies in Form and Function, Alex Klinge and Henrik Høeg Müller (eds.), 247–268. London: Equinox. Vanderveken, Daniel 1990 Meaning and Speech Acts. Cambridge: Cambridge University Press. Van Herreweghe, Mieke 2001 *Motan in the Anglo-Saxon poetic records. Belgian Journal of Linguistics 14: 207–239. Visser, Frederic Th. 1969 An Historical Syntax of the English Language, Part III. Leiden: Brill. Willett, Thomas L. 1991 A Reference Grammar of Southeastern Tepehuan. Dallas: Summer Institute of Linguistics.

Contributors

Karin Aijmer Göteborg University [email protected] Anne Carlier University of Valenciennes [email protected] Alison Cort formerly University of Manchester Hubert Cuyckens University of Leuven [email protected] Kristin Davidse University of Leuven [email protected] Tine Defour Ghent University [email protected] Walter De Mulder University of Antwerp [email protected] David Denison University of Manchester [email protected] Teresa Fanego University of Santiago de Compostela [email protected]

432

Contributors

Lobke Ghesquière University of Leuven [email protected] Hélène Margerie University of Paris III-Sorbonne Nouvelle [email protected] Heiko Narrog Tohoku University [email protected] Scott A. Schwenter The Ohio State University [email protected] Anne-Marie Simon-Vandenbergen Ghent University [email protected] Elizabeth Closs Traugott Stanford University [email protected] Lieven Vandelanotte University of Namur and University of Leuven [email protected] Richard Waltereit Newcastle University [email protected] Anne Wichmann University of Central Lancashire [email protected]

Referees

The editors gratefully acknowledge the help of the following referees in evaluating papers submitted to this volume: Karin Aijmer Ronny Bogaart Wallace Chafe Alex D’Arcy Richard Dasher Hendrik De Smet Regine Eckardt Roberta Facchinetti Ed Finegan Juliane House Brian Joseph Andreas Jucker Alex Klinge Keith Mitchell Tanja Mortelmans Tarja Nikual Carita Paradis Günther Radden Helena Raumolin-Brunberg Anna-Brita Stenström Graeme Trousdale Mark Van de Velde Serge Van Volsem Nigel Vincent Fay Wouk

Author index

Aaron, Jessi, 56–57 Aarts, Bas, 117, 300 Abraham, Werner, 188 Adamson, Sylvia, 18–19, 60, 278, 281, 284, 286–288, 307–308 Aebischer, Paul, 244, 261 Aijmer, Karin, 7, 12, 104, 107, 113– 114, 123, 133, 155, 158–159, 161, 165, 188, 198–199, 209– 210, 315–316, 327–328, 403 Aissen, Judith, 208 Akatsuka, Noriko, 413 Alcaide Lara, Esperanza, 227 Allen, Keith, 32 Altenberg, Bengt, 8, 108 Andersen, Gisle, 316, 325 Andersen, Henning, 38 Anscombre, Jean-Claude, 86, 96, 99 Anttila, Raimo, 40, 54 Athanasiadou, Angeliki, 2, 3, 24 Atlas, Jay, 78 Auer, Peter, 158 Austin, John, 325 Bache, Carl, 278–281, 284, 310 Bäcklund, Ulf, 320 Bakker, Dik, 406 Banniard, Michel, 245, 261 Barrenechea, Ana, 214 Barth, Dagmar, 106 Bauer, Brigitte, 244, 261 Beaver, David, 78, 81 Benveniste, Emile, 2, 13, 29, 31–33, 174 Bergs, Alexander, 56, 70 Bertocchi, Alessandra, 257 Biber, Douglas, 115, 170, 173–174, 207, 230 Blakemore, Diane, 199–200, 215 Blyth, Carl, 328

Boas, Franz, 399 Bolinger, Dwight, 108, 214, 300 Bosk, Charles, 325 Bréal, Michel, 2, 29 Breban, Tine, 16, 60, 268, 278, 280, 284, 296–298, 304, 306, 310 Brems, Lieselotte, 13, 42 Brinton, Laurel, 35, 39, 58–60, 104, 158, 181, 198–199, 205, 227, 316 Brisard, Frank, 23 Brown, Penelope, 320, 410 Brutti, Silvia, 198, 215 Buechel, Eugene, 399, 405, 407 Bühler, Karl, 12 Burridge, Kate, 32, 420 Bybee, Joan, 19–20, 38, 41, 53, 56, 59–60, 149, 395–398, 401, 403– 405, 419 Calboli, Gualtiero, 248 Campbell, Lyle, 52 Canakis, Costas, 2–3, 24 Cano Aguilar, Rafael, 222 Carel, Marion, 99 Carlier, Anne, 12–13, 15–16, 18, 280 Caughley, Ross, 398 Chiang, Chien-Ling, 410–412 Claudi, Ulrike, 107, 198, 205 Coates, Jennifer, 369, 371, 411–412 Company Company, Concepción, 198 Cornillie, Bert, 2–3, 24, 34, 389–390 Cort, Alison, 12–13, 15, 356, 358, 380, 418 Couper-Kuhlen, Elizabeth, 106 Croft, William, 30, 56, 77, 295 Cruse, D. Alan, 295 Cruttenden, Alan, 109, 111, 141, 145 Crystal, David, 329

436

Author index

Culpeper, Jonathan, 157 Dahl, Östen, 41, 53 Danjou-Flaux, Nelly, 213 Dasher, Richard, 2, 4, 9, 11–12, 18, 20, 31–32, 35, 37–38, 54, 56, 59– 60, 75–76, 107, 155–156, 161– 162, 169, 183–185, 198–207, 210–211, 215, 217, 227–228, 230, 243–244, 264, 281, 286, 366, 374, 376–377, 386–387, 390–391, 399–400, 402 Davidse, Kristin, 13, 279–280, 296– 298, 304, 310 Davy, Derek, 329 De Mulder, Walter, 12–13, 15–16, 18, 258, 280 De Smet, Hendrik, 2, 8–10, 30, 35, 288, 386 Declerck, Renaat, 297 Defour, Tine, 12–14, 173 Deloria, Ella, 399 Denison, David, 12–13, 15, 42–43, 315, 331–334, 337–338, 341, 377, 418 Detges, Ulrich, 80 Diessel, Holger, 15, 263, 280, 286 Dietrich, Rainer, 392 Diewald, Gabriele, 106–107, 121, 223 Dirven, René, 405, 417 Dixon, Robert, 278, 282 Downing, Angela, 133 Ducrot, Oswald, 83, 86, 96, 99 Duffley, Patrick, 365 Dykes, Oswald, 360 Eberenz, Rolf, 222 Eckardt, Regine, 51, 52, 79 Enfield, Nicholas, 16, 46 Epstein, Richard, 249–253, 255 Erman, Britt, 105 Espinosa Elorza, Rosa, 94 Evans, Vyvyan, 392 Everett, Daniel, 407

Fanego, Teresa, 7, 12, 17–18, 198 Fernández-Ordóñez, Inés, 217 Ferrara, Kathleen, 108, 209, 210 Finegan, Edward, 1, 4, 207 Finell, Anne, 160, 162–165 Fischer, Kerstin, 104, 199 Fischer, Olga, 220, 365 Fisher, John, 217 Fitzmaurice, Susan, 33, 36, 174 Fleischman, Suzanne, 199, 343 Foolen, Ad, 107, 158 Frader, Joel, 325 Francis, W. Nelson, 157 Fraser, Bruce, 158, 199, 207, 225 Fritz, Gerd, 400 Fruyt, Michèle, 248, 259 Fuentes Rodríguez, Catalina, 212– 214, 227 Gamon, David, 400–401 Garachana Camarero, Mar, 198 Garrod, Simon, 52 Ghesquière, Lobke, 12–13, 15–16, 18–19 Givón, Talmy, 78, 370, 418 Goossens, Louis, 390, 400 Green, Melanie, 392 Greenberg, Joseph, 248 Grice, Paul, 30, 41, 78, 82, 133 Gries, Stefan, 61 Grimm, Jacob, 415 Grimm, Wilhelm, 415 Günthner, Susanne, 106 Gutiérrez-Rexach, Javier, 52 Hall, Fitzedward, 351 Halliday, M. A. K., 10, 17, 31, 34, 60, 115, 121–122, 201, 278, 281, 285–287 Hansen, Maj-Britt Mosegaard, 77, 83, 188, 198, 230 Harris, Alice, 52 Harris, Martin, 241–244, 246–247, 264 Hasan, Ruqaiya, 17, 31, 34, 60, 115,

Author index 201, 278, 285–286 Haspelmath, Martin, 55, 76, 198 Hawkins, John, 247, 297 Heim, Irene, 78 Heine, Bernd, 30, 40, 46, 88, 107, 198, 205, 220, 223, 229, 386, 393, 403 Herman, József, 249, 259 Herzog, Marvin, 45 Hetzron, Robert, 278, 281 Hilpert, Martin, 61 Himmelmann, Nikolaus, 45, 248– 249, 263–264 Hirschberg, Julia, 108, 111 Hoeksema, Jack, 52, 59 Holmes, Janet, 14, 108, 113–115, 119, 127, 133, 136 Hopper, Paul, 11, 30, 38–39, 53, 56, 58, 123, 162, 250, 291, 302, 337, 386, 391 Horie, Karuo, 57 Horn, Laurence, 30 Horne, Merle, 108 Hoye, Leo, 158, 180, 184–186, 188 Huang, Yan, 373 Huddleston, Rodney, 363 Hünnemeyer, Friederike, 107, 198, 205 Iordanskaja, Lidija, 198 Israel, Michael, 50, 51 Jackendoff, Ray, 360 Jacobs, Andreas, 157 Jespersen, Otto, 52, 375, 393 Johansson, Stig, 188 Johnson, Mark, 388, 405 Jucker, Andreas, 37, 156–160, 162– 165, 167–168, 184, 198 Kabatek, Johannes, 217 Kärkkäinen, Elise, 143 Keller, Rudi, 55, 76 Kelly, James, 360–361

437

Kern, Barbara, 407 Kiefer, Ferenc, 392 Kiparsky, Paul, 53, 208 Kleiber, Georges, 249, 258 Kondo, Emi, 57 König, Ekkehard, 11, 14, 51, 250, 316, 409 Kotsinas, Ulla-Britt, 105 Kranich, Svenja, 58 Krifka, Manfred, 51 Kripke, Saul, 81 Krug, Manfred, 11 Kuteva, Tania, 40 Kytö, Merja, 157 Labov, William, 45 Lakoff, George, 2 Lakoff, Robin, 158 Lambrecht, Knud, 78 Langacker, Ronald, 2–3, 9, 16, 19, 29, 36, 58, 279, 283, 297, 385, 387–392, 401, 403, 417, 420 Lange, Deborah, 328 Lehmann, Christian, 41, 105, 205– 206, 262, 342 Lenker, Ursula, 227 Levinson, Stephen, 30, 78, 229, 320, 373–374, 410 Lewis, David, 77 Lewis, Diana, 115–117 Lightfoot, David, 55, 205 Lima, J. Pinto de, 106 Litman, Diane, 108 Lorenz, Gunter, 13 Los, Bettelou, 198 Ludovici, Christian, 360 Lyons, Christopher, 247, 249 Lyons, John, 2, 10, 12, 32–33, 392 Macaulay, Ronald, 205 MacIntosh, Donald, 360 Malamud-Makowski, Monica, 199, 225 Margerie, Hélène, 12–15, 320, 322, 330

438

Author index

Martín Zorraquino, María, 201, 211, 214 Matthiessen, Christian, 121, 122 McGregor, William, 5, 304–305 Meillet, Antoine, 53, 261 Mel’þuk, Igor, 198 Menge, Hermann, 244 Meyer, Charles, 108 Milroy, Lesley, 56 Mindt, Dieter, 418–419 Mitchell, Keith, 350, 355, 363, 365– 366, 368–372, 375–376 Mortelmans, Tanja, 389 Moxey, Linda, 56 Mulac, Anthony, 11 Myhill, John, 407–409, 411, 413– 414, 420 Narrog, Heiko, 3, 12, 19–21, 386– 387, 392–395, 397, 401–402, 406 Nelson, Gerald, 117 Nevalainen, Terttu, 282 Nuyts, Jan, 2, 32, 34, 389–390 Oh, Sun-Young, 198, 209 Onodera, Noriko, 60, 198 Orlandini, Anna, 248 Östman, Jan-Ola, 178 Pagliuca, William, 38, 60 Palander-Collin, Mina, 157, 160, 181 Palmer, Frank, 370, 392, 409 Paradis, Carita, 282, 320, 330 Pelyvás, Péter, 390 Perkins, Revere, 38, 60 Pierrehumbert, Janet, 111 Pinker, Steven, 360 Pinkster, Harm, 244, 257 Plank, Frans, 407 Plungian, Vladimir, 395, 401 Pons Bordería, Salvador, 198 Portolés Lázaro, José, 201, 211, 214 Powell, Mava, 11, 155 Prince, Ellen, 30, 325

Pullum, Geoffrey, 363 Quirk, Randolph, 108, 110, 112– 113, 176, 179, 185, 278, 281– 282, 318, 350 Radden, Günter, 405, 417 Ray, John, 360–361 Recktenwald, Sigfrid, 328 Reh, Mechthild, 386 Renzi, Lorenzi, 245, 248 Rijkhoff, Jan, 406 Rissanen, Matti, 282 Roberts, Craige, 88 Romaine, Suzanne, 328 Rossari, Corinne, 198, 212, 230 Roulet, Eddy, 198 Ruhlen, Merritt, 406 Ruiz Gurillo, Leonor, 198 Sacks, Harvey, 142 Sanford, Anthony, 56 Scheibman, Joanne, 56, 58, 172–174 Schiffrin, Deborah, 14, 29–30, 32, 155, 158–160, 199, 335 Schourup, Lawrence, 156, 158, 160, 162, 166, 199 Schwenter, Scott, 7, 12–13, 17, 52, 57, 81–83, 85–86, 161, 198, 202, 204, 206–207, 209, 212–213, 220 Selig, Maria, 244, 248–249, 257, 259, 261, 263 Shibasaki, Reijirou, 185 Shigeto, Minoru, 404 Siemund, Peter, 409 Silverstein, Michael, 366 Simon-Vandenbergen, Anne-Marie, 7, 12, 104, 107, 113–114, 133, 158–159, 165, 188, 198, 209–210 Sinclair, John, 282–283, 299 Smith, Laura, 407–409, 411, 413– 414 Smith, Sara, 198 Speake, Jennifer, 360 Sperber, Dan, 201

Author index Stefanowitsch, Anatol, 61, 291 Stein, Dieter, 2, 21 Stenström, Anna-Brita, 108–109 Stivers, Tanya, 16 Stoffel, Cornelis, 48 Suzuki, Ryoko, 41, 60 Svartvik, Jan, 158 Sweetser, Eve, 388, 405, 417 Tabor, Whitney, 41, 198, 200, 202, 205–208, 210 Takanashi, Shino, 417 Talmy, Leonard, 388, 417 Thompson, Geoff, 119, 121 Thompson, Sandra, 11, 58 Torres Cacoullos, Rena, 57, 220 Trager, George, 252, 255 Traugott, Elizabeth, 2–20, 24, 29– 33, 35, 37–39, 41–42, 53–54, 56, 59–60, 75–76, 84, 104–107, 117, 123, 155–156, 160–164, 169, 174, 183–185, 198–212, 215, 217, 220, 227–230, 243–244, 250, 264–269, 278, 280–281, 284–288, 291, 302, 308–309, 315–316, 321, 324, 338, 341– 342, 366, 374, 376–377, 379, 385–387, 390–392, 394, 399– 402, 420–421

439

304–305 Vanderveken, Daniel, 413 Vandewinkel, Sigi, 13 Verhagen, Arie, 2, 32–33 Verstraete, Jean-Christophe, 2, 8–10, 30, 35, 386 Visser, Frederic, 353–355, 362, 397– 398, 403–404, 415–416 von Fintel, Kai, 77, 79 Wallis, Sean, 117 Waltereit, Richard, 12–13, 17, 80, 188 Wang, Jenny, 328 Wasow, Thomas, 56 Watts, Richard, 158 Weinreich, Uriel, 45 Wennerstrom, Ann, 108 Westney, Paul, 365–366, 369–370, 372, 378 Weydt, Harald, 188 White, Peter, 107, 114–115, 131– 133 Wichmann, Anne, 7, 12–14, 110– 111, 210 Wierzbicka, Anna, 33 Wilson, Deirdre, 201 Wright, Susan, 2 Yaguello, Marina, 199, 343

Underhill, Robert, 328 van der Auwera, Johan, 395, 401 van der Gaaf, Willem, 353, 355 Van Herreweghe, Mieke, 164, 400 Van linden, An, 280, 296–298, 310 Vandelanotte, Lieven, 11, 278, 287,

Zeevat, Henk, 78, 81 Zubin, David, 40, 58 Zulaica-Hernández, Iker, 81–82, 85– 86 Zwicky, Arnold, 199

Subject index

a pesar de. See pesar de ability. See under modality actually, 203–205 additive particle, 80, 83–84, 86, 92, 94, 96, 204 adjectives. See also classifier; complete; emphasizer; descriptive modifier; total; whole of completeness, 277–310 adverbials. See also conjunct; disjunct sentence adverbials, 200, 205– 210, 222–225, 353 VP adverbials of manner, 216– 222 adversativity, 7, 95, 97, 99, 107, 120, 133, 202, 204, 206, 208–209, 213–214, 223–225, 227–229 advisability, 350, 366–367, 375, 379 analogy, 48, 52–54, 302, 353–354 anaphora, 16, 87–88, 95–96, 98, 121, 242, 244–246, 249–250, 252, 256–262 anche, 81 animacy, 355–356, 366, 379 anyway, 200, 209–210 approximation, 318–319, 325–326, 331–333, 335–337, 341 argumentative contexts, 86–99, 177– 179 auch, 80, 83, 98–99 be to, 418–419 beki, 411, 418 besi, 402, 405, 409 better, 15, 349–380, 418. See also lever; rather bit of, 4–6, 46–49, 51, 53–54 bleaching. See semantic bleaching blend, 300–301, 351–353, 362–363

booster. See under intensifier bridging contexts, 17, 46–47, 88–89, 93–95, 99, 202 Catalan, 242–244 categorization zone in the NP, 283– 284, 310 Chinese, 410 Classical Latin. See under Latin classifier, 283–284, 287, 306, 310 clipped constructions, 358–360, 362 coding, 4–8, 35 Cognitive Grammar/Linguistics, 2, 29, 278, 385, 388–389, 417 cognitive verbs. See mental verbs collocation, 13, 45, 56, 61, 117, 133, 135, 156, 169–185, 212, 217, 292, 294, 296–297, 300, 302– 303, 322 Common Ground, 77–79, 81, 159, 165, 178–179, 189 comparison, 355, 360, 368, 379 complete, 18, 288–309 compromiser. See under intensifier confirmatory function, 7, 214, 220– 229 conjunct (adverbial), 108–113 conjunct (element in conjunction), 81–82, 86, 92, 94, 96, 98–99 constructions, 4–8, 61 Conversation Analysis, 142 conversational implicature, 17, 76– 77, 229, 373–374 conversational maxims. See maxims, conversational counter-expectation, 57–58 counterfactuality, 354–355, 363, 367, 370, 375–377, 378–379 culture. See language and culture

442

Subject index

’d better. See better Dakota, 399 Danish, 83–84 de fait, 198, 212 de hecho, 7, 198, 210–230 decategorialization, 105, 206, 229, 291, 337 definite article, 15–16, 241–269, 297 degree modifier, 42–44, 48, 51, 53, 338, 341 deictification, 298 deixis, 12, 242–243, 247, 250–254, 258, 262, 265, 268, 279–281, 296, 308–310 demonstrative, 16, 121, 241–269. See also anaphora; first mention use deontic meaning, 366–368. See also modality, deontic deontic modality. See under modality description zone in the NP, 284, 310 descriptive modifier, 281 objective descriptive modifier, 281 subjective descriptive modifier, 281, 294–296 determination zone in the NP, 279– 281, 284, 310. See also determiner; predeterminer determiner, 15–16, 279–281, 306– 310, 338. See also predeterminer determiner patterns, 317, 331– 341 primary determiner, 279–281 secondary determiner, 279–281, 296–299, 303, 306–310 diachronic pathways. See pathways of change dialogic/dialogual situation, 84, 92– 94, 99, 131, 144, 146, 148 diminisher. See under intensifier discourse markers, 6–8, 11, 37, 104– 112, 115, 119, 121–123, 127,

138–140, 142–143, 149–150, 197–210, 225–230, 377–378. See also de hecho; of course as a subset of pragmatic markers, 158, 199, 316 elaborative discourse markers, 203–205, 225–230 discourse status, 130, 144 disjunct (adverbial), 108–113 dummy NP, 355–356, 379 Dutch, 33, 59, 389 efectivamente, 227 elaborative discourse markers. See under discourse markers ellipsis, 360–362 emphasizer, 112, 282–284, 299–305, 307–310 wide scope emphasizer, 304–305 en fait, 198, 212–213 encore, 83 English. See also modal verbs in English Middle English, 37, 42–43, 49, 163, 167–168, 203, 217, 288, 290, 292–294, 297, 303–304, 353–354, 364, 390, 415 Modern English, 37, 163, 288, 290–292, 299, 302, 304, 354, 390, 415 Old English, 37, 42, 46, 49, 164, 184–185, 227, 288, 291, 293, 352, 354, 364, 390, 400–401, 403, 415–416, 419 Present-Day English, 42, 288, 293–294, 299, 305, 350, 354, 412 epistemic meaning, 2, 4, 8, 11, 14, 32, 34, 111, 115, 118, 124, 128, 138, 142–143, 161, 166–169, 177, 184–185, 188–189, 214, 369–371. See also modality, epistemic epistemic modality. See under modality

Subject index epistemic parentheticals. See parentheticals event-oriented modality. See under modality evidential meaning, 8, 111, 118, 124, 128, 138 expressive meaning, 17–19, 31, 156, 201, 251, 253, 255, 285, 306, 308–310, 316, 324–331. See also interpersonal meaning extreme subjectification. See under subjectification face (threats), 14–15, 20, 30, 33, 49, 52, 159, 161, 163, 166, 176–179, 280, 315, 320–321, 409–410, 412–413, 420 filler, 15, 113, 128, 159–160, 164, 329. See also hesitation marker first mention use, 16, 249, 259, 262 focus marker, 327–328 force dynamics, 57, 388, 417, 419 French, 16, 52, 83, 198, 212–213, 244–248, 254, 261, 263, 268, 290, 292, 343 future, 20, 36, 375, 388, 395–399, 402–404, 407, 415–416 German, 20, 40, 80, 83, 98–99, 188, 388–389, 395, 400–401, 404, 419 Middle High German, 400, 404 New High German, 404 Old High German, 400–401 grammaticality, 38–39 grammaticalization, 6, 30, 38–42, 45, 48, 50–56, 61, 104–107, 113, 117, 138, 142–143, 148, 156, 160–161, 181, 189, 197–199, 205–210, 228–229, 250–251, 253, 261–262, 265–269, 287, 291, 315–318, 331–343, 354, 372, 374, 380, 386–387, 401– 402, 405–406, 409, 420 grounding predication, 388–390

443

had better. See better hearer-based language change, 75– 80 Hebrew, 407 hedge, 49, 128, 203, 228, 326–327, 329–330, 341–343 hedges, 6, 14–15, 37 heller ikke, 83–84 hesitation marker, 128, 329, 335– 336, 341. See also filler heteroglossia, 107, 114, 131–136, 144–147 hic, 241–245, 257, 262 honorific, 33, 37–38, 268, 387 hope, 369–370, 374, 376 Hopi, 407 host-class expansion, 45, 48, 50 Hungarian, 40 ideational meaning, 10, 17, 31, 61, 84, 99, 121, 161, 201, 243, 285– 286, 386. See also propositional meaning idiom, 185, 350, 366, 373 ille, 15–16, 241–269 illocutionary force, 20, 395–396, 398, 420 immediacy, 365 imperative, 15, 20, 59, 372, 396– 399, 401–402 implicature. See conversational implicature in fact, 198, 203–205, 208–210, 215 indeed, 161, 200, 203–205, 208–209, 215, 217, 227 infatti, 198, 212, 215 intensifier, 281–282, 284, 287, 305, 307–308, 310, 318–324, 329– 331, 338, 341–342 booster, 318, 322–323 compromiser, 318–322 diminisher, 318, 323–324 interpersonal meaning, 10, 17, 31, 60, 84, 98–99, 111, 118, 121,

444

Subject index

123, 126, 128–129, 143, 159– 169, 201, 243, 285–286, 305, 324, 368, 379–380, 386. See also expressive meaning intersubjectification, 2–6, 13–18, 30, 35–38, 41–42, 51–52, 54, 59–61, 84, 107, 125, 127, 148, 156, 161, 243–244, 256–269, 280–281, 286, 306–310, 315–316, 321, 342, 379, 386–387, 402, 420 textual intersubjectification. See textual meaning intersubjectivity, 4, 13–17, 29–34, 84, 125, 127, 148, 167–169, 189, 202–203, 243, 266, 280–281, 286, 306–310, 315, 320–321, 324, 326, 329, 342, 371, 378– 379, 394 textual intersubjectivity. See textual meaning intonational meaning, 111–112, 149 invited inferences, 5, 14, 17, 32, 51, 54–55, 61, 201, 367, 374, 379 ipse, 241–245, 253–262, 265–266, 268–269 is, 242, 244–245 iste, 241–245, 257, 262 Italian, 81, 198, 212, 215, 244 Japanese, 6, 33, 37–38, 41, 54, 57, 60, 202, 387, 395–396, 398, 402, 405, 409–413, 417–418, 420 Jespersen Cycle, 52 joint attention, 16, 280, 286 kind of/kinda, 15, 315–343 Lakota, 399, 405 language and culture, 408, 413–414, 420–421 Latin Classical Latin, 16, 241–245, 256–257, 265–266 Late Latin, 16, 241–269

layering, 38, 123, 128, 137, 162, 167, 183, 354 leftward hypothesis, 18–19, 60, 105, 220, 229, 278, 287–288, 306–310 legal prose, 217–218, 220, 222, 224, 229 let’s, 267 letters (as a genre), 157 lever, 351–354 lexical sources. See necessity, deontic under modality lexicalization, 39, 287 magan, 3 maxims, conversational, 55, 76 of Quantity, 82, 133 may, 419 mechanisms of change, 52–54. See also analogy; metaphor; metonymy; reanalysis mental verbs, 169–183 metaphor, 54, 107 metonymy, conceptual, 54 Middle English. See under English Middle High German. See under German modal idiom, 350 modal particles, 158, 188 modal verbs in English, 2–3, 15, 19, 37, 183–185, 349–350, 356–360, 374, 376–377, 380, 387–392. See also better; may; must; shall; should; will modality. 15, 19–20, 349–380, 385– 421, See also advisability; hope; preference; threat ability, 20, 390, 394, 396, 402– 408 deontic modality, 364–380, 387– 421 epistemic modality, 364–380, 386–404, 408–414, 420 event vs. speaker-oriented, 20, 394–398, 406–408. See also speaker-orientation

Subject index necessity, deontic, 350, 391, 399, 401, 404–419 lexical sources of, 416–419 necessity, epistemic, 408–414 obligation, 15, 20, 350, 370–371, 376, 379, 388, 390–391, 396, 398–402, 404–420 permission, 19, 20, 372, 388, 390, 394, 396–397, 400 possibility, deontic, 394 possibility, epistemic, 19, 20, 185, 396, 398, 402, 418 possibility, root, 20, 396–398, 402, 420 possibility, situational, 393–394, 404–408 prediction, 20, 125, 395–398, 403, 419 probability, 20, 114, 185–186, 396, 398, 403–404, 418 semantic map of, 19, 394–396 volitive/non-volitive, 19–20, 37, 350, 370, 376, 393–404 Modern English. See under English modification zone in the NP, 281– 282, 284, 310 modifier. See descriptive modifier mood, 19–20, 59–60, 387, 392–399, 401–402, 407, 420 motivations for change, 54–56, 75– 80, 161, 205, 420–421. See also hearer-based language change müssen, 400–401 must, 390–391, 399–402, 404–405, 407–412, 419–420 necessity. See under modality New High German. See under German Norwegian, 188 noun phrase, functional zones within the, 278–284, 310. See also categorization zone; descriptive zone; determination zone;

445

modification zone; strengthening zone objectivity, 2–3, 18–19, 34, 117, 125, 128–130, 278–281, 284– 286, 291–294, 299–301, 306– 310, 385, 388 obligation. See under modality of course, 7–8, 14, 112–150 også, 83–84 Old English. See under English Old High German. See under German ought to, 390–391, 399 paradigmatization, 262 parentheticals, 35, 115, 122, 181– 183, 189 partitive constructions, 42–44, 46– 47, 49–50, 54–55, 341 pathways of change, 17–21, 35–38, 84–99, 105, 161, 164, 201–205, 228–229, 268, 284–288, 299– 304, 306–310, 331–343, 379– 380, 414–416, 420 pause filler. See filler performative verbs, 413 perhaps, 208 permission. See under modality person, 366, 399, 401–402, 410–413 pesar de, 57 piece of, 43–45, 55 polarity sensitivity, 42, 50–52, 55– 59, 80, 83 politeness, 32, 37, 41, 61, 107, 114, 157, 177, 188, 268, 320, 327, 387, 410. See also face (threats); hedges polyfunctionality, 402, 404–405, 408–414 polysemy, 7, 32, 34, 46, 48, 50, 92, 123, 128, 149, 189, 209, 218, 387, 405 Portuguese, 84, 242–243

446

Subject index

possibility. See under modality potency, 388–389, 402 pragmatic markers, 113, 155–158, 169, 185, 188, 207, 211, 315– 316, 318, 321, 325–331, 341, 343. See also kind of/kinda; well vs. discourse markers, 158, 199 pragmatic particle. See pragmatic marker pragmatic strengthening, 5, 54, 105, 250, 255, 259, 342, 400 pragmaticalization, 104–106 pragmatics (vs. semantics), 4–6, 9– 10, 17, 30, 32, 35, 42, 54, 76, 79, 123–130, 364, 373–374, 379, 386–387, 391–392, 397–398, 401, 410–411 predeterminer, 279, 284, 303, 310 predicative alternation, 282, 294 prediction. See under modality preference, 350, 374, 376 Present-Day English. See under English present-oriented use of modals, 388– 389, 403. See also extreme subjectification under subjectification presupposition accommodation, 17, 77–78, 80, 82, 84, 88, 91–92, 96, 99 primary determiner. See under determiner probability. See under modality procedural meaning, 18, 31, 199– 203, 207, 225, 228–229, 243, 280, 316, 329 pronoun subjects, 356–358, 378 propositional content, 82, 88, 98, 107, 109, 111, 114, 158, 188, 378 propositional meaning, 10, 16–19, 31, 111–112, 123, 128, 156, 159– 169, 201, 285, 308–310, 315– 318, 324, 329–330, 342. See also ideational meaning

prosodic prominence, 8, 111–112, 137–142, 149 prosody, 7–8, 108–115, 137–148, 210 proverbs, 360–363, 368, 378 pseudo-conditional, 372 quantifier, 42–47, 50–51, 54–55, 341 rather, 351–354 reanalysis, 4–6, 29, 41, 50, 52–54, 60, 80, 207, 243, 302, 337 reported speech or thought, 328 salirse, 56–57 Sardinian, 244 scalarity, 5, 11–12, 45, 48–52, 54, 282 scope, reduction vs. increase in, 41, 105–106, 110, 112, 124, 156, 160, 182, 185, 189, 202–204, 206, 220, 222–223, 229, 282, 304–305, 319, 342 sculan, 2 secondary determiner. See under determiner semantic bleaching, 111, 139, 246– 250, 261, 281–282 semantic field types. See mental verbs semantic map. See under modality semantic prosody, 13, 302 semantic weakening. See semantic bleaching semanticization, 4–6, 10, 29–30, 54, 397 semantics (vs. pragmatics). See pragmatics (vs. semantics) sentence adverbials. See under adverbials shall, 415–416, 419 should, 407, 411, 415–416, 419 shred of, 49–51, 54, 58–59 sinceramente, 200–201

Subject index so, 200 Spanish, 7, 56–57, 80, 83, 85, 94–97, 198, 210–230, 242–243, 389, 404 speaker-hearer interaction, 420 speaker-orientation, 19, 386–387, 394–404, 420. See also modality, event vs. speaker-oriented vs. hearer-orientation, 16, 52, 156, 251, 253 specificity, 365 stance, 8, 11, 107, 109, 111–112, 131, 136–137, 147–148, 165, 188, 200, 286 strengthening zone in the NP, 284, 310 subjectification, 2–6, 8–13, 18, 30, 35–42, 49, 51–52, 54–61, 84, 105, 107, 125, 148, 156, 161– 163, 183, 189, 199–200, 220, 243–244, 250–256, 266–269, 280, 284–288, 306–310, 316, 342, 385–392, 399–404, 420 extreme subjectification, 388– 389, 403 subjectivity, 1–4, 8–13, 29–34, 84, 125, 148, 156, 161, 167–169, 172, 174, 189, 202–203, 243, 253, 280, 284–288, 306–310, 315, 320, 324, 329, 331, 338, 342, 378–379, 385–394 Swedish, 188 también, 80

447

tampoco, 80, 83, 85, 94–97 textual meaning, 16–19, 31, 61, 84, 99, 107, 121, 123, 143, 163–165, 167, 201, 244, 248, 259, 268, 280–281, 285–286, 306–310 threat, 371–373 too, 17, 80–83, 85–94 refutational use of, 88–94 TOO, 80–99. See also additive particle; anche; auch; encore; heller ikke; også; también; tampoco; too topicality, 17, 84–88, 93, 96 total, 18, 288–309 transitivity, 58 unidirectionality, 17–21, 161, 201– 205, 228–229, 284–288, 306– 310, 380 uns, 254 volition. See volitive/non-volitive under modality Vulgar Latin. See Late Latin under Latin Wari’, 407 well, 14–15, 158–190 werden, 404 whole, 18–19, 288–309 wide scope emphasizer. See under emphasizer will, 403