Evidentiality in German: Linguistic Realization and Regularities in Grammaticalization 9783110241037, 9783110240696

Citation preview

Evidentiality in German

Trends in Linguistics Studies and Monographs 228

Editor

Volker Gast Advisory Editors

Walter Bisang Hans Henrich Hock Matthias Schlesewsky Founding Editor

Werner Winter

De Gruyter Mouton

Evidentiality in German Linguistic Realization and Regularities in Grammaticalization

by

Gabriele Diewald Elena Smirnova

De Gruyter Mouton

ISBN 978-3-11-024069-6 e-ISBN 978-3-11-024103-7 ISSN 1861-4302 Library of Congress Cataloging-in-Publication Data Diewald, Gabriele. Evidentiality in German : linguistic realization and regularities in grammaticalization / by Gabriele Diewald, Elena Smirnova. p. cm. ⫺ (Trends in linguistics studies and monographs; 228) Includes bibliographical references and index. ISBN 978-3-11-024069-6 (alk. paper) 1. German language ⫺ Grammaticalization. 2. Grammar, Comparative and general ⫺ Grammaticalization. I. Smirnova, Elena. II. Title. P299.G73D54 2010 435⫺dc22 2010022937

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, Berlin/New York Printing: Hubert & Co. GmbH & Co. KG, Göttingen ⬁ Printed on acid-free paper Printed in Germany www.degruyter.com

Table of contents

Chapter 1. Introduction 1. A first approach to evidential meaning 1.1. Basic notions in the description of evidentials 1.2. Deixis 2. The discovery of evidential meaning and evidentials in (German) linguistics 2.1. The status of evidential meaning in present-day German 2.2. The diachronic development of evidentials in German 3. Hypotheses and theoretical foundations 4. Methodology and corpora 4.1. The data sources of the present study 4.2. Principles and method of the corpus study 5. Structure of the book

1 2 2 9 15 16 20 23 27 28 33 37

Chapter 2. Evidentiality – definitions and delimitations 1. Evidentials, evidential strategies and pre-evidentials 1.1. Terminological distinctions in synchronic perspective 1.2. Terminological distinctions in diachronic perspective 1.3. The dispute on the status of evidential expressions in European languages 2. Intra-evidential distinctions 2.1. Direct and indirect evidentials 2.2. Sub-classifying the indirect branch 2.3. Distinguishing quotatives from mediated evidentials 3. Summary: classification of evidential distinctions

40 40 41 46

Chapter 3. Evidentiality and modality – drawing the lines 1. Prior suggestions on the relation between evidentials and modality 2. The common deictic nature and the different realizations of source concepts 3. Diachronic explication of distinctions between evidentials and epistemics 4. Typological issues

75

50 52 54 59 66 73

76 81 88 90

vi Table of contents 5.

Areas of overlap between evidentiality and epistemic modality

91

Chapter 4. Grammar and grammaticalization 1. Obligatoriness and paradigmatic organization 2. Grammaticalization parameters and German evidentials 2.1. Lehmann’s grammaticalization parameters 2.2. Degree of grammaticalization of evidential drohen – a detailed study 2.3. The degrees of grammaticalization of versprechen, scheinen and werden 3. Summary

97 97 105 105

Chapter 5. Scales and scenarios of grammaticalization 1. Regularities in semantic and functional changes 1.1. Metaphorization and metonymization scales 1.2. Scheme retention and reinterpretation 1.3. Successive semantic-functional stages 2. Dealing with ambiguity and context dependence 3. Stages, contexts and constructions 4. Summary

124 124 124 128 132 136 143 158

Chapter 6. The four evidential constructions in present-day German 1. Werden & infinitive 1.1. Relevant construction types with werden in present-day German 1.2. Semantic diversity of werden & infinitive and its core semantics 1.3. Temporal reference or modal evaluation: the core semantics of werden & infinitive 1.4. What is evidential about the core semantics of werden & infinitive? 2. Scheinen & zu-infinitive 2.1. Relevant construction types with scheinen in present-day German 2.2. Semantic diversity of scheinen & zu-infinitive 2.3. Lexical versus evidential scheinen: some (more) crucial differences

108 116 122

159 159 159 163 171 173 177 177 180 185

Table of contents vii

2.4. 3. 3.1. 3.2. 3.3. 4. 4.1. 4.2. 4.3.

Two types of evidential scheinen and their conceptual relationship Drohen & zu-infinitive Relevant construction types of drohen in present-day German Lexical versus evidential drohen The perceptual inferential value of evidential drohen Versprechen & zu-infinitive Relevant construction types of versprechen in present-day German Lexical versus evidential versprechen Drohen & zu-infinitive versus versprechen & zu-infinitive

189 191 192 196 200 205 206 210 212

Chapter 7. Intermediate summary 1. The inferential evidential paradigm in present-day German 1.1. Formal characteristics of a paradigm 1.2. Conceptual structure of the paradigm 2. German evidential system in a typological perspective

218 218 219 221 225

Chapter 8. Diachronic corpus study 1. Werden & infinitive 1.1. Previous work on the development of werden & infinitive 1.2. Werden in the history of German 1.3. Werden in Old High German 1.4. Werden in Middle High German 1.5. Werden in Early New High German 2. Scheinen & zu-infinitive 2.1. Scheinen in the history of German 2.2. Scheinen in Old High German 2.3. Scheinen in Middle High German 2.4. Scheinen in Early New High German 2.5. Scheinen in the 18th century 3. Drohen & zu-infinitive 3.1. The development of drohen & zu-infinitive: a short survey 3.2. Drohen in Old and Middle High German 3.3. Drohen in Early New High German 3.4. Drohen in the 18th and 19th centuries 4. Versprechen & zu-infinitive 4.1. Versprechen in the history of German and its development up to Early New High German

229 231 232 237 240 243 247 251 252 254 255 257 263 268 268 270 274 276 284 286

viii Table of contents 4.2. 4.3.

Versprechen in the 18th century Versprechen in the 19th century

289 293

Chapter 9. Summary: the diachrony of the four constructions 1. Relevant semantic and functional changes 1.1. Metaphorization and metonymization scales 1.2. Scheme retention and reinterpretation 1.3. Successive semantic-functional stages 2. Stages, contexts and constructions 3. Common grammaticalization path

299 299 299 303 308 311 320

Chapter 10. Summary and outlook

324

Notes References Subject Index

336 348 364

Chapter 1 Introduction1

This book develops a model for the synchronic analysis of evidential markers in present-day German (henceforth: PDG). The study is backed up by a corpus-based investigation of the relevant diachronic developments, using data from Old High German to PDG. It focuses on the four evidential constructions scheinen ‘seem’, drohen ‘threaten’, versprechen ‘promise’ & zu-infinitive and werden ‘become’ & infinitive. Our leading hypothesis is that the development of evidential verbal constructions in German constitutes a grammaticalization process in the course of which evidential distinctions are integrated as grammatical distinctions into the verbal system of German. It is assumed that the category of verbal mood, which is realized by inflectional as well as periphrastic means, is closely related to, but distinct from, the newly developing evidential category. This leads us to the idea that with the rise of the four constructions under analysis to grammatical markers of evidentiality we are witnessing the genesis of a new grammatical category in the German language. Our theoretical background and approach is based on grammaticalization theory, functional theories on language change, construction grammar, and typological research. The research is backed up by extensive empirical work with diachronic as well as modern German data. This first chapter gives a survey of the basic assumptions guiding our endeavour and sketches our central questions and hypotheses. Furthermore, it provides a short review of earlier research on evidentiality and evidential markers and describes the corpora used in the empirical section and the methods applied in dealing with the data. As is inevitable in any overview, many details are omitted, arguments are abbreviated, and background information is heavily condensed. However, all the issues treated in this chapter will be clarified step by step in the following chapters.

2 Introduction 1. A first approach to evidential meaning 1.1. Basic notions in the description of evidentials Evidentiality is a semantic-functional domain having to do with the kinds of “proof” speakers are able to adduce in order to underpin their statements. Thus the study of evidentiality in language is concerned with the linguistic means languages provide for referring to “reasons”, “indications”, “evidence”, or – in short – sources of information speakers have for expressing statements. As can be seen in the following definitions, the coding of an information source is considered the basic meaning of evidentiality, cf.: Evidentials may be generally defined as markers that indicate something about the source of the information in the proposition. (Bybee 1985: 184) Evidentials express the kinds of evidence a person has for making factual claims. (Anderson 1986: 273) Evidentiality proper is understood as stating the existence of a source of evidence for some information; this includes stating that there is some evidence, and also specifying what type of evidence there is. (Aikhenvald 2003: 1)

As Anderson (1986: 274) notes, evidentials encode the kind of justification for a factual claim which is available to the person making that claim. This justification may be derived from direct evidence plus observation, evidence plus inference, inference (where evidence remains unspecified), reasoned expectation from logic and other facts, and it may rely on different perceptual sources, e.g. on auditory, visual or other perceptions. For example, the German construction with the verb scheinen & zu-infinitive is a typical evidential expression which in many cases refers to direct visual evidence2 for the proposition the speaker is asserting. This is illustrated in (1): (1)

Sie scheint ihren Schirm im Büro gelassen zu haben. ‘She seems to have left her umbrella in the office’.

In this example the evidential construction refers to the fact that the speaker derives his/her conclusion from some (visual) evidence she/he has for making the claim (namely ‘She has left her umbrella in the office’). The

A first approach to evidential meaning 3

undeniable fact that this sentence also refers to the uncertainty of the speaker, implying that she/he does not have secure knowledge, is due to a particular trait of German evidentials, namely the fact that German evidentials very often have epistemic modality meanings as concomitant features. This calls for a thorough investigation of the relations between evidentiality and epistemic modality, which will be undertaken as we proceed (see Chapter 3). Many other Indo-European (henceforth: IE) languages such as Russian, Spanish, French, Dutch or English are known to have evidential markers similar to the German constructions as the scheinen-construction illustrated above. Analogous versions of sentence (1) are given under (2) for Dutch (2a), English (2b) and Spanish (2c): (2) a. Ze schijnt haar paraplu in het bureau te hebben gelaten. b. She seems to have left her umbrella in the office. c. Parece haber dejado su paraguas en el despacho. Further IE languages with evidential markers are e.g. Albanian, Bulgarian, West Armenian and Persian (Lazard 1999, Friedman 2000, Johanson 2000, Giacalone Ramat and Topadze 2007). As the expression of sources and types of knowledge belongs among the basic communicative or pragmatic necessities of any language, evidentiality must be seen as an indispensable semantic and functional domain in any language. Thus, although the possibility to express evidential judgments is not part of the grammatical system in all languages, i.e., it is not expressed by default and with grammatical means, speakers can always indicate the “reason” or “evidence” for their statements if they consider this neaessary for some reason. The question to what degree an evidential construction is grammaticalized is one of the main issues of this book. It is therefore essential to provide clear definitions for grammaticalized items expressing evidentials values, in order to distinguish them from ad hoc lexical-semantic circumscriptions of evidential meanings. As this topic will be treated extensively in Chapter 4, some preliminary orientation should be sufficient here. First, by way of terminological clarification it should be noted that only grammaticalized evidential expressions are called “evidentials” or “evidentials markers” here. Second, it will be useful to resort to a non-formalized way of distinguishing evidentials from their lexical, non-grammaticalized predecessors

4 Introduction (cognates): this can be achieved with the help of a feature Anderson (1986) first drew attention to, namely that evidentials (i.e. grammaticalized forms) are always secondary predications operating on a primary predication, so that the secondary predication gives the evidential information which is the basis for making the primary predication.3 A third and last, the term “evidential” is reserved for linguistic expressions which – in the usage under question – have “the indication of evidence as [...] their primary meaning, not only as a pragmatic inference” (Anderson 1986: 274; see Chapter 2 for details). Typological research has shown that a great number of non-IE languages have evidential markers which often form complex evidential systems of high degrees of grammaticalization. Among them are Tariana (an Arawak language spoken in the area of Vaupés, northwest Amazonia) and Wanka Quechua (cf. Aikenvald 2004, 2006), which are illustrated in the following tables. Table 1. Evidential markers of Tariana (Aikhenvald 2004: 2f.) Example Gloss Periphrasis Juse irida di-manika José football 3sgnf-play ‘José has played football’ Juse irida di-manika José football 3sgnf-play ‘José has played football’ Juse irida di-manika José football 3sgnf-play ‘José has played football’ Juse irida di-manika José football 3sgnf-play ‘José has played football’ Juse irida di-manika José football 3sgnf-play ‘José has played football’

Evidential marker Periphrasis -ka -REC.P.VIS

-ka ‘we saw it’

-mahka -REC.P.NONVIS

-mahka ‘we heard it’

-nihka -REC.P.INFER

-nihka ‘we infer from visual evidence’

-sika -REC.P.PASSUM

-sika ‘we infer this on the basis of what we already know’ -pidaka ‘we were told’

-pidaka -REC.P.REP

Tariana provides a five-fold distinction of evidentials values (fused with tense), one of which has to be realized in every sentence. As Aikhenvald (2006: 322) points out, Quechua languages distinguish three evidential values, namely “direct evidence”, “conjectural” and “reported”. Table 2 illustrates this three-fold distinction with data from Wanka Quechua as provided in Floyd 1999. The examples show that evidential constructions are used for the expression and modulation of types, degrees and sources of knowledge. In

A first approach to evidential meaning 5

order to tackle this level of semantic differentiation in evidential distinctions it will be appropriate to say something about the methods applied for isolating and testing the specific meaning of evidential constructions. Table 2. Evidential markers of Wanka Quechua (Aikhenvald 2006:322, data taken from Floyd 1999) Example Gloss Periphrasis trabaja-aña-m li-ku-n work-PURPOSE. Go-REFLEXIVE-3p MOTION-nowDIRECT.EVIDENTIAL ‘He’s gone to work’ (I saw him) chay lika-a-nii juk-ta-chra-a lika-la that seeother-ACCUSATIVEsee-PAST NOMINALISER-1p CONJECTURAL. EVIDENTIAL-TOPIC ‘The witness must have seen someone else' (her house was robbed; she saw someone next to her house, it was not me, I infer it was (-chr) someone else’ Ancha-p-shi wa'a-chi-nki wamla-a-ta Too.much-GENITIVE- cry-CAUSATIVE-2p girl-1pREPORTED.EVIDENTIAL ACCUSATIVE ‘You made my daughter cry too much’ (they tell me)

Evidential marker Periphrasis -mi ‘I saw it’

-chi, -chr(a) ‘I infer’

-shi ‘they tell me’

According to Aikhenvald (2004: 3-4), the exact semantic description of individual evidential forms in general relies on a combination of various methods and procedures, among them native speakers’ intuitions, substitution tests, semantic periphrasis, and what Aikhenvald calls “the possibility of lexical ‘reinforcement’”, which she specifies as follows: [...] an evidential can be, optionally, rephrased with a lexical item, or one can add a lexical explanation to an evidential. A visual evidential would then be rephrased as ‘I saw it’, [...] and a reported evidential with ‘they told me’. (Aikhenvald 2004: 3-4)

Wierzbicka (1986, 1991, 1994, 1996) uses controlled periphrases with a defined vocabulary for circumscribing meaning. Together with her team she has developed a universal semantic metalanguage called “natural semantic metalanguage” (NSM) which consists of a small number of simple, defined and cross-linguistically tested concepts and a syntax for forming semantic paraphrases with these semantic primitives. A very brief exemplification will suffice here: in order to distinguish “sight evidentials” and “hearsay evidentials” in Maricopa Wierzbicka (1994: 103-104) suggests the following paraphrases:

6 Introduction (3)

(4)

“sight evidential” -(k)’yuu: I know this; one could see it “hearsay evidential” -‘ish-‘a: People say this; I don’t say: I know it

In this book we will not use this particular technique. Instead, we will apply the method of controlled, schematized periphrasis in addition to terminological labels. For example, the use of scheinen in its function as an evidential with a visual component like in (5), which today is not its central function (see below), is explicated by a periphrasis. A preliminary version of this is shown in (6) (P = proposition): (5) (6)

Sie scheint ihren Schirm im Büro gelassen zu haben. ‘She seems to have left her umbrella in the office’. The speaker has visual information for P.

As already mentioned in the discussion of example (1) above, the domain of evidentiality is closely linked to but distinct from (epistemic) modality, and the discussion on where to set the boundary between these two domains will take up some space in the following chapters. As a rule of thumb, one can say that while evidentiality indicates the source of evidence a speaker has for making a statement, without necessarily accompanying that with a factuality judgment, epistemic modality, is concerned exactly and exclusively with the latter, i.e. with the degree of factuality a speaker attributes to a proposition. The comparison between the two domains may be illustrated in the following minimal pair. While (7a) represents a set of epistemic modals, (7b) shows evidential constructions. The respective periphrases note the semantic distinctions of factuality judgments and evidential information. These distinctions will be refined and argued for in later chapters. (7) a. Die Lieferung dürfte/könnte/mag größer sein als erwartet. ‘The delivery should/might/may be larger than expected.’ The speaker attributes an uncertain factuality value to P. b. Die Lieferung scheint/droht/verspricht größer zu sein als erwartet. ‘The delivery seems/threatens/promises to be larger than expected.’ The speaker has evidence for P.

A first approach to evidential meaning 7

As will also become clear later, the notions of evidential and modal values intersect in languages like German; nevertheless it is essential to keep them theoretically apart. Beyond its connection to epistemic modality, evidentiality is also closely related to the domains of aspectuality and temporality. Aspectuality comes into play where resultative and ingressive concepts are part of evidential meanings, temporality is touched when aspectual notions become semantically more abstract and thus acquire temporal meaning. These domains are involved in the diachronic development of evidential markers in German, where, e.g. in the case of werden, we find ingressive source constructions of the later evidential constructions in the early stages, which during their development also acquire temporal meaning. This will be treated in the relevant chapters in this book. All these functional-semantic domains, i.e. modality, temporality, aspectuality, and evidentiality, have in common that they can be expressed in a variety of ways in a language: by free lexical expressions, by syntactic constructions or by grammatical markers in the narrow sense (i.e. as bound morphology or periphrastic, i.e. analytic constructions). For example, temporality/temporal concepts in German may be realized by various kinds of linguistic expressions, e.g. by adverbials like heute ‘today’ or später ‘later’ (8a), by nouns like Jahr ‘year’ or Stunde ‘hour’ (8b), by subordinate temporal clauses introduced with temporal conjunctions like als ‘when’ or bevor ‘before’ (8c), and of course by grammatical categories, i.e. tense markers like the German preterite or perfect (8d). (8) a. Später/heute bekommen wir Besuch. ‘Later/today we will have visitors.’ b. Der Harmonisierungsprozess dauert ein Jahr/eine Stunde. ‘The process of harmonization is going to take a year/an hour.’ c. Als/bevor sie ging, betrat der Briefträger das gegenüberliegende Haus. ‘When/before she left the postman entered the house opposite.’ d. Mitten in der Nacht ging er/ist er gegangen. ‘In the middle of the night, he left.’ Analogously, evidentiality and evidential distinctions may be expressed by a whole gamut of linguistic means. As far as German is concerned, we have the following options. First, we may indicate evidential meaning by evidential adverbials like offensichtlich ‘obviously’, anscheinend ‘appar-

8 Introduction ently’, angeblich ‘allegedly’ as in (9a) to (9c) or adjectives like evident ‘evident’ in matrix clauses, like in (9d): (9) a. Sie hat offensichtlich den Bus genommen. ‘Obviously, she has taken the bus.’ b. Anscheinend hat sie den Bus genommen. ‘Apparently, she has taken the bus.’ c. Angeblich hat sie den Bus genommen. ‘Allegedly, she has taken the bus.’ d. Es ist evident, dass sie den Bus genommen hat. ‘It is obvious/evident she has taken the bus.’ Secondly, evidentiality may be expressed by syntactic constructions with semantically suitable verbs, including AcI-constructions with perception verbs like sehen ‘see’, hören ‘hear’, fühlen ‘feel’ (10a-c) or cognitive verbs like schlussfolgern ‘reason’, schließen ‘conclude’ etc. (10d,e): (10) a. Sie sah ihn durch den Garten davonlaufen. ‘She saw him run away through the garden.’ b. Sie hörte ihn hinterm Haus singen. ‘She heard him sing behind the house.’ c. Sie fühlte die Kälte in sich hochsteigen. ‘She felt the cold rising in her body.’ d. Daraus schlussfolgerten sie, dass es keinen Sinn mehr habe, den Antrag zu stellen. ‘From this they concluded that it did not make sense any more to file the application.’ e. Daraus schließe ich, dass es keinen Sinn hat, weiterzusuchen. ‘From this I conclude that it does not make sense to continue searching.’ Furthermore, German has verbal periphrastic forms with the verbs scheinen, drohen and versprechen with zu-infinitive and with werden & infinitive. They express strong evidential values – more particularly inferential evidential values (see Chapter 2) – insofar as they refer to reasons, indications, or evidence, i.e. to the source of information speakers have for expressing statements, cf. (11a) - (11d):

A first approach to evidential meaning 9

(11) a. Sitz gerade auf dem Fahrrad, Ilja! Führe den Lenker nach links, wenn du nach rechts zu kippen drohst! (W)4 ‘Sit straight on the bike, Ilja! Turn the handlebars to the left, when you find yourself falling off to the right.’ b. …und nach längerem Suchen und nachdem er ihn mit den Zähnen getestet hatte, fand Kürenberg einen Reis, der körnig zu kochen versprach. (W) ‘…and after a lengthy search and after testing it with his teeth, Kürenberg found a type of rice which promised to boil grainy.’ c. …und die die freude am risiko scheint sein zwanzigjähriger sohn hans-joachim geerbt zu haben (DSAv) ‘and his twenty-year-old son Hans-Joachim seems to have inherited his readiness to take risks.’ d. Wenn also Franz Beckenbauer und andere der Stiftung Warentest vorhalten, sie wolle sich über die Weltmeisterschaft profilieren, werden sie ziemlich schnell feststellen, dass…(ZEIT) ‘So, if Franz Beckenbauer and others reproach the Stiftung Warentest for trying to promote themselves via the World Cup, they are bound to realize very quickly that ... ’ As has been pointed out, these four constructions, i.e. werden & infinitive, and scheinen/drohen/versprechen & zu-infinitive and their development in the history of German, are the major topic of this book.

1.2. Deixis Another very important issue must be mentioned in this first overview on central aspects of evidentiality: This is the question of deixis and indexicality and its impact on grammatical signs in general and evidentials in particular. One of the defining features of grammatical signs is their indexical potential (cf. Diewald 2006: 414-416).5 They link linguistic elements that vary in size and function in relation to one another or to some relevant nonlinguistic entity. As Jakobson ([1957] 1971) has already shown, a large number of grammatical functions (e.g. tense, mood, nominal determination) are deictic in the strict sense of the term, i.e. they localize the linguistic entity they apply to with respect to the coordinates of the speaker, the deictic origo (Bühler [1934] 1990).6

10 Introduction Starting from the observation that grammatical categories are deictic in nature or display a type of indexicality that ultimately can be deduced from an underlying deictic relation it has been argued in various studies that the grammaticalization of tense and mood markers or of articles and modal particles in its core can be put down to the development of a deictic relational structure from a non-deictic relational structure (cf. e.g. Diewald 1999 for modal verbs and verbal mood).7 This is to say that grammaticalization in these cases is necessarily associated with a shift of the semiotic status of the sign in question from a non-deictic, characterizing function to the function of deictic pointing.8 As we assume here that evidential markers – like tense markers or epistemic modality markers – are deictic signs and that their grammaticalization, too, is an instance of the development of a deictic function in constructions that did not show that function before, it will be necessary to give a brief definition of what is meant here by deixis. Since any linguistically encoded event, i.e. any “event spoken of” in the terminology of Jakobson, has to be located with respect to the speaker, i.e. the origo, deictic anchoring is a basic and inevitable function in language. This is realized by a large variety of linguistic devices, ranging from grammatical to lexical, from backgrounding to foregrounding this very fact of situational anchoring in the speaker. A prototypical instance of deictic anchoring (a deictic process) displays a relational structure in which the speaker is the centre of the communicative universe, i.e. the deictic origo which is the point of departure of any cognitive and linguistic processing associated with a particular communicative event. The other elements of the relational structure are the described entity, the ‘demonstratum’ as Mitchell (1984: 1204) calls it, and the relation between the entity and the origo. All deictic signs share a semantic structure representing the deictic process as their common denominator. This basic structure will be described as a directed relation, a vector, between a source and a goal (cf. Bühler [1934] 1990, Sennholz 1985, Diewald 1991). The basic deictic relation can be conceived as shown in Figure 1. Thus, deictic signs are defined as containing a directed relation between origo and denoted entity as part of their semantic content. The general meaning of this relational structure can be paraphrased as follows: ‘the entity that is denoted with the help of this deictic sign is denoted in its relation to the actual speaker/the origo’. In deictic signs the linking between denoted entity and origo is semanticized in the type; it is actualized, i.e. applied to the specific individual context in every instance of its use. To

A first approach to evidential meaning 11

sum up, deictic signs, i.e. shifters, are defined here as ‘indexical symbols’ that characterize a vectored relation between the origo (the source) and the denoted entity (the goal) as part of their semantic content, and thus represent a specific instantiation of the relational template. described entity

relation/distance between Origo and entity

Origo

Figure 1. Prototypical deictic relational structure

Deictic meaning is not restricted to lexical items like spatial and temporal adverbials, or to free grammatical signs like personal and demonstrative pronouns etc., but is expressed by many inflectional grammatical categories, such as tense markers, markers of definiteness and person markers. However, for the sake of clarity this will be illustrated here by a simple and typical example – the use of the deictic adverbial hier in (12): (12)

Die Katze liegt hier. ‘The cat is lying here.’

The deictic sign hier in (12) represents a vectored relation between two elements of the situational context: The origo, which is the source, and the denoted entity, which is the goal of the deictic process. The latter is, due to further semantic features, specified as a local entity which is ‘close’ to the location of the speaker. Thus the meaning of hier can be paraphrased as a ‘place that from the position of the speaker is regarded by the speaker as near, and is denoted as such’. As obvious from the example, deictic signs do not only characterize the directed relation between origo and denoted entity, they also contain further descriptive information (Bühler [1934] 1990, Diewald 1991, Hanks 1992: 48). First, they characterize the broad semantic domain the denoted entity belongs to, which is called the deictic dimension (Diewald 1991: 30ff.). Hier in (12) denotes a spatial entity; other deictic dimensions are e.g. the temporal dimension (as in an adverb like jetzt ‘now’, or a tense like “past”) and the personal dimension (as in ich ‘I’). And of course, as we claim here, there is also a modal dimension and an evidential dimension of

12 Introduction deictic anchoring. The modal dimension is realized for example by the opposition between indicative and subjunctive, the evidential one by the presence or absence of one of the four evidential markers in question. Figure 2 below may serve as one possible visualization of this dimension: described entity

Origo

local/temporal/personal/modal....dimension

Figure 2. Deictic relation specifiying the deictic dimension realized

Beyond the distinction of different dimensions of deixis, i.e. different types of entities being “pointed at”, a second type of information is very frequently encoded in deictics: the deictic distance between the origo and the deictic object. The basic criterion for distance is the binary distinction between origo-inclusive and origo-exclusive. The origo-inclusive value conceives the described entity as co-present with the origo, while the origoexclusive value conceives the described entity as not co-present with the origo. To illustrate this, compare (12) to (13): (13)

Die Katze liegt dort. ‘The cat is lying there’.

Hier in (12) denotes a place that is conceived as concentric with the location of the origo, i.e. as origo-inclusive, while dort in 13 denotes a place that is conceived as non-concentric with the place of the origo, i.e. as origo-exclusive. This distinction is also responsible for the opposition between dieses ‘this’ (origo-inclusive) and jenes ‘that’ (origo-exclusive) or the opposition between present tense (origo-inclusive) and past tense (origo-exclusive) etc. In Chapter 3 we will show that this distinction applies to evidential and modal meanings as well. The distinction between the origo-inclusive and the origo-exclusive realization of the deictic process is illustrated for hier (‘a place that is conceived as concentric to the place where the origo is located’) and dort (‘a place that is conceived as not concentric to the place where the origo is located’) in Figures 3 and 4.

A first approach to evidential meaning 13 described entity co-present with Origo ‘origo-inclusive’

Origo

local dimension

Figure 3. Relational structure for origo-inclusive local deictic hier described entity not co-present with Origo ‘origo-exclusive’

Origo

local dimension

Figure 4. Relational structure for origo-exclusive local deictic dort

It should be noted that conceptualizing something as origo-inclusive or origo-exclusive is a decision/evaluation made by the speaker. It may be influenced by real-world factors, but nevertheless reflects the speaker’s judgment and has nothing to do with “objective” scales or measurement. Deictic grammatical categories, like tense (which realizes the temporal perspectivization or localization of the scene described with respect to utterance time) and mood (which realizes a speaker-based factuality judgment of the scene described), instantiate a particular variation of the basic relational structure of deictic signs. A grammatical sign modifies another (lexical) sign by relating it to some other element, i.e. to some reference point lying outside both of them. That is, a grammatical sign establishes a link between the linguistic element it modifies and some other entity. Most notably, the deictic distance oppositions in grammatical paradigms are usually organized according to the distribution of formally unmarked versus marked values, whereby the “near”, i.e. origo-inclusive value is expressed by the unmarked form. That is, “formally unmarked” does not mean ‘non-existent’; it usually signifies the basic value/category of a paradigmatic opposition. This can be illustrated by the fundamental deictic distance opposition in the German tense system, the opposition between

14 Introduction present tense and past tense (preterite). While the present, which, if used in a temporal function,9 encodes the origo-inclusive value (coincidence with speaking time) without a formal marker expressing that function, the past tense is formally marked. It realizes an origo-exclusive temporal deictic value, more specifically, it “localizes” the scene as distant from the origo and situated in a time span gone by from the point of view of the speaker (non-coincidence with speaking time). As already pointed out, it is argued here that evidentials are deictic grammatical markers, too, and that the four evidential constructions we are dealing with are a case in point. As can be seen in the following quote, there are several authors who subscribe to this view: These various concerns [...] bring to mind the Jacobsonian notion of “shifters”. […] Since shifters typically involve deixis, evidentiality in part at least may well be thought of therefore as connected to matters of deixis, in that in systems that overtly mark evidentiality, utterances typically include indicators pointing directly to particular sources or away from potential sources, as the speaker takes a particular point of view in describing an action. (Joseph 2003: 308) We are now ready to consider evidentials as a deictic phenomenon. […] The only way in which Wintu verbs locate events with respect to speakers similar to English tense deixis is with the evidential suffixes. […] English uses tense deixis to place a statement with respect to the time its truth depends on; Wintu employs evidential deixis to place an event in the context of the other events which are entailed by it. (Schlichter 1986: 57-58) Evidentiality can also be characterised as a deictic category, one that functions to index information to some point of origin (Bühler’s ‘origo’). […] Evidential markers are deictic because they index information to the conceptualiser who makes an epistemological judgment. In context, the choice of evidential categories (e.g. witness or report) serves to select the deictic origin – the one from which all temporal, spatial and identifying information can be calculated. (Mushin 2001: 33) A recent proposal is to analyze evidentiality not as modal, but as a deictic category. [There are] the connections between spatial deictic elements such as demonstratives, temporal deictic elements, such as tense, and evidential elements. In all cases, the morphemes in question denote the distance between the speaker and: an object (spatial), time (temporal) and the entire proposition (evidential). (De Haan 2005: 51)

The discovery of evidential meaning and evidentials in (German) linguistics 15

The deictic function of evidentials will be dealt with in detail in the following chapters (especially in Chapters 2 and 3). Grammaticalization in the case of German evidentials is an instance of the rise of deictic functions in signs that did not have this function before. Before outlining our hypotheses further, the next section gives a survey of prior research into evidentiality in German.

2. The discovery of evidential meaning and evidentials in (German) linguistics In spite of the fact that marking the source of knowledge, beliefs and convictions is essential to human communication, the study of evidentiality has but a short tradition in linguistics. The first typological studies on the domain and category of evidentiality date from the 1980s (Bybee 1985, Chafe and Nichols 1986, Willett 1988, Matlock 1989), and only as late as in the 1990s did evidentiality, understood as the linguistic coding of speaker’s evidence for making statements, gain attention in German linguistics. Up to then the term evidentiality was not common in descriptions of the German language, and even today, terms like “evidentiality”, “evidential marker” or “evidential meaning” – and their German equivalents – are missing in most textbooks and reference grammars. For example, neither Zifonun et al. 1997, nor Eisenberg [1998] 2004, Helbig and Buscha 2001, Engel 1996,10 or Duden 1995, 2006 list the terms “evidential” or “evidentiality” or some related term in their indices or tables of contents. So, the issue of evidentiality is largely absent from linguistic research of German (both synchronic and diachronic). Traditional grammar up to now does not mention this category and, not long ago, probably did not even know of its existence. This may be due to the fact that it relied too heavily on the set of categories handed down from antiquity. Keeping in mind that even a category like aspect, which is prominent in many languages spoken in Europe, took a long time to be recognized in German grammar books (as a category which, though not grammaticalized in German, is relevant for the discussion of other categories and their diachronic development in some places), this resistance to the introduction of a new grammatical category does not come as a surprise, and should not prevent linguists from introducing this category into their work if it is required. A further possible reason for not mentioning evidentiality might be that the German language does not (yet) express evidentiality in a way that is recognized as a grammatical system in its own right, i.e. that German does

16 Introduction not have a set of clear-cut markers of the same type that may be found in other languages known for their (sets of) evidential markers. The German system is – as we contend – in the process of evolving, i.e. its grammaticalization has not yet been completed. However, the absence of a comprehensive system should not be treated as an argument for the absence of the category altogether. In the last two decades general linguistic research into evidentiality has gained momentum and there have been important steps towards an understanding of the cognitive and universal foundations of evidential systems. A great and still growing number of typological case studies has shown that evidentiality is a relevant category in the grammatical systems of numerous languages (Bybee, Perkins and Pagliuca 1994, Wierzbicka 1994, Sumbatova 1999, De Haan 1999, Johanson and Utas 2000, Dendale and Tasmowski 2001, Faller 2002, Aikhenvald and Dixon 2003, Aikhenvald 2004, Speas 2004). In addition to the grammatical dimension there have been promising findings on the semantic foundation of evidentiality (Plungian 2001) and on ways to analyse the distinctive values of evidential markers of a particular language (see for instance Cornillie 2004, Dendale and Tasmowski 2001, Diewald 2004, Mortelmans 2000). In contrast to the upsurge of interest in evidentiality in general linguistics and typology over the last few decades, the development in German linguistics has been much slower. There exists a number of synchronic studies concerning particular German verbs and constructions which are connected to the study of evidentiality to some extent. In particular, the verb scheinen and its development towards an auxiliary has received the attention of linguists in recent years (cf. Askedal 1989, 1997, 1998, 1999; Ebert 1976, Olsen 1981, Pafel 1989, Diewald 2000, 2001). For drohen and versprechen, there are also several synchronic studies, particularly concerned with the syntactic distribution patterns and the (semi-)auxiliary status of the two verbs in present-day German (cf. Askedal 1997, 1999; Gunkel 2000, Kokutani 2000, Diewald 2004, Reis 2005). However, up to now, there has been no explicit discussion on evidentiality as a grammatical category of German, nor is there a comprehensive study on the expression of that category in present-day German or on its diachronic development.

2.1. The status of evidential meaning in present-day German Current opinions on the place of evidentiality in German vary widely among linguists, and we may distinguish the following three general

The discovery of evidential meaning and evidentials in (German) linguistics 17

trends: we will call them here the Sceptics, the “Conversational Implicatures” and the Advocates. The Sceptics deny the presence of evidential meaning in German verbs, and consider them to be ‘modal verbs’ or ‘modality’ verbs (see especially Askedal 1997, Engel 2004, Zifonun et al. 1997, Eisenberg 2004, Heine and Miyashita 2004, 2008). Scholars promoting this opinion treat the infinitive constructions with drohen, versprechen and scheinen as markers of (epistemic) modality. By taking this position they implicitly disclaim that evidentiality has a status of a (grammatical) category in the German verbal system. Consider the following quotations: In the relevant literature there is fairly wide agreeement that the main component of functional drohen is to express epistemic modality. (Heine and Miyashita 2008: 59) Sie [drohen and versprechen] drücken somit eine epistemische Modalität (und eine Bewertung) aus. [They express epistemic modality (and an evaluation)]. (Gunkel 2000: 111) […] die subjektive Modalität bei den Modalverben [ist] eine nicht lexikalisch festgelegte Interpretationsmöglichkeit, bei scheinen mit Infinitiv aber ein inhärentes lexikalisches Merkmal. [The subjective modality of modal verbs is a possible intepretation that is not lexically anchored, for scheinen with infinitive however it is its inherent lexical property.] (Askedal 1998: 62)

Scholars treating werden as a modal verb (see especially Vater 1975, Itayama 1993, Engel 2004; more details are presented in Chapter 6.1) can be seen as proponents of this tendency. The second group of German linguists accepts that there are grammaticalized linguistic devices in present-day German expressing evidential meanings. These evidential values, however, are considered by this group to be context-induced interpretations, i.e. conversational implicatures arising on the basis of their primary non-evidential meanings, e.g. epistemic modal meanings. In addition, evidential readings are often seen as derived from aspectual or temporal basic meanings of linguistic expressions. So, for instance, Reis 2005 suggests that the primary meaning of the auxiliary drohen is temporal-aspectual and that its evidential components are derivative of this meaning; cf.: Dafür spricht bereits, dass bei den allermeisten drohen2/versprechen211 Belegen eher/nur die temporal-aspektuelle Deutung als Bericht über vorliegende Indizien für das (gerade beginnende oder unmittelbar bevorstehen-

18 Introduction de) Geschehen p greift; Inferenz und subjektive Einschätzung der Wahrheit von p spielen so gut wie keine Rolle. [This is already hinted at by the fact that in the vast majority of cases of drohen2/versprechen2 there is basically/only the temporal-aspectual interpretation in the form of a report on existing evidence for the event (which is beginning or is about to begin) that holds; inferences and subjective evaluations of the truth of p do not play a noteworthy part in this matter.] (Reis 2005: 129)

Mortelmans 2004 treats both the German modal verb müssen and the verb werden as grammatical expressions of epistemic modality. According to her, they both are inferential in the sense that the proposition is presented as being based on some evidence available to the speaker. She sees this – evidential – value of these two verbs as a contextually induced interpretation of their core epistemic modal meaning, cf.: It should be noted that epistemic werden is generally associated with a rather strong conviction on the part of the speaker regarding the future actuality of the state of affairs. This strong epistemic assessment must be related to werden’s outspoken speaker involvement: the verb is typically found in contexts in which the speaker expresses an emotional, highly personal attitude. Very often, the speaker’s assessment is based on her own observation or perception of things, on her personal experience, which in a rather immediate way gives rise to a particular prediction. In contrast to müssen, German werden is not strongly inferential: the speaker does not so much express a logical conclusion as a result of particular reasoning process, but rather a subjective prediction that follows naturally from the speaker’s own observations or personal conviction. (Mortelmans 2004: 46)

Fritz 2000 renders the evidential reading of the construction werden & infinitive as one of several possible interpretations of its primary nature as an explicit sign of the speaker’s personal involvement (‘Sprecherverweis’), cf.: Die beschriebenen evidentiellen Lesarten von werden stellen in der Gegenwartssprache systematisch die grundlegenden Bedeutungen der Fügung dar. Dies heißt freilich nur, dass sie im Vergleich zu weitergehenden Bezugsmöglichkeiten des Sprecherverweises einen kürzeren Interpretationsweg aufweisen, da sie näher an der semantischen Markierung ‚SPR‘ [Sprecher] angesiedelt sind. [The described evidential readings of werden – seen from a systematic perspective – are the basic meaning of the construction in the language of today. This however does not mean more than that they display a shorter interpretative trajectory as compared to more subtle applications of the speaker index as they are closer to the semantic feature ‘SPR’ [speaker].] (Fritz 2000: 153)

The discovery of evidential meaning and evidentials in (German) linguistics 19

Dieling (1982: 328-330) describes the construction werden & infinitive as an expression of ‘a special type of hypothesis’, thus rendering it as an expression of epistemic modality as well. However, he states that, using werden & infinitive, … drückt der Sprecher aus, dass er nicht allein gute Gründe für die Annahme von p hat, sondern auch subjektiv daran glaubt, dass p. [… the speaker indicated that not only does he have good reason for assuming p, but that moreover he himself believes that p.] (Dieling 1982: 328)

Hence the speaker has certain ‘grounds’ for the particular hypothesis that is formulated with werden & infinitive. Such grounds can also be understood in terms of evidences on which the speaker bases his/her reasoning procedure (a similar view is hinted at in Brinkmann 1971: 398-399). To sum up: according to the general opinion of the scholars of the ‘second group’, evidentiality is not a functional domain of its own in present-day German. Rather, evidential meanings are expressed as (contextual, derivative, etc.) meaning extensions of linguistic forms which have inherently non-evidential base-meanings pertaining primarily to other functional domains, especially to epistemic modality. The Advocates regard evidentiality as a regular, independent grammatical category that has been evolving in present-day German. According to this view, there are certain linguistic constructions for coding evidentiality in present-day German, among them the infinitive constructions with werden, scheinen, versprechen and drohen, which are highly grammaticalized as evidentials and form a grammatical paradigm for coding evidentiality. This is to say that the constructions in question constitute a closed set of paradigmatic distinctions (oppositive values) which are united under the category label of evidentiality and distinct from other categories (e.g. epistemic modality/mood). The essential feature is the close association of the categorial distinctions with particular constructions – the four infinitive constructions. The fact that these constructions are morphologically not entirely homogeneous (e.g. afford infinitives with or without the infinitive particle zu) and display slightly diverging syntactic behaviour (cf. Chapter 4.2), is not a counterargument against their membership in one grammatical category. On the contrary, it is quite common for grammatical categories to incorporate category members of diverging formal characteristics (e.g. the category of tense in German which unites inner inflection, affixation, and periphrasis). The same applies to the fact that some members of the evidential category in their syntactic behaviour ressemble some members of other categories (e.g. the epistemic modals): this observation, too,

20 Introduction holds for other categories as well. The essential criteria evaluating the (degree of) paradigmaticization are the combinations of functional, semantic and formal characteristics that will be discussed in detail in Chapter 4. The main proponents of this approach are the authors of this book. Preparatory work on this issue, including the introduction of the notion of evidentiality into German linguistics has been achieved by Diewald 2000, 2001, 2004, 2005, who has argued that the German language indeed has specific linguistic expressions for coding evidentiality. Since this book is devoted to justifying our fundamental assumptions and central hypotheses, we will not go into further details at this point, and postpone the treatment of these issues to later chapters. Further studies focussing on diachronic aspects, and laying the ground for this book are mentioned in some detail in the next section.

2.2. The diachronic development of evidentials in German As to the diachronic development of evidential expressions, it must be stated that in the literature on the history of German and especially on grammaticalization phenomena, the notion of evidentiality is virtually absent, except in Diewald 2000, 2001, focussing on the diachrony of the grammaticalization of the construction scheinen & zu-infinitive. In Diewald (2000) it is argued that scheinen is an auxiliary in present-day German, arising via a specific grammaticalization channel. This channel is described as a path leading from a (main) verb denoting a visual effect through a marker expressing visual evidentiality toward a neutral inferential evidential auxiliary. Furthermore, this path conforms to the crosslinguistic observation that indirect inferential evidentials often arise from verbs of appearance (cf. Anderson 1986, Willett 1988, Aikhenvald 2004). Moreover, this developmental path is argued to be analogically related to the grammaticalization paths of German modal verbs, involving the shift in meaning from deontic to epistemic modality. Thus on the basis of the observations on the development of scheinen & zu-infinitive, Diewald 2000, 2001 formulates the hypothesis that the grammaticalization of evidential markers in German is closely interrelated with the – chronologically earlier – grammaticalization of the German modal verbs into periphrastic moods (i.e. expressions denoting epistemic modal values), which has an important impact on the whole domain of German evidentiality. This fact may explain, for instance, why German evidential markers develop epistemic modal overtones.

The discovery of evidential meaning and evidentials in (German) linguistics 21

In Smirnova (2006, 2007) the development of the evidential meaning of werden & infinitive has been described in the context of the development of the construction würde & infinitive. It has been shown that the temporal function of the construction werden & infinitive evolved as one of several possible interpretations of a more abstract semantics that werden & infinitive had acquired in the Middle High German (henceforth: MHG) period. The grammaticalization channel for werden & infinitive is sketched as the development from an aspectual (ingressive) verb through a marker of a sequential relation towards an inferential evidential, which marks the proposition as anchored in the speaker’s inference. This path can be noted as follows: ingressive meaning > indicating a relation of sequence > inferential evidential This developmental path of a German evidential construction also has cross-linguistic validity: it is generally accepted that inferential evidentials arise from perfective aspectual forms (cf. Anderson 1986, Willett 1988, Aikhenvald 2004). Finally, the paper by Diewald and Smirnova [in press] is an explicitly preparatory study to this book as it roughly sketches the historical development of each of the four German evidential constructions and ascribes to them a common developmental path leading to a system of inferential evidentials in German. It is argued, first, that the inhomogeneity of the present-day evidential system of German is due to different ages and grammaticalization degrees of the four evidential constructions: whereas werden & infinitive and scheinen & zu-infinitive are relatively old and highly grammaticalized as evidentials, the infinitive constructions with drohen and versprechen started their development towards evidentials only in the course of the 18th century. Therefore, the latter two still show (syntactic) restrictions and (semantic) nuances which are due to the preservation of their lexical semantics, which is a quite common phenomenon in language change and is referred to by the term “persistence” (Hopper 1991). Second, the paper delineates the development of the four German inferential evidential constructions as a process, in which each one of the four constructions started out from a different source, successively converging with the others in one common and general path of grammaticalization for inferential evidentials. This development will be treated extensively in the sections of this book that deal with diachronic aspects.

22 Introduction Though there are no other studies explicitly concerned with the development of evidential expressions in German, there is some work taking up related questions, like the diachronic development of erstwhile aspectual constructions, e.g. the ingressive infinitive constructions with gistantan or biginnan & infinitive (e.g. Wilmanns 1906, Schmid 2000, Westvik 2000). In connection with werden & infinitive, they may be seen as direct ancestors of evidential constructions, and thus may be generally seen as potential source constructions of evidential markers (see Chapter 8). Furthermore, several studies deal with the diachrony of the German construction werden & infinitive and – though often not explicitly – recognize the strong evidential semantics of this construction. Kotin 2003, for example, defines werden & infinitive as an essentially prognostic construction with its prognostic meaning having arisen from the aspectual ingressive semantics of werden. Lerch 1942 classifies the construction werden & infinitive as the “Futurum des zu Erwartenden” [expected future], emphasizing that the expectation is always anchored in the present experience. Hence it can be said that the strong speaker-oriented semantics of werden & infinitive and its close relation to the experience and knowledge sphere of the speaker are generally recognized in German linguistics. However, the opinions about the exact nature and origin of this link (to the speaker) differ significantly: most often the speaker involvement is treated as an epistemic modal component arising historically from the reinterpretation of the temporal meaning of the construction. As will be expounded in later chapters, we do not agree with the proposals that the core meaning of werden & infinitive is an epistemic modal or a temporal meaning. Rather, we contend that the temporal value of werden & infinitive has developed parallel to its evidential meaning (see Saltveit 1962 for a similar view). At this point, it is necessary to say some words on the role of the notion of subjectivity or subjectification in the diachronic development of evidential constructions, although this topic, too, will have to be treated later in more detail. Most of the studies dealing with the cognates of the English verbs promise and threaten (see especially Traugott 1997 for English, De Haan 1999 for Dutch, Cornillie 2004, 2005 for Spanish) emphasize the role of subjectification in the semantic change of these verbs. Subjectification is conceived by these authors as a major type of semantic change in the initial stages of grammaticalization and is considered the motivating force for the shift in meaning occurring in both verbs. Traugott 1997 describes this as follows:

Hypotheses and theoretical foundations 23 The motivation for the development of the raising uses of promise and threaten is part of a general process whereby epistemic meanings gradually shift from more “objective” possibility based in general beliefs and attitudes to a more “subjective” possibility based in the individual speaker’s belief or attitude. In other words, the assessment of truth becomes “relativized to a speaker”. […] It is an epistemic meaning that, as we have seen, involves the speaker’s tentative judgment of likelihood based on appearances as well as expectations and knowledge. (Traugott 1997: 198-199)

We agree that the meaning of evidential versprechen and drohen can be conceived as a “subjective” meaning as compared to their “objective” lexical meanings: as evidential markers, they relate the proposition to the inner world of the speaker, whereas as main verbs they describe a situation occurring in the “world out there”. Seen from this angle, the former variants of the verbs may be called more “subjective” than the latter.12 However, the notion of subjectification is taken to be a highly problematic one here, and we will not use this term as a central concept in our own investigation. Summing up this section on the research history on evidentiality in German we can make the following points: 1. It is under dispute whether there is a grammatical category of evidentiality in present-day German. 2. There is no agreement on the question whether these verbs and constructions should be seen as pertaining to the domain of epistemic modality, aspectuality, temporality, or evidentiality. 3. Though German is acknowledged to have verbal constructions with evidential meaning which emerged in the history of the German language, a synchronic as well as diachronic study on them in the light of the grammaticalization is still missing. 4. There is no extensive study that concentrates on the grammatical system of evidential markers of German (i.e. their interrelations, similarities, paradigmatic organization), or looking for a common diachronic developmental path.

3. Hypotheses and theoretical foundations This book intends to close the gap described at the end of the previous section and sets out to give a comprehensive description of the semanticfunctional domain of evidentiality in German, as far as verbal constructions are concerned. The hypothesis steering this endeavour is exactly that Ger-

24 Introduction man does have an evolving system of evidential markers. Thus, this book is intended to provide a study of the evidential system of German and its development with special attention to the semantics and frequency of particular evidential markers. The central hypothesis can be broken down into more specific hypotheses which we divide into three groups: 1. Hypotheses concerning the synchronic evidential system of PDG. 2. Hypotheses concerning the diachronic development of the four constructions into evidential markers. 3. Hypotheses concerning more general theoretical issues. Table 3 summarizes the hypotheses underlying this study: Table 3. Hypotheses of this study Central hypothesis: German is about to develop a system of evidential markers by grammaticalization of four constructions with infinitives and the verbs scheinen, versprechen, drohen and werden. Derived hypotheses: 1. Hypotheses concerning the synchronic evidential system of present-day German − The constructions werden & infinitive, and scheinen/drohen/versprechen & zu-infinitive are evidential markers in present-day German. − The four evidential constructions are grammaticalized to a relatively high degree. − They build a paradigm in German and represent distinctions within the domain of indirect inferential evidentiality. − German has developed a new grammatical category, the category of evidentiality, which is integrated into the growing system of verbal periphrastic paradigms in German. 2. Hypotheses concerning the diachronic development of the four constructions − Their diachronic development since OHG can be described as a process of grammaticalization. − The grammaticalization paths of these constructions can be traced with reference to known paths for the development of evidentials. − In German, there are two basic source constructions from which these inferential evidential meanings arise diachronically. As will be shown in the course of this book, one of them represents a successive relationship between two events and the other codes a simultaneous relationship between two events. 3. Hypotheses concerning more general theoretical issues

Hypotheses and theoretical foundations 25 − − − −

There is a clear conceptual distinction between evidential markers and epistemic modality markers (i.e modal verbs). Evidentials overlap with epistemic modality markers in particular areas in line with general cognitive principles. Evidentials, like epistemics, are deictic grammatical markers. The grammaticalization of evidentials in German can be described as a change of the semiotic status of the lexical non-deictic source items to grammatical deictic markers.

As we propose to treat the development of evidentials in German as a typical case of grammaticalization, grammaticalization theory, by necessity, is the dominant theoretical backing of this book (Hopper and Traugott [1993] 2003, Bybee 1994, Lehmann [1982] 1995, Diewald 1997). From a diachronic perspective, grammaticalization is a process whereby lexical entities develop grammatical functions in the course of time, or where elements which already display grammatical functions develop further or more central grammatical functions. As Lehmann (2004: 155) points out, the essence of the process can be characterized by two general features: the loss of autonomy of the linguistic material involved and the integration into the obligatory rules of the grammatical system, i.e. into a paradigm. Furthermore, linguistic constructions with an increasing grammatical functionality tend to converge into a common grammaticalization path or channel. As has been shown in typological and grammaticalization studies, grammaticalization channels are a universal phenomenon, resting on the fact that particular grammatical categories typically derive from particular source lexemes or source structures and evolve via particular chronologically ordered stages which display specific characteristics (Bybee, Perkins and Pagliuca 1994, Heine, Claudi and Hünnemeyer 1991, Lehmann [1982] 1995, Diewald 2002). The establishment of grammaticalization paths is one of the major concerns of diachronic grammaticalization studies. In accordance with this – and as mentioned above – one of the central aims of this book is to establish possible grammaticalization channels for evidential markers in German. As grammaticalization is a particular type of language change, the general tendencies and motivations for change in linguistic systems constitute the overall background of this study. The position taken here is a clearly functional one, which is the only one compatible with grammaticalization theory and with the claim to integrate universal and category specific characteristics with the factual particulars of the German language.

26 Introduction Furthermore the perspective taken here is in part compatible with notions developed in the last decades under the label of construction grammar. The issues susceptible to constructional notions are the impact of contextual factors in linguistic changes, and the notions “context types” and “constructions” which play a role in the description of the successive diachronic stages that are associated with grammaticalization processes (cf. Bybee, Perkins and Pagliuca 1994: 11, Bisang 1998: 20, Traugott 2003, Fillmore, Kay and O’Connor 1988, Fillmore and Kay 1995, Goldberg 1995, 2004, Kay and Fillmore 1999, Michaelis 2004, Diewald 2002, 2006). It will be argued that context types are diachronic variations of specific constructions and that some of the central notions of construction grammar can be applied in order to develop a more systematic and general way to describe these types of context, and to help solve some notorious problems of investigating diachronic change in general and the development of evidential markers in German in particular. As mentioned above, evidentiality has been first discovered by typologists working with non-Indo-European languages, and the impressive body of typological work on this topic is the substantial basis of our classification of evidential distinctions in German (see in particular Anderson 1986, Willett 1988, Plungian 1998, 2001 and Aikhenvald 2003, 2004). Furthermore, there is a close link between grammaticalization studies and typology, and there are known paths of grammaticalization of evidential systems leading from diverse lexical sources and their particular constructions to different subtypes of evidentials (cf. e.g. Anderson 1986, Aikhenvald 2003, 2004). For instance, direct evidentials often develop out of verbs of perception and locative markers; indirect evidentials evolve from perfects, resultatives and past tenses; reportive evidentials come from verbs of speech etc. The essence of all these studies is that evidentiality is a much more important and prominent category than scholars used to assume. In the following we will show that, in their diachronic development, the German evidential constructions derive from characteristic lexical sources and follow specific patterns of change. In addition to the synchronic evidence these diachronic facts strongly support our view that the four German constructions under consideration here are evolving evidentials. Typological support for our hypotheses comes from the fact that the development in German is not unique. As recent studies on different European languages have revealed, similar (lexical) expressions have been developing evidential functions, most often the cognates of the English verbs seem, promise and threaten (see especially Traugott 1997 for English, Cornillie 2007 for Spanish, Squartini 2001, 2004 for Italian and other

Methodology and corpora 27

Romance languages, Giacalone Ramat and Topadze 2007 for Italian, Verhagen 2000 and Vliegen 2006 for Dutch etc.). These verbs have been developing towards evidential (semi-)auxiliaries indicating indirect, i.e. inferential, evidentiality. Such parallel developments among several languages suggest that there is a common developmental path for evidential markers (which is productive at least in European languages) that has not yet been recognized and sufficiently described so far. We will draw on these studies whenever it is illuminating for our purpose. By conducting a large-scale corpus-based study employing different text types and/or registers, it is possible to determine to what extent German has grammaticalized the expression of evidentiality in its diachronic stages, and to show which roads the development has taken. With these new data and results, it will be possible to compare German to other languages with evidential systems as to the degrees of grammaticalization, the patterns of use and meaning extensions of evidential markers. Moreover, as the German system is a complex one and displays some interesting idiosyncrasies, its thorough exposition and description will provide a valuable in-depth case study for typological comparison, which will lead to further refinement of general assumptions on the developmental paths of evidential markers. In particular we hope that the hypotheses and arguments presented in this book will contribute to the general discussion on evidentiality, especially on evidentiality in European languages. Finally, if our hypotheses prove right, grammatical descriptions as well as diachronic accounts of German will have to be supplemented by a chapter on evidentials markers.

4. Methodology and corpora As sketched in the last section, this study combines several theoretical strands of modern linguistic research, i.e., grammaticalization theory, functionalist models of language change, construction grammar and typology, in order to set up a scenario for the development of verbal evidential constructions in German. Furthermore, the theoretical approach is underpinned with extensive empirical work done as corpus analyses, including a diachronic study and a synchronic analysis of the system of evidentials in present-day German. The task of the corpus studies is two-fold: for one part, they are to test the validity of the hypotheses put forward, for the other part, the corpora supply valuable material for illustrating the richness of the category of evidentiality in German and thus demonstrate the impor-

28 Introduction tance of this neglected category. The following paragraphs give an overview of the corpora used, the principles of the empirical work, and some general information on how the data where processed.

4.1. The data sources of the present study Data for this study were drawn from a number of German text corpora. As the focus of this study is on both the historical development of German evidential constructions and their synchronic status, the corpora used represent various historical periods of German as far back as the Old High German period. The core of the empirical work is a corpus based investigation on the rise and development of the evidential markers from periphrastic verbal constructions in the history of the German language, whereby the full span of the history of German, i.e., from Old High German of the 8th century to PDG is considered. The focus is on the verbal periphrastic constructions with the verbs werden, scheinen, drohen, versprechen. Apart from them, additional forms such as AcI-constructions with perception verbs, ingressive participial and infinitive constructions in earlier stages of the language will be occasionally included in the scope of the study, as they in some cases served as predecessors of the present evidential markers, but were not (fully) integrated into the grammatical category of evidentiality in German. For the most part we refer here to the traditional division of the German historical periods into four larger segments, four idealized successive diachronies: (i) Old High German (Althochdeutsch, henceforth OHG) covers the period from 750 to 1050, (ii) Middle High German (Mittelhochdeutsch, henceforth MHG) covers the period between 1050 and 1350, (iii) Early New High German (Frühneuhochdeutsch, henceforth ENHG) is defined as the period between 1350 and 1650, and (iv) New High German (Neuhochdeutsch, NHG) is the term used for the most recent period in the history of German, starting from 1650. For the purposes of this study, we divided the last period into two subperiods. On the one hand, data up to the 19th century were collected and analyzed separately (when needed, see below) in order to discover relevant changes that occurred to the verbs scheinen, drohen and versprechen. On

Methodology and corpora 29

the other hand, data from the 20th and 21st centuries were taken as the basis for the synchronic analysis. Text material from present-day corpora of German was investigated in order to detect the current degree of grammaticalization of each of the four constructions and, furthermore, to analyze the present-day system of evidential distinctions. In the following, this period will be referred to as present-day German (henceforth PDG). The following paragraphs introduce the corpora that were used to obtain data for the diachronic as well as synchronic analyses. We will also briefly outline the motivations behind the use of these corpora. The OHG data analyzed in this study were drawn from the KALIcorpus. The KALI-corpus is a diachronic corpus of German containing 25 representative texts from the OHG and MHG periods (roughly 250,000 tokens). The earliest text dates from around 790. Some texts are completely glossed using an adaptation of the Leipzig Glossing Rules13 in order to make at least some of the texts accessible to researchers less familiar with the diachronic stages of German. The corpus has been designed for the study of grammaticalization phenomena in German over the last years. As setting up a diachronic corpus is a large and laborious task, it is in place here to spend some words on the reasons for venturing on that undertaking. Although many historical texts already have been digitized, e.g., in TITUS,14 Bibliotheca Augustana,15 MHDBDB16 etc., the present situation of diachronic corpora of German is not yet satisfactory. Indeed, there exist a number of text collections of different sizes and formats. However, there are neither common standards for digitization, meta-information, or text annotation; nor are there commonly used, unified search interfaces. Furthermore, many of the digitized texts are not available to the general public. A large project, DeutschDiachronDigital (DDD), has been planned during the last years. It aims at creating a generally available, unified resource with common standards and search programs (cf. e.g. Lüdeling et al. 2004). However, the project is still at its initial stage and its resources are not yet available. For these reasons we decided to compile a (new) diachronic text corpus which is focussed on our particular research interests on the one hand, and is oriented towards common standards and formats of digitizing and annotation on the other.17 All texts in the corpus are partially morphologically annotated. As we aim to study the grammaticalization of German evidential constructions, the annotation of the corpus data, which constitutes the fundamental empirical basis of the study, has been restricted to verbal units (finite and infinite) in their occurrences through all documented periods. The morpholog-

30 Introduction ical annotation parameters determine a hierarchical tag-set containing the grammatical categories of finiteness, inflection, tense, mood, person, and number (cf. Lehmberg, Diewald and Smirnova 2007). Each verbal token is thoroughly analyzed by tagging it with information regarding person, number, tense and mood. In addition to this the matching headwords and translation terms are added to each token. Examples from the KALI-corpus cited in this study will be indicated by the author and the position of the token within the corpus (which is mostly the position of one of the four verbs werden, scheinen, versprechen and drohen).18 As the amount of the MHG data gained from the KALI-corpus is relatively small, we collected further MHG data from another corpus of this period of German. The Mittelhochdeutsches Wörterbuch19 (hereafter MHDWB) is an online dictionary of MHG containing the German vocabulary from the period between 1050 and 1350. It is based on a text corpus composed of MHG texts of various types. Though the text corpus itself is not available for search and reading purposes directly, a list of examples from the corpus for each lexeme is obtainable via the Belegarchiv (‘examples data base’). That means that the lexemes stored in the dictionary are connected to a digitalized text archive containing the data from the text corpus. Thus by searching for a particular lexeme one can also receive a list of examples from the corpus illustrating the contexts of use of this lexeme. Unfortunately, there is no information concerning the overall number of tokens in the corpus. The corpus consists of around 1300 texts. The complete list of the texts is available on the project webpage. Examples cited in this book will be indicated according to the (internal) abbreviation standards of the MHDWB which identify the text title and the position of the example in the text. By collecting the data from the ENHG period, we primarily consulted the Bonner Frühneuhochdeutschkorpus (hereafter B20). This corpus contains 40 representative texts of the ENHG period, which are arranged according to dialect areas. All tokens in the texts are annotated regarding their word class membership and (partially) their morphological form. There are roughly 608,000 tokens in the corpus (an exact count is unavailable). As every example can easily be traced back in the online available corpus, and as the texts in the corpus do not exceed the size of about 30 pages, examples cited in this study will be indicated only by the text number (e.g. B 141).

Methodology and corpora 31

Additional ENHG data were drawn from the Bibliotheca Augustana21 and the Glonings Corpus.22 Examples taken from these corpora will be marked by ‘BA’ (for Bibliotheca Augustana) or by ‘G’ (for Glonings Corpus) and by the text title. As noted above, for the purposes of this study, we divided the New High German period into two sub-periods. This is due to the following considerations: First, the development of the evidential construction with the verb scheinen started in the ENHG period. In PDG, however, the construction is already highly grammaticalized as an evidential marker, as has been revealed from the analysis of the present-day data. In order to reconstruct the intermediate stages in the development of this construction, we looked at the data from the 18th century (and partly from the 19th century) to detect relevant changes leading to the establishment of its evidential function. Second, the infinitive constructions with drohen and versprechen are the youngest members of the today’s evidential paradigm in German. Their development towards evidential markers did not start until the 18th century. For this reason, the data from the earlier periods of German could not be used to trace back the exact developmental path of these two constructions. Therefore, data of the 18th and 19th centuries were collected and analyzed more carefully. For the 18th century data we consulted the webpage of the Projekt Gutenberg-DE.23 Examples cited will be indicated by the abbreviation ‘GDE’, by the author and by the text title. The data of the 19th century originate from the sub-corpora available as parts of COSMAS-II (Corpus Search, Management and Analysis System) through the Institut für deutsche Sprache Mannheim24. We used the HiK sub-corpus (i.e. historical corpus) consisting of the texts by the authors Karl Marx and Friedrich Engels. This corpus contains ca. 2,900,000 tokens. Moreover, we used the texts by the authors Johann Wolfgang von Goethe and Jacob and Wilhelm Grimm containing roughly 2,100,000 tokens.25 Citations will follow the (internal) COSMAS-II standards. The present-day German data come from the DWDS corpora. The DWDS26 (Digitales Wörterbuch der deutschen Sprache ‘Digital Dictionary of the German Language’) provides a sample of roughly 100,000,000 tokens of spoken and written German of the 20th and 21st centuries. Different text types are represented in this corpus. The indication ‘Kernkorpus’, which will be used to mark the citations in this book, refers to the subcorpus of written data, whereas ‘Gesprochene Sprache’ refers to the subcorpus of spoken data.

32 Introduction Given the overwhelming size of the corpus, only random samples of the four verbs were used in this study. The samples contain 1,000 randomly selected examples per verb. Apart from the DWDS corpora, some additional sources have been used to collect more incidental data. So, we have occasionally made use of language material found on the Internet (i.e., we ‘googled’ some expressions), which is, if not a very reliable, certainly an extremely rich resource of highly informal and unmonitored language use. We have tried to make sure that Internet examples could be traced to native speakers. For convenience we conclude this section with Table 4 sketching the most important databases that were used for the present study. The right column gives the numbers of examples that were extracted from the corpora and analyzed. Table 4. Main sources used in the present study and the overall number of examples analyzed historical period of German

database (number of tokens, if available) Old High German (OHG) KALI (ca. 250,000) Middle High German KALI (ca. 250,000) (MHG) MHDWB (---) Early New High German B (ca. 608,000) (ENHG) 18th + 19th centuries G-DE (---) IDS (ca. 5,000,000) present-day German DWDS (ca. 100,000,000) (PDG)

examples extracted and analyzed ca. 500 ca. 1,900 ca. 4,500 ca. 4,500 ca. 4,000

It should be noted here that, starting from the period of the 18th century, in which a huge amount of digitalized language data is available, we have tried to get a balanced set of data. To that end, we restricted the data subsets of each verb to about the same size. In doing so we adjusted the number of examples to be analyzed to the amount of examples we gained for the ENHG period, i.e., to ca. 4,500. As concerns the earlier periods, i.e., OHG and MHG, we investigated all the examples we could extract from the corpora.

Methodology and corpora 33

4.2. Principles and method of the corpus study As we have mentioned above, the theoretical considerations presented in this study will be underpinned with extensive corpus analyses, including a diachronic study as well as a synchronic analysis of the system of evidentials in PDG. The empirical part of this study consists thus of two basic parts: a clearly diachronic corpus study, on the one hand, and a synchronic corpus study, on the other. Surely, these two studies are supportive of each other, so that we can speak rather of a combination than of a division of the two approaches. In other words our approach here combines diachrony and synchrony. Our investigation rests upon the following general assumptions. First, the grammaticalization of the verbal constructions under investigation proceeds diachronically in certain local linguistic contexts. As is generally known, a continuous interplay between the grammaticalizing construction and the context may lead to the development of new constructional meaning (cf. e.g. Diewald 2002, Heine 1993, 2002). Constructions which are free syntactic structures at the beginning of their grammaticalization coalesce over time into verbal periphrases. Hence, periphrastic constructions develop and become grammatical. It is of particular importance that a whole periphrastic construction consisting of a single verb plus its infinitive (or participle) complement undergoes a process of change. Therefore, investigating the grammaticalization of verbal evidential constructions in German, we have to look at these verbs and their syntagmatic relations to other constituents of the context. Second, we believe that diachronically observable changes in morphology and syntax of the elements under consideration follow or parallel their semantic reinterpretation processes, which can abe characterized in a summarizing way by the notions of desemanticization and increased noncompositionality. Therefore, we may reconstruct semantic shifts by retracing formal, i.e., morphological or syntactic shifts, diachronically. That means that shifts on the formal “surface” which can be retraced diachronically offer valuable clues to shifts on the “deeper” conceptual-semantic level. In what follows a few paragraphs will be spent on the principles of our corpus study which are in accordance with the general assumptions sketched above. The diachronic corpus-based approach is realized here by the combined use of different historical corpora and one present-day corpus of German. For each historical period the data are first analyzed independently. That is, a synchronic analysis is carried out for the linguistic

34 Introduction elements under investigation in each of the different historical periods of German taken here into account. In doing so the overall scope of the construction types of the element in question can be described: variable readings, different morphosyntactic properties, and individual patterns of use are discovered for each element and its (synchronic) variants. The term “construction type” is used as a non-technical, theory-independent term in order to describe the usage patterns of each item, which are composed of any kind of linguistic features (among them valency frames, topological patterns etc.). Sections 8.2.3 to 8.2.6 of this book, for example, will deal with the verb scheinen in the history of German. They are devoted to presenting our findings of such synchronic analyses for each individual historical period. The construction types of the verb scheinen with their characteristic properties and in their interconnections are described in these sections. Additionally, prototypical contextual configurations in which the discovered construction types were used are given particular attention. In this way, the descriptions of the whole paradigm of each verb (i.e., werden, scheinen, drohen, and versprechen) in each particular historical period of German will be offered in the individual sections of Chapter 8. In the next step the results of the synchronic analyses of the consecutive historical periods are compared to each other in order to figure out whether any relevant changes in the overall system had occurred. Here the focus is on the diachronic issues. The aim is to discover specific contextual configurations in which existent construction types most probably received new interpretations and in which new construction types arose. It should be noted here that only new (or unusual) contexts are considered as relevant contexts, whereby we regard only those contexts as unusual and new, which (i) trigger a new interpretation of the whole construction, and, what is more important, (ii) may be seen as precursors of further new and unusual constructions, i.e., which can be considered to have influenced the overall development of the element in question. The results of such comparative work will be presented in the final sections of Chapter 8. Section 8.2.7, for example, gives a summary of the relevant changes which occurred to the verb scheinen in the course of the history of German. The aim of this kind of representation is two-fold. On the one hand, the grammaticalization path of the evidential construction scheinen & zu-infinitive will be extracted and presented as an independent development of scheinen. On the other hand, the grammaticalization process will be integrated in the changes concerning the whole constructional scope of the verb scheinen. The diachronic development of each verb will be presented in this manner in Chapter 8.

Methodology and corpora 35

In the final step, the common developmental path for the German evidential markers will be set up. Here, not the individual constructions and their diachronic developments are in the focus of our interest. Instead, the entire paradigm of German (inferential) evidentiality will be accounted for in terms of its grammaticalization. In this vein we attempt to establish a common grammaticalization path for the German evidential markers, whereby the peculiarities of the individual constructions will be abandoned. In so doing, we will be concerned (i) with the general properties of the (lexical) sources which gave rise to the establishment of evidential meanings, (ii) with regularities of grammaticalization of the whole paradigm, and (iii) with main mechanisms of change being at work in the grammaticalization of the grammatical category of evidentiality in German. By establishing a new grammaticalization path for a previously neglected German category we will contribute to the overall research on grammaticalization and typological research. In addition to the extensive diachronic corpus research a comprehensive analysis of the present-day data will be undertaken. The aims of this analysis differ from the purposes of a diachronic account in many ways. While investigating synchronic data of PDG we are not particularly interested in the overall distribution of different construction types of each verb. Instead, the focus will be on the evidential constructions only. We will be concerned with the contexts of use in which the evidential meaning is predominantly found. To that end, we first ascertain the degree of grammaticalization of each evidential construction. For this we contrast the newly grammaticalized variants of the verbs (within infinitive constructions) with the original lexical variants of the verbs, which – for all verbs – have been preserved till PDG. Second, we investigate how the present paradigm of inferential evidentiality is structured, i.e., how the evidential constructions under investigation are linked. For instance, in Section 6.2 we will deal with the synchronic status of the evidential construction scheinen & zu-infinitive. A brief outline of the construction types of the verb scheinen will be given at the beginning (see Section 6.2.1), in order to set the boundaries between the evidential and other synchronic variants of this verb. In the following sections, semantic and morphosyntactic properties of the evidential construction, its specific contexts of use and their distribution, and the interconnections between different alternative readings of scheinen & zu-infinitive will be discussed

36 Introduction in greater detail. Other evidential constructions will be described in a similar manner in Chapter 6. A summary of our synchronic corpus analysis will be presented in Chapter 7, where the internal organization of the German evidential paradigm will be described and where this paradigm will be compared to other evidential systems known from the typological studies. Language-specific observations will be thus integrated into the broader typological perspective. Some general remarks concerning the principles and methods of our corpus research are in place. It should be noted that the corpus analysis conducted here is mainly qualitative. This means that it is oriented towards careful investigation of contextually induced changes leading to the development of evidential functions. Therefore, the corpus studies do not focus on statistical issues, but on context-sensitive thorough analysis of the items in question as they appear in our material. We are convinced that grammaticalization studies require intensive in-depth linguistic work, which is not restricted to a quantitative assessment of given linguistic forms and structures, but goes on to qualitative text interpretation in certain linguistic contexts. To make plausible statements on the basis of the mainly qualitative corpus analysis is thus the main intention of this study. Our concentration on qualitative analysis is also due to the small amount of texts passed down from earlier periods (especially from OHG). Diachronic corpora cannot hope to be representative in the full sense of the word, so that a quantitative analysis of such corpora would not be representative either. The texts and text segments collected from different corpora were fuly investigated and every token of the items in question was taken into account. Still, the numbers, in particular in the OHG period, are small and surely not apt for quantitative statements. Analyzing the data obtained from different corpora, we concentrated particularly on the following issues: (i) the chronology of appearance of each relevant construction was ascertained; (ii) the different distribution patterns of single construction in their diachronic development were investigated; and (iii) the changes on all levels were interpreted in the sense of grammaticalization theory (the general underlying principles will be presented in Chapter 4). The actual work on the corpora – anachronistically – started with the present-day data, as they manifest the actual stage of grammaticalization, i.e., the system of evidential marking that we have in German today. From this present-day scenario we formed our hypotheses on possible developments as well as on the possible age of the relevant forms. These forms

Structure of the book 37

were then investigated in the diachronic corpora in the manner described above. We started from the OHG period and went forwards on the time line. As already mentioned above, though the focus lay on the development of the four verbs which form the lexical core of the evidential constructions, we also – in single cases – investigated constructions which are important for the development of our forms although they only played a subordinate role for a small period of time and did not develop into grammaticalized evidentiality markers. These are, in particular, the ingressive constructions of OHG with the verbs gistantan ‘to stand’ and beginnan ‘to begin’, which are combined with infinitives in sentences like the following: (14)

do ih riuuon gestuont die sunda (Notker II,15,21 taken from Behaghel 1924 [II]: 311) ‘as I began to regret the sins’

Our decisions which of these evanescent, extinct forms that cropped up in the history of German should be treated were guided by our own observations of spectacular patterns and constructions of semantically or structurally relevant features that we made while working with the corpora. Last but not least, it has to be mentioned that while the data have been thoroughly investigated and all evidential forms appearing in them have been considered in this study, in some places additional linguistic data found outside the corpora are drawn on if they reveal phenomena not represented well enough in the data obtained from the sources used. Citations from other works will be explicitly indicated.

5. Structure of the book While this first chapter is intended as an overview, Chapters 2 to 5 form the theoretical core of this book, which is steered by the following necessities. First, it will be necessary to elaborate a comprehensive definition of the semantic-functional domain of evidentiality in order to reach clarity and explicitness in the linguistic metalanguage used to describe evidential distinctions. This is done in Chapters 2 and 3, which offer a detailed review of current definitions of evidentiality on the basis of typological findings, and argue for a clear separation from other domains. The results of prior research evaluated in these chapters are measured against the actual

38 Introduction situation in German and the requirements of this study. At the end of these two chapters we reach a typologically based definition and classification of evidentials, which at the same time is adapted to the peculiarities of German in order to serve as the theoretical blueprint of the empirical study. Thus, the outcome of these two chapters is a set of semantic labels or markers and a classification of relevant distinctions by which the German system of evidential distinctions can be described in its synchronic state as well as in its development through time. The second major untertaking is to provide an overview of the current system of evidential marking in PDG, to establish relevant distinctions (to verify the hypotheses of the previous chapters) and to prepare the ground for the diachronic chapters. As this topic is closely connected to the extent of grammaticalization of evidential markers in PDG it calls for an elaboration on the indication of the degrees of grammaticalization for different constructions; semantic paraphrases, internal distinctions as well as the integration into the paradigm of modal verbal constructions and the category of verbal mood. Chapters 4 and 5 are devoted to these tasks, i.e. marking our positions concerning language change and grammaticalization. Here, the major theoretical issues that have to be decided on in order to pursue our aim are discussed and applied to our research object. For the purposes of assessing degrees of grammaticalization Lehmann’s grammaticalization parameters are applied to exemplary evidential constructions. The regularities of semantic change and paths of grammaticalization are expounded and again projected onto our object of study. Finally, a model of relevant constructions and their succession is developed on the basis of a combination of grammaticalization paths and central notions of construction grammar. Chapters 6 to 9 are devoted to the empirical investigation of the four evidential constructions in German. Chapter 6 presents the synchronic data gained by the investigation of the present-day German corpus. It provides a sketch of the system of evidential distinctions as it is found in German today and thus presents a picture of the current state of the target category of the process of grammaticalization of evidential markers in present-day German. Chapters 7 and 8 delineate the details of the development of each marker as it can be extracted from the diachronic corpus. This part of the book summarizes the development of the whole system from its OHG beginnings to the situation today: a scenario of the diachronic events and determining factors of the grammaticalization channels involved in the grammaticalization of the German evidential constructions. The last chapter summarizes our findings.

Structure of the book 39

It should be noted that one of our methodological principles is to continually adjust empirical and theoretical aspects to each other. This means that although there is a preponderance of theoretical issues in the first half of the book and a preponderance of empirical data in the second half, both aspects are closely intertwined, and on several occasions the theoretical chapters anticipate empirical data in order to model a position which will be justified by corpus data only in the second half of the book. Vice versa, the empirical chapters contain theoretical discussions and further reflection which are directly derived from the discussion of the data presented.

Chapter 2 Evidentiality – definitions and delimitations

As the field of evidentiality is a relatively new one, not all terminological conventions are accepted by everyone and there are overlapping distinctions in some areas. In order to avoid misunderstandings, this chapter takes up the discussion on labels and distinctions used for describing semantic evidential values in typological literature.1 The purpose of this endeavour is to finally come up with those distinctions that are useful for a correct and consistent naming and classification of evidential meaning distinctions (and ultimately evidential markers) in German. However, it should be noted that though some terminological choices may be arbitrary in the sense that it in fact does not matter which term to choose, there are others where the terminology chosen is indicative of an underlying theoretical conception.

1. Evidentials, evidential strategies and pre-evidentials It is common ground among most linguists that evidentiality is a semantic and functional domain, which, like temporality or modality, may be expressed by lexical or grammatical means. That is, evidential distinctions may form a grammatical category in one language, while in the other they are not obligatory and may be expressed by a variety of linguistic options. As there are several grammaticalization channels leading towards grammaticalized evidential systems we should be prepared to be confronted with all sorts of intermediate stages. Thus, evidential expressions may belong to the lexical as well as to the grammatical layer of a language, or to some intermediate stage between the two. As mentioned in Chapter 1.1, we follow the convention of talking of evidentials or evidential markers only in those cases which we regard as grammatical categories or grammaticalizing expressions of evidentials (see below for further discussion). In doing so we follow Anderson (1986: 273274), who restricts the term “evidentials” to linguistic expressions representing “a special grammatical phenomenon”. In the last decades typologists have described numerous languages of various non-related families which realize evidentials in the strict sense of

Evidentials, evidential strategies and pre-evidentials 41

the term. These evidential markers form grammatical paradigms, interact with other grammatical categories (notably mood and aspect) and show a diachronic development along the lines of known, category-specific channels of grammaticalization (Aikhenvald and Dixon 2003, Chafe and Nichols 1986, Johanson and Utas 2000, Nuyts 2001, van der Auwera and Plungian 1998). A large number of examples of fully grammaticalized systems of evidential markers is given in Aikhenvald 2004 (see Chapter 1.1). Irrespective of the existence of a grammaticalized system of evidentials, every language has a variety of non-grammaticalized expressions and constructions carrying evidential meaning. Therefore, in order to be able to talk about our topic in an unambiguous way, this section is devoted to drawing the line between evidentials and all other kinds of linguistic material, which may be used to express evidential meaning in certain surroundings.

1.1. Terminological distinctions in synchronic perspective The term evidential expression is a neutral label (a hypernym) used to denote any kind of linguistic string with evidential meaning in a particular context, regardless of its linguistic structure and degree of grammaticalization. On the next level we distinguish between evidentials (evidential markers), i.e. grammaticalized evidentials (secondary predications denoting the source of knowledge for a proposition, Anderson 1986: 274; see Section 1.1.1) on one hand, and what Aikhenvald 2004 calls evidential strategies on the other. These latter are constructions which do have evidential meaning but do not pertain to a grammaticalized system. This fundamental distinction between evidentials and evidential strategies is given in Figure 5.2 The distinction may be illustrated by examples from PDG. As we have already claimed in Chapter 1 and as we will prove in the course of this book, the constructions werden & infinitive and scheinen/drohen/versprechen & zu-infinitive are evidentials in present-day German. (1) a. Der Wasserspiegel wird steigen. ‘The water level is bound to rise.’ b. Der Wasserspiegel droht zu steigen. ‘The water level threatens to rise.’

42 Evidentiality – definitions and limitations c. Der Wasserspiegel verspricht zu steigen. ‘The water level promises to rise.’ d. Der Wasserspiegel scheint zu steigen. ‘The water level seems to rise.’ evidential expression any linguistic items being used in a particular context to express evidential meaning

evidential strategies non-grammaticalized linguistic devices for expressing evidential meaning Figure 5.

evidentials = evidential markers expressions for evidential meaning which are grammaticalized (to a certain degree)

Terminological distinctions in the domain of evidentiality I (selected aspects)

These constructions constitute the grammatical core of evidentials in German, i.e. the automatised, routinized, backgrounded method of expressing evidential distinctions in German. The tests to determine the degree of grammaticalization of the more advanced stages of developing evidentials are the grammaticalization parameters set up by Lehmann (1985), which will be explained in detail and applied to the four evidential constructions in Chapter 4. Here, it may suffice to state that the verbal periphrases3, i.e. the four infinitive constructions we are primarily dealing with, show a high degree of grammaticalization in comparison to other constructions being used as evidential expressions. As to the open class of evidential strategies, there is a whole range of expressions belonging there; this has been pointed to before (in Chapter 1.1) and is illustrated here again by the following examples: (2) a. Sie hört ihn die Treppe heraufkommen. ‘She hears him come up the staircase.’ b. Sie fühlt den Schmerz hinter den Schläfen pochen. ‘She feels the pain throbbing behind her temples.’

Evidentials, evidential strategies and pre-evidentials 43

c. Sie fängt an, ihren Umzug ins Ausland vorzubereiten. ‘She starts preparing to move abroad.’ d. Sie ist dabei, ihren Umzug ins Ausland vorzubereiten. ‘She is about to prepare to move abroad.’ e. Offensichtlich hat er es getan. ‘Obviously, he has done it.’ Among evidential strategies, there are the AcI-constructions (2a, 2b) as well as ingressive and progressive constructions (2c, 2d), and there are adverbials acquiring evidential meaning in certain contexts/constructions. The AcI-constructions, which are the most obvious peripheral evidential constructions of PDG, have stayed in the language since the OHG period, but have been reduced in frequency (type as well as token frequency). As to the ingressive and progressive constructions, there have been several of them in the history of German, some of which have been extant to the present day, some of which disappeared in the course of time. These constructions have to be classified as peripheral evidential constructions of their respective periods. They represent the recognizable part of the reservoir of potential sources for grammaticalization towards evidentials at a certain stage, which, however, have not (yet) been exploited for that purpose, i.e. have not been grammaticalized (further). As we will see later, these constructions may play an important role in the genesis of evidential systems as they function as a kind of catalyst. They, too, are evidential expressions in PDG, having two predications and naming the perceptual sources due to which the scene described in the core proposition is presented the way it is. As will be shown later, they are typical peripheral constructions, which are/were potential competitors but are not integrated in the present-day grammaticalized system. Constructions like (2d) show that evidential functions may not only be realized by single morphemes or periphrastic constructions, but show up at the interface of syntax and semantics in particular, highly specific contexts or constructions, like the “Verlaufskonstruktion” dabei sein, etw. zu tun ‘to be about to do something’. Furthermore, there are constructions using semantically suitable evidential adverbials like offensichtlich ‘obviously’ in (2e). Examples like (2e) have a constructional component, as the position of the adverbial is relevant for its most likely meaning, compare (2f) to (2g):

44 Evidentiality – definitions and limitations (2) f. Er hat es offensichtlich ja getan. ‘Obviously, he has JA done it.’ g. Er hat es ja offensichtlich getan. ‘He has JA done it in front of everybody.’ While offensichtlich in (2f) has scope over the modal particle ja and thus clearly evidential meaning, (2g) may be given a different reading: as the MP ja has scope over offensichtlich the latter can be interpreted as a verb phrase adverbial ‘visible to everyone’/‘so that everybody can see’. This observation can be linked to the universal, topological restrictions concerning evaluative, evidential, and epistemic adverbials and modal particles, which where first observed by Cinque 1999 and can be related to grammaticalization data on evidential constructions in the verbal domain (cf. Speas 2004, Pittner 1999: 178). In (2e), on the other hand, there is no ambiguity concerning different readings: the evidential reading of offensichtlich is the only one possible, as long as it does not bear contrastive stress. As the evidential strategies comprise a very large group of diverging forms, it will be necessary to subdivide this class further. In order to be able to talk about synchronic differences among the evidential strategies, we need to distinguish peripheral evidential constructions, which have already been mentioned from evidential periphrasis. While evidential periphrases (see Note 3) are expressions used ad hoc to encode evidential meaning (via conversational implicature), peripheral evidential constructions are those which are identifiable constructions stereotypically expressing evidential meaning which might develop/have developed into evidential markers, but are/have stayed peripheral to the systems, or have been reduced to a peripheral status in the course of history. They are not predecessors/source constructions of the evidential markers of the synchronic system, but potential or former competitors to them. An excellent example for peripheral evidentials are the the AcI-constructions of PDG. Examples (2a-b) are taken up here again as (3a-b) and supplemented by further AcI-constructions. While (3c) with the visual perception verb sehen ‘see’ is a common and stereotypical construction (as are 3a-b), (3d) with the olfactorial verb riechen ‘smell’ is less acceptable, and (3e) with the verb bemerken ‘remark’, ‘notice’ is not accepted as a correct construction by most speakers of German today. (3f) finally, with verbs of thinking and assuming is clearly inacceptable in the language of today.

Evidentials, evidential strategies and pre-evidentials 45

(3) a. Sie hört ihn die Treppe heraufkommen. ‘She hears him come up the staircase.’ b. Sie fühlt den Schmerz hinter den Schläfen pochen. ‘She feels the pain throbbing behind her temples.’ c. Sie sah ihn die Treppe hinablaufen. ‘She saw him run down the staircase.’ d. ???Sie roch die Kartoffeln anbrennen. ‘She smelt the potatoes burning.’ e. *?Sie bemerkte die Kartoffeln anbrennen/ihn die Treppe hinablaufen. ‘She noticed the potatoes burning/him run down the staircase.’ f. *Sie vermutete/glaubte/wusste ihn die Treppe hinablaufen. ‘She assumed/believed/knew him run down the staircase.’ any linguistic item being used in a particular context to express evidential meaning

evidential strategies

evidential periphrases any ad hoc constructions expressing evidential meaning

peripheral evidential constructions stereotypical constructions expressing evidential meaning/ “would be” evidentials

evidentials

Figure 6. Terminological distinctions in the domain of evidentiality II (completed view)

In sum AcI-constructions in PDG are a fixed set of constructions with heavy semantic and syntactic restrictions and a clear evidential meaning, which however is not fully grammaticalized. As far as the history of AcIconstructions is concerned, they were much more frequent and had a much wider distribution and fewer restrictions as to the matrix verb in OHG and MHG. Considering their lexical source semantics and the potential of the whole construction, they in principle would have been apt candidates for grammaticalizing into full-fledged evidential markers. However, they de-

46 Evidentiality – definitions and limitations clined in the course of the development of the German language and today are restricted to more or less stereotypical usages. In short, peripheral evidential constructions are “would be” evidentials which, however, are not integrated into the present system of evidential marking. Therefore, we can complete our diagram as illustrated in Figure 6 above.

1.2. Terminological distinctions in diachronic perspective As we are also concerned with the diachronic rise of evidential meaning, and as diachronic development is typically gradual (if not in its onset than at least in its spread, like e.g. word order change), we need to distinguish between different chronological stages as well, i.e. we need a further label to name the predecessors of evidential markers. We call them preevidentials. The distinction between pre-evidentials and evidentials is one of chronological succession and descent: pre-evidentials are the known predecessors of evidentials, which do not yet display the relevant grammaticalization features, but do have an appropriate semantics to develop evidential meaning later on.4 In the German case, the diachronic lexical sources of evidentials are mostly verbs or verbal constructions. Thus, while verbal evidential markers are secondary predications which specify the information the speakers relies on for making his statement, verbal pre-evidentials are not; instead they are the lexical origin of verbal evidential markers. This means that the decision whether a construction qualifies as a preevidential can be made only ex post, i.e. from a retrospective when its grammaticalization was successful. The following OHG example of scheinen shows a pre-evidential construction with the original lexical meaning of the verb scheinen. (4)

Óstar filu férro/so scéin uns ouh ther stérro. (KALI, Otfrid, 12892) ‘In the east, very far … so shone for us also the star.’

Depending on the peculiarities of linguistic change in a particular instance, pre-evidential constructions may die out and leave the field to the grammaticalized constructions, or they may persist as normal lexical items, thus leading to a lexical split of the sign in question into two coexisting items: a lexical sign and a grammaticalized sign. This latter alternative has happened to all four evidential constructions we are dealing with here. A case

Evidentials, evidential strategies and pre-evidentials 47

in point is present-day scheinen, which exists as an evidential marker (like in (5a)) as well as a lexical verb continuing the old pre-evidential, i.e. lexical meaning as in (5b). (5) a. Sie scheint abgereist zu sein. ‘She seems to have left.’ b. Die Sonne scheint. ‘The sun is shining.’ In (4) and (5b), the verb scheinen is used as a verb of visual appearance. It constitutes the unique verbal centre of the sentence, i.e. it is used as a primary predication and thus characterizes the verbal scene, which is predicated of the subject. Evidential constructions, on the other hand, as already mentioned, form a secondary predication on a primary predication, i.e. the scene which is evaluated (Anderson 1986: 274, cf. also the analysis in terms of grammaticalization parameter in Chapter 4). The state of co-existence of lexical and grammaticalized usages (or constructions) illustrated here for scheinen also applies to the other three verbs in questions (werden, drohen, versprechen, see later chapters). The situation is similar for the peripheral evidential constructions. They, too, can be associated with pre-evidentials, and they may show the same type of split between lexical (pre-forms) and grammaticalized usages. This can be seen in the case of the verbs hören ‘hear’ and sehen ‘see’. The OHG examples in (6) show both verbs used as verbs of perception. This usage is present in modern German as well, as exemplified in (7). (6) a. Wir sáhun sinan stérron/thoh wir thera búrgi irron (KALI, Otfrid, 12863) ‘We saw his star, but we missed the town.’ b. wanta er ni hórta man thaz, Thaz io fon mágadburti/man gibóran wurti (KALI, Otfrid, 12813) ‘Because one never heard that before that ever a man was born by a virgin.’ (7) a. Wir sehen den Stern. ‘We can see the star.’ b. Wir hören die Worte. ‘We can hear the words.’

48 Evidentiality – definitions and limitations As noted before, these verbs, used in AcI-constructions like in (3), have developed into (peripheral) evidential markers. AcI-constructions with these verbs are known in OHG as well; number (8) illustrates this usage for see: (8)

Tien fórderên rédôn nemág íh uuídere sîn. chád ih. únde dísa sîderûn sího íh ín nôte fólgên. (KALI, Notker, 5897) ‘I am not able to argue against these earlier speeches, I said, and I see this later one follow them by necessity.’

This is to say that these peripheral evidential constructions have been in the language since its first attested documents. They have had an evidential function from their beginning, but did not develop into (members of) the major evidential category. Accordingly, the AcI-constructions in PDG have a lower degree of grammaticalization than the evidentials proper. Peripheral evidential constructions display a clear evidential meaning (and a certain degree of grammaticalization), other construction types, which we class among pre-evidentials here, show readings with ambiguity between evidential and non-evidential meanings, like the following ones:5 (9) a. Das klingt wie eine Novemberdepression. ‘This sounds like a November depression.’ b. Das hört sich ja recht gut an. ‘This sounds REFL.PRON JA quite good VERBAL PARTICLE.’ c. Das sieht wie eine echte Verschwörung aus. ‘This looks like a real conspiracy VERBAL PARTICLE.’ These sentences have only one verb (one predication) and therefore do not qualify as evidential constructions. Nevertheless, the perception verbs used here are ambiguous between perceptual meaning and evidential meaning. The verb klingen, which usually has the meaning of ‘making an acoustic impression’/‘emitting sound’ may also be understood here as ‘implying that’/‘giving reason to conclude’, as paraphrased in (9a1-2): (9) a1. ‘This gives an acoustic impression/emits sound like (I imagine) a November depression does.’ a2. ‘This is evidence for the proposition: ‘This is a November depression’’.

Evidentials, evidential strategies and pre-evidentials 49

The verb hören in the construction sich anhören wie (9b) is equally ambiguous between ‘hear’ in the sense of ‘perceive acoustically’ and in the sense of ‘getting to know’/‘being told’. Thus we get as alternative readings: (9) b1. ‘The accoustic quality of this is quite pleasant/good.’ b2. ‘The acoustic (linguistic) event is evidence for the proposition: ‘This is good’’. And, finally, the verbal construction aussehen wie ‘to look like’ is in principle ambiguous between ‘to produce the visual effect of x’ or ‘to give evidence for x happening/being the case’. (9) c1. ‘This gives the visual effect of a true conspiracy.’ c2. ‘There is evidence for the proposition: ‘This is a conspiracy’’. In short, there is ambiguity in the source lexical items, which is usually resolved by context and which we would usually put down to conversational implicatures (see also Chapter 5 on semantic change). Returning to the major aim of this chapter, laying out and defining useful terminology, we may note the following: the path from a pre-evidential to an evidential usually leads via an evidential strategy, i.e. a communicatively motivated pattern of usage in various stages of stereotypicality (see Chapter 5). As Figures 5 and 6 take a synchronic-systematic, i.e. stative perspective and do not display the dynamic dimension, the diachronic implications have to be captured by a separate diagram. This is done in Figure 7, whereby “t” represents an (idealized) stage, i.e. a distinct period, an assumed synchrony in diachronic succession, while the arrow represents the gradualness of the passage through time as well as of the development of linguistic signs. With these distinctions it is possible to capture the general relevant features of linguistic forms with evidential meaning under the synchronic as well as the diachronic perspective. Time line

t1

t2

t3

Linguistic status

pre evidentials >

evidential strategies >

evidentials

Figure 7. Diachronic development

50 Evidentiality – definitions and limitations 1.3. The dispute on the status of evidential expressions in European languages It will be appropriate here to briefly take up two issues raised by Aikhenvald (2004: 3-11), who argues vehemently against the existence of evidential markers in European languages. Her arguments concern first the relations between epistemic modality and evidentiality, and, second, varying degrees of grammaticalization in particular expressions. As we equally vehemently argue for the existence of evidentials in (at least) one European language, this matter deserves some attention. In her discussion on the distinction between evidentiality and epistemic modality Aikhenvald comments on the fact that modal verbs in European languages often have evidential overtones and usages. She points out that – contrary to the opinion of many scholars – this is neither an argument for treating evidentiality as a subcategory of epistemic modality nor for claiming that European languages have evidentials. True as this is, the following should be kept in mind: the fact that modals are not evidentials does not mean that European languages lack evidential systems, or evidential constructions (see Chapter 3 for a detailed discussion of the relations between modality and evidentiality). Furthermore, Aikhenvald emphasizes that the existence of adverbials with evidential meaning, which are the object of research of many scholars, is by no means proof of a grammatical category for evidential distinctions (2004: 5-7).6 Again, we perfectly agree with this view, and will elaborate on it in the following sections. However, we plead for a closer and more differentiated look at European languages, and in particular, we plead for a proper consideration of the fact that grammatical systems develop, and thereby give rise to systems which are not yet fully grammaticalized, but display varying degrees and stages of a grammaticalization process. Though Aikhenvald (2004: 11) admits this, she insists that only fully grammaticalized evidential systems are worth being treated under the label of evidentiality, and consequently proposes to restrict the term evidentials to Non-European languages which do have a fully grammaticalized inflectional system.7 The reason for this position (which in our view, if adhered to, would hamper linguistic research on evidentials), may lie in the fact that Aikhenvald herself does not always distinguish properly between structural issues and semantic domains. This becomes clear in the following quotation:

Evidentials, evidential strategies and pre-evidentials 51 In about a quarter of the world’s languages, every statement must specify the type of source on which it is based – for example, whether the speaker saw it, or heard it, or inferred it from indirect evidence, or learnt it from someone else. This grammatical category, whose primary meaning is information source, is called ‘evidentiality’. (Aikhenvald 2004: 1) [our emphasis]

Equating the grammatical category with the name of the domain, i.e. “evidentiality”, is analogous to equating the grammatical category “tense” with “temporality”.8 This tight and impermeable association of domain and grammatical category prevents the adequate evaluation of European evidential expressions (with their epistemic overtones, i.e. their semantic conflation) as well as an unbiased look at the chronological development, i.e. at the fact that grammaticalization by necessity implies forms not yet fully grammaticalized. Therefore, we would like to emphasize the following two positions as important for our work: Firstly, as evidentiality is a functional category, its investigation should not be restricted to inflectional morphology but consider all types of formal realization of the functional category. This is to say we should take care not to neglect the form-function distinction at any time. Secondly, grammaticalization is a gradual process. Therefore, in trying to investigate its details, it would be counterproductive to restrict attention to fully developed grammaticalized systems. In other words: Though Aikhenvald is absolutely right in emphasizing the necessity to distinguish between evidential grammatical systems on one hand, and optional expressions of evidential meaning on the other, it is a gross omission to neglect the gradualness of grammaticalization and to exclude all items with only intermediate degrees of grammaticalization from further consideration in the context of research in evidentials. The constructions we are investigating here, i.e. werden & infinitive and scheinen/drohen/verprechen & zu-infinitive, like many analogous constructions in other languages found in the Indo-European family, clearly are of an intermediate stage as concerns the degree of grammaticalization, i.e. they are not yet full-fledged grammaticalized evidential systems as compared to those systems invoked by Aikhenvald, which have inflectional or clitic evidential markers, but they are instances of evidential systems on the rise (e.g. Cornillie 2004, Dendale and Tasmowski 2001, Diewald 2004, Mortelmans 2000, Plungian 2001). The following sentences, which are partly repeated here once more from Chapter 1.1.1, illustrate several periphrastic constructions expressing evidential meanings in European languages:

52 Evidentiality – definitions and limitations (10) a. German Sie scheint ihren Schirm im Büro gelassen zu haben. b. Dutch Ze schijnt haar paraplu in het bureau te hebben gelaten. c. English She seems to have left her umbrella in the office. d. Spanish Parece haber dejado su paraguas en el despacho. As already explained (cf. Chapter 1.1.1), the evidential verbal constructions in these examples refer to the fact that the speaker derives his/her conclusion from some (visual) evidence s/he has for making the claim (namely ‘She has left her umbrella in the office’). From a semantic point of view, they fulfil the requirement of having “source marking” as their central feature; from a structural point of view, they are secondary predications, i.e. without testing them in detail on the basis of the grammaticalization parameters, we can state here that they are somewhere between full verbs with infinitive complements and auxiliary verbs (see Chapter 4.2). It is plausible to assume a relatively advanced degree of a grammaticalization, which means that the German constructions treated here show a medium degree of grammaticalization towards evidential markers.

2. Intra-evidential distinctions While Section 1 is concerned with formal distinctions between linguistic items of different degrees of grammaticalization, this section discusses the semantic distinctions and the possible subdivisions of evidentials meanings. We do not intend to give a full report on terminological problems here (cf. Jacobson 1986), but concentrate on working towards a sound terminological decision for describing the German system. As mentioned before, the basic semantic requirement for some form to be classified as evidential, is “source marking”. This is clearly stated by Aikhenvald: To be considered as an evidential, a morpheme has to have ‘source of information’ as its core meaning; that is, the unmarked, or default interpretation. Evidence for such an interpretation comes from various quarters, not least native speakers’ intuitions, and the possibility of lexical ‘reinforcement’. (Aikhenvald 2004: 3)

Intra-evidential distinctions 53

Beyond this common core, evidential expressions do not only signify the fact that the speaker has evidence, but also specify the type of evidence the speaker makes use of. They answer the question of “how the speaker obtained the information on which s/he bases an assertion” (Willett 1988: 55). From what has been said so far, we may derive the ingredients for a general description of a linguistically expressed evidential setting: 1. The necessary components of a systematic description of the linguistic

structure of evidential meaning are the speaker, the described event, the event functioning as evidence and the relation between these three elements (cf. Jakobson [1957] 1971). 2. Furthermore, the distinction between direct evidence, i.e. evidence based on direct and personally acquired sensual perception, and indirect evidence, i.e. evidence, based on “hearsay”, “conclusion” etc., is the foundation of any further classification. As already mentioned in Chapter 1 we treat the evidential markers as deictic elements, the structure of which is based on the general deictic process, which has already been explained in Chapter 1.1.2.9 The evidential relation is slightly more complicated than simple deictic localizations, because the speaker sees the primary event through the secondary event, i.e. the evidence can be equalled to a pair of glasses through which the primary event is perceived. The existence of a pair of glasses is a given fact in any perception and any linguistic description of a scene. However, the defining crucial feature of evidentials is that this fact, i.e. the existence of a pair of glasses, and even the specification of what kind of glasses this is, is the central semantic feature of evidentials. We have to take this into account in our depiction of the evidential deictic process, which is given in Figure 8. primary event

secondary event functioning as evidence

evidential dimension

Figure 8. Evidentials as a deictic category

Origo

54 Evidentiality – definitions and limitations As can be seen in comparison to the other deictic dimensions illustrated in Chapter 1.1.2, the evidential dimension has the secondary event, which is logically situated between origo and primary event, as a third point in the vectored relation. Furthermore, we need a kind of “reflexive” relation pointing back from the evidence to the origo and investing the origo with the “pair of glasses” to perceive the primary event.

2.1. Direct and indirect evidentials Before going into further detail on deictic distinctions, we will take a look at the basic typological distinctions of types of evidential meanings. Though the classification of universal evidential values and evidential systems is by no means a simple task (Plungian 2001), the hierarchically topmost distinction, which has a wide distribution cross-linguistically and is acknowledged by most scholars working in the field, is the distinction between direct and indirect evidence (see e.g. Plungian 2001, Aikhenvald 2004). Most classifications go back to the one proposed by Willett (1988), which is given in Figure 9 below as a first orientation. Direct evidentials express the fact that there is/was direct access to the event described, and this access is mostly of perceptual nature (visual, other sensory). Indirect evidentials, by contrast, indicate absence of direct access to the event. The speaker either infers the situation from other information pieces (traces, results, general knowledge etc.), or s/he reports about the situation because s/he learnt the information from someone else. In the first case it is a different situation that provides evidence for the described event; in the second case it is another speaker who provides evidence for the event described. In both cases the event is not accessible in the current speech situation, thus indirect. In the following the distinction between direct evidentials and indirect evidentials will be discussed in detail where necessary. The feature of “direct access to the described event” can be equated with the co-presence of the speaker and the event in the same deictic field, i.e. a “near” value, which we called “origo-inclusive”, while “indirect access to the described event” corresponds to a “far” value, i.e. to the “origo-exclusive” stage, which signals that the described event is not copresent with the speaker (cf. Chapter 1.1.2). From this, the relational structures for direct evidentials versus indirect evidentials can be derived (see Figure 10).

Intra-evidential distinctions 55 Types of Source of Information

Indirect

Direct Attested

Reported

Inference

Visual Auditory Other Sensory

Secondhand Thirdhand Folklore

Results Reasoning

Figure 9. Types of evidence (Willett 1988: 57) primary event co-present with Origo

secondary event functioning as evidence co-present with Origo

speaker’s communicative location during utterance Origo

Figure 10. Direct evidentials

In the case of direct evidentials, all relevant points of the relation are treated as co-present in the same deictic field: the secondary event, functioning as evidence for the primary event, as well as the primary event are co-present with the origo, i.e. the marker encodes an origo-inclusive value. This is different in indirect evidentials, as can be seen in Figure 11. Here, the primary event is not co-present with the origo (and the secondary event): the marker expresses an origo-exclusive value.

56 Evidentiality – definitions and limitations primary event

not co-present with Origo

secondary event speaker’s communicative location functioning as during utterance evidence co-present with Origo Origo

Figure 11. Indirect evidentials

It should be added here that, in our view, the information source of the speaker, i.e. the secondary event functioning as evidence, is always copresent with the (deictic) position of the speaker. This means that it is always the speaker who possesses evidence for the described event. So, the information source cannot be separated from the speaker. In the following section, we are going to describe this fundamental distinction in another, more formal way. Let us call the primary event P and the secondary event functioning as evidence Q. If there is no difference between P and Q, i.e. if P is co-present with Q, we are dealing with direct evidence. The act of perceiving or performing P (i.e., Q) and P itself are not separated from each other; they are located in the same deictic field. If there is a difference between P and Q, i.e. if P is not co-present with Q, we are dealing with indirect evidence: “The speaker has not perceived P directly, being separated from P in space or time” (Plungian 2001: 352). Thus, Q and P are not located in the same deictic field. The basic opposition between direct and indirect evidence may be thus formulated as follows: direct evidence:

P=Q

indirect evidence:

PQ

Having established these two primary branches, we can go on to establish subdivisions in the direct as well as in the indirect branch. We start with distinctions in the direct domain. In the class of direct evidentials, it is common to draw further distinctions by isolating and encoding linguistically various perception channels, like the visual, acoustic, auditory channel etc. (Anderson 1986: 274, Wil-

Intra-evidential distinctions 57

lett 1988: 57, Bybee, Perkins and Pagliuca 1994: 323-324, Aikhenvald 2004, De Haan 2005). This tradition is also adhered to by Plungian (2001: 353), who divides the branch of direct evidentials as shown in Figure 12. Direct evidence

Visual

Non-visual

Sensoric

Endophoric

Figure 12. The classification of direct evidential values (Plungian 2001:353)

According to Plungian (2001: 351) the visual value, i.e. the most direct one, is realized when “the speaker has observed P directly (which yields a ‘Visual’ value: ≅ ‘P [described situation] in my sight’)” [our emphasis]. The conditions for the non-visual values are described by Plungian as follows: The speaker has perceived P directly, but not visually [our emphasis]. Here, a distinction between two important cases can be drawn: when the observer’s eyes are not used and when they are not needed. In the first case, the perception of P is actually non-visual, but could in principle be visual. For example, when somebody is running into the house, we may hear or see it – depending on how we observe the situation: similarly, a flower may be smelt or seen, etc. This leaves us with a ‘Sensoric’ value, encompassing all remaining human senses and corresponding to a ‘default’ perception. Interestingly, the finer distinctions within the sensoric domain (e.g. those between hearing and smelling) are rarely (if ever) grammaticalized. Thus, what is frequently called ‘Auditive’ is in fact the same undifferentiated ‘Sensoric’ value, which takes over whenever sight is ‘switched off’, and which is by no means restricted to hearing. The second case refers to those situations where visual (as well as any sensoric) perception is not possible at all, i.e., when unobservable things are described, such as one’s own intentions, desires, or other inner states (both mental and physical). This yields an ‘Endophoric’ value, which has specialized markers (at least) in Tibetan languages. (Plungian 2001: 351-352)

In the relational diagram, these distinctions can be represented by modifying the shape and size of the secondary event, and/or by specifying the arrow pointing from the secondary event to the speaker.

58 Evidentiality – definitions and limitations primary event

secondary event functioning as evidence

co-present with Origo

co-present with Origo

speaker’s communicative location during utterance Origo

Visual

Figure 13. Direct visual evidentials primary event

secondary event functioning as evidence

co-present with Origo

co-present with Origo

speaker’s communicative location during utterance Origo

Auditory

Figure 14. Direct auditory evidentials

As illustrated before in Chapter 1, in German, direct evidentiality is expressed either by lexical means (e.g. sichtlich, sichtbar ‘visible’) or by syntactic constructions (e.g. matrix verbs, AcI-constructions, parenthetical uses). Typical AcI-constructions are given in (11), where the verb triggering the AcI-construction is a perception verb encoding visual perception in (11a), acoustic perception in (11b),10 and internal sensory perception in (11c), which is called “endophoric” by Plungian (2001: 353): (11) a. Er sah sie über das Schneefeld heraufkommen. ‘He saw her come up the snow field.’

Intra-evidential distinctions 59

b. Er hörte sie über das Schneefeld heraufkommen. ‘He heard her come up the snow field.’ c. Er fühlte Kälte in sich aufsteigen. ‘He felt the cold rising in his body.’ The paraphrases of the channels of direct information are given in (12):11 (12) a. P and Q [Origo sees P] b. P and Q [Origo hears P] c. P and Q [Origo feels P]

2.2. Sub-classifying the indirect branch While there are elaborate systems of direct evidential distinctions in many languages of the world (Aikhenvald 2004, Wierzbicka 1994), European languages are said to mostly display expressions of indirect evidentiality, leaving the direct evidentiality formally unmarked (also see above). This view is confirmed in the recent survey by De Haan (2005a): […] languages with only indirect evidentials are common in Europe. […] The languages with indirect evidentials in Europe are mostly Germanic (though not English). Romance languages do not seem to have evidentials, with the exception of French, probably under influence of Germanic. (De Haan 2005a: 315)

Examples for indirect evidential marking in European languages are the “meditative” in Bulgarian and Western Armenian (Lazard 1999: 100), standard Estonian -vat, which is a hearsay marker (Wiemer [to appear]), the invariant participle for the present and past (or perfect) tense (-ma, -ta), which is used in specific constructions as an inferential evidential (Wiemer [to appear]), the usage of the French conditional to express indirect evidentiality (Squartini 2001), and the use of the German and Dutch modals sollen and zou to express reportive evidential values (Diewald 1999a, Mortelmans 2009). Plungian (2001: 351-352) describes the indirect branch as follows: “The speaker has not perceived P directly, being separated from P in space or time. However, s/he has had access to other kinds of information about P.” As it is the indirect branch that is heavily marked in European languages and is expanded in their history, a closer look at possible internal dis-

60 Evidentiality – definitions and limitations tinctions in this branch will be useful. Beyond the first basic distinctions given above, there are some classificational difficulties:12 The most frequent classifications provide three possibilities: either the speaker has (directly) observed another situation which s/he interprets as pointing towards P (captured by an ‘Inferentive’ or more commonly ‘Inferential’ value), or s/he simply knows something which suggests that P is probable (a ‘Presumptive’ value), or else s/he got the information from other persons (a ‘Quotative’ value). In the latter case, some additional distinctions may or may not be applied, such as reported speech (presupposing a known author) vs. generalized, second-hand information (presupposing an unknown author or non-definite anonymous author) vs. tradition or common knowledge (where no personal author is invoked). (Plungian 2001: 352)

One of the suggestions Plungian offers in terms of an overall classification is the following, which nicely renders the tripartite distinction of the quote above: Indirect evidence Reflected evidence Direct evidence Mediated evidence (= inferentials and pre(= Quotatives) sumptives) Personal evidence Figure 15. Different types of evidentials oppositions (Plungian 2001: 535)

As mentioned before, we choose the distinction between direct and indirect as the basic criterion of classification. A further sub-classification along the lines of personal involvement (which by the way does not have a clear oppositive value in Plungian’s diagram) in our view is not necessary. Therefore, we will not make use of the term “personal” and the segmentation achieved by it. This decision rests on the fact that evidential meaning is deictic, and thus inherently personal: the speaker is always – in one way or another – personally involved in the information processing. This should be briefly explained: the co-presence of some direct evidence (e.g. sensory input, personal involvement in doing something, etc.) with the (deictic position of the) speaker is fairly obvious. The speaker and his/her perception or his/her own activity at the time of speaking cannot be separated from each other: they pertain to one and the same deictic field (see above). The same holds also for indirect evidence. It is the speaker to whom some evidence (e.g. visible results, general knowledge, some previously communicated information, etc.) is available. That means that the information

Intra-evidential distinctions 61

source has always the feature ‘personal’, since it has to be identified with the speaker. As this aspect is controversially discussed in literature on evidentials, it is in place here to render some strands of this discussion. Many authors working on evidentiality have suggested that evidential values may be classified in (at least) two different ways (see e.g. Plungian 2001 on ‘personal’ vs. ‘(in)direct’ evidence, Faller 2002 on ‘the personal evidence cline’ vs. ‘the mediated evidence cline’, De Haan 2001 on ‘deictic component’ vs. ‘witnessing component’). These two ways are said to be reflected in two different evidential ‘systems’ found in the languages of the world. On the one hand, there are evidential systems which seem to prefer the parameter of speaker’s involvement and thus have only one opposition of quotative (‘mediated’) vs. non-quotative (‘personal’) forms (e.g. Lezgian, Latvian, Quechua). On the other hand, there are systems where the most prominent opposition is that of direct vs. indirect access, so that quotative uses are only a pragmatic (or contextual) variety within a larger class of indirect evidence markers. Typical representatives of the latter kind are found among ‘inferential’ (and ‘prefectoid’) Balkan, Caucasian, or Ob-Ugrian systems. (Plungian 2001: 353) There are thus two conflicting forces at work in the inferential. By choosing which of the two features is basic, languages decide whether the deictic component or the witnessing component is the underlying factor that drives the evidential system. As we have seen, languages like Kashaya Pomo and Hualapai highlight the deictic component, represented by means of the feature direct, [DIR]. They do this by making the inferential responsible for part of the sensory information, usually the nonvisual, nonauditory ones. Other languages, such as Takelmy and Sherpa, prefer to view the inferential as denoting a nonwitnessed event. In these languages, the feature firsthand, [1ST] underlies the evidential system, with inferentials being [-1ST]. The inferential evidential in such languages typically has a quotative (the quintessential nonwitnessed evidential) interpretation as well. (De Haan 2001: 218)

The two different ways in dividing the conceptual space of evidentiality may be summarized as follows. On the one hand it is claimed that evidential values may be distinguished due to the criterion of speaker’s personal involvement (i.e., the feature [1ST] in the terminology by De Haan; ‘personal evidence’ in the terminology by Plungian). In this regard, types of evidence such as performative, visual, auditory, (other) sensory, inference, reasoning, assumption, etc. are placed within the subgroup of ‘personal evidence’. Other values such as secondhand, thirdhand, folklore, hearsay,

62 Evidentiality – definitions and limitations etc. pertain to the subgroup of ‘non-personal evidence’. On the other hand, it is suggested that evidential values may be distinguished due to the criterion of the speaker’s access to the information source. Here, the opposition ‘direct’ vs. ‘indirect’ evidence seems to play a crucial role. Direct evidence is associated with types of evidence such as firsthand, witnessed, sensory, etc. Indirect evidence is represented by values such as secondhand, thirdhand, hearsay, inference, reasoning, assumption, etc. In our view, however, there is no need to distinguish between two separate ways of dividing the conceptual space of evidentiality. In this respect, our general classification of evidential distinctions (and of German evidential distinctions in particular) does not follow two different criteria but uses only one basic classification criterion: direct vs. indirect access. The first sub-branch of indirect evidentiality is inferential evidentiality. Grammatical elements indicating this type of evidential value are usually called inferential evidential markers. Other terms, e.g. “inferred” (Aikhenvald 2004), “conjecture” (Floyd 1997), “reasoning” (Willett 1988), “assumed” (Aikhenvald 2004) have been proposed so far for this type of evidential value in the typological literature. In the last decades the term “inferential” has been mostly preferred (e.g. de Haan 2001, 2005, Plungian 2001, Willett 1988). […] the evidential category of inference, which is the grammaticalized way of showing that the speaker makes his or her statement based on a deduction from facts, and not on a direct observation of the action itself. (de Haan 2001: 1) […] two main types of indirect evidence are clearly seen: evidence via verbal report and evidence upon which an inference is based. (Willett 1988: 57) [original emphases] The domain of ‘inference’ is subdivided differently in different systems. A major distinction appears to exist between an inferred evidential covering inferences made on the basis of visible or tangible results, and an assumed evidential involving general knowledge and assumption based on reasoning. (Aikhenvald 2004: 64)

One last terminological note must be made here before we go on with our classification. The fact that inference may be based on several distinct sources is caught in Aikhenvald’s distinction between “inferred” and “assumed” which she explains as follows:

Intra-evidential distinctions 63 The difference between the ‘assumed’ evidentials […] and the ‘inferred’ […] lies in access to visual evidence of something happening and to the degree of ‘reasoning’ involved. The less obvious the evidence and the more the speaker has to rely on reasoning based on common knowledge or on common sense, the more chance there is that the assumed evidential will be used. An inferred evidential refers to something based on obvious evidence which can be easily observed (even if the event itself was not seen). This illustrates two types of inference – the one based on visible results, and the other based on reasoning, general knowledge and, ultimately, conjecture. (Aikhenvald 2004: 2-3)

Instead of the terms “inferred” and “assumed” proposed by Aikhenvald (2004), we use “perceptual” versus “conceptual” inferential, because these terms specify better the type of premises the speaker bases her/his inference on. At this level, we distinguish between inference which is perceptually founded, i.e. based on perceptual input, and inference which is conceptually founded, i.e. based on internal reflection (cognition) without concomitant current perceptual input. As shown above for the further distinctions within the subcategory of direct evidentiality, the relationship between P (primary event) and Q (secondary event) may (but need not)13 be further specified. This also holds for the indirect branch. The paraphrases below illustrate different possibilities to specify the indirect relationship between Q and P (cf. Plungian 2001: 354 for a similar classification), which result in different indirect evidential values (for the mediated branch see below): perceptual inferential value: conceptual inferential value:

P, because Q [Origo can observe some signs/traces of P] P, because Q [Origo knows Q, and Origo knows that Q entails P]

Thus, inferential evidentials primarily denote the speaker’s reflection of some evidence, i.e. they indicate the relation between the described situation and some other situation, which is treated by the speaker as evidence for the former. As in the case of the direct evidentials, the various types of sources (secondary events) may be illustrated by variation of the arrow, as in the Figures 16 and 17 below. As will be exposed later, the German infinitive constructions with drohen and versprechen represent perceptual inferential values, while the construction with werden represents the conceptual inferential value (see Table 5, Chapter 3.5).

64 Evidentiality – definitions and limitations primary event

not co-present with Origo

secondary event speaker’s communicative location functioning as during utterance evidence co-present with Origo Origo

perceptual inferential

Figure 16. Indirect evidentials of the perceptual inferential value primary event

secondary event speaker’s communicative location functioning as during utterance evidence not co-present with co-present with Origo Origo Origo

conceptual inferential

Figure 17. Indirect evidentials of the conceptual inferential value

The second sub-branch of the indirect evidentiality is the reportive branch. Though this is not in the focus of this study, we will have to take a look at it in order to be able to situate the indirect evidentials properly within the full system. While indirect inferential evidentials indicate that the speaker has indirect access to the event through his/her own reflections, indirect reportive evidentials express that the speaker has indirect access based on someone else’s perceptual and cognitive capacities, i.e. what he has heard from someone else. This communicative source may be specified in various ways leading to sub-classes of mediated evidentials, like hearsay (cf. thirdhand), folklore, rumours, report by someone (i.e. second-hand), etc. Linguistic expressions of mediated information can be basically described as having the relational structure of indirect evidentiality and can be rendered by the same type of diagrams.

Intra-evidential distinctions 65 primary event

secondary event speaker’s communicative location functioning as evidence during utterance not co-present with co-present with Origo Origo Origo mediated evidence

Figure 18. Relational structure of markers of mediated information

In German the modal verb sollen as it is used in (13) and (14) is an instance of mediated evidentials. (13)

(14)

Nach Auskunft der Veranstalter soll es derzeit kaum mehr kurzfristige Stornierungen geben als sonst. (DWDS, Kernkorpus) ‘According to the convener there are presently hardly more cancellations at short notice than usually.’ Was ich über 2003er aus der Bourgogne lese, weckt wenig Erwartung. Es soll aber, und das kann man sich gut vorstellen, ein gutes Jahr für Beaujolais gewesen sein. (DWDS, Kernkorpus) ‘What I have read about the 2003 Bourgogne vintages raise few expectations. However, they say that it has been a good year for Beaujolais, and it is easy to imagine that to be the case.’

As the examples show, sollen indicates that the speaker learned about the described event from someone else. Among the languages having markers of mediated information in the sense explained are Georgian, Kham, Queshua, etc., cf.: […] in Italian the conditional is reportive, not inferential: this means that it conveys “a true mediated knowledge” (Plungian 2001: 253), the source of information being external to the speaker. The speaker/writer is simply reporting the information acquired without any overtones of unreliability. Admittedly, contexts may be found with epistemic extensions of uncertainty, since by pragmatic inference less direct or mediated information may be taken to be less reliable (a similar inference has been grammaticalized in Balkan systems, see Plungian 2001: 253). (Giacalone Ramat and Topadze 2007: 17)

66 Evidentiality – definitions and limitations The reported evidential may simply not be used if the source of information is explicitly stated, as in Kham. […] The reported -sh(i) in Quechua never indicates who the author is. […] Similarly, in Tsafiki the reported evidential -ti- indicates that the information comes from someone else’s report but does not mark quotations. The verb marked with the reported can be a complement to the verb ti ‘say’ indicating who the author is. (Aikhenvald 2004: 177-178)

Beside sollen, also German wollen may be used with a mediated evidential meaning. However, as this – i.e. the mediated – branch of indirect evidentiality is not treated in this book (nor is the copious literature on reported speech), these items are not treated here. Furthermore, the exact evidential status of the relevant sollen and wollen uses with respect to distinctions like mediated, reportive and quotative is not yet fully clear so that their treatment would call for an extensive digression. At this point it will be necessary to discuss the controversial issue of how to draw a relevant distinction in the area of mediated evidentials, reportive evidentials and quotatives, as there is a lot of terminological confusion here.

2.3. Distinguishing quotatives from mediated evidentials As typological research indicates, the notions of “evidentials”, “reportives” and “quotatives” are difficult to keep apart in a number of languages (e.g. Sumbatova 1999, Plungian 2001). While some authors subsume quotatives among evidentials (e.g. Palmer [1986] 2001: 52), there are reasons to distinguish them from evidentials proper. The following pages will argue the case for this distinction. Among the most important sources of evidence and knowledge are linguistic utterances of other speakers and knowledge acquired via some type of language-driven transfer. At the same time, reference to and re-use of the communicative acts of others has implications concerning authorship, identity, possession/ownership of ideas, fictionality etc.14 Thus it does not come as a surprise that in every language there is not only a gamut of linguistic devices representing those various ways of “speaking about things other people have said”, but also a gamut of functions these indications are applied to. The former, the linguistic devices, range from direct quotes via reported speech marking to subtle traces of intertextual connections (e.g. using an expression which is known to be used only by a particular (group of) speaker(s)). The latter, the functions of “speaking about things other people have said”, range from register driven

Intra-evidential distinctions 67

issues like journalistic work (reporting important news), scientific aims (connecting knowledge and demonstrating one’s membership in the scientific community) or fictional writing (writing a novel, designing dialogues for a foreign language course), to more fundamental communicative issues like gaining authority or declining responsibility etc. These issues cross-cut several areas of linguistic meaning and structure – in particular those of evidentiality and modality but also issues of discourse structure (like dialogic elements, polyphony etc.). Therefore, in a book on evidentials like this one, it is essential to circumscribe the area under investigation, to define one’s categories and to make some terminological decisions. A brief survey of the terminological options offered in the literature on evidentiality and modality shows that this latter point, terminological clarification, is not a trivial one. Among the terms frequently found in literature are the following: − quotatives and quotative markers (De Haan 2001) − reportives, reportatives, and reported evidentials (Aikhenvald 2004, Willett 1988, Faller 2002) − reported speech representation and reported speech markers (Mushin 2001) − mediatifs/mediatives (Lazard 2001, Guentcheva 1996) − (general) hearsay (evidentials) (Anderson 1986, Willett 1988) − second-hand evidentials/third-hand evidentials (Willett 1988, Aikhenvald 2004) Most authors suggest some type of internal classification between secondary information with reference to a quoted source and secondary information without such reference to a source. So, Aikhenvald (2004) proposes to draw a distinction between ‘quotatives’ and ‘hearsay’ evidentials. Both indicate that information was acquired from what someone else said. The difference between them is, however, that the former do make overt reference to the source (i.e. to the quoted speaker) while the latter do not: HEARSAY: for reported information with no reference to those it was reported by. QUOTATIVE: for reported information with an overt reference to the quoted source. (Aikhenvald 2004: 64)

Moreover, a clear boundary has usually been drawn between reportive evidentials and other linguistic expressions of reported speech (which may include evidential expressions as well). This distinction is mostly based on

68 Evidentiality – definitions and limitations functional considerations, i.e. whether an expression pertains to the class of (grammatical) evidentials or not. Reported evidentials and reported speech do essentially the same job: they indicate that the information was acquired from someone else. (Aikhenvald 2004: 135) […] reported speech and reported evidentials complement each other. One may be used instead of the other to avoid unwanted epistemic or other overtones. Reported speech may allow the speaker to be precise about who told what. Both reported speech and reported evidentials may be employed as stylistic devices in discourse. (Aikhenvald 2004: 140)

In a similar manner, this distinction is described by Mushin (2001): […] b. reported speech representation typically involves the explicit mention of the reported speaker while use of the reportive clitic simply reports the information as the product of some prior utterance unconnected to any purported original speaker; c. reported speech representation typically involves some shift in deictic coding to the time, place and identity of the reported speech act (and reported speaker) while use of a reportive clitic does not motivate any shift in the deictic centre. (Mushin 2001: 71)

If we look at the two oppositions (i.e. reportives vs. quotatives and reported evidentials vs. marking of reported speech) introduced above, a common point will not remain unnoticed. In both cases, the fact whether the source may or may not be explicitly mentioned serves as a main distinctive feature. Taking this into account we suggest that we are dealing here not with two separate oppositions, but with one and the same crucial difference. In the remainder of this section we will argue for this view. Given the broad variety of linguistic phenomena and procedures referring to what other speakers said, we refrain from discussing whether these elements as a whole should be classified as evidentials, epistemic expressions or something else. It is very likely that this domain does not fit into one grammatical category and that we find markers and distinctions having to do with the speech of others distributed among several categories. Instead we would like to set up some essential notional distinctions, which allow us to decide which linguistic categories are relevant for the phenomena found in German, and how this fits into our system of evidentiality and epistemic modality. The first and most important distinction we would like to make is the one between the following two groups: On the one hand, there are linguis-

Intra-evidential distinctions 69

tic devices used for shifting the deictic origo from the present speaker to a secondary speaker. As will be explained later, this does not necessarily have to do with evidentiality; it is rather a sort of communicative role play. In the following, this linguistic category will be called Quotative, the content of the category will be called quotative meaning, and various linguistic expressions of this content will be called ‘expressions’, ‘strategies’, or ‘markers’, depending on the grammatical status of linguistic items in question (see Chapter 2.1 for the similar terminological distinctions proposed for the domain of evidentiality). Quotative expressions embrace all types of the reported speech marking, including direct and indirect citations, the use of embedding speech act predicates, etc. On the other hand, there are linguistic devices used to mark one’s source of evidence (which consists of some communicated content) with no obligatory shift to a secondary speaker. This may be done in a variety of ways and with a variety of specifications, including hearsay, folklore, mediated communication, reported speech representations etc. All these meanings and functions fall into the domain of (indirect) evidentiality. As a category name we chose mediated information here. Before we go on to a closer inspection of both groups, we would like to give an overview of our (not merely) terminological decisions. Let any other member of the linguistic community beside the actual speaker be called “alter-ego” and let any type of linguistic condensation of a communicative act be called “linguistic utterance”; the whole domain may be superscribed as “linguistic utterance of alter-ego” and we get the following divisions: linguistic utterance of alter-ego

origo shift = quotative (non evidential)

no origo-shift = mediated information (evidential)

Figure 19. Quotatives and mediated information

The right branch, called “mediated information”, belongs among evidential distinctions, the evidential source being a linguistically construed one. The left branch, called “quotative” here, does not belong among evidential distinctions (though quotative expressions may often serve as diachronic

70 Evidentiality – definitions and limitations sources for evidentials). A word of caution: terms like “reported speech” are too general to give a clear indication as to what type of marker and what type of linguistic procedure is addressed. Therefore we will use this and similar terms only in a non-technical way and do not imply any further specifications by it. The category of Quotative, as we have briefly defined it above, concerns the representation of a linguistic utterance of alter ego, whereby the origo is shifted from the actual speaker to a secondary speaker. That is, quotative expressions constitute a special communicative strategy, whereby some other communicative event is embedded into the actual one. By using a quotative expression, the speaker brings some previous speech situation into play. The actual communicative scene, i.e. the here-and-now of the actual speech situation, no longer serves as the only vantage point. Instead, the reported information is anchored in the here-and-now position of the embedded communicative event, which is distinct from the actual one, i.e. the reported information is anchored in a shifted origo. In this view, the category of Quotative is a matter of what Bühler called “imagination-oriented deixis”. It is a form of deictic displacement (“Versetzungsdeixis”, Sennholz 1985: 224). It differs from deixis proper because, rather than using the actual deictic field as context, it uses an imagined deictic field instead. The relational structure is displaced from the actual speaker to another imagined origo (Rauh 1978: 101). An example is given in (15): (15)

We spent our Easter vacation in Italy. Here the sun shone all day and we went for long walks across the hills.

The origo of here obviously is not identical with the present here-and-now position of the speaker. Instead, it is displaced to the location the speaker is talking about, i.e. Italy mentioned in the preceding sentence. Imagination-oriented deixis is not restricted to the spatial domain only. It operates in other domains as well, such as temporal, personal, modal, textual, situational etc. The most important areas of deictic displacement are however the different forms of reported speech, in which the category of evidentiality, mood and the modals are highly relevant. In our view Quotative constitutes a communicative strategy which manifests itself in various linguistic techniques. That means that Quotative is a sub-group of the type of displaced deixis. It does not belong to the category of evidentiality (nor to those of epistemic modality, tense, mood, per-

Intra-evidential distinctions 71

son, etc.). It is the indication of a shift to an embedded communicative scene. More particularly, Quotative is the indication of a quote (citation) of another speaker. This communicative strategy may appear in languages in a variety of ways ranging from simple to highly artificial: A simple way to shift to another speaker is to use a direct quote. The purported exact words of the “linguistic utterance of alter-ego” are represented directly, without any modifications. This results in a complete shift of deixis and deictic markers to the deictic coordinates of the secondary origo. (16)

Yesterday, Mary said: “I am going to visit my mother this afternoon”.

A more complicated way to shift to another communicative scene is to use an indirect quote. In this case, only the gist of the reported “linguistic utterance of alter-ego” is normally introduced. Indirect quote (or: indirect speech) result in a more complex system of shift of deictic pointer, dependent on meaning but also on intra-linguistic rules and conventions. So, first person is always shifted and occurs as a third person in the reported utterance. As for the dependence on linguistic peculiarities, consecutio temporum in English indirect speech provides an excellent example. (17)

Yesterday, Mary said (that) she was going to visit her mother that afternoon.

The most complex way to displace the communicative scene is represented by literary techniques such as “stream of consciousness”, or “erlebte Rede”. A mixture of different processes for shifting the perspective is characteristic for such modes of displacement. Often, even both techniques described above, i.e. source marking (i.e. evidentiality) and origo shifting (i.e. Quotative), are mixed in such operations. The wide array of ways to shift the actual origo, or to linguistically code quotative meanings, obviously, goes far beyond the range of evidentials. Among other things, this coding technique often makes use of expressions of many grammatical categories, including mood, tense, person, etc. Therefore, the strategy of origo displacement is not restricted to a particular grammatical category or to a particular functional domain. Instead, it serves to fulfil an important communicative function, which can be generally described as “referring to things other people have said”. From this, it is not surprising that there is a large area of overlap between reported speech marking and epistemic modality as well as evidentiality. The close

72 Evidentiality – definitions and limitations interconnection of the quotative category with epistemic modality is e.g. reflected by the fact that in German the subjunctive mood figures as a quotative marker. However, as we already pointed out above, we hold the opinion that origo shift concerns a different level of linguistic structure than the level of factuality judgments, i.e. the level of epistemic modality. Origo shift opens up a new communicative scene with its own origo and its own factuality judgments. By a pure marker of origo shift, the actual speaker asserts the fact that the communicative scene is shifted. S/he does not judge the factuality of the embedded quotes. Indeed, the interpretation of the proposition as ‘uncertain’ or ‘unreliable’ often accompanies the shifting of the communicative scene. However, such interpretations always have the status of conversational implicatures and cannot be considered an inherent meaning component of quotatives (at least in German). The adoption of a reportive epistemological stance can have the rhetorical effect of making the speaker seem less responsible for the information. In English, the function of reportive coding has been described ‘distancing’ or ‘detaching’ the actual speaker from the information described as reportive. […] Although the adoption of a reportive epistemological stance necessarily codes information as at least one step removed from the speaker’s direct experience, it does not necessarily code the information as unreliable or uncertain. The extent to which information expressed from a reportive epistemological stance can be considered reliable lies in the identity of the source and not in its reportiveness per se. (Mushin 2001: 73)

Among the languages having quotatives in the sense explained is Georgian, which according to Giacalone Ramat and Topadze (2007) uses quotative particles to express what the authors (rather mistakenly) call “reportive evidentials”. In our view, the Georgian particle -o is a prototypical example of a quotative marker rather than of a (reportive) evidential marker. Quotative particles as reportive markers are frequently used in oral speech. They occur especially in dialogues of various narrative genres (folk tales, stories, etc.) to report the speaker’s words exactly. (Giacalone Ramat and Topadze 2007: 16-17)

Other languages having quotative markers in the above described sense are e.g. Comanche (Uto-Aztecan), Dakota (Siouan), Chemehuevi (UtoAztecan): The quotative particle me [in Comanche; GD & ES] occurs when there is a direct quotation, as in 2.65. suti= patsi 2.65. hãã me-se

Summary: classification of evidential distinctions used here 73 yes QUOT-CNTR that.one older sister ‘The older sister said, “yes”’ […] ‘Reported’ and ‘quotative’ are distinguished in Dakota […]. The particle Ğk`a` indicates ‘statements known by hearsay’ and ‘when the statement or thought of a definite person is quoted, the quotation may end with lo (le), ye, cè (or k`ǐ). (Aikhenvald 2004: 50-51)

To summarize: quotative expressions shift the origo to a secondary speaker and quote, i.e. cite what that secondary speaker has said, often together with the factuality judgment of the secondary speaker. Thus, quotative expressions, as we define them here, are neither evidential nor epistemic modal – they constitute a category of their own.

3. Summary: classification of evidential distinctions used here On the basis of typological research and earlier studies on the German system referred to above, we suggest as a hypothesis the following paradigm of German evidential distinctions. This overall classification of evidential values will guide us through the investigation of the German data in the next chapters; it will be refined and adapted as necessary when further aspects need to be accommodated.

P and Q [Origo can see/hear etc. P] Example (visual) Sie hört ihn die Treppe heraufkommen. ‘She hears him come up the staircase.’

paraphrase

Figure 20. Overview of general classification of evidential values

P, because Q [Origo knows Q, and Origo knows that Q entails P] see Table 5, Chapt. 3.5

conceptual

P, because Q [Origo has information pointing towards P] Example: In Hannover wird es regnen. ‘It will be raining in Hannover.’

inferential

P, because Q [Origo can observe some signs/ traces of P] see Table 5, Chapt. 3.5

visual, auditory, etc.

Level II

P is non co-present with Q

paraphrase

P is co-present with Q

description

‘origo-exclusive’

perceptual

‘origo-inclusive’

deictic value

indirect

Level III

direct

Level I

evidential values

P, because Q [Origo was told that P] Example: Er soll abgereist sein. ‘He is said to have left.’

mediated information

74 Evidentiality – definitions and limitations

Chapter 3 Evidentiality and modality – drawing the lines

Evidentiality and epistemic modality are closely related concepts and domains, so it is of particular importance to clearly draw the lines between them, to define their distinctive features, to delineate their respective borders, to explore their common denominator, and to indicate the areas of potential overlap as well as the reasons for this. These tasks are dealt with on the following pages, and the focus lies in a particularly important and difficult area: the dividing line and the area of overlap between epistemic modality and evidentiality. The problem of distinguishing between evidentiality and modality has two aspects which have to be kept in mind. One is the theoretical, notional aspect, i.e. the question whether and how we should in principle distinguish between the two concepts independent of their linguistic realization. The other is the empirical fact that the two concepts may overlap and be conflated in their linguistic realization in a number of languages in different ways. From this observation arises the question of how to relate language specific overlap to universal distinctions, or, formulated differently, if a language fuses both notional domains, how should we treat this? It should be noted that these questions cannot be reduced to terminological decisions. Reverting to a broad notion of epistemic evidentiality and postulating evidentiality as a sub-category (cf. Palmer 1986, Leiss 2000, Diewald 2004) hides the problem rather than solving it, as does of course the converse procedure, i.e. defining evidentiality in a broad way and declaring epistemic modality as a subcategory thereof (Mushin 2001). We clearly favour a succinct definition of both domains as a basis for elucidating potential similarities, overlaps and neutralisations. To justify this, let us compare the situation with the one of the domains of temporality and aspectuality. The discrimination and regions of overlap of these two categories has received much more attention, as both are highly prominent categories in European languages. While at the beginning of modern linguistics aspectual distinctions were often treated as temporal ones, it can be stated that, meanwhile, linguists agree that both categories are in principle independent of each other but have systematic areas of overlap, and also a relative chronological ordering, whereby tense distinctions develop from as-

76 Evidentiality and modality – drawing the lines pectual distinctions (e.g. PAST from PERFECT cf. e.g. Bybee, Perkins and Pagliuca 1994, Heine and Kuteva 2002). We suggest treating the categories of evidentiality and epistemic modality in an analogous way. We argue that both categories are in principle independent of each other, though closely related. We put down their similarity to a common abstract relational template, which, too, accounts for their functions as deictics. Variations in the “filling” of the positions of the template account for semantic and functional differences of the forms and categories in question, whereby the specification of the source features of a relational structure play a decisive part. Only after drawing the categorical distinction between evidentiality and epistemic modality, it is possible to describe neatly the areas of overlap and the reasons for this. As the issue of the categorical distinction between evidentiality and epistemic modality touches the nerve of this study, it deserves an extended explication, which, however, will focus mainly on German. In order to show how the categorical difference as well as the overlap in specific forms are manifested in the German system, our line of argument touches the following issues which will be treated in the order given. The most important factor for a distinction of both areas is the differing specification of the concept of source (see Section 2). The distinctions in the make-up of the source features can be explicated by diachronic facts, which, in addition to that, provide further insight into the fundamental difference of both categories (see Section 3). Additional arguments can be derived from typological issues, and from the behaviour of the items with respect to scope, negation and embedding (see Section 4). Finally, a study of the areas of overlap between evidentiality and epistemic modality will help to weigh the arguments in favour of a separation of both categories (see Section 5). Before discussing these issues, the next section is devoted to a brief review of existing opinions and models on the connection between evidentiality and epistemic modality.

1. Prior suggestions on the relation between evidentials and modality The fact that both categories are closely intertwined in many languages may be the major reason for many linguists conflating both notions into one category. In most of these approaches evidentiality is treated as a subcategory of epistemic modality, e.g. in Gunkel (2000: 111), PalanderCollin (1999: 27-28), Traugott (1997), Sumbatova (1999: 74) who defines

Prior suggestions on the relation between evidentials and modality 77

evidentials as follows (with respect to the South-Caucasian language Svan): “Evidentials presuppose that the speaker does not possess direct knowledge of the situation”. This definition foregrounds a feature which, in our view, is essential for epistemic modality and not evidentiality. Diewald (2004), too, subsumes evidentials in German under epistemic modality: Modalität wird im Deutschen u.a. durch Infinitivkonstruktionen zum Ausdruck gebracht. An erster Stelle sind hier die klassischen Modalverben, also die sechs Verben dürfen, können, mögen, müssen, sollen und wollen zu nennen, dann aber auch die Verben scheinen, drohen und versprechen, […]. [‘In German, modality is expressed by infinitive constructions among other means. Here, the classic modal verbs, i.e. the six verbs dürfen, können, mögen, müssen, sollen and wollen have to be mentioned frist, with the verbs scheinen, drohen and versprechen following after them’]. (Diewald 2004: 230)

Evidentiality in this paper is regarded as an additional semantic feature further differentiating the system of epistemic distinctions. Da nun die Faktizitätsbewertung die gemeinsame Eigenschaft aller Modalund Modalitätsverben und das distinktive Merkmal der Moduskategorie ist, eignet sich die nur bei einigen Faktizitätsmarkern auftretende evidentielle Komponente zu einer weiteren semantischen Subklassifikation dieses Paradigmas. [‘As the common feature of all modal and modality verbs and the distinctive feature of the mood category is the feature of factuality judgement, the evidential component, the appearance of which is restricted to some factuality markers only, is suitable as a means of further subclassifying he paradigms in semantic terms.’] (Diewald 2004: 238)

This results in the scheme given in Table 1 that unites epistemic modal periphrasis and evidential periphrases with scheinen, drohen, versprechen under the categorial function of expressing uncertain factuality. This scheme postulates the deictic nature of epistemic modals and verbal mood, i.e. the deictic anchoring of the grammaticalized factuality judgment expressed by verbal moods and grammaticalized epistemic modals (cf. Diewald 1999, 2001). Furthermore, it suggests that the three inflectional verbal moods of German and the periphrastic modal constructions (i.e. the grammaticalized uses of the modals) form a highly differentiated deictic grammatical category in PDG, which also contains the evidentials as a sub-group. As this topic has been already touched upon, a very brief description of the structure of the category will suffice.

78 Evidentiality and modality – drawing the lines Table 1. Semantic distinctions of modal verbs and the so-called “Modalitätsverben” (Diewald 2004: 253) uncertain factuality

deictic können müssen

-evidentiality modal verbs quotative sollen wollen

phoric dürfte mag

-ingressive scheinen

+evidentiality modality verbs +ingressive -desired +desired drohen versprechen

As is well-known, the distinctive semantic core of epistemic modality is a speaker-based, i.e. deictic, factuality judgment (“subjective” epistemicity in the tradition of Lyons 1977, Diewald 1999). An epistemic marker of this type ascribes a factuality value to the proposition, whereby the speaker is the source of this evaluation process, the proposition is the goal. The specific semantic value of this judgment, i.e. the degree of factuality ascribed to the proposition, is measured against the speaker’s own epistemic status. The basic opposition is, as expected, the one between the origo-inclusive, unmarked value and the origo-exclusive, marked values. The former is represented by the indicative; all other forms belonging into the paradigm (the subjunctive I, the subjunctive II, and the deictic modals), express origoexclusive, marked values.1 This basic distinction between origo-inclusive and origo-exclusive factuality judgments is rendered in the well-known relational diagram below: described entity: proposition co-present with Origo ‘origo-inclusive’

Origo

modal dimension

Figure 1. Relational structure of origo-inclusive modal value

Prior suggestions on the relation between evidentials and modality 79 Origo

described entity: proposition not co-present with Origo ‘origo-exclusive’

modal dimension

Figure 2. Relational structure of origo-exclusive modal values

Illustrations for each distinction are given below in this order: indicative, subjuncive I, subjunctive II, purely deictic modals, deictic modals with phoric component, and deictic modals with quotative/mediated component: (1)

(2)

(3)

(4)

Die Gießkanne steht im Garten. ‘The water can is in the garden.’ [indicative: origo-inclusive factuality judgment] Die Gießkanne stehe im Garten. ‘The water can is said to be in the garden.’ [subjunctive I: origo-exclusive factuality judgment] Die Gießkanne stünde im Garten/würde ... stehen. ‘The water can would be in the garden.’ [subjuncitve II: origo-exclusive factuality judgment] Die Gießkanne muss/müsste/kann/könnte/dürfte/mag/soll im Garten stehen. ‘The water can must/might/should/may/is said to be in the garden.’ [epistemic uses of modal verbs: origo-exclusive factuality judgment]

As concerns the place of the evidentials in the paradigmatic organization, the scheme suggested in Diewald (2004), and given above in Table 1, incorporates the evidentials as a sub-group of the overarching category of factuality marking, where the epistemic modals, too, form a sub-group. This scheme, which we are going to revise in the course of this section, reflects the fact that evidentials in German as in most IE languages do have epistemic overtones, i.e. factuality and evidentiality overlap in these languages (cf. also Willett 1988: 51, 66 and Bybee 1985: 192). In this connec-

80 Evidentiality and modality – drawing the lines tion, De Haan (1999, 2001) holds the opinion that most Germanic evidentials have their origins in epistemic modality. Evidentials from modals is mainly a Western European feature. It occurs in most Germanic languages, as well as in Finnish. In these languages evidentiality is another interpretation of modal verbs. (De Haan 2005a: 320)

This being the case, we nevertheless opt for a clear distinction. As we will argue in the course of this chapter, these overlaps are the results of conversational implicatures and not inherent semantic components of these verbs. Therefore, the above scheme of Diewald (2004) merely serves as a first approach which will be modified in the following in order to clearly show the categorical difference between the two domains and from there to be able to explain their overlap. While many researchers working with IE languages – due to the empirical data in those languages – state that evidentiality usually is concomitant with epistemic overtones, typologists working with non-IE languages from the beginning of research in evidentials have warned against bluntly subsuming both domains under one broad heading without distinguishing between these concepts on a notional level. These linguists usually plead for clear distinctions; e.g. Anderson (1986) (see Chapter 1) and Willett (1988), who defines the distinctive feature of evidentials as opposed to true epistemics by the answer to the question “how the speaker obtained the information on which s/he bases an assertion” (Willett 1988: 55). With the growing body of data from languages with distinctly evidential categories without modal overtones, more and more researchers have pleaded to draw a sharp line between evidentiality and epistemic modality (Wierzbicka 1994, De Haan 1999, van der Auwera and Plungian 1998, Aikhenvald 2003, 2004, Giacalone Ramat and Topadze 2007, cf. also Diewald and Smirnova [to appear]). Some quotes illustrate the span of argumentation used here: […], evidentiality, deals with the evidence the speaker has for his or her statement, while the other, epistemic modality, evaluates the speaker's statement and assigns it a commitment value. This evaluation is obviously done on the basis of evidence (which might or might not be expressed overtly, or which might or might not be expressed by means of evidentials), but there is nothing inherent in evidentials that would compel us to assign an a priori epistemic commitment to the evidence. (De Haan 1999: 98-99) [our emphasis]

The common deictic nature and the different realizations of source concepts 81 The evidentials, however, signal only the way the speaker arrived at knowledge about the event. (Bybee 1985: 182) Evidential markers may indicate a speaker’s attitude towards the validity of certain information but do not have so. This is why evidentiality should not be considered as part of the ‘linguistic coding of epistemology’. (Aikhenvald 2003: 13)

2. The common deictic nature and the different realizations of source concepts The deictic anchoring of the modal and evidential dimensions has been discussed in several places in this book. In this section we argue that the deictic nature of both categories and the notion of “source” (which is inseparable from the notion of deixis) are responsible for the parallel structuring as well as the notional distinctness. Epistemic modality concerns the factuality judgment of the proposition in relation to the situation of the “epistemic” origo. Evidentials concern the indication of a specific source used by the speaker (origo) to make the statement. The central distinctive feature between evidentials and epistemics lies in the existence of a secondary event (the evidential source) in the former. As we have seen in previous chapters, this is what is meant by stating that evidential markers indicated that the speaker has proof, hints, evidence for his statement, and it is directly expressed in the definitions for evidentiality rendered in Chapter 1 (see also Bybee 1985: 184, Anderson 1986: 273, Aikhenvald 2003: 1). Thus, a parenthesis like wie ich höre in (5) has a purely evidential meaning, which contrasts with the meaning of an epistemic periphrastic evaluation like (6): (5) (6)

Sie ist – wie ich höre – in München.2 ‘She is – as I am told – in Munich.’ Sie ist – wie ich vermute – in München. ‘She is – as I assume – in Munich.’

The parenthesis wie ich höre in (5) denotes the existence of acoustic evidence, more exactly, evidence in the form of “hearsay”, which induces the speaker to utter the statement Sie ist in München ‘She is in Munich’. An epistemic judgment concerning the degree of factuality of this proposition

82 Evidentiality and modality – drawing the lines (i.e. Sie ist in München ‘She is in Munich’) is not expressed in the parenthesis in (5) as opposed to the one in (6). In (5), the speaker does not attribute any amount of uncertain factuality to this statement; s/he just indicates the source of the knowledge on which it rests. The same can be said of other evidential constructions in German, like scheinen in the following example, which is paraphrased by an adverbial as well as by our formalized paraphrase: (7)

Sie scheint beim Friseur gewesen zu sein. ‘She seems to have been at the hairdresser’s.’ Offensichtlich war sie beim Friseur. Obviously P P, because Q [Origo can observe some traces of P]

If we imagine the utterance of this sentence in a situation, where the speaker sees the person spoken of crossing the street after the lunch break with a much shorter haircut as compared to the one she had in the morning when they both met in the office, it becomes quite clear that what the speaker wants to express by the scheinen-construction is the fact that s/he has (visual) evidence for the statement Sie ist beim Friseur gewesen. The attribution of an uncertain epistemic value is not part of the inherent meaning of this statement. The use of a modal verb instead of an evidential produces a different meaning, as can be tested by inserting muss or kann instead of scheinen in (8) and (9). The respective factuality value of each modal verb can be made explicit by paraphrases containing a sentence adverbial that conveys the meaning of the modal, or a superordinate clause: (8)

(9)

Sie muss beim Friseur gewesen sein. ‘She must have been at the hairdresser’s.’ Ich halte es für wahrscheinlich, dass P. ‘I regard it as probable that P.’ Bestimmt P ‘certainly P’ Sie kann beim Friseur gewesen sein. ‘She may have been at the hairdresser’s.’ Ich halte es für möglich, dass P. ‘I regard it as possible that P.’

The common deictic nature and the different realizations of source concepts 83

Vielleicht P ‘perhaps P’ Important in the context of contrasting the epistemic modals with evidentials is the fact that in these cases the focus is not on evidence for the proposition but on the fact that the speaker is not entirely sure whether the statement is true or not, i.e. these sentences represent an uncertain factuality judgment of the speaker. That is: the deictic value in the modal dimension is the degree of factuality of the proposition as it is judged by the speaker. The point of reference for the degree of factuality of the proposition is the degree of factuality that the speaker attributes to himself/herself. As the speaker regards his/her existence, his/her state of being at the moment of utterance as factual, this state represents the near value: a narrated event that is judged by the speaker to be on the same level of existence receives the value [factual], while any narrated event that is not judged that way receives a remote value (cf. Rauh 1984: 52). A remote factuality value can be encoded by a range of lexico-syntactic expressions (like an epistemic parenthesis as wie ich vermute ‘as I suppose’) and by grammatical markers of epistemic modality. In German, these are the verbal moods (the indicative, subjunctive I and subjunctive II including the würde-construction) and the epistemically used modal verbs. The six relevant modals of German are dürfen ‘to be allowed to’, können ‘can, to be able to’, mögen ‘to like, may’, müssen ‘must, to have to’, sollen ‘shall, to be to’ and wollen ‘to want’.3 The paraphrases in (8) and (9) spell out the different degrees of doubt or certainty the speaker has concerning the factuality of the proposition.4 Stating that one regards something as probable, likely, possible and so on, amounts to stating that one – in fact – does not know, and therefore does not commit oneself to the truth of the proposition. The modals in this use signal that the speaker cannot attach a definite value of factuality to the proposition (Lyons 1977: 800). From the speaker’s perspective the proposition cannot be assessed as factual or nonfactual: it may be either. Thus, the meaning of the grammaticalized modals is that the speaker has tried to evaluate the factuality of the proposition but has not reached a definite conclusion: the speaker cannot (or does not want to) determine the “location” of the proposition in the deictic modal space. This indetermination of the speaker concerning the factuality of the proposition, which is the result of an evaluation procedure, is the distinc-

84 Evidentiality and modality – drawing the lines tive value of the epistemic modals. It is noted here as [factual/nonfactual] or [+/-factual], i.e. ‘no definite value attritubted’. These modals form a paradigm expressing fine-grained distinctions, which have been explored in Diewald (1999, 2001) and will be briefly touched upon later as far as they constitute a necessary background for the system of evidential distinctions. Epistemic modals imply a “modal” source, which indeed is one of the defining features of all modal concepts and every possible usage of modal verbs (Bech 1949, 1951, Diewald 1999, see next section). In the epistemic usages, this source is located in the speaker: it lies in the deictic origo. It is not a fact or phenomenon of the outside world, which is perceived and used by the speaker as evidence. The modal source in epistemic modal judgments is nothing but the speakers own judgment, which of course, may not be unconditioned; but these potential motivating conditions remain unexpressed by the modal marker itself (they may be expressed in the surrounding text, of course).5 That is, while the evidential source is an information-based source, the modal source is a judgment-based source. We mark this difference in source features by distinguishing a modal source from an evidential source (i.e. information source). The following test sentences once more illustrate in some detail that in pure epistemics (here expressed by müssen) any type of evidential source may be explicated in the textual surroundings, thus justifying the speakerbased factuality judgment, which, however, in itself does not contain any hint of an evidential source. (10) (11)

(12)

Sie muss ihre Doktorarbeit abgegeben haben. ‚She must have handed in her doctoral thesis.’ Ich habe sie gestern auf der Party gesehen. Sie muss ihre Doktorarbeit abgegeben haben. ‘I saw her at the party yesterday. She must have handed in her doctoral thesis.’ [Situational setting: The speaker is attending a party and talking to the hearer. Both are able to see the person talked about] > Sie muss ihre Doktorarbeit abgegeben haben. ‘She must have handed in her doctoral thesis.’

The mirror images are given in the following examples with a pure evidential (here exemplified by scheinen). In principle, evidentials do not contain

The common deictic nature and the different realizations of source concepts 85

any information on a judgment concerning the degree of factuality allotted by the speaker. But, of course, this information may be added in the surrounding text or situation. (13) (14)

(15)

Sie scheint ihre Doktorarbeit abgegeben zu haben. ‘She seems to have handed in her doctoral thesis.’ Sie scheint ihre Doktorarbeit abgegeben zu haben. Ich bin (ziemlich/fast/sehr) sicher, dass sie das noch rechtzeitig geschafft hat. ‘She seems to have handed in her doctoral thesis. I am (rather/almost/dead) sure she’s made it in time.’ A: Sie scheint ihre Doktorarbeit abgegeben zu haben. A: ‘She seems to have handed in her doctoral thesis.’ B: Dann hat sie es also noch rechtzeitig geschafft? B: ‘She’s made it in time, you mean?’ A: Naja, ich weiß es nicht genau, aber ich habe sie gestern auf der Party gesehen. A: ‘Well, I am not sure, but I saw her at the party yesterday.’

In these cases the proposition “sie hat ihre Doktorarbeit abgegeben” is inferred by the speaker on the basis of source information. This reference to source information is encoded by the evidential scheinen, which, however, does not make any factuality judgment in itself. The proposition can be combined with epistemic judgment of different strength, i.e. evidentials are neutral to a factuality value in the same way that epistemics are neutral to source information. That is, evidentials do not inherently imply a particular factuality judgment and, on the other hand, epistemics do not inherently imply reference to a particular information source. This may be further elucidated by the following simplified test sentences with scheinen and their paraphrases. They show that evidentials agree with any factuality value, any degree of “certainty”, which thus cannot be part of the semantic component of the evidential itself. (16)

Sie scheint ihre Arbeit abgegeben zu haben. ‘She seems to have handed in her doctoral thesis.’ Ich vermute das, (weil ich sie gestern auf der Party gesehen habe.) ‘I assume that (because I saw her at the party yesterday).’ Ich bin mir (dessen sehr) sicher, (weil ich sie gestern auf der Party gesehen habe.)

86 Evidentiality and modality – drawing the lines ‘I am (dead) sure (because I saw her at the party yesterday).’ Ich bin davon überzeugt, (sie war doch gestern auf der Party.) ‘I am convinced (because I saw her at the party yesterday).’ As evidentials per se are neutral with regard to a factuality value, they cannot be part of the system of epistemic distinctions as expressed by the modal verbs. The opposite is true of epistemic modal verbs, which may be contextually associated with any type of information source, which again, is proof that they do not contain this feature as part of their semantic structure. (17)

Sie muss die Arbeit abgegeben haben. ‘She must have handed in her doctoral thesis.’ Katarina hat es mir erzählt. ‚Katarina told me.‘ Ich habe sie gestern auf der Party gesehen. ‚I saw her at the party yesterday.‘ Ich habe sie mit einem großen Paket Richtung Dekanat gehen sehen. ‘I saw her heading for the dean’s office with a large parcel.’

Table 2. Epistemic and evidential paraphrases Epistemic modality: Die Lieferung dürfte/könnte/mag größer sein als erwartet. ‘The delivery might/could/may be larger than expected.’ Origo attributes an uncertain degree of factuality to P. Evidentiality: Die Lieferung scheint/droht/verspricht größer zu sein als erwartet. ‘The delivery seems/threatens/promises to be larger than expected.’ Origo has an information source for P.

Summarizing this discussion we may state the following: while epistemic markers, which are usually treated as the grammatical core of modality, instantiate speaker-based factuality judgments, i.e. the degree of certainty, with which the speaker is willing to assert the proposition, evidential markers serve to indicate the information source, which is the speaker’s basis for his/her statement. While epistemics judge modal values on a scale of certainty/factuality, evidentials refer to different kinds of information sources. Thus, the gist of the source distinctions is captured in the exam-

The common deictic nature and the different realizations of source concepts 87

ples and their paraphrases in Table 2. In particular the following points should be noted: 1. Epistemic modality and evidentiality are semantic-cognitive domains which are concerned with the speaker-based evaluation of the ontological status (the kind of existence) of a described event/situation. 2. Both are deictic categories, i.e. they make a speaker-based evaluation of the proposition concering each dimension’s particular semantic domain. 3. Both share the concept of a deictic source, the origo. 4. Epistemic modality concerns the linguistic encoding of a deictic factuality judgment of the speaker concerning the described event. 5. Evidentiality concerns the linguistic encoding of the information source, i.e. the type of source the speaker adduces for the described event. 6. The essential contrast between them is the fact that the modal source in epistemics is not explicitly expressed, while it is encoded in the evidentials as the secondary predication. Table 3. Epistemic and evidential distinctive features Common feature (deictic relation)

Distinctive features

Speaker-based evaluation of ontological status of event + speaker-based factuality judgment – speaker-based reference to information source Ð epistemic modality

+ speaker-based reference to information source – speaker-based factuality judgment Ð evidentiality

The abstract common feature of both categories, i.e. the speaker-based evaluation of the ontological status of a described event, as well as their distinctive features which concern the nature of a source concept, may be put down as given in Table 3.

88 Evidentiality and modality – drawing the lines 3. Diachronic explication of distinctions between evidentials and epistemics Modal and evidential sources have a lot in common but also have some important, category-distinctive differences. The last two sections have shown the notional source distinctions, which make a difference, and the deictic relation, which makes a commonality. The present section will show that both classes – i.e. evidential and epistemic verbs – in German have fundamental differences in their diachronic lexical ancestors (the grammaticalization “source”, which is not to be (con)fused with the modal source or any source in a relational structure), a fact which in our view is a major argument to keep the two categories apart, at least, if one wants to explore their diachronic grammaticalization paths. As an overview, the diachronic difference in the origin of both types of verbs – later evidentials and later epistemic modals – may be summarized in a list. The antecedents of modal verbs (for an extensive treatment see Diewald 1999) 1. are IE preterito-presents (i.e. perfecto-presents, old IE perfects which in Germanic languages have present meaning); 2. are stative, resultative verbs (due to former perfect meaning); 3. are pseudo-transitive verbs (i.e. combined with accusative complement without being transitive); and 4. have an experiencer subject as the carrier of a state (which is encoded in the verbal predicate) expressing psycho-physical or spatial location (cf. Meid 1983: 335; Meid 1971: 18, 39; Birkmann 1987: 87). The diachronic development of modal verbs from “preterito-presents” or “psycho-physical resultative verbs” implies a number of complex changes on the semantic as well as the structural level. The most important ones are: 1. the complementation by an infinitive which results in a condensed stative predicate; 2. the reinterpretation of the resultative semantic structure into a structure involving a modal source; 3. the development of a “secondary” experiencer (which is the encoded subject in non-deictic, narrow-scope modal uses);6 4. the development of the central modal construction, i.e. a stative verb predicating a state concerning the infinitive complement on the subject,

Diachronic explication of distinctions between evidentials and epistemics 89

whereby this state is induced by a modal source; this can be paraphrased as “(the subject of the sentence) is in the state of having been given a directive (by the modal source)”; 5. the differentiation of the modal meanings into deontic, volitional, dynamic (and possibly other) modal values via specification and feature variation of the modal source (see Bech 1949, 1951, Calbert 1975, Lyons 1977, Diewald 1999); 6. the development of the central deictic meaning via (inter alia) the reinterpretation of the modal source as the deictic origo; and 7. the reinterpretation of non-deictic modal values as various kinds of uncertain factuality judgments (the modal sentence represents the result of a factual judgment procedure by the speaker). Of most importance is the fact that the modal source has always been distinct from the subject of the sentence which is the experiencer of a result of an event/situation instigated by the modal source. The diachronic source items of evidentials in German show a remarkably different make-up, and also their development into evidentials differs markedly from the development of epistemic uses in the modals. The major points are given here as lists, too. It should be kept in mind, however, that many of the issues addressed here will come up in the discussion of the diachronic data in the chapters to come. The relevant features of the antecedents of the four evidential verbs are the following: 1. the ancestors of evidential verbs are dynamic verbs, denoting some kind of change or transformation; 2. the verbal scene is conceptualized as implying some cause leading to some consequence; 3. as opposed to the concept of a modal source in modal verbs and their antecedents, the implied cause in antecedents of evidentials is not specified as an external source; 4. the antecedents of evidential verbs are not restricted to animate subjects; 5. the antecedents of evidential verbs denote some real world, “external” phenomena (as opposed to internal psychophysical or local states in preterito-presents); 6. they are intransitive verbs (werden, scheinen) or fully transitive verbs (drohen, versprechen) with typically ingressive semantic features; and 7. they do not constitute a homogeneous group comparable to the preterito-presents in the case of the modals.

90 Evidentiality and modality – drawing the lines The development of evidentials from pre-evidentials is highly deviant from the development of the modals; it implies 1. the complementation by an infinitive, which – as opposed to the modals – takes an idiosyncratic chronological path in each evidential verb; 2. the reinterpretation of a dynamic verb into a stative verb (resp. auxiliary); 3. the reinterpretation of a “cause” of movement or effect into an evidential source causing the speaker to make a statement, i.e. the evidential source is not the deictic origo (as opposed to the modal source in the deictic use of the modals). This process is called here the relocation of the source entity. The last feature points to the most important fact that the evidential source is conceived of as distinct from the speaker (but co-present with the speaker). By using the evidential the speaker points to this source, i.e. s/he locates the source and only from there makes a statement concerning the (primary) proposition. In other words: in evidentials the speaker points to the existence of the evidential source and makes the statement on the basis of the factuality of this evidential source.

4. Typological issues In addition to the arguments brought up so far a look at evidential distinctions in languages with sophisticated and fully grammaticalized systems of evidential markers, like that of Tariana, which has already been presented in Section 1.1.1, must lead to the conclusion that encoding the information source does not necessarily entail concomitant epistemic marking. Observations supporting this view are the following: First, evidential and modal markers may be combined in a sentence. This observation supports the notion of two independent categories, with independent functions and meanings, which are freely combinable in a sentence. In a number of languages, evidentiality marking is mutually exclusive with mood and modality, as is the case in Abkhaz and in many Samoyede languages. However, in Western Apache, Jarawara and Tariana, mood and modality markers can occur together with evidentials. (Aikhenvald 2003: 16)

Second, evidential expressions may be free of any epistemic overtone. This, too, is a crosslinguistic obversation:

Areas of overlap between evidentiality and epistemic modality 91 For Kashaya Pomo, a Hokan language spoken in Northern California, Oswalt says that ‘it might be noted that … all propositions with the Kashaya evidentials are presented by the speaker as certain and true’ (1986: 43). For Iquito, a Zaparoan language from the Andes, Eastman and Eastman say, ‘regarding the reportative suffix -na and free form kináhá “so it is said”, [their use] does not mean to cast doubt as to the truth of the statement, but merely implies that it is a reported statement (1963: 191). Finally, in Coos, an Oregon Penutian language, the evidential particle cku denotes evidence based on inference. Frachtenberg (1922: 388) analyses this morpheme as being composed of cΩ ‘slight surprise’ and ku ‘dubitative’. Regarding the status of this evidential, Frachtenberg notes that it is used ‘whenever the speaker wishes to state a fact that occurred beyond doubt, but whose causes are not known to him (1922: 388). […] The hypothesis I am using is that evidentials are in fact a priori unmarked with respect to a commitment to the truth of the speech utterance on the part of the speaker. Evidentials merely assert that there is evidence to back up the speaker’s utterance. Any connection between the two […] is secondary in nature. They encode different things (source of information vs. attitude towards that information). Although they are closely enough related to cause overlap in some languages, this overlap is not universal. (De Haan 1999: 87)

This is to say that the behaviour of müssen and scheinen which we have illustrated in the previous section can be found in many languages (cf. De Haan 1999 for Dutch) and seems to represent a universal tendency.

5. Areas of overlap between evidentiality and epistemic modality Evidentiality and epistemic modality being two distinct categories (which we hope to have shown by now), does of course not prevent them from having areas of overlap (otherwise this chapter would not be written). Van der Auwera and Plungian (1998) (see also Faller 2002, Nuyts 2001, 2005, and Mushin 2001) suggest the following classification of this part of the semantic map of both categories. As can be seen in Figure 3, it is the regions of epistemic necessity and inferential evidentiality that overlap. [...] the inferential reading amounts to epistemic modality and more particularly epistemic necessity: for both categories we are dealing with the certainty of a judgment relative to other judgments. From this point of view it also causes no surprise that inferential evidentials often receive an English translation with epistemic must. Inferential evidentiality is thus regarded as an overlap category between modality and evidentiality. (van der Auwera and Plungian 1998: 85-86)

92 Evidentiality and modality – drawing the lines Necessity ...

Deontic necessity

Epistemic necessity = Inferential evidentiality Evidentiality

Quotative evidentiality

...

Figure 3. Overlap of evidentiality and modality (Inferential evidentiality = epistemic necessity) (van der Auwera and Plungian 1998: 86)

In contrast to van der Auwera and Plungian (1998) we do not assume a total overlap of epistemic necessity and inferential evidentiality. Instead, we assume that the sub-domains of epistemic necessity and inferential evidentiality can and must be distinguished, for the following reasons: 1. not every inferential statement needs to have an epistemic modalization; 2. not every statement expressing epistemic necessity needs to represent the result of an inferencing process. The following quote supports our opinion:7 In particular, an indirect evidential, which indicates that the speaker has only indirect knowledge concerning the proposition being asserted, implies that the speaker is not totally committed to the truth of that proposition and thus implies an epistemic value. (Bybee, Perkins and Pagliuca 1994: 180) [our emphasis]

Thus we agree that both sub-categories are related via implicational procedures. Furthermore, we contend that this implicational relation is not obligatory. Keeping this in mind, it is not surprising that the two categories tend to be mixed and expressed by the same linguistic entities. At the same time, however, one cannot rule out that both sub-categories are conceptualized and expressed independently of each other. Empirical argument: in languages with overlapping categories of epistemic modality and evidentiality or “polysemous” linguistic meaning expressing these categories, the area of overlap/polysemy lies exactly in the domain specified above (this concerns all European languages, including Baltic and Slavic languages). The inferred in a three-term system may acquire epistemic extensions of uncertainty and probability – as in Quang and in Tsafiki. In Shipibo-Konibo, both inferred evidentials (-bira and -mein) have overtones of uncertainty and

Areas of overlap between evidentiality and epistemic modality 93 doubt. In some four-term systems, only nonvisual acquires epistemic extensions, since it may imply that the speaker is not in control. But this is by no means universal. Many multi-term systems require subtle precision in indicating how the information was obtained, which leaves little leeway for uncertainty. Thus, epistemic meanings are expressed evidentials. Moreover, languages with multi-term evidentials generally tend to have a multiplicity of other verbal categories, especially ones that relate to modalities – examples include Tsafiki, and Tariana. (Aikhenvald 2003: 15)

Logical argument (implicational relations): usually, it is not possible to be sure 100%, whether the conclusion is valid even if the premise is true (that is the case with deduction, induction, and abduction); the conclusion always has the status of a (strong) hypothesis. On the other hand, it may be assumed that the higher the degree of certainty a speaker assigns to a statement the more likely it is that the speaker has “good reasons” (i.e. evidence) for it. From this it follows that in German all evidentials in the field of inferential evidentiality usually seem to display epistemic implicatures. This, too, is the opinion of most scholars working in this field. As the paraphrases in earlier sections have shown, the relation between the two domains is one of converse implicatures and converse inherent semantics. German evidentials stereotypically display epistemic conversational implicatures. These epistemic nuances are not part of the inherent semantics of the evidentials. And vice versa: German epistemic modals (müssen, können, mögen, dürfte) may often have evidential – inferential – interpretations (in particular if inflected with the dental suffix), which, however are not part of their inherent semantics. That is, we do not consider modals to be evidential expressions in PDG. Confer also Plungian (2001), who mentions possible textual conditions for an evidential implicatures in epistemics: […] an epistemic marker contains more evidential properties when the source of the speaker’s hypothesis is specified (it may be based on a priori as well as a posteriori evidence, etc.). (Plungian 2001: 354)

As mentioned before, the evidential meaning of German epistemic modal expressions is assumed to arise via pragmatic strengthening of particular conversational implicatures. The speaker marks the epistemic modal status of the proposition characterizing it as more or less sure, i.e. indicating his/her degree of commitment to the truth of the proposition. In this case, on the hearer’s side, the implicature may easily arise that the speaker’s attitude towards the proposition (and the proposition itself) is based upon some evidence available to him/her. This implicature may be gradually

94 Evidentiality and modality – drawing the lines conventionalized, or “semanticized” (Hopper and Traugott [1993] 2003), becoming part of the semantic content of the originally epistemic modal form.8 Conversational implicatures are not part of the inherent semantics, therefore, as already Anderson (1986: 274) emphasizes, evidential implicatures do not justify speaking of evidentials, and vice versa, epistemic implicatures do not justify classification as epistemics. Finally it may be noted in passing that the fact that both notions – evidentiality and epistemic judgment – are compatible with each other, is a further argument for our point that they are distinct categories, with distinct meanings: it is possible that the speaker wants to express both meanings (epistemic judgment and source information) in the same sentence: however, s/he is not forced to do so, as we are dealing with two independent categories. We may represent the situation as a converse distribution of semantic features and conversational implicatures in the modal epistemic and evidentials systems respectively. This relationship is illustrated in Table 4. Table 4. Converse distribution of semantic features and conversational implicatures Evidentials Inherent semantic feature Conversational implicature Epistemics Inherent semantic feature Conversational implicature

Source information +

Factuality judgment –

–

+

Source information –

Factuality judgment +

+

–

Summarizing this section (and this chapter) we conclude that we share the conviction of many linguists that it is absolutely necessary to draw a clear theoretical distinction between the two concepts, between evidentiality and epistemicity/factuality. It is only on the basis of a sharp distinction that overlaps and gradation between the two domains can be properly integrated into a linguistic description. Thus, we postulate two distinct paradigms for PDG, one modal-epistemic, one evidential. The two systems are 1. semantically coherent, 2. formally homogeneous,

Areas of overlap between evidentiality and epistemic modality 95

3. derived from their own specific grammaticalization paths respectively, 4. and compatible with typological classifications. The epistemic system is closely integrated into the paradigm of verbal moods, i.e. it encodes deictic values concerning speaker-based factuality judgments (see Table 1 above). The evidential system, on the other hand, encodes speaker-based information source and is in principle independent of the system of factuality markers. This distinction has been put forward here on the basis of synchronic, diachronic and typological arguments. As concerns the original classification given in Diewald (2004) (see Section 1), the following revisions have been made: 1. German evidentials may have epistemic overtones. However, they are not inherently epistemic. Instead, they are neutral towards any epistemic value and therefore allow combinations with any degree of epistemic certainty. 2. The facts call for a separation of evidentials and epistemics into two distinct systems with converse distribution of inherent semantics and conversational implicatures. 3. The labels for features will be adapted with reference to typological studies and continued empirical studies with our corpora. 4. The features “+/-ingressive” will be omitted as they render earlier stages of the grammaticalization path. For the German system the decisive part of the universal classification of evidential values elaborated in Chapter 2.2, is the inferential evidential section (see Chapter 2.2.2). As a first attempt, it is extracted and provisionally supplemented with the respective German constructions in Table 5. Table 5. Preliminary classification of inferential evidential values in German inferential evidentiality conceptual

werden & infinitive

conceptual/ perceptual

scheinen & zuinfinitive

perceptual non-desired

desired

drohen & zu & infinitive

versprechen & zu & infinitive

96 Evidentiality and modality – drawing the lines The following minimal pairs are meant to exemplify these distinctions and semantic values. (18)

(19)

(20)

(21)

In Hannover wird es regnen. ‘It will be raining in Hannover.’ [inference based on general knowledge] In Hannover scheint es zu regnen. ‘It seems to be raining in Hannover.’ [inference based on general knowledge and/or on observable evidence] In Hannover droht es zu regnen. ‘It threatens to start raining in Hannover.’ [inference based on observable evidence, propositional content non-desired] In Hannover verspricht es zu regnen. ‘It promises to start raining in Hannover.’ [inference based on observable evidence, propositional content desired]

As this system stands now, it is of course a hypothesis the plausibility of which has to be proved in the rest of the book.

Chapter 4 Grammar and grammaticalization

As we regard the development of evidentials in German as a process of grammaticalization, it is necessary to provide the basic theoretical assumptions of grammaticalization theory that guide our study. Therefore this chapter is devoted to laying out the relevant concepts and tenets of grammaticalization theory. Beyond this introduction to theoretical considerations this chapter provides exemplary results from our corpus study, i.e. it intertwines theoretical explication and corpus based exemplification. Section 4.1 is devoted to definitions of grammaticalization and grammar, Section 4.2 discusses the parameters of grammaticalization as set up by Lehmann, and applies them to our data in an investigation of the degree of grammaticalization of drohen & infinitive in PDG. Section 4.3 provides a rough sketch of the respective degrees of grammaticalization of all four verbs when checked against Lehmann’s grammaticalization parameters.

1. Obligatoriness and paradigmatic organization Grammaticalization theory today is a broadly accepted and well-introduced linguistic model for describing and explaining the rise of grammatical formatives and grammatical paradigms in language. This broad acceptance, however, also means that there are several conflicting notions of relevant concepts. Therefore, it seems necessary to sketch briefly the central concepts and their definitions as used in this study. As most influential definitions of grammaticalization agree, grammaticalization is a process which leads to the rise of new grammatical markers and grammatical oppositions from source items which did not have a grammatical function (or that particular grammatical function) before. This is reflected in most definitions of grammaticalization, e.g. the one by Lehmann:1 Grammaticalization is a process leading from lexemes to grammatical formatives. A number of semantic, syntactic and phonological processes interact in the grammaticalization of morphemes and of whole constructions. (Lehmann 1995 [1982]: viii).

98 Grammar and grammaticalization It has been common knowledge from the very beginning of modern grammaticalization theory that grammaticalization processes are of a composite nature, which is to say that there is no single process constituting a necessary and sufficient condition for talking about grammaticalization (for an overview of the literature see Diewald 2004, 2008). Instead, we have to deal with a range of processes which interact in grammaticalization. Traugott (2003: 644) states that “early grammaticalization can therefore be seen as a complex set of correlated changes”, which she specifies as follows: i. structural decategorialization; ii. shift from membership in a relatively open set to membership in a relatively closed one (i.e., from lexical category to syntactic operator category) in the context of a specific construction; iii. bonding (erasure of morphological boundaries) within a construction; iv. semantic and pragmatic shift from more to less referential meaning via invited inferencing and phonological attrition, which may result in the development of paradigmatic zero [...]. Traugott (2003: 644)

Summarizing literature on this issue,2 Diewald and Wischer (2005) state: The data which have been examined cross-linguistically constitute convincing evidence that the mechanisms and principles involved in grammaticalization conform to a complex process of coding and organization of language. It is not enough to know the individual mechanisms, because none of them is confined to grammaticalization. They can all be involved in other processes of language change. Only in their interaction, they make up a gradual and directed path that leads to the evolution of grammatical forms. (Diewald and Wischer 2005, manuscript)

In short, irrespective of the fact that the exact descriptions and classifications of the sub-processes vary among individual authors, grammaticalization scholars have acknowledged the multifactorial nature of grammaticalization from the very beginning. Consequently, the distinctive and unique feature of grammaticalization is generally seen in its particular combination and serialization of several processes and stages, which – among other things – are reflected in grammaticalization scales and paths and complex scenarios of successive contexts and constructions. Beyond the features conjointly defining the process of grammaticalization, there are others which are derived from the structural organization of grammatical expressions in a language. Of particular relevance in this context – though not undisputed – are the notions of obligatoriness and para-

Obligatoriness and paradigmatic organization 99

digm (or turned into dynamic concepts: “obligatorification” and “paradigmaticisation”). Referring to Roman Jakobson and the notion of “Ars obligatoria” of the medieval grammarians, Lehmann (2004: 155) states that grammar is “essentially a system of constraints”, and consequently, the “grammaticalization of a linguistic sign is a process in which it loses in autonomy by becoming more subject to constraints of the linguistic system”, i.e. an obligatory choice determined by linguistic rules and oppositions.3 As grammatical paradigms are constituted by obligatory choice among a defined set of members, the notions of paradigmaticity and obligatoriness are two sides of the same coin. A good description of the connection between obligatoriness and paradigmatic organization which also explains the notion of zero-marking is given in Bybee, Perkins and Pagliuca (1994): Once a gram or class of grams has come to be used in all appropriate contexts, redundantly or not, the lack of a gram of that class in the appropriate context becomes meaningful. Thus if a past tense gram develops and comes to be used in both redundant and non-redundant situations the cases where it does not appear will be interpreted as signalling meaning other than past. The tense category in that language will have become obligatory, with an overt gram for past and zero marking for present (Bybee 1990b). (Bybee, Perkins and Pagliuca 1994: 9)

As obligatoriness is usually taken to be the ultimate criterion of membership in the realm of grammatical signs, it is legitimate to formulate as a rule of thumb that “Grammatical categories are obligatory”, which marks a prototypical situation.4 And consequently, the feature of obligatoriness, i.e. subjection to grammatical rules within a paradigmatic arrangement, is treated as an abstract high-level criterion for defining the target area of grammaticalization. However, beyond the prototype, obligatoriness and paradigmaticity are a matter of degree. They are dependent on empirical verification and on actual usage, i.e. they have to be determined on a language-specific basis. Furthermore, as grammaticalization is the process of building and rearranging paradigms with signs becoming obligatory, it is of course to be expected that there is an infinitely fine-grained scale of intermediate phenomena. Consequently, there have always been authors, like Lehmann (1995 [1982]: 12) and Plungian (1998), who state that both criteria refer to phenomena which are matters of degree, i.e. grammatical categories can form more or less clear-cut paradigms, and the choice among their members can be more or less obligatory.

100 Grammar and grammaticalization On one end of the scale, there are paradigms the choice among whose members is 100 percent obligatory and subject to grammatical rules. These are rules for the choice among a morphological paradigm like the choice in gender concord in adjectives or the conditions for case marking in noun phrases (Det. & N) in German. Usually, grammatical items subject to this sort of obligatoriness are members of inflectional paradigmatic oppositions, i.e., notwithstanding periphrastic forms, at the core of such paradigms there is bound morphology in the form of inflection, i.e. old grams with heavy semantic and formal reduction. The selection of one element in paradigms of this sort is obligatory and governed by language-internal rules, which means that it is concomitant to higher level decisions. For example, in the case of gender concord in German, the choice of a particular nominal lexeme, which in German has a fixed gender, automatically determines the choice of gender concord in the attributive adjective and the determiner. Any deviant realization would produce incorrect utterances (e.g. ein silberner Löffel (masc.), *eine silberne Löffel, *ein silbernes Löffel ‘a silver spoon’). This type of obligatoriness is called here language-internal obligatoriness. As it is steered language internally and thus subject to formal triggers, its mechanism of choice can be represented by the following conditional sentence: If form x, then form y. Language-internal obligatoriness is contrasted with a second type of obligatoriness, for which the term communicative obligatoriness is chosen (see also next section under the parameter of paradigmatic variability). Communicative obligatoriness concerns the behavior of those linguistic items, which do function as grammatical closed-class items, but which are still not obligatory in the sense illustrated above. The term “communicative obligatoriness” is intended to embrace the fact that many categories are obligatory in the sense that they have to be realized in the relevant position, i.e. the speaker cannot leave them unspecified if s/he does not want to produce incorrect utterances, but the choice among the paradigmatic members of the category is not determined by language-internal features but by the communicative intentions of the speaker.5 A fine example of communicative obligatoriness is the voice distinctions in German, i.e. the choice between the active and the two passive constructions, the werden-passive, and to the so-called dative passive.6 Both passives are realized as periphrastic constructions which can be

Obligatoriness and paradigmatic organization 101

grouped into a paradigm together with the active verb form as the unmarked member. The three voice constructions of German are given in Table 1, examples with the verb überweisen ‘transfer’ are given in Table 2: Table 1. The voice constructions in German Active Agent Fin.Verb Recipient W-Pass. Theme Fin.Aux werden Recipient B-Pass Recipient. Fin.Aux bekommen PrepPhr.Agent

Theme PrepPhr.Agent Theme

PP main verb PP main verb

Table 2. Examples for voice constructions with the verb überweisen ‘transfer’ Active Das Unternehmen The company

überweist transfers

dem Verein to the society

die ganze Summe. the full amount.

W-Passive Die ganze Summe The full amount

wird is

dem Verein to the society

vom Unternehmen by the company

B-Passive Der Verein The society

bekommt gets

vom Unternehmen by the company

die ganze Summe the full amount

überwiesen. transferred.

überwiesen. transferred.

The voice distinctions obviously make up a grammatical paradigm in German (and are treated as such in current grammars). However, there is no linguistic context in which a speaker would be forced to use a passive in German, i.e. there are no contexts in which a passive construction is obligatory in the strict sense, meaning that another choice of one of the paradigmatic members would be grammatically wrong. The choice is controlled communicatively; it is a question of speaker perspective and not determined by linguistic structure.7 By analogy to the rule for language-internal obligatoriness, communicative obligatoriness can be formulated in a conditional clause which refers to a speaker’s needs. If intention x, then form y. According to the category concerned, this rule may be expanded as needed. For the choice of the werden-passive as in the example above, the conditions motivating its use may be formulated as in Table 3. This type of communicatively given obligatoriness is exactly what we find in the case of the evidential constructions of German (the paradigmat-

102 Grammar and grammaticalization ic organization of which will be described in detail in Chapter 7). The proposed paradigm of the four evidential constructions – which of course is seen to have the pure indicative form of the main verb as its unmarked base – is “obligatory” in a similar way as the choice of a passive form is. There is no structural, language-internal need to use any of these forms in a particular context, i.e. there is no language-internal obligatoriness. Speakers of German are not forced to express marked values of evidential meanings by using these forms. They may leave this information unspecified (note, however, that the indicative may be used to express an unmarked evidential value as well as a neutralized value not sensitive for evidentiality at all, cf. Section 2.2). Table 3. Conditions motivating the use of a werden-passive If the speaker wants to put the theme into the privileged topological position at the beginning of the sentence, and keep the lexical verb of the corresponding active sentence, and avoid a marked linear order (which in this case would be a topicalized object) then the speaker must use the werden-passive (given it is possible at all).

Also, as shown before (Section 1.1.1), there is a range of lexical, phrasal, syntactic alternatives to express evidential meanings in German as is once more illustrated in the following examples: (1) a. Anscheinend/offensichtlich scheitern die Verhandlungen. ‘Obviously/seemingly the negotations are failing.’ b. Es sieht so aus, als ob die Verhandlungen scheitern. ‘It looks as if the negotations are failing.’ However, if the speaker intends to express evidential meaning in a routinized way which triggers clear paradigmatic distinctions, s/he will use one of the evidential constructions according to the particular value to be expressed. (2) a. Die Verhandlungen scheinen zu scheitern. ‘The negotiations seem to fail.’

Obligatoriness and paradigmatic organization 103

b. Die Verhandlungen drohen zu scheitern. ‘The negotiations threaten to fail.’ c. Die Verhandlungen versprechen zu scheitern. ‘The negotiations promise to fail.’ d. Die Verhandlungen werden scheitern. ‘The negotiations are bound to fail.’ That is to say that by using this construction the speaker highlights the fact that s/he clearly wants to mark the evidential status of the proposition and that this information is not intended as the focus of the message, but as backgrounded information. Therefore, it is encoded by one of the grammaticalized evidential constructions which form an evidential paradigm with distinct values that could not be substituted by one another without change of meaning. Thus, while there is freedom to choose lexical material to encode evidential meaning in PDG, if the speaker decides to use a grammatical marker of evidentiality, one of the four evidential periphrases will be used obligatorily according to communicative intentions in the sense that the choice within the paradigm is systematically constrained. By analogy to the passive-rule in Table 3, we can formulate the following conditional rule for the use of evidentials in German:8 Table 4. Conditions motivating the use of an evidential periphrasis If the speaker wants to indicate that s/he has evidence for his or her statement, and keep the lexical verb of the corresponding indicative sentence, and does not want to foreground the evidential meaning then the speaker must use an evidential construction.

Summing up what has been discussed so far: the feature of obligatoriness (regularity or rule-boundedness), which is usually treated as one of the main indicators of grammar, is a matter of degree (cf. grammaticalization parameter 5, paradigmatic variability or obligatorification below). Very often, it is difficult to decide the degree of obligatoriness of a certain item. Many linguistic items which are classified as genuine grammatical categories of a language, and which are rightly classified as such, because they

104 Grammar and grammaticalization share many of the features of grammatical categories (a sufficient degree of formal and semantic reduction, paradigmatic association in a closed class), do not pass the strict test of obligatoriness. Therefore, language-internal obligatoriness, in the way defined above, is a sufficient but not a necessary condition for the status of a gram, i.e. a grammatical marker. Instead, many grammatical distinctions, many grammatical categories, are employed according to communicatively controlled conditions; they are subject to communicative obligatoriness. The rules of usage for the evidential constructions of German are of this type of communicative obligatoriness. Connected to the problem of obligatoriness and paradigm formation is the question of membership of periphrastic forms in a paradigm, i.e. the question by which criteria and when a particular periphrastic form should be inserted into or excluded from a grammatical paradigm. As this pertains to the topic treated here, the case may be illustrated by some examples from periphrastic verb forms with infinitive constructions, which closely touch the topic of this book. There are some infinitive constructions, which are traditionally seen as pertaining to the tense and mood paradigm – these are the periphrases with werden and würde like in wird fragen, wird gefragt haben (which are usually dubbed “future” and “future perfect” respectively), and würde fragen, and there are others, e.g. the infinitive constructions with modal verbs like mag fragen, dürfte fragen, which are not included, although many of them in frequency as well as in some of their functions equal verbal periphrases with werden (cf. Diewald 1999, Smirnova 2006). This is illustrated in Table 5. Table 5. Future tense/mood paradigms: periphrastic forms with infinitives Infinitive constructions typically inserted into the future tense/mood paradigm: wird & infinitive: wird fragen würde & infinitive: würde fragen wird & perfect infinitive: wird gefragt haben würde & perfect infinitive: würde gefragt haben Infinitive constructions typically not inserted into future tense/mood paradigm: modal verbs & infinitive mag/dürfte ... fragen modal verbs & perfect infinitive mag/dürfte ... gefragt haben

Again, there have been broad and long-standing discussions on the question of whether German does have an analytic future or not, and in what place the würde-construction falls,9 and to what degree modal verbs plus infinitives are realizations of a complex periphrastic mood system (Diewald 1999, 1999a). But as in the case of the voice distinctions, all

Grammaticalization parameters and German evidentials 105

these discussions so far are hardly reflected in the standard descriptions of verbal paradigms and categories in grammars and reference books. Without going into the details of this discussion, the gist of this example can be summarized as follows. As far as semantic, functional and structural aspects, as well as frequency and pragmatic factors are concerned, there is no convincing argument for drawing the line between members and non-members of grammatical paradigms where it is traditionally drawn. Periphrastic verbal constructions are a constant though often suppressed problem for any strict paradigmatic organisation. This statement of course is relevant for evidential constructions as well. They, too, up to now never appeared as a paradigm in a grammar or reference book, although they fulfill the same functions as other constructions which are treated as members of a paradigm. Therefore, the next section takes pains to prove the high degree of grammaticalization of these constructions.

2. Grammaticalization parameters and German evidentials In order to compare degrees of grammaticalization in a reliable way that allows quantitative and qualitative comparisons and judgements on all levels of linguistic structure involved in the process, one needs a testing procedure which can be used for all possible grammatical categories, all stages of the process and also for diachronic as well as synchronic comparison. The grammaticalization parameters set up by Lehmann serve this task and have proved a sharp tool for evaluating degrees of grammaticalization by a combination of formal and semantic features in both the syntagmatic and the paradigmatic dimensions. As the verification of the degrees of grammaticalization for each item treated is an a priori to further argumentation, and as the parameters constitute a tight-knit and complex system of distinctions, the discussion of the parameters has to be devoted some space here.

2.1. Lehmann’s grammaticalization parameters The parameters categorize the notions discussed in the last section, e.g. obligatoriness and paradigmaticity (cf. in particular parameters 2, 3, 5) and combine them with inherent features of the sign in question (i.e. its phonological and semantic features). By applying the parameters, it is possible to compare particular linguistic items precisely as to their respective beha-

106 Grammar and grammaticalization viour in each parameter, and thus decide on their relative degree of grammaticalization. This comparison is possible between items of the same synchronic period – e.g. one periphrastic construction as compared to another one – as well as on the diachronic level, i.e. comparison of a construction X1 at time t1 with a construction X2 at time t2 with the purpose of estimating the increase in grammaticalization of that linguistic entity in the course of time. The fundamental property relevant for breaking down the parameters is the autonomy of a sign, which by definition is high in lexical and low in grammatical signs. Applied to the syntagmatic and paradigmatic axis of language and split into the three properties of weight, cohesion and variability, the relative degree of autonomy of a sign can be measured by the following six parameters. The minus behind a parameter label means that this feature is reduced in grammaticalization, the plus means that this feature is increased under grammaticalization. The terms in italics are Lehmann’s names for the diachronic processes involved. Table 6. Grammaticalization parameters adapted from Lehmann (1985: 306) Paradigmatic

Syntagmatic

weight

(1) integrity (–) phonetic/semantic attrition

(2) structural scope (–)10 condensation

cohesion

(3) paradigmaticity (+) paradigmaticization

(4) bondedness (+) coalescence

variability

(5) paradigmatic variability (–) obligatorification

(6) syntagmatic variability (–) fixation

The weight of a sign concerns its distinction from the other members of its class. The more distinct it is, the less it is grammaticalized. Its paradigmatic realization leads to the first grammaticalization parameter, integrity, which is reduced by phonetic and/or semantic attrition. Its syntagmatic expression renders the second grammaticalization parameter, structural scope, which is reduced by condensation. The concept of cohesion captures relational aspects of a given sign and its neighbouring signs, i.e. the degree of dependence and sticking together. On the paradigmatic axis we get the third grammaticalization parameter, i.e. paradigmaticity; on the syntagmatic axis we get bondedness (parameter 4). Cohesion increases in grammaticalization on both the syntagmatic as

Grammaticalization parameters and German evidentials 107

well as the paradigmatic axis, which is called paradigmaticization and coalescence respectively.11 Bondedness and loss of integrity are the two parameters which come out most prominently in the grammaticalization scales which are commonly used in grammaticalization studies. Thus, Bybee, Perkins and Pagliuca (1994: 40) suggest a scale representing “degrees of fusion”, which proceeds from syntactic realizations to non-bound grams to inflection, derivation and lexical realization, thereby spanning all levels of linguistic structure. A category-specific example of such a scale is Lehmann’s cline for the development of case affixes, which has the following shape: relational noun > secondary adposition > primary adposition > agglutinative case affix > fusional case affix Figure 1. Grammaticalization scale for case affixes from relational nouns (Lehmann 1985: 304)

As clines like this, which are central to grammaticalization processes, represent typical stages of otherwise continuous, gradual shifts through several levels of linguistic structure, they presuppose a view of language that does not draw a sharp line between syntactic structure and lexical entities, or between words (i.e. free forms) and morphemes. Finally, Lehmann introduces the feature of variability, i.e. the mobility or shiftability with respect to other signs, which generally decreases with increased grammaticalization. Its paradigmatic aspect, called paradigmatic variability, is associated with obligatorification in the case of continued grammaticalization; its syntagmatic aspect, called syntagmatic variability, shows increasing fixation in correlation with increasing grammaticalization. Lehmann’s parameters have been successfully applied in many grammaticalization studies in the last years. They have proved particularly fruitful in the studies of the development of several verbal periphrastic categories, which makes them highly attractive for this study as well.12 The functioning of the parameters can best be demonstrated by an example. Therefore, in the next section each parameter is described in some detail, and its functioning is exemplified by data from our corpus with the construction drohen & zu-infinitive in PDG being used as an exemplar case.

108 Grammar and grammaticalization After that, grammaticalization data from the other three constructions in present day German are given in the form of a summary.

2.2. Degree of grammaticalization of evidential drohen: a detailed study In the German linguistic literature, drohen within infinitive constructions has been called a Modalitätsverb ‘modality verb’ (Askedal 1997, Engel 2004), a Halbmodalverb ‘semi-modal verb’ (Zifonun et al. 1997) or a raising verb (Reis 2005), without specifying its grammatical status. Some grammarians argue for treating drohen as an auxiliary (e.g. Jung 1988, Diewald 2004), others reject such a view (e.g. Eisenberg 2004, Engel 2004). Still others place the verb somewhere in between (e.g. Helbig and Buscha 1988). Finally, there are some authors (Gunkel 2000, Kokutani 2004) who claim that there is no distinction in status (between lexical versus grammatical drohen) at all, arguing that the difference between two readings of the verb is a matter of lexical semantics rather than of different – lexical versus grammatical – status. Heine and Miyashita (2004) is the first attempt to describe drohen within the framework of grammaticalization theory and, in doing so, have convincingly shown that the functional – i.e. grammatical – drohen exhibits a high degree of grammaticalization. For their investigation they applied the following four parameters of grammaticalization established by Heine and Kuteva (2002) which – as obvious – can easily be associated with the respective parameters of Lehmann’s system: a. extension (or context generalization) b. desemanticization (or semantic bleaching) c. decategorialization, i.e. loss in morphosyntactic properties characteristic of lexical or other less grammaticalized forms d. erosion (or phonetic reduction) The analysis of drohen within infinitive constructions with regard to these four parameters of grammaticalization has the result that drohen is a canonical instance of auxiliarization (see Heine and Miyashita 2004, 2008 for further details). The two comparanda are the lexical and functional uses of drohen & zuinfinitive in PDG – the division into these two uses being common practice among linguistics. These usages are illustrated by the following examples. (3) represents drohen as a (lexical) main verb with agentive semantics, in

Grammaticalization parameters and German evidentials 109

which case extraposition is what is to be expected; (4) shows functional drohen as an auxiliary (or: semi-auxiliary) verb with evidential semantics. Not surprisingly, it is assumed that the lexical usage is less grammaticalized than the functional usage, and this difference should come out in the grammaticalization parameters as well. We will see later, that – on the basis of our own corpus analysis – it is reasonable to assume (at least) three different readings of the verb drohen. For the time being, and in order to illustrate in principle the functioning of Lehmann’s grammaticalization parameters and the way we put them to use in this book, it may suffice to work with the two fundamental distinctions illustrated above. (3)

(4)

... worin diese drohen, sofort die Verbindung abzubrechen (DWDS, Kernkorpus) ‘in which they threaten to cut off the connection at once.’ Wenn man also merkt, daß eine Diskussion (etwa über Politik) in einen Streit auszuarten droht. (DWDS, Kernkorpus) ‘This means that as soon as you realize that a discussion (say about politics) threatens to turn into a quarrel.’

Before starting, one brief remark concerning the discriminatory power of the parameters is in place. The parameters are particularly distinctive in advanced stages of grammaticalization with ample formal changes having taken place. In less advanced grammaticalization processes, one or the other of the parameters may not produce notable differences between the two items compared. Thus, in our data on the four German evidential constructions, we found that the phonological aspects of parameter 1 (attrition) are not distinctive, i.e. there is not yet a sign of phonological attrition (loss of phonological substance) in the evidential verbs in PDG. Concerning their phonological substance both lexical and functional uses of the four verbs are equal. Similarly parameter 6, syntagmatic variability, is not distinctive as to degrees of grammaticalization, which is to be expected on typological grounds (see below). Still, even non-distinctiveness in a given case at a given stage is a valid and interesting result of the comparison. Beyond these two reservations, the parameters are highly distinctive. Parameter 1: integrity. There are some crucial differences between the lexical and functional variant of drohen, which strongly suggest that the semantics of functional drohen has undergone desemanticization. The lexical variant of drohen has strong intentional semantics; it normally combines with human or animate subjects. Frequently, drohen serves as a

110 Grammar and grammaticalization speech-act verb denoting the announcement of an action that is undesirable for the hearer; the action announced is typically to be carried out by the speaker, who is also the one making the announcement in oral form. The functional variant of drohen, on the other hand, shows the complete loss of both lexical semantic components: the intentional semantics as well as the component of threatening (i.e. speech act meaning). Only one semantic feature has been retained, i.e. the component of undesirability (i.e. negative evaluation) of the described event. The presence of some crucial semantic differences together with the absence of phonological traces of attrition nicely reflects the young age of this grammaticalization process, with semantic change starting the process of grammaticalization and therefore being clearly present and phonological attrition not yet being noticeable. Parameter 2: syntagmatic weight or structural scope. Lexical drohen realizes a wide range of its valence relations, only one of them being the option with an infinitive. As the sentences below show, it can take dative objects (b), direct- (c) and indirect-speech sentences (d) as well as dasscomplement clauses (e) and further complement clauses introduced by damit (f) as complements (see also Askedal 1997 and Heine and Miyashita 2004, 2008 for a similar analysis). Additionally, lexical drohen can take facultative prepositional complements as in (5g). The functional variant of drohen, on the contrary, occurs only with zu-infinitives (6). (5) a. Sie droht, mich zu verlassen (, wenn ich mich nicht verbessere). ‘She threatens to leave me (if I am not going to improve myself).’ b. Sie droht mir, mich zu verlassen. ‘She threatens [me(DAT)] to leave me.’ c. Sie droht: „Ich werde dich verlassen“. ‘She threatens: « I am going to leave you. »’ d. Sie droht, sie werde mich verlassen. ‘She is threatening to leave me.’ e. Sie droht, dass sie mich verlassen wird. ‘She threatens that she is going to leave me.’ f. Sie droht damit, mich zu verlassen. ‘She threatens with leaving me.’ g. Sie droht mit dem Finger/mit der Trennung. ‘She threatens with her finger/with separation.’

Grammaticalization parameters and German evidentials 111

(6)

Sie droht zu fallen/in Tränen auszubrechen/in Vergessenheit zu geraten. ‘She threatens to fall /to break into tears /to be forgotten.’

Furthermore the zu-infinitive complement of functional drohen cannot be replaced by any other constituents and cannot be pronominalized by damit ‘with it’, cf. (7) versus (8): (7) a. Sie droht, mich zu verlassen. ‘She threatens to leave me.’ b. Sie droht damit. ‘She threatens with it.’ c. Sie droht mit Trennung. ‘She threatens with separation.’ (8) a. Sie droht in Vergessenheit zu geraten. ‘She threatens to be forgotten.’ b. *Sie droht damit. *‘She threatens with it.’ c. *Sie droht mit Vergessenheit. *‘She threatens oblivion.’ These modifications show that lexical drohen has a valency relation with the subject, and thus serves as a predication about the subject of the sentence. The grammatical variant of drohen, on the contrary, does not function as a predication about the subject of the sentence. Instead, it is a secondary predication and modifies the whole proposition that is introduced by the infinitive. For this reason, the infinitive complement cannot be replaced by any other complement. Moreover, the lexical variant of drohen allows for adverbial modifications, while this does not hold for grammatical drohen, cf.: (9) (10)

Sie droht jetzt/weinend, mich zu verlassen. ‘She sobbingly threatens now to leave me.’ Sie droht *jetzt/*weinend in Vergessenheit zu geraten. *‘She sobbingly threatens now to be forgotten.’

The use of jetzt ‘now’ as in (10) is only acceptable if this adverb does not refer to the drohen-event but modifies the verbal event introduced by the

112 Grammar and grammaticalization infinitive. Hence, any further specification of functional drohen is blocked. If there is a modification, it does not concern drohen itself but the infinitive verb combined with drohen. The evidential variant of drohen does not show any restriction with regard to its subject: the latter can be animate (11) or non-animate (12); it may be a complement clause (13) or even an impersonal subject (14), cf.: (11)

(12)

(13)

(14)

Mein Mann droht krank zu werden. (taken from Heine and Miyashita 2008) ‘My husband risks falling ill/threatens to fall ill.’ Eine der größten politischen, sozialen und wirtschaftlichen Katastrophen drohte Deutschland zu vernichten. (DWDS, Gesprochene Sprache) ‘One of the biggest political, social and economic disasters threatened to destroy Germany.’ Dass wir das Heil in der Abwehr suchen, droht ein Desaster zu werden. (taken from Gunkel 2000, 113) ‘That we are looking for salvation in defense, threatens to turn into a disaster.’ Es droht zu regnen. (taken from Gunkel 2000: 114) ‘It threatens to rain.’

By way of conclusion it can be stated that the functional variant of drohen forms one constituent with the obligatory zu-infinitive. The lexical variant of drohen, on the other hand, serves as an independent constituent which may combine with other constituents at the clausal level. Hence the syntagmatic weight of evidential drohen has significantly decreased in contrast to its lexical counterpart, so that it can be considered an auxiliary (cf. also Askedal 1997: 13). Parameter 3: paradigmaticity or paradigmatic cohesion. The evidential construction drohen & zu-infinitive together with the constructions scheinen/versprechen & zu-infinitive and werden & infinitive forms a (grammatical) paradigm of inferential evidentiality in PDG, where drohen & zuinfinitive serves as a perceptual inferential marker. The perceptual inferential evidential value of drohen & zu-infinitive is the distinctive value of this construction within the paradigm. It also has an additional semantic component, i.e. it evaluates the described event negatively. Sharing the same general property, i.e. the evidential inferential function, the four constructions differ with regard to the specifications of the common basic meaning

Grammaticalization parameters and German evidentials 113

they express. This fact suggests the high degree of integration of each construction into the paradigm, insofar as each construction has its own distinct property within the paradigm (for further details see Chapter 7). As for the lexical variant of drohen, it shows only a very low degree of paradigmatic cohesion in an open lexical field including other expressions (verbs, lexical expressions or syntactic structures), which are centered around such semantic cores like ‘warn’, ‘risk’, ‘announce something negative’ etc. Thus, the degree of paradigmatic cohesion of lexical drohen is very low, whereas functional drohen shows a higher degree of grammaticalization with respect to this parameter. Parameter 4: bondedness or syntagmatic cohesion. There is no difference between lexical and functional drohen with regard to their cliticization status (and univerbation as well): they both occur as free morphemes and are not cliticized. However some symptoms on the syntactic level may be observed which suggest that the functional variant of drohen is grammaticalized to a higher degree. The phenomenon in question is the coherence of the infinitive construction. As Bech (1955) showed, there is a systematic difference between verbs admitting coherent and/or incoherent infinitive constructions. While a coherent construction has the infinitive within the verbal bracket in the middle field, an incoherent construction is one in which the infinitive clause is extraposed to the right of the right verbal bracket (cf. Zifonun et. al 1997: 2191). Verbs restricted to coherent constructions show a higher degree of bondedness (as the verbal elements are part of one verbal phrase) and are more auxiliary-like. As shown in (15), when drohen is used in an incoherent construction it can only be given its lexical reading. (16a) illustrates the evidentials reading which is associated with the coherent construction. As to the possibility of examples like in (16b), there are conflicting opinions as to whether extraposition of the infinitive complement is possible (or acceptable). Zifonun et al. (1997) assume that extraposition is acceptable, whereas e.g. Askedal (1997) and Heine and Miyashita (2004, 2008) do not agree with this analysis: Ich weiß, dass sie droht, mich zu verlassen. ‘I know that she threatens to leave me.’ (16) a. Ich weiß, dass sie in Vergessenheit zu geraten droht. ‘I know that she threatens to be forgotten.’ b. ?? Ich weiß, dass sie droht, in Vergessenheit zu geraten. ?? ‘I know that she threatens to be forgotten.’

(15)

114 Grammar and grammaticalization As regards this question we agree with the majority of the authors in assuming that extraposition is essentially not possible (even if not fully blocked). This may be seen as proof of the increasing degree of bondedness between the grammaticalized item drohen and the lexical material (zuinfinitive) to which it bears syntagmatic relationship. Parameter 5: paradigmatic variability. The lexical variant of drohen does not show any signs of obligatory use; it can be freely substituted by other lexical means depending on the communicative context. The use of evidential drohen is considerably constrained: the freedom of substitution by other linguistic entities is to a large extent restricted by the intended – notably grammatical – value which is ascribed to the proposition. As concerns drohen & zu-infinitive, as well as infinitive constructions with scheinen, versprechen and werden, they form a closed paradigm in PDG (see Chapter 7). That is, given that the speaker wants to mark his/her statement as based on his/her inference, s/he will use one of the four evidential constructions (cf. Section 4.1 on communicative obligatoriness). From this it follows that the freedom of choice is restricted to the four members of the paradigm, from which alternative values are selected. In this connection, evidential drohen shows a decreasing degree of paradigmatic variability and thus a higher degree of grammaticalization than its lexical counterpart. A second aspect of paradigmatic variability has to do with the availability of the range inflectional forms of each verb. As indicated above, “morphological degeneration”, i.e. the loss of the ability to inflect may result as an effect of attrition (Lehmann 1995 [1982]: 132), and, thus, under certain conditions, might be also treated under Parameter 1. However as the verbs in question do not lose their inflectional potential as a result of a phonologically accompanied attrition process, but just do not realize certain inflectional forms in their grammaticalized use, this property is treated here under Parameter 5. The reduction of inflectional paradigmatic variability, too, can be observed in the case of evidential drohen. It lacks a number of inflectional forms which are available to its lexical counterpart. For instance, it does not occur in periphrastic tense, mood and passive forms (see (17) to (19)), it does not appear as a present participle (20), and it cannot be used itself as infinitival complement as shown in (21) - (22). If drohen occurs in such forms, it can only be given its lexical readings, cf.: (17)

Da ist er sauer geworden und hat mir gedroht, daß er den längeren Arm hätte. (DWDS, Kernkorpus) ‘Then, he got angry and threatened me that he has more leverage.’

Grammaticalization parameters and German evidentials 115

(18)

(19)

(20)

(21)

(22)

Ich würde hierüber kein Wort verlieren, wenn mir nicht vor kurzer Zeit, allerdings von sehr unmaßgeblicher Seite, mit Enthüllungen aus jener Zeit gedroht worden wäre. (DWDS, Kernkorpus) ‘I would not lose a word on it, if I had not been threatened recently – though from a quite unimportant quarter – with disclosures from that time.’ Sie ist ein Damoklesschwert, aufgehängt über den Spitzen der Regierung und Verwaltung, und es würde manches nicht geschehen und manches geschehen, wenn dieses Schwert nicht über den Häuptern drohen würde. (DWDS, Gesprochene Sprache) ‘.... if this sword did not threaten them above their heads.’ Doch der Mann hielt drohend ein Gewehr im Anschlag und brüllte. (DWDS, Kernkorpus) ‘But the man threateningly levelled a gun at him and shouted.’ Damit wollen Sie uns drohen, das werden Sie büßen! (DWDS, Kernkorpus) ‘Do you think you can threaten us like that ? You will pay for it if you do!’ Uns zu drohen, ist idiotisch. (DWDS, Kernkorpus) ‘Threatening us would be idiotic.’

It should also be noted that lexical drohen occurs in the imperative forms, whereas an imperative form of evidential drohen is impossible (the following examples are taken from Askedal 1997: 13), cf.: (23) (24)

Drohe den Arbeitern gefälligst nicht, sie zu entlassen! ‘Don’t you dare and threaten the workers with firing them!’ *Oh, ihr langwierigen Verhandlungen, droht euch dennoch nicht zu zerschlagen! *‘Oh, you tedious negotations, don’t threaten to break up though!’

Hence evidential drohen has lost to some extent its capacity to realize “normal” inflectional categories of main verbs. It functions as a grammatical formative with its own inflectional paradigm which is typical of other auxiliaries as well. In the light of all these observations we can postulate a certain degree of reduction of paradigmatic variability. Parameter 6: syntagmatic variability. There is no difference between lexical and more grammaticalized variants of drohen in respect to this pa-

116 Grammar and grammaticalization rameter. This fact originates from the overall structural characteristics of the German language, where the finite verb always takes a fixed position in the clause. Hence both main verbs and auxiliaries show the same degree of syntagmatic variability, as both take the same position (of finite verb) in sentences. To summarize, the application of the six parameters of grammaticalization has shown that in five out of six parameters there are significant changes in status of the verb drohen from a lexical main verb towards a grammaticalized evidential auxiliary. Thus, we were able to reveal convincingly the relatively high degree of grammaticalization, i.e. of auxiliarization, of the verb drohen.

2.3. The degrees of grammaticalization of versprechen, scheinen and werden The investigation of the other three verbs in PDG results in an analogous diagnosis: the evidential constructions, i.e. werden & infinitive, scheinen & zu infinitive and versprechen & zu-infinitive show a distinctly higher degree of grammaticalization than other constructions with these verbs if tested by the grammaticalization parameters. In order to avoid redundancies, the following list mentions only peculiarities of these verbs, while all traits that are analogous to the ones illustrated for drohen above, are left out here. As Parameters 3 and 6 have already been treated, we concentrate on the remaining ones. As will be spelled in out in more detail in later chapters, versprechen shows a high degree of grammaticalization and has been acknowledged to function as an auxiliary.13 However, there are some aspects that prove that – as compared to drohen – it displays a lower degree of grammaticalization (for details see Section 6.4). For Parameter 1 the situation is completely analogous to drohen – for the semantic as well as the morphonological aspects (semantic attrition from speech act verb to evidential marker, phonological indifference). The same applies to Parameter 2. Like drohen the evidential variant has lost all its combinatorial variability concerning the complement. As the range of complements differs from that on lexical drohen, it is illustrated here for lexical vs. grammatical(ized) versprechen by (24) and (25), respectively: (24) a. Er verspricht (uns) (hoch und heilig), bei der Arbeit zu helfen. ‘He promises (us) (solemnly) to help us with the work tomorrow.’

Grammaticalization parameters and German evidentials 117

b. Er verspricht (uns) Hilfe/, dass er helfen wird/, er werde helfen. ‘He promises (us) assistance/that he will help us/that he helps us.’ c. Er verspricht (uns) das/es. ‘He promises this to us.’ d. Das verspricht er (uns). ‘This he promises us.’ (25) a. Das Wetter verspricht (*uns) (*hoch und heilig) besser zu werden. ‘The weather promises (*us) (*solemnly) to get better.’ b. Das Wetter verspricht *Verbesserung/, *dass es besser wird/, *es werde besser. ‘The weather promises *improvement/*that it will get better/*it will get better.’ c. *Das Wetter verspricht das/es. *‘The weather promises this/it.’ d. *Das verspricht es. *‘This it promises.’ While the lexical reading of versprechen has several options with regard to its complements – including an infinitival construction –, versprechen in its evidential reading has lost the capacity to take different complements. Parameter 4 shows the same as for drohen. As noted by Askedal (1997: 13), the extraposition of the infinitive complement of evidential versprechen is not possible, see (26), whereas the infinitive complement of lexical versprechen may be extrapositioned, as shown in (27). Due to this fact versprechen (as well as drohen) is considered to be similar to the typical auxiliary verbs in German. (26) (27)

* …weil dieser Streit versprach, große Erfindungen hervorzurufen. ‘because this quarrel promised to evoke great inventions.’ … weil er versprach, uns morgen zu helfen. ‘because he promised to help us tomorrow.’

Parameter 5 has in part been treated in Section 4.2.2 when discussing this parameter for drohen. The following examples give an overview of the reduced choices of inflectional forms in the more grammaticalized reading of versprechen:

118 Grammar and grammaticalization (28) a. *Das Wetter legt die Hoffnung nahe, besser zu werden zu versprechen. *‘The weather gives reasons to hope to promise to get better.’ b. *Das Wetter hat besser zu werden versprochen. *‘The weather has promised to get better.’ c. *Das Wetter würde besser zu werden versprechen. *‘The weather would promise to get better.’ d. *Oh Wetter, versprich besser zu werden! *‘Oh, weather, promise to get better!’ e. *Vom Wetter wurde versprochen, besser zu werden *‘It was promised by weather to get better.’ Evidential versprechen lacks a number of inflectional forms: it does not have infinitive forms (a); it does not occur in periphrastic tense and mood forms (b) - (c); it further lacks imperative forms (d) and passive forms (e). If versprechen occurs in such forms, it can be taken on its lexical reading. Hence evidential versprechen has to some extent lost its capacity to realize “normal” inflectional categories of main verbs and functions as a grammatical formative with its own inflectional paradigm. Thus we can postulate a relatively high degree of grammaticalization of evidential versprechen, as compared to its lexical counterpart. Scheinen in constructions with zu-infinitive today is mostly (though not exclusively) used in its evidential reading (for details see Section 6.3). This construction is acknowledged to display a high degree of grammaticalization. Many scholars have pointed out the intermediate status of scheinen between main verb and auxiliary (e.g. Pafel 1989, Ebert 1976, Zifonun et al. 1997, Askedal 1998, Diewald 2000, 2001). Askedal (1998: 70) treats scheinen explicitly as an auxiliary, and calls it a “Modalitätsauxiliar der epistemisch-futurischen Klasse 1 im deutschen Prädikatssyntagma” [‘modal auxiliary of epistemic-future class 1 in German predicate syntagm’]. As regards Parameter 1, evidential scheinen does not show any loss in phonological substance, but does show significant semantic attrition in comparison to its lexical use. As an evidential auxiliary it only indicates the fact that there is some accessible information (which may be of visual or other origin) which is connected (via inference) to the described event. This can be seen as a clear instance of desemanticization: neither the visual nature of information source nor the fact of perceiving the described event/object is a distinctive feature in the meaning of evidential scheinen. What has remained in the semantics of the grammaticalized scheinen is the

Grammaticalization parameters and German evidentials 119

relational meaning component, i.e. the relation between some information source and the described situation. As to Parameter 2, there are some important facts which point to the high degree of grammaticalization of scheinen within infinitival constructions. As Askedal (1998: 59f.) noted, the zu-infinitive complement of scheinen cannot be replaced by any other constituent and cannot be pronominalized by das ‘this’ or es ‘it’, cf.: (29) a. Sie scheint allein zu leben. ‘She seems to live alone.’ b. *Sie scheint das/es. *‘She seems that/it.’ c. *Das scheint sie. *‘That/it[ACC] seems she.’ d. *Das scheint. *‘That seems.’ e. *Diese Aussage scheint. *‘This statement seems.’ As regards Parameter 4, evidential scheinen only appears in coherent construction, i.e. displays increased boundedness in the sense explained above (cf. Diewald 2000: 341, from where the following examples are taken). (30) a. … dass sie alles selbst zu schreiben scheint. ‘… that she everything on her one to write seems.’ b. *… dass sie scheint, alles selbst zu schreiben. *‘… that she seems, everything on her own to write.’ (31) a. … dass sie alles selbst zu schreiben wünscht/hofft/verspricht. ‘… that she everything on her own to write wishes/hopes/promises.’ b. … dass sie wünscht/hofft/verspricht, alles selbst zu schreiben. ‘… that she wishes/hopes/promises to write everything on her own.’ The infinitive construction with scheinen thus belongs into the class of coherent infinitive constructions. The infinitive is part of the whole construction which forms a linguistic unit, and for this reason the infinitive complement cannot be extrapositioned as a separate constituent.

120 Grammar and grammaticalization The Parameter 5 is to be treated analogously to the other verbs discussed so far, i.e. as far as the choice of a certain grammatical marker within the German evidential paradigm is concerned, scheinen may be seen to be used increasingly obligatorily. Further proof of the increasingly obligatory status of evidential scheinen lies in the fact that it lacks a number of inflectional forms which its lexical counterpart has. For instance, evidential scheinen does not occur in periphrastic tense forms. Consider the following examples which due to the periphrastic tense form (perfect) are restricted to the lexical reading only: (32)

Und vielen hat es damals auch so geschienen. (DWDS, Gesprochene Sprache) ‘And at that time many people did have the same impression.’

Moreover, evidential scheinen does not occur itself as infinitive complement, while lexical scheinen does: (33)

Es mag dem Außenstehenden so scheinen, als wenn die Zahl dieser Betriebe nicht so groß wäre, .... (DWDS, Gesprochene Sprache) ‘The outsider may have the impression that the number of shops might not be as big, ...’

As werden & infinitive is an acknowledged auxiliary verb in German, the degree of grammaticalization of this construction is generally accepted to be high. Still, for the sake of comprehensiveness, the following brief remarks point to some details in order for werden & infinitive to be compared with the other evidential constructions. We do not deal with the question of which category werden & infinitive belongs to here, i.e. we do not answer the question whether it is a future marker, a modal marker or an evidential. We just look at its degree of grammaticalization with respect to the evidential paradigm we are assuming. Parameter 1 shows exactly the same results as for the other evidential constructions, i.e. there is no phonological reduction but heavy semantic reduction, the details of which will be treated in Section 6.1. Therefore a few brief remarks will suffice here. The lexical variant of werden exhibits strong aspectual – more precisely: ingressive – semantics. It denotes the initial phase, the onset of a state of affairs. Lexical werden designates that an event has already started and is developing further towards an intermediate or even final stage, cf.:

Grammaticalization parameters and German evidentials 121

(34)

Unsere Beziehungen werden zu einer ständigen Gegebenheit der internationalen politischen Landschaft. (DWDS, Kernkorpus) ‘Our relations become a permanent factor in the international political landscape.’

As for werden within infinitive constructions, its ingressive semantics has bleached to a large extent. Werden does not designate a change which happens to the subject of the sentence, nor does it denote an initial step of an event. What has remained in the semantics of werden in constructions with infinitive is a highly abstract relation between two points – information source and the described event. As compared to the lexical variant of werden which denotes an observable change towards a goal, the evidential werden expresses neither a changing process from an old towards a new state of affairs (which is/has been observed by the speaker) nor is this process bounded in its onset. Instead, being an inferential evidential, grammaticalized werden & infinitive introduces a situation as being inferred by the speaker from known information. As regards Parameter 2, the situation is similar to the other verbs. Lexical werden can take different constituents as complements, i.e. the valence relations of lexical werden allow for connections with e.g. nominal and pronominal objects with and without preposition, and with adjectives, cf.: (35) a. Sie wird traurig/so/wie ihr Vater. ‘She becomes sad/like that/like her father.’ b. Sie wird Lehrerin/eine gute Lehrerin/zur Lehrerin/zu etwas/das. ‘She becomes a teacher/a good teacher/like a teacher/something/ that.’ c. Das Eis wird Wasser/zu Wasser. ‘The ice turns into water.’ d. Das wird sie/er. ‘That becomes she/he.’ None of the sentences in (35) can be given a grammatical reading. Grammatical werden does not permit such a variety of complements: it combines only with infinitive complements and together with them forms one constituent. The following paraphrases show that the infinitive complement cannot be replaced by any other constituents and cannot usually be pronominalized by das ‘this’ or es ‘it’ either:

122 Grammar and grammaticalization (36) a. Sie wird das Buch lesen. ‘She will be reading/read the book.’ b. *Sie wird das/es. *‘She will this/it.’ c. *Das wird sie. *‘This will she.’ d. *Das wird. *‘This will.’ Parameter 4 for werden is analogous to the other evidentials verbs. First, lexical as well as grammaticalized werden occur as free morphemes, and second, grammaticalized uses of werden imply coherent constructions only, thus pointing to an increased amount of bondedness. As regards Parameter 5, the two variants of werden exhibit some crucial differences. Lexical werden can be used in all possible inflectional forms. The inflectional spectrum of evidential werden, on the contrary, is restricted to present tense and to the würde-form, which is morphologically the preterite subjunctive of werden.

3. Summary The application of the grammaticalization parameters has proved that all four evidential constructions, i.e. drohen/versprechen/scheinen & zuinfinitive and werden & infinitive, display a high degree of grammaticalization as compared to their other usages. This could be shown in particular by their respective reactions to grammaticalization parameters 1 (as far as semantic change is concerned), 2, 3, 4, and 5. Only Parameter 6 – expectedly – did not produce relevant results here, nor did Parameter 1 as far as phonetic reduction is concerned. Figure 2 gives an overview of these findings.

versprechen

scheinen

werden

high degree of semantic attrition in grammaticalized variant: loss of intentional loss of intentional loss of visual loss of semantics and of semantics and of speech semantics ingressive speech act act component; however, semantics component still particular semantic restrictions as opposed to drohen significant condensation (loss of structural scope) in grammaticalized variants of all four verbs, which have developed into auxiliaries combined with an infinitive (depending on the item concerned with or without zu) high degree of paradigmaticization between the four constructions, which form a paradigm of inferential evidential values high degree of bondedness (coalescence) in grammaticalized variant: only coherent infinitive construction (no morphological fusion, though) certain degree of reduction of paradigmatic variability and thus, a certain degree of obligatorification fixation is not distinctive between lexical and more grammaticalized variants of all four verbs for typological reasons (fixed position of verbal elements in German sentence structure)

not (yet) distinctive between lexical and more grammaticalized variants of all four verbs

drohen

Figure 2. Degrees of grammaticalization of the four evidential constructions (according to Lehmann 1985: 306)

(5) paradigmatic variability (–) obligatorification (6) syntagmatic variability (–) fixation

(2) structural scope = syntagmatic weight (–) condensation (3) paradigmaticity (+) paradigmaticization (4) bondedness (+) coalescence

Grammaticalization parameter and diachronic process involved (1a) integrity (–) phonetic attrition (1b) integrity (–) semantic attrition= desemanticization

Summary 123

Chapter 5 Scales and scenarios of grammaticalization

This chapter is devoted to the dynamic, procedural aspects of grammaticalization. It takes up questions concerning semantic and functional change (see Section 5.1), the problems of polysemy (see Section 5.2) and, in Section 5.3, focuses on contextual and constructional factors in grammaticalization processes, thereby, though only in passing, taking into account recent developments in construction grammar.

1. Regularities in semantic and functional changes Grammaticalization studies have elaborated several regularities concerning semantic and functional changes that are typically found in grammaticalization processes. The ones important to our topic are the following three developmental tendencies: 1. metaphorization and metonymization scales, 2. the tendencies of scheme retention elaborated by Sweetser (1988, 1990), and 3. the model of successive semantic-functional stages suggested by Traugott (1989) and Traugott and Dasher (2002). They are described in the order given above in the following three sections.

1.1. Metaphorization and metonymization scales One of the best investigated topics in semantic change accompanying grammaticalization is the relative chronology of metaphorical and metonymic reinterpretation, which is understood as a cognitive process based on the construal of similarity or contiguity respectively. There is no room here to discuss the commonalities and differences of metaphorization versus metonymization (cf. Diewald 1997). Instead, it is taken for granted that both processes work in an analogous way in grammaticalization processes and, therefore, for the purpose of this study, need not be explicitly distinguished.

Regularities in semantic and functional changes 125

As Heine, Claudi and Hünnemeyer (1991) have shown, a large percentage of the semantic changes accompanying grammaticalization processes can be described in terms of successive metaphorization procedures which, in accordance with general cognitive principles, unidirectionally evolve from local and/or concrete to temporal to more abstract meanings. The following is a simplified metaphorization scale encompassing the basic semantic-cognitive domains:1 local/concrete > temporal > abstract (e.g. causal, adversative, copulative)

Conceived in this way in broad and general terms for the respective cognitive domains, the metaphorization scale – with suitable adaptations and refinements – has been shown to underlie many different instances of grammaticalization cross-linguistically, e.g. the English going-to future, French sentence negation, case marking systems in several languages, epistemic meanings in modal verbs etc. (cf. e.g. Heine, Claudi and Hünnemeyer 1991, Stolz and Jeßing 1994, Heine and Kuteva 2002). A quite straightforward example of the operation of the metaphorization scale, which demonstrates one way of concretisation of the abstract scale, is the semantic development of the German conjunction aber. Its origin lies in the comparative form of an adverb meaning ‘farther way’, which, in addition to its use as a local preposition meaning ‘behind’ in Gothic (afar), also developed the temporal meaning ‘after that’, ‘later’ (DWB; Paul and Henne 1992). In OHG, the adverb afur/abur had the temporal meaning of ‘after that’, ‘later’, and later developed the more abstract, iterative meaning of ‘again’, ‘once more’. In the following example from the 9th century, the adverbial abur is ambiguous between the temporal and the iterative reading, and also supports a reading that merges the temporal and iterative meaning to ‘later again’: (1)

giuuelîh dê dar trinket fon uuazzare thesemo, thurstit inan abur. (Tatian 87,4) Lat. : omnis qui bibet ex aqua hac sitiet iterum. ‘Whoever drinks from this water will be thirsty later again.’

The iterative meanings of aber, which developed an attenuated copulative meaning, persisted up to the 18th century. Today, it is found in fossilized and idiomatic uses such as aber und abermals (‘again and again’) or abertausend (‘thousands and thousands of’).

126 Scales and scenarios of grammaticalization The next step, which already started in OHG, was the rise of an adversative meaning (‘however’, ‘yet’). The adverb abur in (2) exemplifies this: (2)

Ther geist giuueso funs ist, thaz fleisc ist abur ummahtîc. (Tatian 181,6) Lat.: Spiritus quidem promptus est, caro autem infirma. ‘The spirit indeed is ready, the flesh, however, is weak.’

This adversative meaning is the dominant meaning of aber in PDG. It is present in all its uses, but owing to functional differentiation (and contextual influences) it appears attenuated or modified towards only mildly adversative or even copulative meanings in many instances. Thus, even in this abbreviated diachronic sketch, it can be seen that the semantic development of aber is a specification of the metaphorization scale given above. Table 1. Metaphorization of German aber local/concrete ‘farther away’/ ‘behind’

>

temporal ‘later’/‘again’

> abstract ‘adversative’/ ‘juxtaposition’/ ‘copulative’

Analogous scales of semantic development may be set up for the four German evidential constructions treated here. The proof for these paths will be given in later chapters using diachronic corpus data (see Chapter 8). Here, it may suffice to illustrate that the semantic development of the verb drohen is in line with the scale of abstraction described above. In OHG and MHG drohen was a speech act verb, and this meaning variant of drohen has been preserved till now. It denotes the declaration of an intention (of someone) to do something unpleasant to someone else (mostly the addressee). In contrast to versprechen, drohen has speech act meaning but is not used performatively: (3)

Alse er daz gehôrte, daz si im drouten, dô sprach Pylatus. (Frau Ava, Leben Jesu, 10568, KALI-Corpus) ‘After he had heard that they threatened him, Pilatus said.’

In this meaning, drohen refers to a speech act being performed by the subject of the sentence, whereby the subject is always a human referent. This human referent (the agent of the threatening act) predicts that something unpleasant for the recipient will happen as a result of his intervention if the

Regularities in semantic and functional changes 127

recipient does not comply with certain conditions intended by the threatening person. The latter one is normally able to carry out the intended action that is unpleasant in some way. In the course of the grammaticalization process, the meaning of drohen became more abstract: it gradually changed from denoting a particular type of a speech act towards expressing a more abstract notion of a threat. This shift in meaning most probably happened as a result of metaphorical transfer, namely via the process of personification, i.e. the use of non-human subjects with verbs that usually allow human subjects only. Thus via the feature [+human] that is inherent in the verb, the personification of the non-human subject is provoked. Drohen began to be used with non-human subject referents, whereby the overall situation introduced with drohen was given the status of a threat, or more generally: the status of a dangerous situation. (4)

Wie köndte man der Uns drohenden dienstbarkeit nachdencklichere zeichen finden? (BA Leib 1) ‘How might it be possible to find more ominous signs of the subservience threatening us.’

In this example, the abstract noun dienstbarkeit ‘subservience’ constitutes the entity emitting a threat, which is perceived by the recipients of that threat (uns ‘us’) via ‘ominous signs’. This development towards a more abstract meaning due to stereotypical and frequent personification of abstract concepts started in the ENHG period and continued until the 18th century, when the non-agentive variant of drohen (implying a nonintentional subject) established itself as an independent lexical variant of the verb. Finally, in the 18th and 19th centuries, an even more abstract meaning of drohen – the inferential evidential meaning – developed, but this was restricted to the construction type drohen & zu-infinitive exclusively. The inferential evidential meaning can be described as the expression of a particular mental state of the speaker. The evidential variant of drohen means that (i) the speaker has access to some information, which (ii) s/he interprets as pointing towards the described event (which is negatively evaluated by the speaker). Thus the speaker’s mental operation of inference has become the primary meaning of the grammaticalized verb drohen. (5)

Nach einem so langen Glücke, nach einer so glänzenden Reihe von Siegen, nach so herrlichen Eroberungen, nach so viel unnütz vers-

128 Scales and scenarios of grammaticalization pritztem Blute sieht sich der österreichische Monarch zum zweitenmal an denselben Abgrund geführt, in den er beim Antritt seiner Regierung zu stürzen drohte. (Schiller, Geschichte des 30jährigen Krieges, DE-Gutenberg) ‘... for the second time, the Austrian monarch finds himself led to the very same abyss into which he threatened to fall at the accession of his rule.’ The semantic development of the verb drohen from denoting a speech act (performed by the subject of the sentence) to referring to a particular mental state/operation (of the speaker) constitutes a particular realization of the metaphorization scale, and may be summarized as follows (for further details, see Chapter 9): Table 2. Metaphorization of drohen concrete action/speech act > ‘declaration of an intention to do something undesirable’

psychophysical observation > ‘designation of an observable undesirable situation (caused by something)’

mental state ‘marking of (perceptual) inferential evidentiality’

In the following chapters it will be shown that the other evidential constructions display a similar semantic development. Therefore, we may sum up this section by stating that, in the development of the German evidential constructions, the general developmental tendency, referred to as metaphorization and metonymization scales above, has played a very prominent role.

1.2. Scheme retention and reinterpretation The grammaticalization parameters discussed in Section 4.2 aim at bringing out the differences between two comparanda, thus attesting different degrees of grammaticalization. However, as grammaticalization is a gradual phenomenon, there must be some aspects linking the different stages, i.e. some kind of common denominator that remains stable from the source entity through all stages up to the grammaticalized functions. This common denominator is seen here as an abstract semantic structure, a relational template which will be specified in this section.

Regularities in semantic and functional changes 129

It is assumed that this basic concept is responsible for the “conservative” aspects of the diachronic processes involved in grammaticalization. Without this common denominator it would be impossible to account for the existence of ambiguous instances and borderline cases, or for the possibility of gradual development through time. Sweetser (1988, 1990), who also works with metaphorization chains, claims that in grammaticalization processes, we can make out a basic semantic template, which Sweetser calls ‘image scheme’, that is retained in its fundamental relational structure through all developmental stages. In the course of semantic change, this relational template in its fundamental relational aspects is successively transferred to different semantic domains. The domains are characterized, on the one hand, by increasing abstraction, but, on the other hand, also provide new semantic features that had not been present before. As an example, Sweetser invokes the development of the be-going-to future in English, arguing that while lexically rich features are bleached (e.g. the notion of movement on foot which typically is part of the verb go), the relevant diagrammatic, i.e. relational feature, is kept in all uses, and is employed via the space-to-time-metaphor in the new domain of temporal meaning. This is made quite clear in the following quote referring to the be-going-to construction. In this mapping, we lose the sense of physical motion (together with its likely background inferences). We gain, however, a new meaning of future prediction or intention - together with its likely background inferences. We thus cannot be said to have merely “lost” meaning; we have, rather, exchanged the embedding of this image-schema in a concrete, spatial domain of meaning for its embedding in a more abstract and possibly more subjective domain. (Sweetser 1988: 392)

Sweetser’s model has been successfully deployed in the investigation of various grammaticalization processes in German (cf. Diewald 1999b for modal verbs; Diewald 2006 for modal particles; Smirnova 2006 for the verb würde). As will be shown in Chapter 8 in greater detail, German evidentials, too, develop along the reinterpretation and abstraction of a relational semantic template. In advanced stages of grammaticalization, this development results in the indexical-grammatical interpretation of that template. To give a rough idea of this process, the development of drohen is used here again for illustration.

130 Scales and scenarios of grammaticalization Though the details of the semantic development of drohen will be treated only later (see Section 8.3) we can note in passing that Sweetser’s postulate is relevant here as well. We observe the retention of a relational template consisting of a vectored relation between a source, a path, and a goal as inherent semantic features of the verb drohen (compare Section 1.2 for a discussion of the inherent relational structure of deictic signs). This inherent relational template may be sketched as follows: source

>

path

>

goal

This relational structure originally applied to the concrete domain of the speech act meaning of drohen, and in the course of the diachronic development was transferred to successively more abstract domains expressing temporal, aspectual, and logical relations. Thereby, i.e. in all later stages, the basic relational scheme was retained. The original lexical verb drohen indicates a communicative act performed by a human subject that functions as a source entity in the relational scheme above. Usually, the content of the threat, which takes over the position of the goal in the semantic relational template, is explicitly present in the sentence (by means of infinitive complements, direct or indirect speech clauses, prepositional complements with mit, rarely also via an accusative object), or it can be easily reconstructed from the context. This prototypical semantics of lexical drohen may be visualized in the following abstract relational template: Table 3. Relational template of lexical drohen source human subject Sie She

> >

path > drohen > drohte threatened

goal propositional content of the threat die Polizei zu rufen to call the police

The same semantic relational template with three basic positions underlies the non-agentive variant of the verb drohen that denotes an observable negatively evaluated situation. The original abstract scheme remains unchanged; the individual positions are transferred to another semantic domain and thus are taken over by other entities (see Table 4). The position of the starting point, or source, is represented by the subject of the sentence, which – as opposed to the original lexical meaning of drohen – is mostly an inanimate entity. The verbal relation, i.e., the position of a path, is no longer restricted to a verbal announcement, i.e., a

Regularities in semantic and functional changes 131

speech act, but is transferred to an (observable) threat emanating from the source entity. The goal position has remained roughly the same and marks the propositional content of the threating situation. Table 4. Relational template of non-agentive lexical drohen source some entity Die Mauern The walls

> path > > drohen > drohten threatened

goal propositional content of the threatening situation den Einsturz collapsing

As regards the evidential construction drohen & zu-infinitive, the same abstract relational scheme may be used as a base. The abstract positions in the scheme are once again transferred to another domain and are now realized as follows. The source position, i.e. the entity to which the construction drohen & zu-infinitive links the described event, is now reinterpreted as the speaker’s information source. The relational component (‘path’), which holds between the source and the goal position, can now be described as the designation of the speaker’s inference process. The goal position is taken by the proposition, which is modified by the evidential value expressed by drohen. Hence, two significant changes may be stated that occur in the semantic relational template of drohen in the course of its grammaticalization towards an evidential auxiliary. First, the position of the source has been externalized, i.e., it has been transferred from the textual plane to the layer of the speech event and now comprises the information pieces available to the speaker. Second, the position of the goal has been widened and consists now of the whole proposition. Drohen thus modifies the whole proposition by anchoring it in the actual deictic field (i.e., the origo cf. Sections 1.1.2 and 2.2). This reinterpreted relational structure of the evidential construction drohen & zu-infinitive may be represented as follows: Table 5. Relational template of evidential drohen source > (speaker’s) information >

path [drohen

> >

goal proposition]

Hence, for the inferential evidential construction drohen & zu-infinitive, we can state that the abstract relational scheme of the lexical verb drohen has been preserved, which will also be shown for the verbs werden, scheinen and versprechen in the next sections.

132 Scales and scenarios of grammaticalization The concrete linguistic realization and semantic content of the positions have changed, however: they have been transferred to other domains. Table 6 summarizes the three developmental stages sketched above and specifies the relevant domains. Table 6. The abstract relational template of drohen in different domains Domain 1 concrete level: performative situation/speech act source (human subject of sentence) > path > goal Domain 2 abstraction from human sphere to non-human, mostly abstract entities: personification source (non-human abstract entity) > path > goal Domain 3 deictic anchoring in speaker: evidential meaning source (speaker’s information) > path > goal

1.3. Successive semantic-functional stages The concept of cognitively motivated chronological ordering of metaphorization and metonymization processes and the principle of scheme retention are supplemented here by the model of successive semantic-functional stages suggested by Traugott (1989) and developed further by Traugott and Dasher (2002). Traugott’s well-known findings on general tendencies of semantic change (accompanying grammaticalization) have been demonstrated to be relevant for many categories in a number of languages. In the original version, Traugott (1989) postulates three diachronic tendencies, which she describes as follows: Tendency I: Meanings based in the external described situation > meanings based in the internal (evaluative/perceptual/cognitive) described situation. Tendency II: Meanings based in the external or internal described situation > meanings based in the textual and metalinguistic situation. Tendency III: Meanings tend to become increasingly based in the speaker’s subjective belief state/attitude toward the proposition. (Traugott 1989: 34-35)

Regularities in semantic and functional changes 133

In subsequent studies (Traugott 1995/1997, 1999, Traugott and Dasher 2002) this model was expanded, refined and generalized in order to account for as many instances of semantic change as possible. The modifications concern, first, the notion of subjectification, which has been complemented by the notion of “intersubjectification” in order to account for social deixis (Traugott and Dasher 2002: 23f.). Second, possibly concomitant structural changes have been integrated, especially the change of scope and topological features in the development of discourse markers (Traugott 1999: 177ff., Traugott 1995/1997). Both the chronological ordering of the three stages and the description of the respective semantic and cognitive domains have been retained in its essence. Therefore, we refer to the original version of the three tendencies of semantic change accompanying grammaticalization, leaving aside the notion of intersubjectification. In addition to the phenomena emphasized by Traugott, we argue that beyond the change in the semantic domains involved these tendencies indicate a functional change concerning the semiotic status of the item in question (including word class variation). In our view, it is this change of semiotic status which is the essence of grammaticalization. The development runs from mainly referential to mainly textual and connective functions, and further evolves to indexical grammatical functions.2 Therefore, we use Traugott’s model with some shifts of focus and adaptations (e.g. merging the external and internal situation into one function, i.e. into the ‘referential function’). The development, as we understand it, may be sketched as follows: (i) (self-contained) referential function > (ii) text-integrative/connective function > (iii) indexical-grammatical function

That is, we postulate that the last stage in Traugott’s model coincides with what we described as one of the defining features of grammatical signs, i.e., their indexical potential (see Sections 1.1.2 and 2.2). In Section 2.2 we have already indicated to what extent the evidentials are indexical signs. In what follows we give an example of the developmental stages elaborated here using the verb drohen as an exemplary case again. The original lexical verb drohen serves to announce an event which is intended by the (human) subject. That means that drohen introduces a connection between two events, i.e., between the announcement itself (i.e., the act of speaking) and the announced action (i.e., the content of the threat):

134 Scales and scenarios of grammaticalization (6)

… gab die feurige Eleonora endlich selbst Gelegenheit, daß ich meine hefftigen Seuffzer und Klagen kniend vor derselben ausstieß, und mich selbst zu erstechen drohete, woferne… (Schna4) ‚… finally, passionate Eleonora herself gave me a chance to kneel down in front of her and heave my heavy sighs and complaints, and I threatened to stab myself, if …’

At the second stage (from the ENHG period onward), especially in specific contextual configurations, i.e. in combination with accusative objects or with infinitive complements, the verb comes to mark a successive relationship between two situations. The speech act semantics of drohen is already largely bleached and only the relational connection between textually represented situations is promoted, whereby the event introduced by drohen is conceptualized as following others. In these particular contexts, drohen receives a kind of temporal-aspectual interpretation without being a grammaticalized marker. (7)

Spuren der Plünderung ließen sich bemerken in Gefolg innerer Feindschaft. hohe Mauern drohten den Einsturz, Türme standen unsicher. (IDS-Corpora, GOE/AGA.01784) ‘Traces of the looting could be noticed togehter with internal animosity: high walls threatened the collapse/to collapse, towers stood insecure.’

This passage puts three events/states into chronological perspective, which are however expressed in a condensed way (partly using nominalizations). The first event – the looting – leads to the state where traces of the former event may be relized (Spuren ... ließen sich bemerken). One of these traces is the reference to an imminent third event/state, namely the collapse of the walls, which is expressed in a pseudo-transitive construction using drohen with the inanimate subject Mauern and the accusative object Einsturz, which is a nominalization refering to the imminent event. Finally, the textually represented connection between the relevant situations changes into a meaning that is close to the final inferential evidential meaning, indicating that there is a logical connection between events one of which is conceived of as evidence for another. This ongoing development can be witnessed in PDG:

Regularities in semantic and functional changes 135

(8)

Wenn man also merkt, daß eine Diskussion (etwa über Politik) in einen Streit auszuarten droht… (DWDS, Kernkorpus) ‘This means that as soon as you realize that a discussion (say about politics) threatens to turn into a quarrel …’

Here, there in no longer an explicit reference to an initiating event, instead the passage refers to a cognitive act of the subject, i.e an act of gathering evidence (Wenn man also merkt) for a situation which is about to evolve in a foreseeable direction (in einen Streit auszuarten droht). The evidential construction drohen & zu-infinitive serves indexicalgrammatical functions. Thus we can state that, during its grammaticalization process, drohen has followed the developmental path formulated above. This development concerns different aspects of the linguistic status of drohen, as shown below: Table 7. Changes at different levels (as regards drohen) functional level: referential function > text-connective function > indexical-grammatical function word class: (agentive) lexical verb > (non-agentive) lexical verb > auxiliary for the verbal category of evidentiality semantics: concrete action of subject > observation of imminent change > mental operation expressing information source (inferential evidential)

In summary, we have argued that the diachronic development of German evidentials, exemplified here by the development of the verb drohen, can be ascribed to the interplay of general grammaticalization scales concerning the semantic development, the functional development, and the inherent semantic core structure of the item in question. The core structure remains stable in its essence, but is reinterpreted for its broad semantic or functional domains, which become more and more abstract, less referential, and more grammatical. This semantic change, or more precisely the shift of a core structure to various domains, is accompanied by a functional change corresponding to the general tendencies set up by Traugott.

136 Scales and scenarios of grammaticalization 2. Dealing with ambiguity and context dependence Grammaticalization typically is associated with semantic change, and semantic change by definition raises the question of gradualness and ambiguity. In order to deal with this pervasive semantic indeterminacy, we need to define some central concepts which we will use in order to describe meaning components and semantic change. Ambiguity. The term ambiguity is used here in its general meaning referring to any kind of unclear, indeterminate linguistic item, and not as a term for a special sub-type of functional and/or semantic underdetermination. That is to say that the term ambiguity covers structural ambiguity as well as semantic ambiguity. Furthermore, it should be noted that semantic ambiguity as we use the term here is synonymous to “(semantic) vagueness”, which, too, is a common way of referring to this phenomenon in literature. Context dependence. Analogous to “ambiguity”, we use the term “context dependence” as a general notion which is relevant for the resolution of any kind of semantic indeterminacy, i.e. the linguistic and non-linguistic context is the decisive factor in the resolution of all of the variants of ambiguity described below, i.e. of heterosemy, polysemy, gradience as well as conversational implicature. Only the degrees of the impact of contextual clues as well as the types of contexts differ and thus provide criteria for distinguishing between the respective types of ambiguity. Heterosemy. Heterosemy as opposed to polysemy refers to the existence of distinct meanings in a linguistic item, which are associated with different word classes (Diewald 2006, Autenrieth 2002). Heterosemy typically is the outcome of an advanced stage in a grammaticalization process, when a lexical (i.e. functional) split has taken place and the grammaticalizing item (one of the heterosemes) has been at least partially integrated into its new grammatical paradigm (i.e. it is associated with Stage 4 of the grammaticalization scenario, cf. Section 5.3). Most important for the reconstruction of the diachronic development is the fact that the functional spectra of the coexisting heterosemes, though seemingly unrelated in a purely synchronic perspective, retain the successive gradient steps of regular grammaticalization. This situation can be easily illustrated by the class of particles in German, which, as is well-known, often have functionally defined ‘heterosemes’ in other word classes. For example, aber ‘but’ can be used as a discourse particle as in (9), as a modal particle as in (10), and as a conjunction as in (11):

Dealing with ambiguity and context dependence 137

(9)

K: und das wird dann da so seitlich draufgeschraubt oder? K :‘and that’s going to be screwed there to the side this way, isn’t it? I: ja genau, aber mach das erstmal so. I : ‘yes exactly, ABER do it this way first.’ (Sagerer et al. 1994, cf. Diewald and Fischer 1998)

Aber in (9) operates on the thematic plane of discourse. After the first interlocutor K has asked for the next step in constructing a toy airplane, the second interlocutor I uses aber to relate his utterance to the preceding utterance of the partner and simultaneously to indicate that he or she wants to change the topic (Diewald and Fischer 1998: 87). Example (10) shows the classic use of aber as a modal particle, where the utterance takes up a given proposition and modifies it in an adversative manner. (10)

Das ist aber keine gute Idee. ‘That is ABER not a good idea.’

In (11) finally the conjunctional use of aber has the function of connecting two linguistically expressed propositions and of integrating them into one complex sentence. (11)

Sie wollte telefonieren, aber sie hatte kein Netz. ‘She wanted to make a call, but she didn’t get a signal.’

These uses of the linguistic item aber belong to different word classes in PDG and therefore are classified as heterosemes (cf. Diewald 2006). While the status of particles like aber above as heterosemes is fairly uncontroversial, matters become more complicated when we look at verbs and their auxiliarized cognates, like e.g. haben ‘have’ in German; which has a range of usages in PDG, among them the full verb use (12) and the use as a perfect auxiliary (13): (12) (13)

Ich habe mindestens zehn Stadtführer von Berlin. ‘I have at least ten tourist guides to Berlin.’ Ich habe die Reise nach Berlin gebucht. ‘I have booked the trip to Berlin.’

138 Scales and scenarios of grammaticalization Though it is quite unproblematic to talk about heterosemy for these two types of usage of haben, there are others, where it becomes difficult to establish a clear-cut change of word classes, so that the distinction between heterosemy and polysemy blurs. These are cases like the so-called modal periphrastic use of haben & zu-infinitive like in: (14)

Ich habe zehn Pferde zu versorgen. ‘I have ten horses to attend to.’

This usage of haben & zu-infinitive is clearly different in meaning from the full verb usage of haben; still, most linguists would not want to state a difference in word class here, and rather classify this case as polysemy of the lexical entity haben. Polysemy. As has already become clear, we speak of polysemy (or polysemous signs), if the following conditions are met (cf. also Fischer 2006): 1. there are distinct meanings in one single, phonological/orthographic form, 2. these meanings are closely related, and 3. they do not correlate with a change in word class. Coates (1983: 10) lists three criteria, which can be used in practice as a polysemy test for a particular linguistic item. According to her, semantic ambiguity in an item – which per definition must have at least two readings – is to be classified as a case of polysemy, if:3 1. the two readings appear in different linguistic contexts; 2. the two readings need different semantic paraphrases; 3. there are ambiguous cases, i.e. sentences which allow one or the other reading. The employment of this test can be illustrated with the German modals. They display polysemy in the sense explained in examples like the following: (15)

Sie muss in die Schule gehen. ‘She must go to school.’

Dealing with ambiguity and context dependence 139

The modal muss in this sentence can receive either a deontic interpretation, as in (15a), or an epistemic reading as in (15b): (15) a. Sie ist genötigt in die Schule zu gehen. ‘She is forced/obliged to go to school.’ b. Sie geht bestimmt/sicher/wahrscheinlich in die Schule. ‘She probably/certainly goes to school.’ Both meanings, the deontic meaning of obligation and the epistemic meaning of a high degree of probability of the proposition to be factual, clearly fulfill condition two above: they call for different paraphrases, i.e. they are clearly distinct and discernible meanings. The recipient has to make a decision in interpreting this sentence: s/he chooses one of the two possible meanings, and even if s/he cannot make up his or her mind, s/he can define these two meanings as alternatively possible (Criterion 3). That is, this is a clear case of polysemy, whereby, usually, the further context helps to make a definite choice between the two readings (cf. Coates 1983: 15f.). It is important to note that ambiguity of this type, which leads to two distinct options in an either-or-decision in one context – is an excellent indicator that both meanings are indeed independent meanings, i.e. that the item in question is truly polysemous. Therefore, this is an important feature in the polysemy test. For investigations into degrees of grammaticalization, which is always associated with meaning change and thus ambiguity, Criterion 1 receives additional importance. As will be explained in Section 5.3 the existence of mutually exclusive linguistic contexts for one or the other reading (isolating contexts) is a decisive step towards isolating the new, more grammatical meaning of the relevant item as a distinct, independent reading. Again, another well-known example for context-dependent distinct readings in modal verbs may help to make this point. While (16) illustrates the isolating context for the lexical (dispositonal) reading of müssen, (17) illustrates the isolating context for the epistemic or deictic reading of müssen, i.e. as a grammaticalized marker of a factuality judgement. (16) (17)

Sie muss lachen, wenn er seinen Fahrradhelm auf hat. ‘She can’t help laughing, when he wears his cycling helmet.’ Sie muss als letzte im Raum gewesen sein. ‘She must have been the last person in the room.’

140 Scales and scenarios of grammaticalization The exact morpho-syntactic and semantic features of these two isolating contexts will be explained later. Here, it suffices to note that there are linguistic contexts which favor one of the alternative readings respectively to the exclusion of the other one, which means that Coates’ Criterion 1 is fulfilled and we have an instance of polysemy here. Gradience. Beside polysemy, there is a further case of semantic indeterminateness, which, following Coates (1983: 13), we call “gradience”. Here, several meanings in one item shade into each other, i.e. they may be activated at the same time, without excluding each other. As Coates (1983:12) points out, this phenomenon can be described by using the concept of gradient membership of prototype theory. We speak of gradient membership when one element at the same time belongs to more than one category (to diverging degrees). Coates illustrated this with some of the meanings of English can, as in the following sentences (1983: 14f.): (18) (19)

(20)

I can only type very slowly as I am quite a beginner. (‘ability’) These young assistants [...] can give the pupils valuable practice in understanding and speaking the foreign language. (gradience between ‘ability’ and ‘possibility’) You can’t see him because he’s having lunch with a publisher. (‘possibility’)

The meaning ‘ability’ in (18) is characterized by the fact that the subject’s potential to act is determined by inherent properties, and that the main verb of the sentence is an action verb. The ‘possibility’ reading in (20) is highlighted when external enabling circumstances are present. The many intermediate cases between ‘ability’ and ‘possibility’, which are illustrated by sentence (19), are characterized by the fact that it is difficult or impossible to decide whether or not the enabling factor is inherent or whether there is a combination of internal and external enabling factors. That is, the features may combine without contradiction. The recipient need not or cannot decide between alternatives which means that if there are only gradient meanings in one item the notion of polysemy may not even come up in that case, as speakers are only aware of mutually compatible nuances. In contexts of semantic change gradience may be the trigger for a development towards polysemy in an item. Diewald (1999: 71) shows that an analogous scale of gradience is found in the German modal kann. As we will see in the empirical chapters of this book, this type of semantic vagueness does play an important role in the

Dealing with ambiguity and context dependence 141

development of evidential meanings as well. It is in particular in the beginning of a process of semantic change that gradience – i.e. low-level difference in meaning – plays an important role as a kind of trigger for further, more salient meaning changes. In any case, we subscribe to Coates’ summary of her study on the meaning of English modals that “a correct description of modal meaning must reconcile categorial and non-categorial approaches” (1983: 10) as an insight which may be generalized to other instances of semantic change accompanying grammaticalization. Conversational implicature. With reference to Grice (1975/1989), who first introduced the concept of conversational implicature into linguistics and also showed its impact on semantic change, we define this process as a pragmatic procedure undertaken by the partners of the communicative exchange attributing meaning to a linguistic construction which does not have this meaning potential inherent in itself. This procedure may be employed for example, if the “literal” meaning of an utterance is irrelevant, uninformative or for any other reason calls for additional processing. Conversational implicatures are not (yet) a particular type of linguistic meaning (meaning of a linguistic item), but meaning achieved by pragmatic enrichment (cf. Traugott and König 1991). Conversational implicature may be applied to resolve any kind of semantic ambiguity. Again, the German modals provide fine examples here. In particular sentences with a generic or indefinite animate subject easily lead to semantic ambiguity correlated with scope ambiguity (see also Gamon 1993). (21) gives an example: (21)

Das muss man alles erst mal wissen. that-ACC must one-NOM all-ACC first know narrow scope, lexical: ‘You have got to realize all that first.’ wide scope, lexical: ‘It is necessary: you realize all that first.’

For a sentence like (21) there are contexts that allow a narrow-scope reading as well as a wide-scope lexical reading. While the narrow-scope reading encodes an obligation for the subject man to acquire a certain state of knowledge, the wide-scope reading expresses the general necessity that the proposition ‘Man weiss das alles erst mal’ is true. For the present study it must be borne in mind that while the tests for polysemy usually lead to a distinct periphrasis for each meaning, there are cases where the distinction between true polysemy and conversational implicature blurs. This problem is solved here not by deciding once and for all which usage falls under the heading of polysemy and which under the head-

142 Scales and scenarios of grammaticalization ing of contextually triggered enrichment, but by providing a unified descriptive template that incorporates all information relevant for a particular interpretation. This seems to be a reasonable solution in the light of the following facts, some of which have been mentioned before: − the answer to the question of whether one usage is merely pragmatically triggered or should instead be regarded as an independent and distinct meaning inherent in the semantic make-up of the lexeme depends, among other factors, on frequency, stereotypicality, and register, which have to be explored for every item via more comprehensive empirical data; − because diachronic change is ubiquitous and pervasive, and because it continually leads to a semanticization of stereotypical context-induced interpretations, there may not be a clear cut-off point between contextual variation and polysemy. As has been elaborated in Section 5.1.2, semantic change of grammaticalizing items like the verbs treated in this book, consists in the transfer of a lexeme-inherent semantic template to increasingly abstract semantic and functional domains. Thus, diachronic change as well as synchronic ambiguity of an item is due to the reinterpretation of the basic semantic template. Depending on the factors mentioned above as well as on frequency, stereotypicality, distance of semantic domains, and functional specifications, this reinterpretation can result either in a distinction of word classes (i.e. different ‘heterosemes’) or in polysemy within a particular word class or in contextually triggered ‘pragmatic’ interpretations. As diachronic change continually leads to a semanticization of stereotypical context-induced interpretations (Traugott and König 1991), the decision whether to treat a particular reading as part of the semantic content, i.e. as one meaning of a polysemous item or as a pragmatic function legitimated by its lexical meaning without being part of it, can only be made individually for each item and each diachronic period, because it is certain to change. As far as the evidentials and their source constructions are concerned, we hypothesize that the distinction between grammaticalized evidential meaning and non-grammaticalized evidential meaning is one of polysemy at least, which is supported by the distinction of relevant linguistic context types (see next section), while meaning distinctions within the range of non-grammaticalized usages usually are (but need not be) examples of gradience, and under certain conditions may be resolved via conversational implicature. This hypothesis can be based on independent factors (like the

Stages, contexts and constructions 143

grammaticalization parameters) and will be substantiated in the chapters to follow. An overview is sketched in Table 8, which, as already mentioned several times, will gain full plausibility after the concept of different context types has been explained in the next section. It should be noted that the chronology suggested in this table does not capture the full development of a lexical item in all its variants, but only that one strain of development that – ex post, i.e. if there has been grammaticalization into a particular category – can be retraced as the relevant stages in its diachronic rise. In any case there may be parallel semantic changes concerning the lexical source item in any of its variants. Table 8. Meaning types and grammaticalization stages time arrow

t1

t2

t3

t4

stage of grammaticalization type of meaning

I

II

III

IV

lexical meaning of source item, possibly various types of gradient meanings

lexical meaning & conversational implicatures in stereotypical pragmatic contexts

polysemy constituted in oppositive contexts; later grammaticalized meaning appears as one of the polysemous readings

heterosemy supported by paradigmatic integration of grammaticalizing item; lexical split or survival of the grammaticalized meaning/form only

3. Stages, contexts and constructions From the very beginning grammaticalization studies pointed to the fact that it is not single linguistic items but linguistic items in particular contexts that are the locus for grammaticalization. A new grammatical function does not arise homogeneously in all uses of the linguistic item concerned, but is bound in its origin to specific linguistic “contexts” or “constructions” (Bybee, Perkins and Pagliuca (1994: 11), Bisang (1998: 20), Himmelmann (2004: 31), Lehmann (1992: 406), 1995 [1982], Traugott 2003, Traugott 2008, Diewald 2006).4 In the last years this has lead to an integration of concepts used in construction grammar models into grammaticalization studies,5 and to the de-

144 Scales and scenarios of grammaticalization velopment of detailed models for describing the various context types in grammaticalization (e.g. Heine 2002). Without discussing these models and suggestions here, we focus on a description of the model used in our investigation, a model which has proved to be useful in a number of grammaticalization studies and has been revised and refined continually (Diewald 2002, 2006, 2008). It integrates semantic, morphological and structural aspects into the definition of context types, and emphasizes the role of paradigmatic relations among constructions at a certain historical stage in a particular language as well as the influence of paradigmatic oppositions in the target category of a grammaticalization process. While earlier versions assumed three chronologically ordered stages in the diachronic rise of grammatical functions, the latest version contains four stages each of which is associated with a particular type of context. Table 9 gives an overview. Table 9. Context types in grammaticalization: revised version. Context

Meaning/Function

Construction types

Stage I preconditions of grammaticalization Stage II triggering of grammaticalization

untypical contexts

conversational implicature

no particular construction type

critical context

multiple opacity

“extragrammatical idioms”

Stage III reorganization & differentiation

isolating contexts

polysemous / heterosemous items

“formal or lexically open idioms”

Stage IV paradigmatic integration

paradigmatic paradigmatic opposicontext tions (choices) with reduced relational meanings, i.e. grammatical meanings

paradigmatic choices within an abstract constructional schema

The first column renders the label of the four context types, the second column gives a hint on the most important defining linguistic processes involved in the ongoing change, and the last column indicates possible interrelations with constructional notions, whereby the concepts of “extragrammatical idioms” and “formal or lexically open idioms” as defined by Fillmore, Kay and O’Connor (1988) have proved useful (for relevant con-

Stages, contexts and constructions 145

structional concept see as well Fillmore and Kay 1995, Goldberg 1995, Kay and Fillmore 1999, Michaelis 2004). In the following sections each stage will be explained and illustrated by examples from the German modals and by the development of the evidential construction with drohen. Stage I. In the first stage, the preconditions of the grammaticalization process develop. It shows an unspecific expansion of the distribution of the lexical unit to contexts in which it had not been used before. These contexts are called untypical contexts. In them, the new meaning, which may be grammaticalized in the further development, arises as a conversational implicature, i.e. this meaning is contextually and pragmatically triggered and not explicitly encoded in the linguistic items themselves. Untypical contexts may persist after grammaticalization has taken place. Very often, untypical contexts are those that do not display definite clues/indications as to a possibly preferred reading, or that display conflicting clues. Examples of untypical contexts from the modal verbs of German are given in (22) and (23): (22)

(23)

Man mac si morgen mehelen einem andern man. (NL 1928,1) ‘One can her tomorrow marry to another man’ narrow scope, lexical: One is able/has the opportunity to marry her to another man tomorrow. wide scope, lexical: It is possible/There exists the opportunity: one marries her to another man tomorrow. wide scope, deictic: Perhaps one will marry her to another man tomorrow. (Sênu thô uúîb thaz thâr bluotes fluz tholêta zuelif íâr inti uuas managu tholênti fon uuola managên lâhhin inti gispentôta allu irâ,) noh fon irô niheinîgemo mohta uuesan giheilit. (T 60,3) ‘and could not be cured by any of them’ wide scope, lexical: There was no opportunity for any of them: he cures her wide scope, deictic: *Maybe, nobody cured her.

Example (22) illustrates a context that results in gradience in the sense of Coates 1983 (see Section 5.2). While the narrow scope reading is the older, more lexical one, various wide scope readings typically arise via conversational implicatures in the sense of Grice [1975] 1989. That is, it is the result of a reasoning procedure of the hearer in a specific communicative situation. The context exemplified in (23) triggers structural reanalysis

146 Scales and scenarios of grammaticalization without semantic change (passive infinitive of the main verb). In this type of context, the modal obligatorily (due to structural factors) has a wide scope and at the same time retains its old lexical meaning. That is, the deictic meaning is virtually excluded in spite of wide scope (cf. Nordlinger and Traugott 1997, Gamon 1993, Diewald 1999, 2006). Throughout the history of the German modals there are cases of this type of semantic ambiguity and variable scope as well as examples with obligatorily wide scope and the old lexical meaning.6 This is the first stage of grammaticalization. It is characterized by two principally independent types of contexts that show clusters of contextual features that were previously untypical. One context implies semantic generalization (generic, indefinite or expletive subjects). The other context triggers structural reanalysis without semantic change (passive infinitive of the main verb). An analogous stage, i.e. untypical contexts from the perspective of the contemporary language use, can be found – at least this is our claim – in all grammaticalization processes (see Diewald and Smirnova to appear). The following examples show the untypical context of drohen. (24)

Spuren der Plünderung ließen sich bemerken in Gefolg innerer Feindschaft. hohe Mauern drohten den Einsturz, Türme standen unsicher. (IDS-Corpora, GOE/AGA.01784) ‘Traces of the looting could be noticed togehter with internal animosity: high walls threatened the collapse/to collapse, towers stood insecure.’

In (24) we have semantic extension leading away from the original meaning. The original lexical verb drohen, being a speech act verb, serves to encode the announcement of an event which is intended by the (human) subject. That means that drohen introduces a connection between two events, i.e., between the announcement itself (i.e., the act of speaking) and the announced action (i.e., the content of the threat). And it always needs an animate subject. In this construction the verb comes to mark a successive relationship between two situations which are expressed in the sentence by nominalizations (Plünderungen, Einsturz). The existence of usages like this is a preparatory condition for the grammaticalization process to start. This state, i.e. the existence of untypical contexts without further grammaticalization, could continue forever without any necessity for further development whatsoever. This is to say that Stage I does not initiate a

Stages, contexts and constructions 147

grammaticalization process; it merely marks a necessary though not sufficient precondition of the grammaticalization process. As untypical contexts make use of existing constructions and can be easily interpreted due to their compositional structure, their role in the grammaticalization processes is less relevant than those of critical and isolating contexts. Stage 2. The second stage describes the actual triggering of the grammaticalization process. It is associated with a highly marked construction, called the critical context here. This is characterized by multiple structural and semantic opacity, thus inviting several alternative interpretations, among them the new grammatical meaning. The critical context functions as a kind of catalyst; it is found only during Stage II and disappears in the later development. Generally, in contrast to the other context types, the existence of the critical context is restricted to a fairly narrow time span in history and does not exist before or after that critical period (see later on for modification of this over-generalizing statement). For the grammaticalization of the modals it is found in MHG, around 1200. It is the construction shown below: modal with dental suffix -t- & (nominal object) & haben/hân/sîn & past participle

The critical context consists of the modal with the dental suffix -t- plus an optional nominal object plus haben/hân/sîn plus a past participle of a lexical verb. Although this construction is not attested before the middle of the 12th century, it is found with all six modals as soon as around 1200 with a relatively high frequency (cf. Westvik 1994, Deeg 1948), which means it spread rapidly. Most important for our topic is the fact that this construction is only attested for the modals. There are no other verbs found in this construction at that time (Paul, Wiehl and Grosse 1989: 295f.). (25) gives an example with the modal können, which will be used here for illustration:7 (25)

von Veldeke der wîse man! der kunde se baz gelobet hân. (Parz 8, 404,29f.) he could her better praised have ‘He can/could have praised her better.’

The interpretation of this construction is difficult as it is highly ambiguous (which is not rendered fully in the semantic paraphrase in (25)). Thus, disregarding the further linguistic context, the second line

148 Scales and scenarios of grammaticalization of (25) – i.e. der kunde se baz gelobet hân – can be translated into PDG in at least the following three ways (compare Deeg 1948 and Westvik 1994): (25) a. Der hätte sie besser loben können (pluperfect subjunctive) ‚He could have praised her better.’ b. Der konnte sie besser als Gelobte haben (past participle as a predicative adjective) ‚He was better able to have her as a praised one.’ c. Der könnte sie besser gelobt haben/Der hat sie vielleicht besser gelobt (deictic reading) ‚Perhaps he has praised her better.‘ The opacity of this MHG construction is not only a semantic one. The morphological forms that build it are highly opaque themselves. First, the non-finite structure hân & past participle is ambiguous between a reading in the sense of the perfect infinitive of PDG, in which the participle is the main verb and hân or sîn is the auxiliary, and a reading as a complex predicative structure where ‘have’ is the main verb and the past participle functions a predicative adjective, like in (25b). Second, the morphological form of the modal itself is ambiguous, too. For MHG modals with a dental suffix it is not possible to distinguish between the indicative and the subjunctive of the past (Westvik 1994), because the formal marking of the subjunctive by Umlaut, which today provides a clear and systematic opposition between mochte vs. möchte, konnte vs. könnte, had not yet developed at that time and there was a high amount of formal syncretism (see Birkmann 1987: 194). Therefore it is reasonable to assume that the MHG modals with a dental suffix express a distal value which – depending on context – can be interpreted as modal and/or as temporal distance (cf. Thieroff 1992). Thus, the critical context is characterized by the coincidence – or rather the clash – of two verbal forms that were both morphologically and morpho-syntactically ambiguous. Therefore, these forms could not mutually disambiguate each other. Being confronted with this construction, the recipient had several possibilities for interpreting it without getting a clear indication from the construction itself. It is proposed here that it is this structure, the critical context, in which the deictic reading and its necessary wide scope was not only one alternative interpretation among equally plausible others, but the most likely, the

Stages, contexts and constructions 149

favored, reading. Table 10 very briefly summarizes the different possibilities of resolving the opacity of the critical context: Table 10. Options for interpreting the critical context of the MHG modals 1. Lexical, non-deictic reading: 1.1. modal as indicative preterite (i.e. temporal distance only): predicates a past modal state on the subject; participle II is interpreted as modifying the direct object. Result as in (25b): Der konnte sie besser als Gelobte haben. ‚He was better able to have her as a praised one.’ 1.2. modal as subjunctive preterite (i.e. temporal and modal distance): predicates a modal state, which existed as a possibility in the past, on the subject; particle II is interpreted as modifying the direct object; Result as in (25a), corresponds to PDG pluperfect subjunctive: Der hätte sie besser loben können. ‘He could have praised her better.’ 2. Deictic reading: modal has wide scope; predicates a factuality value on the whole proposition; infinitive & PII are interpreted as infinitive II. Result like in (25c), corresponds to PGD deictic reading: Der könnte sie besser gelobt haben. ‘Perhaps he has praised her better.’

The critical context for drohen appeared in the 18th century, and has the structural and semantic features of inanimate subject referents and infinitive complement denoting an action: inanimate subject & inflected drohen & infinitive complement denoting an action

(26)

schon loderte die Flamme, und drohte das herrliche Werk in die Asche zu legen. (Klinger, DE-Gutenberg) ‘the flame blazed and threatened to burn up the lovely work’

This instance of the critical context of drohen has the following three readings: (26) a. main verb, metaphor: ‘the flame threatened the destruction’

150 Scales and scenarios of grammaticalization b. temporal-aspectual, ingressive: ‘the flame was about to burn up the lovely work’ c. perceptual inferential evidential: ‘some observable evidence points to the undesired event of the flame burning up the lovely construction’ Table 11 gives an overview of the interpretative choices available for this sentence. Table 11. Options for interpreting the critical context of drohen 1. main verb reading (metaphor): the action verb lodern triggers the metaphorical interpretation of the subject; die Flamme is conceived as an animate agent being able to threaten someone, i.e. drohen is a full verb here; an extended paraphrase for this interpretation is: ‘the flame (like an animate being) already performed the speech act of threatening to burn up the lovely work.’ 2. temporal-aspectual reading (ingressive): the subject is not conceived as being an animate entity; the semantics of the original full verb drohen is changed from that of a speech act verb to a socalled phase-verb marking the content of the infinitive as being about to happen; in this case, drohen is no longer interpreted as designating a speech act announcing an action; instead it is understood as a verb indicating the onset of an event. an extended paraphrase for this interpretation is: ‘the process of the flame burning up the lovely work was imminent’ 3. inferential evidential reading: the evidential meaning consists in the interpretation of drohen as a secondary predication (and auxiliary verb) having the whole proposition in its scope and referring to a piece of information which the speaker adduces as evidence for the proposition. an extended paraphrase for this interpretation is: ‘...some observable evidence points to the undesired event of the flame burning up the lovely construction’

As already mentioned above, it is proposed here to treat the multiple structural and semantic opacities showing up in the critical context in a grammaticalization process as an idiomatic construction in the sense of Fillmore, Kay and O’Connor (1988). More exactly, we are confronted here with what the authors call an extragrammatical idiom, and which they define as follows: “Such expressions [i.e. extragrammatical idioms] have

Stages, contexts and constructions 151

grammatical structure, to be sure, but the structures they have are not made intelligible by knowledge of the familiar rules of the grammar and how those rules are most generally applied” (Fillmore, Kay & O’Connor 1988: 505). This construction cannot be treated as a simple case of polysemy, because the different meanings correlate with different structural analyses. Instead, we are confronted with a complex structure containing several potential meanings, each of them associated with a distinct structural analysis. This raises the question which semantic and structural analysis should be assigned to this construction in the first place. Without being able to provide a complete answer to this question, we would like to suggest the direction a possible solution could take. As for the structural description, we propose to model it closely according to the attestable distinctions of the surface forms without overinterpreting any morphosyntactic structure. As far as the meaning of the construction is concerned, there are indications that this construction may have had a specific stylistic function, namely that of reinforcement or emphasis. Thus, it seems reasonable to assume a “pragmatic” meaning which is assigned holistically to the construction. The content of this pragmatic meaning might be circumscribed as an iconic relation between the structural and semantic layer on one side and the pragmatic and stylistic layer on the other.8 This suggestion rests on the assumption that the extraordinary semantic and structural opacity (which surely originates in conscious creative choices of the authors) is meant to iconically indicate an extraordinary stylistic salience, and thus serves as an instruction to the hearers or readers to recognize and evaluate the interpretational options and choose a suitable one themselves. In short, the notion of the critical context as it is suggested here avoids any attempts to assign a particular categorical (functional, semantic, structural) identification. Instead, it explicitly treats critical contexts as semantically and structurally ambiguous structures which allow for several alternative interpretations, whereby one of them is the predecessor of the newly developing grammatical function. Stage 3. Stage three shows the consolidation of the grammaticalization process, i.e. the re-organization and differentiation of the grammatical formatives and the paradigm that is the target category of the ongoing grammaticalization process. In this phase, the new grammatical meaning is isolated as a separate meaning from the older, more lexical, meaning. This separation of the two meanings is achieved by the development of isolating

152 Scales and scenarios of grammaticalization contexts for both the lexical and the grammaticalized readings, i.e. specific linguistic contexts that favour one reading to the exclusion of the other. As soon as the opposition between the isolating contexts is established, the process of grammaticalization can be said to be irreversible as it cannot be reduced to an earlier stage. The new grammatical meaning is no longer dependent on conversational implicature, as the linguistic element under grammaticalization has become truly polysemous (cf. polysemy test in 5.2). Usually, isolating contexts for the grammaticalized and for the non-/less grammaticalized variants show complementary distribution with regard to their relevant structural properties. Hence, isolating contexts for non-/less grammaticalized variants of the elements in questions may be defined negatively with respect to the isolating contexts for grammaticalized variants. The isolating contexts for the modal verbs in PDG are the following: the grammaticalized modals do not allow for periphrastic tenses; this means that, if a modal is used with a periphrastic tense, it can only have the less grammaticalized meaning. This is shown in (27), which allows only for a lexical reading of hat müssen, etc.):9 (27)

Er hat/habe/hatte/hätte/wird ... erst das Semester zu Ende bringen müssen. ‘He has been/had been/would have been/will be … obliged to finish the semester first.’

On the other hand, there is a type of context in which the lexical reading is virtually excluded and the grammaticalized reading is highly favoured. This is the combination of the modal with an infinitive perfect, as shown in (28): (28)

Ich muss mich getäuscht haben. ‘Perhaps I was mistaken.’

Table 12 gives an overview of the two isolating contexts of the modal verbs in PDG. The isolating contexts for drohen are illustrated in (29) and (30), which both allow only the reading of drohen & zu-infinitive as an inferential evidential marker, i.e as an evidential auxiliary.

Stages, contexts and constructions 153 Table 12. The isolating contexts of German modals 1. Isolating context for the less grammaticalized, lexical reading: periphrastic modal verb (e.g. haben as auxiliary) & modal verb as past participle (realized as the so-called “Ersatzinfinitiv”, i.e. a past participle looking like an infinitive) & lexical verb as infinitive (complement of the modal) (see (27)) Meaning of the whole construction: ‘a modal state being predicated on the subject of the sentence’ 2. Isolating context for the grammaticalized, deictic reading: finite modal verb & perfect infinitive of the main verb (see (28)) Meaning of the whole construction: ‘uncertain factuality value being attributed to the whole proposition by the speaker’

(29)

(30)

Sitz gerade auf dem Fahrrad, Ilja! Führe den Lenker nach links, wenn du nach rechts zu kippen drohst! (IDS-Corpus) ‘Sit straight on the bike, Ilja! Turn the handlebars to the left, when you threaten to topple to the right.’ aber die Sonnenhitze brennt gar zu stark, daß mir das zarte Kraut zu welken droht…(GRI/KHM.00122, IDS-Corpus) ‘but the heat burns too strongly, so that my fragile herbs threaten to fade‘

The relevant contextual features of this isolating context are the following ones: 2nd and 3rd person subjects (animate/inanimate) & drohen in present/preterite (i.e. non-periphrastic) tense form & infinitive verbs denoting a non-action

It is suggested here that the isolating contexts in the case of the modals as well as in the case of drohen, like the critical contexts, are constructions in the construction-grammatical sense. More precisely, they belong to a subtype of idioms which in Fillmore, Kay and O’Connor (1988) are classified as “formal or lexically open” idioms. The authors define these idioms as “syntactic patterns dedicated to semantic and pragmatic purposes not knowable from their form alone” (Fillmore, Kay and O’Connor 1988: 505). Formal idioms are constructions whose compositionality is reduced, i.e. at least some part of their form-meaning correspondence has to be treated holistically and cannot be derived in its totality from other constructions or from a combination of other constructions; in short: they are characterized

154 Scales and scenarios of grammaticalization by non-compositionality. However, they are still fully productive, as their syntactic positions are not filled with lexically fixed items. Thus, though there are partial similarities with other constructions which indicate partial inheritance relations, the isolating contexts display a unique form-meaning correspondence which cannot be reduced to combinations of other constructions. They therefore qualify as partially productive, idiomatic constructions in the constructional sense of the term elaborated above. Stage 4. The fourth and final stage, which ultimately defines a given development as grammaticalization is the paradigmatic (re)integration or ‘paradigmaticization’ of a new grammaticalized element. We argue that we need this stage of paradigmatic (re)intergration in order to account for the full process of grammaticalization, and also to distinguish grammaticalization from other types of language change. As already discussed in Section 4.1, paradigmatic organization and obligatory use (with the relevant reservations and restrictions) are essential features of grammatical categories and signs (independently of their morpho-syntactic expression as inflection). Paradigmaticity is the very property that distinguishes grammatical items from lexical ones, which may conjoin into word fields (i.e. loosely organized, open classes) but not into paradigms. Thus, beyond morpho-syntactic and semantic changes, the most important factor of grammaticalization is the increase in paradigmaticity. By the process of paradigmatic (re)integration the new grammatical(ized) meaning is stabilized, as it (i) comes to be confronted with opposing members, on the one hand, and (ii) is gradually integrated within a more abstract grammatical meaning serving as a common denominator for each member of the paradigm, on the other hand. The resulting structure is a tightly integrated paradigm, as it is described by Lehmann: On the semantic side, the members of a paradigm have a common semantic basis with varying differentiae specificae. This would be brought out by a componential analysis and is reflected in traditional terminology by the fact that there is a generic category name for the whole paradigm and opposite names for the specific subcategories. Such paradigmacity is gradually reached in the process of grammaticalization. Categories grammaticalized very little do not constitute such tightly integrated paradigms. (Lehmann 2002: 120)

In other words, in Stage 4 the new grammatical meaning becomes subject (or: dependent or non-autonomous) to the meaning of the whole paradigm it pertains to and at the same time dependent on the meanings of other members of the paradigm. This contrasts sharply with instances of lexicali-

Stages, contexts and constructions 155

zation, i.e. the inventarization of an item as part of the lexicon, or mere semantic change, where there is (usually) no abstract meaning serving as an integrating label as is the case in a paradigm. If there is paradigmatic integration which at the same time is accompanied by persistence of other – more lexical – uses of the relevant item, then we may talk of the rise of heterosemy, i.e. the split into different functional classes. It should be emphasized here that we treat the very tight semantic organization of a grammatical paradigm as its most relevant property. As concerns the formal characteristics, the members of a paradigm may (and usually do) show variation, such as variation between inflectional and periphrastic forms within one paradigm (e.g. German tense and mood). Nevertheless, it can be assumed that the more developed a grammatical paradigm, the more homogeneous the formal characteristics of its members are. For this reason, describing ‘construction types’ of the fourth stage in grammaticalization, we call it ‘abstract constructional schema’ (cf. also Langacker 1987, 1991, 2008). An abstract constructional schema consists of the common semantic basis of a paradigm (e.g. tense marker) and the abstract structural properties of its participating constructions (e.g. verbal inflection or periphrasis), thus representing the type of non-reducible association of form and meaning crucial for classifying a given linguistic entity as a construction. In a first approach we suggest that a notational convention of abstract constructional schemas includes the following information: the category/paradigm label – the abstract formal characteristic of the members of the category – the common semantic feature constituting the general meaning/function of the whole paradigm. A few examples will help to clarify what is meant here. The abstract constructional schema of German tense markers in an abbreviatory way may be noted as follows: CATEGORY LABEL: FORMAL CHARACTERISTICS:

CATEGORY MEANING:

tense marker modifies the main verb of the clause: either by an inflectional morpheme (e.g. the German preterite affix -te) or ablaut (schreibe – schrieb) or a periphrastic form (haben/sein + perfect participle) temporal location of the described event as related to the speech time

The fourth stage – paradigmatic integration – thus refers to a whole set of constructions, i.e. the constructional network building the paradigm, on the one hand, and every single construction within it, on the other. It is the

156 Scales and scenarios of grammaticalization conceptualization of the paradigmatic oppositions building a grammatical category as a specific type of form-meaning pairing, and thus a specific type of construction. It represents the process of integration of existing constructions established during the three preceding stages into a closely interrelated network. The process of consolidation into a paradigm – the establishment of paradigmatic relations between constructions with the result of a new paradigm (a new type of construction) – is what actually happens at this stage. The process of paradigmatic integration of the German modals has been described by Diewald (1999, 2002). Although the details cannot be rendered here, and notwithstanding the fact that it may be necessary to adapt single values and oppositions in the scheme, Table 13 may serve as a summary showing the importance of Stage 4. Table 13. The paradigmatic organization of the German modal markers CAT. LABEL: modal marker F. CHAR.: modifies the main verb CAT. MEANING: factuality status of the proposition

semantic oppositions within the paradigm (distinctive values of individual members)

factual formal realizations verbal mood markers: - inflectional - periphrastic MV & infinitive

non-factual purely deictic

indicative

können müssen

phoric

quotative

subjunctive II würde & infinitive

subjunctive I

dürfte mag

sollen wollen

A further important aspect should be addressed here: if there is an already existing paradigm which functions as the target category of the process, this paradigm is the paradigmatic context (cf. “renovation” or “renewal”

Stages, contexts and constructions 157

Lehmann 1995 [1982]: 21). This is the case for the development of periphrastic tenses, grammaticalized uses of modals etc. If there is no existing paradigm, i.e. if a new category is arising in a language (e.g. the modal particles, the article system, the evidential system in German), the members of the newly arising category build a paradigm among themselves, whereby the older members are the blueprint and give the pattern for analogous developments etc. (cf. “innovation” Lehmann 1995 [1982]: 21). This, as we contend, is the case with the German evidential constructions, which developed into a system like the one illustrated in Table 5 in Section 3.5, and which is repeated here (in a slightly modified form) for convenience. Table 14. Paradigmatic oppositions of inferential evidentials in PDG CAT. LABEL: evidential marker F. CHAR.: modifies the main verb CAT. MEANING: information source for the proposition

semantic oppositions within the paradigm (distinctive values of individual members)

unmarked evidentiality formal realizations unmarked form

auxiliary & infinitive

inferential evidentiality conceptual

conceptual/ perceptual

perceptual

werden

scheinen

drohen versprechen

indicative (zeromarking)

It should be noted, furthermore, that the stage of paradigmatic integration is not a sharply delimited chronological period, but rather a continuous process. While at the beginning of a process of the paradigmatic integration there are typically formally inhomogeneous paradigms with relatively high numbers of members and relatively rich, overlapping, non-distinctive semantic features, moving towards the end the members become more homogeneous formally and at the same time more distinctive semantically.

158 Scales and scenarios of grammaticalization The smaller the paradigm, the more distinct the semantic contrast between its members and the more dominant the grammatical function. This, of course, can be witnessed in all diachronically old inflectional paradigms, which have only a few members which, however, have a greater distinctive impact than members of larger, more loosely structured paradigms.

4. Summary This chapter laid out the dynamic aspects of grammaticalization, i.e. factors steering the process of change itself, which – in addition to the more stative, classificatory and comparative concepts presented in Chapters 2 and 3 – are necessary for the construction of a complete diachronic grammaticalization scenario. These dynamic concepts are first the metaphorization and metonymization scales, which capture essential cognitive procedures triggering change in single lexical items, second, the concept of scheme retention and reinterpretation, and third, the successive semantic-functional stages, whereby the two latter ones provide the matrix for possible steps and direction of semantic change relevant in grammaticalization. Furthermore, definitions of ambiguity, polysemy, heterosemy and related concepts have been elaborated as far as they are necessary for a description of grammaticalization. The model of successive stages, contexts and construction types generalizes over single instances and relates processes of change to states in linguistic systems. It distinguishes four stages, which are associates with specific processes and features and which are assumed to describe the general developmental stages of grammaticalization. Both perspectives – the one of tracing a change through time, and the one segmenting processes into discernible stages, must be taken account of and must be combined in order to model a grammaticalization process as realistically as possible. The remaining chapters will proceed along these guidelines.

Chapter 6 The four evidential constructions in present-day German

This chapter is devoted to the presentation of our synchronic corpus findings about the four evidential constructions in PDG. We first present an overview of the relevant construction types in which the verbs werden, scheinen, drohen and versprechen occur in PDG, irrespective of their semantics and degrees of grammaticalization. Next, our focus will be on the infinitive constructions with these verbs. A survey of the grammaticalization status of each construction with respect to the parameters of grammaticalization has been given in Chapter 4. In the following we will concentrate on the evidential meaning of each construction and provide corpus evidence for treating these verbal periphrases as relatively highly grammaticalized linguistic devices for expressing the category of (inferential) evidentiality in German.

1. Werden & infinitive The construction werden & infinitive is highly polysemous in PDG: it has several different readings if it is used in different contextual surroundings. In the following we will argue that all readings of werden & infinitive in PDG can be dervied from one core meaning, which has inferential evidential semantics. The next section presents a detailed analysis of the core meaning and its context-induced interpretations. The different chronological stages of the diachronic development will be laid out in Section 8.1.

1.1. Relevant construction types with werden in present-day German In PDG the verb werden occurs in several variants. Its functional status ranges from main verb to auxiliary, i.e., it displays several degrees of grammaticalization. The following classification gives the relevant variants of werden-constructions in PDG:

160 The four evidential constructions in present-day German 1. 2. 3. 4. 5.

werden as a main verb; werden as a copula; werden with perfect participle, where werden serves as a passive auxiliary; werden with infinitive, where werden has ambiguous interpretations1; würde with infinitive, where würde serves as a (preterite) subjunctive auxiliary.

The variants listed above are illustrated in the following examples: (1) (2) (3) (4) (5)

Was nicht ist, kann noch werden. ‘What isn’t yet may well still be [become].’ Sie wird traurig. ‘She is getting sad.’ Das Haus wird gebaut. ‘The house is built.’ Sie wird schlafen. ‘She will be sleeping/will sleep.’ Sie würde schlafen (, wenn das Licht aus wäre). ‚She would sleep (, if the light was off).

Werden as a main verb is very rarely used in PDG. In this function it means ‘to happen, to rise, to occur, to come about, to come into being’ etc. It carries aspectual semantic features and is ingressive,2 i.e., it refers to an event which is bounded in its onset. This meaning of werden has remained relatively stable since OHG. Werden as a copula combines with nouns and adjectives and expresses ingressive aspectual meaning as well (see (2)). In this function werden shows the whole range of inflectional forms, i.e., it occurs in different – synthetic as well as periphrastic – tense and mood forms. Werden as a passive auxiliary is the most frequent use of this verb in PDG. In this function werden serves as a highly grammaticalized passive auxiliary. It is constitutive for the formation of the so-called werdenpassive (“Vorgangspassiv”), which forms a paradigmatic opposition to the active voice, cf.: (6)

Der Mechaniker repariert den Motor. > Der Motor wird vom Mechaniker repariert. (examples taken from DUDEN 2005: 553)

Werden & infinitive 161

‘The mechanic repairs the motor.’ > ‘The motor is repaired by the mechanic.’ As this variant of werden is not decisive for our further analysis it will suffice to say here that it serves as a usual grammatical device to form passive sentences. It combines with most German verbs, transitive as well as intransitive. As already mentioned, the construction werden & infinitive is ambiguous, even polysemous in PDG, and the functional status of this construction is still a matter of considerable debate in German linguistics. In particular, it has been conceived of as a future tense form, on the one hand, or as a modal construction, indicating the speaker’s epistemic assessment of the likelihood of the described situation, on the other. The former account is put forward by the so-called temporalists (e.g. Matzel and Ulvestad 1982, Fabricius-Hansen 1986, Thieroff 1992 et al.), while the so-called modalists (e.g. Vater 1975, Itayama 1993) favour the latter view. Taking a position in between, DUDEN 2005 classifies werden as a temporal-modal auxiliary (“temporal-modales Hilfsverb”; DUDEN 2005: 424) that is inherently ambivalent between the temporal and the modal interpretations. So, in examples like the following werden is considered to be a temporal auxiliary marking the future time reference of the described event: (7)

Bereits am 1. April wird das Strandbad Müggelsee seine Pforten öffnen. (DWDS, Kernkorpus) ‘The Müggelsee Lido will open as early as April 1st.’

On the other hand, in contexts like in (8) werden is usually identified as a modal auxiliary indicating the factuality status of the described event: (8)

Ein bisschen wird man wohl nachdenken dürfen. (DWDS, Gesprochene Sprache) ‘Well, one should be allowed to give it some thought.’

The quotation below illustrates that the (exact) modal value of the construction werden & infinitive is difficult to define: Diese modale Verwendungsweise wird oft als Ausdruck einer Vermutung bezeichnet. Sie scheint sich im Hinblick auf die modale Stärke eher neutral zu verhalten. [‘This modal use is often considered an expression of an as-

162 The four evidential constructions in present-day German sumption. It seems to be rather neutral in regard to the modal strength of the speaker’s assessment.’] (DUDEN 2005: 515)

This, as we think, is a rather problematic issue. Werden is said to have a modal value, hence it is identified as a factuality marker in PDG. At the same time it is admitted that the modal value of werden & infinitive is neutral so that it cannot be attributed a particular distinctive value within the paradigm of the German factuality markers (see Chapter 3). As we will see in the following sections neither the purely temporal nor the purely modal interpretations nor the mixed temporal-modal identification are able to cover the functional properties of the construction werden & infinitive. In contrast to these descriptions we are going to suggest a core evidential semantics which allows for a derivation of all contextually induced interpretations. As far as the würde & infinitive construction is concerned, it has been shown in Smirnova 2006 that this construction has two main variants in PDG: first, it serves as an analytical variant of the synthetic preterite subjunctive (as exemplified in (9) and (10)); second, it functions as a preterite counterpart of the indicative construction werden & infinitive. (9) (10)

Ich würde es kaufen (, wenn ich Geld hätte). ‘I would buy it (if I had money).’ „Das war, als hätte man einen Teddybär dabei“, erzählt die Managerin, „oder als würde man mit einem Menschen sprechen.“ (DWDS, Kernkorpus) ‘”It was like having a Teddy bear with you“, the manager told, “or like talking to a real person“.’

Würde & infinitive can replace the preterite subjunctive forms of most verbs in German, in particular in counterfactual and irreal conditionals. Moreover it serves as an obligatory substitute of the preterite subjunctive forms of German regular verbs which are isomorphic to the past indicative (see (9)). It is uncontroversial that this reading of würde is highly grammaticalized: Mit dem Grundtempus des Konjunktivs II (Konj. Prät.) bezieht der Sprecher sich auf Gegenwärtiges oder Zukünftiges (d.h. Nichtvergangenes). [...] Die würde-Form wird in diesem Funktionsbereich des Konjunktivs II weitgehend gleichbedeutend mit der Grundform gebraucht. [‚With the basic tense of the Konjunktiv II the speaker refers to the present or the future (i.e. nonpast). […] In this functional domain of the Konjunktiv II the würde form is

Werden & infinitive 163 largely used in a synonymous way’]. (Duden 2005: 523) (see also Helbig and Buscha 2001, Flämig 1991, Eisenberg 2004, Weinrich 2003, Smirnova 2006 for similar opinions)

The second reading of the construction – the preterite of werden & infinitive – is exemplified in (11) where würde kommen cannot be substituted by the synthetic preterite subjunctive of kommen, i.e., käme: (11)

Flavio hatte gewusst, dass diese Stunde einmal kommen würde. (taken from Smirnova 2006: 137) ‘Flavio knew that this moment would come at some time.’

The construction würde & infinitive in both readings – which has been mentioned here for the sake of complete coverage – will not be dealt with in the course of this study.

1.2. Semantic diversity of werden & infinitive and its core semantics The functional status of the construction werden & infinitive has not yet been precisely identified due to its extensive polysemy. In this section we deal with five distinct readings, which we label as follows: (a) future time reference 1; (b) future time reference 2; (c) performative reading; (d) textconnecting function, (e) evidential reading. Our survey is based on our corpus analysis and on the observations found in the relevant literature on this topic. As we are mainly concerned with the notion of evidentiality, our central intentions are, first, to prove that werden & infinitive indeed has inherent evidential meaning and, second, to investigate how this meaning is connected to the whole range of readings werden & infinitive displays in PDG. (a) werden & infinitive with future time reference 1 It is uncontroversial that the construction werden & infinitive can be used to introduce events whose occurrence is temporally located in the future. A few selected examples from our corpus illustrate this use: (12)

Der König wird… Donnerstag früh nach Biarritz abreisen. Er wird eine oder zwei Nächte in Paris verbringen und dort, bevor er seine Reise fortsetzt, mit dem Präsidenten Fallières zusammentreffen. (DWDS, Kernkorpus)

164 The four evidential constructions in present-day German

(13)

(14)

‘The king will depart for Biarritz on Thursday morning. He will spend one or two nights in Paris and, before he continues his journey, he will meet President Fallières there.’ Wir werden das nachher erläutern. (DWDS, Gesprochene Sprache) ‘We will comment on it later.’ Wir werden heute wie üblich fünf Bücher besprechen. (DWDS, Gesprochene Sprache) ‘Today we will discuss five books as usual.’

It should be noted that, while having analyzed our corpus data, we identified only those instances of werden & infinitive as purely temporal which allowed for a strict temporal interpretation without any additional (modal, epistemic, evaluative, etc.) nuances. This means that whenever werden & infinitive might also have been interpreted as an indication of the speaker’s assessment of the likelihood of the (present or future) described event, we did not consider it to be purely temporal. Therefore uses like the following, which are identified by e.g. Thieroff (1992: 126-127) and Matzel and Ulvestad 1982 as temporal, are excluded from our group (a). (15)

Er weiß, was geschehen wird. ‘He knows what will happen.’

In our view these uses of werden & infinitive differ from the purely temporal reading (a) of this construction in that they always convey an additional – speaker-oriented – meaning component which is joined with the temporal one. For this reason we classify them as pertaining to the reading represented under (b) below. Although in the examples above werden & infinitive serves to mark the (future) temporal location of the described event, it is not the exclusive function of the construction. Throughout the future uses of the construction there are also (contextually given) factors that have an impact on this particular interpretation and, as we think, should not be ignored. As illustrated in (12) it is some decision, resolution, law or some other fact that causes the existence or appearance of the situations introduced by werden. In (13) and (14) there is an announcement made by a person who exactly knows what is going to happen (because s/he is moderating the discussion and knows the schedule of the program). Similar uses are frequently found in fiction where the author announces what is going to come in the next pages

Werden & infinitive 165

or chapters of a novel. Thus, we can state that there is always some reference point (apart from the temporal one, i.e., the speech time) to which a sentence with werden is anchored: it may be (knowledge about) a prior event or knowledge about the planned occurrence of events. This fact, i.e. the presence of a relevant reference point, may explain, for instance, why werden & infinitive is not used when introducing events grounded in calendar time, cf.: (16) (17)

* Morgen wird Dienstag sein. VERSUS Morgen ist Dienstag. *‘Tomorrow will be Tuesday.’ VERSUS ‘Tomorrow is Tuesday.’ * Am nächsten Freitag werde ich meinen Geburtstag haben. VERSUS Am nächsten Freitag habe ich meinen Geburtstag. *‘Next Friday I will have my birthday.’ VERSUS ‘Next Friday I have my birthday.’

In our view such uses of werden & infinitive are blocked because there is no reference point in the sense described above, i.e., there is no such thing as a prior event or knowledge about what is coming next, to which werden usually relates the described event. Instead, it is always a statement about a fact, and this fact is neither dependent on some prior event nor on someone’s knowledge of it: it is just a fact. For this reason werden is not acceptable in such contexts. Only if the described event implies a link to some (knowledge about) facts or arranged plans can werden be used (example (18) is taken from Thieroff 1992: 132): (18)

A: – Was machst du nächsten Freitag? A: – ‘What are you doing next Friday?’ B: – Nächsten Freitag werde ich meinen dreißigsten Geburtstag feiern. B: – ‘Next Friday I will celebrate my thirtieth birthday.’

Here the fact that the birthday party has been arranged for the next Friday serves as a reference point (in the sense described above). Therefore, the use of werden & infinitive is possible. (b) werden & infinitive with future time reference 2 Werden & infinitive is often used together with lexical expressions referring to an epistemic stance of the speaker towards the proposition. That is, not only a (future) event is introduced, but also an epistemic evaluation of

166 The four evidential constructions in present-day German this event is provided in a sentence. In this case we do not distinguish between various degrees of certainty of the epistemic judgment. The degree of the (speaker’s) certainty may range from sureness to doubt; in most cases however it has a neutral status in the sense that the speaker’s epistemic stance is explicitly expressed without being further specified in terms of degree (e.g. by such lexical expressions as Aussicht ‘outlook’, Erwartung ‘expectation’, prognostizieren ‘to predict’ etc.). (19)

(20)

(21)

(22)

Damit eröffnet sich die Aussicht, dass das Gesetz auch im Reichstage die notwendige Mehrheit finden wird. (DWDS, Kernkorpus) ‘So the expectation appears that the law will get the necessary majority in the Reichstag.’ Nach seinem Eindruck sei Polen in riesigen Schwierigkeiten. Der kommende Winter werde sehr schwer werden. Die EG werde Lebensmittelhilfe an Polen geben müssen. (DWDS, Kernkorpus) ‘His impression is that Poland is in big trouble. The next winter will be very hard. The EU will have to provide Poland with food assistance.’ Ich sage Ihnen die Diskussion mit den Vereinigten Staaten von Amerika voraus. Sie werden sich isolieren. (DWDS, Gesprochene Sprache) ‘I predict that there will be discussions with the United States of America. She will isolate herself.’ Ich glaube, ich werde auch Recht haben. (DWDS, Gesprochene Sprache) ‘I believe I will be right as well.’

Matzel and Ulvestad 1982 distinguish between two types of werden & infinitive that they term “Zukunftsfutur1” and “Zukunftsfutur2”. The first is labeled [+Gewähr ‘warranty’] and is said to allow substitution by the present tense of the infinitive verb because the future reference is somehow given in the context. This type comprises also uses described in (a) above. The second is “subjective” and is labeled [–Gewähr]. We think however that in both cases, irrespective of whether the epistemic assessment of the proposition is explicitly coded in the text (or may be reconstructed from the context) or not, werden & infinitive fulfills one and the same function: it refers back to a significant reference point, be it the speaker (together with his/her world of knowledge) or the person talked about.

Werden & infinitive 167

(c) werden & infinitive with performative meaning Performative uses of werden & infinitive have been usually seen as derived from its future meaning (cf. DUDEN 2005, Kotin 2003, Welke 2005 et al.): they have been explained as a particular interpretation of the future time reference that arises in contexts with first and second person human subjects. In such contexts werden is said to express intention (promise or threat) or request (command). On the other hand, the modalists (especially Vater 1975, Amrhein 1996) have claimed that such uses may be considered “non-inferential” (or deontic) uses of the “modal verb” werden. In these cases, the semantics of werden is very close to the semantics of the modal verb wollen ‘want’, while the combination with the second person subject corresponds to that of sollen ‘shall’. The following examples from our corpus illustrate the performative use of werden & infinitive: (23)

(24)

(25)

Wir werden uns dieser Herausforderung stellen. (DWDS, Gesprochene Sprache) ‘We will meet this challenge.’ Wir werden neonazistische Gruppen in Berlin sich nicht mehr entfalten lassen. Wo immer möglich, werden wir sie zerschlagen, weil das unsere moralische und nationale Pflicht ist. (DWDS, Kernkorpus) ‘We will not let neonazi groups unfold in Berlin. Wherever possible we will break them up because this is our moral and national obligation.’ Nein, nein, nein! Das werden Sie mir nicht befehlen. (DWDS, Gesprochene Sprache) ‘No, no, no! You will not order me to do this.’

We argue that in its performative reading the construction werden & infinitive implies a strong link to a relevant reference point to which the proposition is related. Mostly this reference point may be easily reconstructed from the context: this is either a decision which results in the (speakerinternal) intention or a (external) status of the speaker that allows him/her to order something from others. (d) werden & infinitive with text-connecting function While investigating our synchronic corpus data we have paid special attention to the contextual configurations in which werden & infinitive occurs (the first attempt in this direction has been done in Smirnova 2006). What

168 The four evidential constructions in present-day German we have found is the large number of werden uses which undoubtedly are very closely interlinked with their contextual environment. Werden & infinitive has a particular text-integrating function which has been paid very little attention in linguistic research so far. Namely, werden & infinitive indicates a strong link to the (preceding) text (segment) in that it introduces situations which are seen as a consequence from what has been described before. Consider the following examples: (26)

(27)

(28)

Über zweitausend Menschen verlassen die DDR, siedeln in den Westen über. Und das wird sich noch weiter vergrößern. (DWDS, Gesprochene Sprache) ‘More than two thousand people leave the GDR, move to the West. And this number will continue to grow.’ Wenn Sie gütigst erwägen, daß mir der Spaß schon zu teuer gekommen ist, so werden Sie wohl sich sagen müssen, daß ... (DWDS, Kernkorpus) ‘If you kindly consider that the fun has cost me dear you will no doubt have to say yourself that…’ Wer sich gegen die verfassungsmäßige Ordnung betätigt, wird durch das Bundesverfassungsgericht ausgeschaltet werden. (DWDS, Kernkorpus) ‘People who act against the constitutional order will be eliminated by the Federal Constitutional Court.’

The text passages preceding the clause with werden & infinitive contain information that can be rendered as a fact that is (in the sense of a causeeffect relationship, empirically, logically, hypothetically, etc.) followed by the situation formulated with werden & infinitive. The future temporal location of the described event seems to be not the exclusive function of the construction. Though the state of affairs is usually located in the future, this need not necessarily be the case. Instead, the temporal meaning mostly results from the interpretation of the whole context of the utterance. Rather, and this holds for all examples above, werden & infinitive serves as a means to mark the described event as related to the preceding – textually present – information and as resulting from it. Further examples from our corpus illustrate the close interconnection between the construction werden & infinitive and the textually present information:

Werden & infinitive 169

(29)

(30)

Wenn die Reformen in Polen und Ungarn scheiterten, wären sie auch in der DDR zum Scheitern verurteilt. Deshalb werde er in drei Wochen nach Ungarn reisen. (DWDS, Kernkorpus) ‘If the reforms in Poland and Hungary failed they would also be doomed to failure in the GDR. Therefore (he said) he would travel to Hungary three weeks later.’ Also bitte! Dann wird ein Buch wegfallen. (DWDS, Gesprochene Sprache) ‘Very well then! Then one book will be omitted.’

As these examples show the clause with werden often contains an anaphoric device that explicitly refers to the preceding text (deshalb ‘therefore’, dann ‘then’, we also found darum ‘for that reason’, daraus ‘hence’, damit ‘therewith’ in our corpus). These devices mark the proposition as a (logical, causal, conditional, etc.) consequence from the preceding text due to their inherent semantics. Taking into account the examples (26) - (28) discussed above we can conclude that it is not only the anaphoric elements in the text which establish a (sequencial) relationship between two text portions. Rather it is the construction werden & infinitive which establish this link, and the anaphoric elements serve to emphasize, i.e., to textualize this relationship. Our diachronic findings support this suggestion (for more details see Chapter 8.1): chronologically, the constructions of the type presented in (26) - (28) are older than the constructions with anaphoric elements. In this connection the occurrence of anaphoric linguistic elements can be seen as a (textual) strengthening of the already established function of werden & infinitive. In sum, our corpus analysis has revealed a relatively frequent use of werden & infinitive that is characterized by a strong text-connective function of this construction. Werden indicates that the described event is related to the event(s) introduced in the preceding text. Furthermore werden characterizes the described event as a result or consequence that follows from the preceding situations. An important property of werden in such contexts is that it does not necessarily establish a temporal reference. What werden & infinitive of this type shares with the uses described earlier in this section is that there is a relevant reference point to which the clause with werden is anchored. This reference point is represented here by the textually present information pieces as far as they serve as a fact or a situation from which the werden-event results.

170 The four evidential constructions in present-day German (e) werden & infinitive with evidential meaning In the literature on werden & infinitive it is generally agreed that the construction has modal nuances when it introduces present or past situations. In such cases werden is said to express the speaker’s (subjective) attitude towards the proposition. The modalists claim that this is the core meaning of the construction werden & infinitive (see above). Others (e.g. DUDEN 2005, Kotin 2003, Fritz 2000 et al.) see it as a particular interpretation of the core temporal meaning of werden & infinitive: since the described event cannot be located in the future (which is obvious from the context), the future temporal meaning of werden is said to be blocked and therefore reinterpreted as epistemic meaning. As we have demonstrated above, the temporal value of werden & infinitive is not necessarily present in all of its uses. Even when this construction is used to introduce future events it has a more fundamental semantic component which can be found in every use of this construction. Therefore we suggest that the temporal interpretation itself results from the core semantics of werden & infinitive and serves as one possible interpretation of this core meaning. It is to say that we do not restrict so-called epistemic interpretations of werden & infinitive to present and past situations. Rather, such uses are commonly spread through all temporal perspectives, cf.: (31) a. Er wird jetzt in seinem Büro sitzen. He WIRD now in his office sit.INF ‘He is bound to be in his office right now.’ b. Er wird gestern in seinem Büro gesessen haben. He WIRD yesterday in his office sit.PP have.INF ‘He must have been in his office yesterday.’ c. Er wird morgen in seinem Büro sitzen. He WIRD tomorrow in his office sit.INF ‘He is bound to be in his office tomorrow.’ We claim that the meaning of werden & infinitive in all three examples in (31) is the same, irrespective of the different temporal location of the described event. In our view, this meaning is clearly evidential and not epistemic (see Chapter 2). In the next section we will argue that werden & infinitive has a core semantics which underlies all its various interpretations.

Werden & infinitive 171

1.3. Temporal reference or modal evaluation: the core semantics of werden & infinitive Several attempts have been made in the last few decades to subsume different readings of the construction werden & infinitive under a core meaning. We have already mentioned two opposite approaches: the temporalists (Gelhaus 1975, Fabricius-Hansen 1986, Thieroff 1992, Welke 2005 et al.) and the modalists (Vater 1975, Engel 2002, Itayama 1993, Weinrich 2003 et al.). The former see werden & infinitive as a future tense form. The latter treat werden as a modal verb. Further conceptions have been offered that argue for a core meaning of werden & infinitive which is neither pure temporal nor pure modal. Lipsky (2002: 106), for example, argues for a pragmatic view and claims that the main function of werden & infinitive is to change the actual speech situation in order to influence the hearer. Fritz 2000 proposes to treat werden & infinitive as an explicit expression of the speaker’s orientation towards the proposition. He defines the core semantics of werden & infinitive as Sprecherbezug ‘reference to the speaker’. This means that, using the construction werden & infinitive, the speaker explicitly refers to him/herself as a producer of the utterance. This very abstract semantics, following Fritz 2000, is the basis for every use of werden & infinitive in PDG as it allows for different interpretations in specific contexts. This view is partly in line with our suggestion about the core semantics of werden & infinitive. Kotin 2003 terms the core semantics of werden & infinitive “prognostic”. This view is rather problematic. First, the verb werden is attributed a two-fold grammatical status: it is classified as a modal verb and at the same time as a ‘prognostic’ auxiliary. Second, it is hard to see the difference between the notion of prognosis and the future semantics, as it is argued that there is always a temporal distance between the fact of uttering a prediction and the fact of testing the validity of this prediction. The first attempt to describe the core semantics of werden & infinitive in evidential terms has been done by Diewald 2005 and has been further developed by Smirnova 2006. This approach will be described in more detail in the next section. At this point the abstract semantics of werden & infinitive will be introduced. We propose that one crucial semantic property can be attributed to all relevant readings of werden & infinitive mentioned above. This semantic feature can be sketched only in very abstract terms (which is not surprising if we take into consideration the high degree of grammaticalization, and

172 The four evidential constructions in present-day German thus of semantic bleaching, of this verb within infinitive constructions). Werden & infinitive anchors the proposition to some relevant reference point. The particular nature of the reference point may vary: it can be either the preceding text (d) or the knowledge about certain facts (a), or the general (shared) knowledge about the outside world (b), (e), or even the social status of the speaker (c). In any case the described situation is linked to some (textual, metatextual, “outside world”, mental, etc.) entity that lies outside the described situation. This core semantics of werden & infinitive can be represented as an abstract relational structure with two obligatory elements: − −

the reference point (A) – or the starting point, i.e., the entity to which the described event is anchored (not part of situation described by the sentence containing werden) the end point (B), i.e., the event introduced by werden & infinitive (the proposition).

The third element in this abstract semantic structure is the vectored relation between the starting point and the end point: this pure relational component can be defined as a sequential relation. By way of illustration this semantic relational structure may be visualized as follows: A [starting point] > reference point >

sequential relation [werden

> >

B [end point] proposition]

Figure 1. Abstract semantic relational structure of werden & infinitive

In the werden-construction the infinitive (together with its complements) serves as a primary predication and werden functions as a secondary predication adding an evidential value to the proposition (see also Chapter 1.1). The proposed abstract semantic schema of werden & infinitive underlies every instance of use of this construction. This abstract schema has two prototypical realizations: the future temporal reading on the one hand and the evidential reading on the other. The future temporal meaning of werden & infinitive implies the following realization of the proposed schema: the starting point is interpreted as (the speaker’s knowledge about) certain past or present facts. Locating the described situation in the future the speaker relies upon his/her current knowledge about some other facts which are connected with the described event. The connection between known facts and the described event is a sequential relation in terms of temporal development from past/present towards the future. Thus the

Werden & infinitive 173

proposition is linked to the reference point which is construed as the speaker’s knowledge about certain facts which are seen to be temporally developing towards some end point that lies in the future. The evidential reading, on the other hand, represents a different realization of the basic schema: the reference point is interpreted as the speaker’s information source. Saying that the described situation is bound to happen/to have happened the speaker relies upon his/her personal knowledge about some facts which are connected with the described event. The relation between known facts and the described event is an inferential relation between them. In this point the evidential reading of werden & infinitive differs from its temporal reading: the latter represents the temporal sequential relation between the starting and the end point of the basic schema whereas the former represents the logical sequential relation between them. In this manner both temporal and “modal” interpretations of werden & infinitive can be traced back to the same core meaning of this construction. Note that the presence of a relevant reference point – which is conceived of as pieces of knowledge – is an indispensable part of the meaning of werden & infinitive in both cases. This particular characteristic of the reference point justifies the evidential nature of the construction werden & infinitive, as we will argue in the next section.

1.4. What is evidential about the core semantics of werden & infinitive? In the previous section we have argued for a unified view of the semantics of the construction werden & infinitive. We hope to have shown that we can indeed establish a core – though a very schematic – semantic structure of werden & infinitive that underlies its various interpretations. Here we advocate the view that the evidential meaning is inherent to the core semantics of werden & infinitive. As described above werden & infinitive expresses a sequential relation between some reference point and the proposition. That is, the proposition, even in the case of the future reading of werden & infinitive, is construed as a (temporal, logical etc.) consequence from some known facts. Note that if no sequential relation can be assumed the use of werden is normally blocked. So, if the future occurrence of a situation is anchored in calendar time (such as birthdays, public holidays, etc.) and thus cannot be inferred from some other facts, werden cannot be used and the present tense form of the verb is preferred (see Section 6.1.2).

174 The four evidential constructions in present-day German Now, if we look for a grammatical category into which such a function would fit, the category of inferential evidentiality appears to be the most appropriate candidate. It is quite obvious that the sequential relation between some reference point and the proposition can be conceived of as a logical operation carried out by the speaker. As for the reference point which we have only vaguely defined in the abstract semantic schema of werden & infinitive, it can easily be interpreted as the speaker’s information source. Bringing these two observations together – the speakeroriented nature of the reference point and the sequential relation it bears to the proposition – the inferential evidential value of the construction werden & infinitive can be easily deduced. The speaker’s knowledge about some facts serves thus as the information source from which the proposition is inferred, i.e., the information source on which the statement is based. Our view is in line with the conceptions offered by Fritz 2000 and Mortelmans 2004. As has been argued by Diewald 2005 and Smirnova 2006, our conception differs from the others in that we define the speakeroriented, or the speaker-centered, character of werden & infinitive in evidential terms. That is, the inherent link to the speaker is – in our conception – specified as the link to the speaker’s information source. Moreover the notion of a sequential relation established by werden & infinitive is also central in our approach. The semantic component of the sequential relationship whose development may be also observed diachronically for werden & infinitive (see Chapter 8.1 for details) is decisive for the particular – inferential – evidential value of the construction werden & infinitive. The interpretation of the sequential relation in terms of the speaker’s inference process can be seen as a further proof of the development of the inferential evidential function by this construction. By way of illustration let us present some examples from our corpus which are best explained as carrying inferential evidential value, cf.: (32)

(33)

Und ich bin auch immer bestrebt, daß alle, die von Dir kommen, wieder zurückgehen. Du wirst sie schon gut aufheben. (DWDS, Kernkorpus) ‘And I am always anxious that everyone, who comes from you, goes back again. No doubt you will take care of them well.’ Einer solchen kleinen Dame wird es nicht ganz leicht werden, die Richter an ihre Unbescholtenheit glauben zu machen. (DWDS, Kernkorpus)

Werden & infinitive 175

(34)

‘For such a small lady it won’t be quite easy to make the judges believe her respectability.’ Also, die in Stockholm lieben die nicht so hervorragende Ware, aber wir wollen hier nicht über die Stockholmer Akademie reden. Grass wird ihn schon kriegen. (DWDS, Gesprochene Sprache) ‘Thus, those in Stockholm love the cheaper goods but we don’t want to speak here about the Academy of Stockholm. Grass will get it.’

Here the statements formulated with werden & infinitive are all anchored in the speaker’s information source that may be further specified for each example due to the given context: in (32) it is the speaker’s personal knowledge about the person addressed with du ‘you’, in (33) it is part of general knowledge about the situation and people’s behavior, and in (34) it is the specific information pieces about the writer Grass and the current situation. Moreover these statements are derived from known facts via the inference process of the speaker. Werden & infinitive marks the proposition as inferred from the (internalized, known) speaker’s information. This is exactly what inferential evidential elements typically do. Further examples from our corpus may be interpreted in the same way, cf.: (35)

(36)

(37)

(38)

Schon einmal geriet der zuständige Gesundheitsminister unter den erbarmungslosen Druck organisierter Klientelinteressen. Sie werden sich daran erinnern. (DWDS, Gesprochene Sprache) ‘The responsible secretary of health has already got under the pressure of the organized clientele interests. You will surely remember it.’ Als ich gesehen habe, es sind 550 Seiten, habe ich gewusst, es sind hundert zu viel. Das wird er nicht durchhalten. (DWDS, Gesprochene Sprache) ‘When I saw that there were 550 pages I knew (that) there were a hundred too many. He won’t stand it.’ Aber immer wieder erinnert man sich an die Konstruktion und denkt, das kann nicht stimmen. Niemand wird einer solchen Geschichte zuhören. (DWDS, Gesprochene Sprache) ‘But over and over again one remembers the construction and thinks: it cannot be true. Nobody will listen to such a story.’ Da gibt es eine Szene in einem Berliner Kaufhaus. Sie werden sagen, die ist sehr konstruiert. (DWDS, Gesprochene Sprache)

176 The four evidential constructions in present-day German

(39)

(40)

‘There is a scene in a store in Berlin. You will say: it is very artificial.’ In jeder neuen Frau wirst du Entsetzliches erleben. (DWDS, Gesprochene Sprache) ‘In any new woman you will experience terrible things.’ Ja, also, Lieber. Wir werden uns hier heute nicht gegenseitig überzeugen. Möge jeder bei seiner Ansicht bleiben. (DWDS, Gesprochene Sprache) ‘Yes, then, my dear. We won’t convince each other here today. Let’s agree to disagree.’

Note that the temporal location of the described situations is rather irrelevant in these examples. What is being expressed here is the inferential evidential status of the proposition. This means, first, that the speaker bases his/her statement on his/her own knowledge, and, second, that the statement is always represented as resulting, i.e., inferred, from this knowledge. Such – pure evidential – readings show a prototypical interpretation of the abstract semantic schema of the construction werden & infinitive presented in the previous section: the starting point is interpreted as the speaker’s knowledge, i.e., the speaker’s own information source, and the proposition results from this knowledge via the path of inference process. Note furthermore that the evidential status of werden & infinitive, though conceptually very closely related with the notion of epistemic modality, cannot be equaled with any modal epistemic value (for a detailed account see Chapter 3). Werden does not indicate a particular factuality degree of the proposition. Instead, different specific epistemic evaluations may be added to the statement formulated with werden, cf.: (41)

Udo wird jetzt möglicherweise/wahrscheinlich/ganz gewiss in Leipzig arbeiten. (example taken from Dieling 1982: 330) ‘Udo will possibly/probably/certainly be working in Leipzig now.’

It can thus be concluded that werden & infinitive does not attribute any factuality degree to the proposition; it cannot be associated with any particular modal epistemic value and is compatible with various modal epistemic values. Therefore, it does not pertain to the domain of epistemic modality. To summarize, we have shown that the construction werden & infinitive can be best described as an inferential evidential construction. The core

Scheinen & zu-infinitive 177

semantics of werden & infinitive may be described using the paraphrasis introduced in Chapter 2: P, because Q [Origo knows Q, and Origo knows that Q entails P] In other words: the construction werden & infinitive is used if the speaker has access to some information (i.e., personal and/or general knowledge) which s/he interprets as pointing towards the described event. In short: the speaker infers the proposition from premises known to him/her. Hence werden & infinitive exhibits the conceptual inferential evidential value. This is the default realization of its basic semantics.

2. Scheinen & zu-infinitive In this section we deal with the construction scheinen & zu-infinitive in PDG. In doing so, we first briefly outline the common construction types of the verb scheinen in PDG, irrespective of their grammatical status. Next, we focus on the construction scheinen & zu-infinitive and its semantic and structural properties in more detail. This analysis will result in some new crucial semantic as well as morphosyntactic criteria for distinguishing between lexical and evidential scheinen. Finally, we will propose that the construction scheinen & zu-infinitive has two different evidential readings in PDG: the direct perceptual evidential value on the one hand and and the indirect inferential evidential value on the other.

2.1. Relevant construction types with scheinen in present-day German In PDG the verb scheinen displays several different variants. Specific readings of this verb are usually identified with specific construction types. They show different semantic and syntactic features and exhibit different degrees of grammaticalization. Following the suggestions put forward by Askedal 1998 and Diewald 2000, 2001 we summarize the various uses of the verb scheinen in the following manner: 1. 2. 3. 4.

scheinen as a main intransitive verb, scheinen as a copula, scheinen as a compound copula with added infinitive of the verb sein, scheinen with zu-infinitive,

178 The four evidential constructions in present-day German 5. 6.

scheinen with a complement clause, and the parenthetical scheinen.

It should be noted at this point that, synchronically, the variants 3 and 4 are subtypes of the same construction. However, they can be differentiated on diachronic grounds. In order to be able to describe the diachronic development of scheinen in Chapter 8.2, this distinction is set up in this part of the study as well. The above listed variants of scheinen are exemplified in (42) - (47) respectively: (42) (43) (44) (45) (46) (47)

Die Sonne scheint. ‘The sun is shining.’ Sie scheint traurig. ‘She seems sad.’ Sie scheint traurig zu sein. ‘She seems to be sad.’ Sie scheint sich sehr darüber zu freuen. ‘She seems to be very happy about it.’ Es scheint (mir), dass sie sich sehr darüber freut. ‘It seems (to me) that she is very happy about it.’ Sie ist, wie es scheint/so scheint es (mir), sehr traurig. ‘She is, (as) it seems (to me), very sad.’

Scheinen as a main intransitive verb has the meaning ‘to shine, to give off light’ and belongs to a large group of verbs denoting visual effects. It does not have intentional semantics and is not restricted to animate subjects. This basic meaning of visual effect has remained stable since OHG. Scheinen as a copula verb is syntactically associated with the other German copula verbs sein ‘to be’, werden ‘to become’, and bleiben ‘to remain’, and occurs with various types of subjects and predicative expressions. (48)

Sie ist/bleibt/wird/scheint traurig. ‘She is/remains/becomes/seems sad.’

Scheinen as a copula has the meaning ‘to be seemingly, to appear (as)’. In this meaning scheinen already contains an evidential meaning component, as it describes not only a visual effect produced by a subject (like Sonne

Scheinen & zu-infinitive 179

‘sun’ or Mond ‘moon’) but rather an impression which is caused by the subject referent by means of its visual (or other) appearance. There is usually an observer who experiences this particular impression. Hence the copula verb scheinen often occurs with an additional (free) dative complement referring to the observer/experiencer. The third variant of the verb scheinen is associated with its uses as “compound” copula with the added zu-infinitive of sein ‘to be’. This variant of scheinen is semantically very close to its second type. Constructions of this type are very frequently found in our corpus data. The structural difference between this variant of scheinen and its uses as a copula verb is very important because the connection of scheinen with the infinitive of sein most likely made it possible for further infinitives to be connected with scheinen. This is among others a crucial step for the initialization of the grammaticalization process of scheinen towards an evidential auxiliary. This process will be discussed in detail in Chapter 8.2. The fourth variant of scheinen, which is found in constructions with zuinfinitives of any verb, represents the highly grammaticalized variant of this verb (see Chapter 4). Mostly the meaning of scheinen & zu-infinitive is evidential, i.e., it indicates the information source for the described event, cf.: (49)

(50)

Er scheint zu hoffen, daß ich mit den Kindern in den Tod gehe… (DWDS, Kernkorpus) ‘He seems to hope that I and the children will meet our deaths…’ Dieser Roman scheint doch deutlich autobiographische Züge des Autors zu haben. (DWDS, Gesprochene Sprache) ‘This novel seems to have clear autobiographical traits of the author.’

This variant of scheinen will be analyzed in more detail in the next sections. The fifth variant of scheinen is found in syntactic constructions containing complement clauses with dass ‘that’ and als ob ‘as if’, as exemplified in (46). We assume that this syntactic use of scheinen is semantically very closely connected to its other uses, particularly to the second and the third variants described above. What scheinen seems to express in constructions with complement clauses is the impression which is produced (on some observer) by some entity or situation. The impression may be more or less concrete, and it may be more or less plausible. Every actual interpretation

180 The four evidential constructions in present-day German of the evaluative and epistemic status of the represented impression is supported by further contextual devices and does not arise from the verb scheinen alone. So, for instance, if the complement clause is formulated in the subjunctive mood, the expressed proposition is considered to be nonfactual and thus less plausible. The so-called parenthetical scheinen (i.e., syntactically not integrated into the sentence as exemplified in (47)) is treated here as a separate variant. Though semantically very close to the other types, especially types 2, 3 and 4, this variant of scheinen differs considerably from other types with respect to its distributional properties. The construction types with scheinen represented here so far have been distinguished on the basis of their syntactic properties. As for the semantic nature of the established types, three main semantic variants of scheinen may be distinguished at this stage: 1. 2. 3.

the meaning of visual effect, i.e. ‘to shine, to give off light’, including construction type 1; the meaning of (visual) impression, i.e. ‘to be seemingly, to appear (as)’, including construction types 2, 3, 5 and 6; and the evidential semantics of scheinen (construction type 4).

2.2. Semantic diversity of scheinen & zu-infinitive The construction scheinen & zu-infinitive has been shortly described in Chapter 3 as indicating the concenptual/perceptual inferential evidential value of the proposition. This meaning represents the distinctive value of scheinen & zu-infinitive within the system of German evidential markers (for further details see Chapter 7.2.2). This meaning is exemplified in the following examples from our corpus: (51)

(52)

Man scheint nämlich vollkommen vergessen zu haben, dass … (DWDS, Gesprochene Sprache) ‘One seems to have totally forgotten that… Aber die geheimen Vorbereitungen zu Putschen, wie sie die Nationalsozialisten jetzt zu beginnen scheinen, können wir unter keinen Umständen dulden. (DWDS, Kernkorpus) ‘We can under no circumstances tolerate the secret putsch preparations which the National Socialists now seem to be initiating.’

Scheinen & zu-infinitive 181

Here scheinen indicates that there are some information pieces available to the speaker on which s/he bases his/her statements. The nature of information is not explicated in text: only the fact that the proposition is based upon the speaker’s inferences is marked by the infinitive construction with scheinen. In the following examples the evidence available to the speaker is explicitly mentioned in the text: (53)

(54)

Fritz hat hier und da von sich hören lassen. Es scheint ihm gut zu gehen. (DWDS, Kernkorpus) ‘Fritz has been in touch here and there. He seems to be doing well.’ Sie scheint mit ihm Orgien zu feiern, das belastende Material liegt vor. (DWDS, Gesprochene Sprache) ‘She seems to celebrate orgies with him, we have the incriminating material at our disposal.’

From these examples it is obvious that the evidence, i.e., pieces of information on which the statements are based, can belong to different types: the information may be conceived of as a testified fact, as in (54), or as learned from others, as in (53). Some further examples from our corpus provide evidence that the information pieces being referred to by scheinen & zuinfinitive are not restricted to facts known to or heard by the speaker, but can comprise personal knowledge of the speaker as well, cf.: (55)

(56)

Ich glaube, manches Stück von ihm könnte man retten durch völlig neue Inszenierungen, aber dem scheinen sich ja die Erben Brechts, soviel ich weiß, zu widersetzen. (DWDS, Gesprochene Sprache) ‘I think some of his plays could be rescued by staging completely new productions, but Brecht’s the administrators of Brecht’s estate seem to be set against that, as far as I know.’ Der Herr Minister wird dieses Problem sicherlich einer Untersuchung zuführen und er scheint, so weit ich im Bilde bin, bereits irgendwelche Wege gefunden zu haben. (DWDS, Gesprochene Sprache) ‘The Minister will surely guarantee an investigation of this problem and he seems, as far as I am in the know, already have found some ways to do it.’

Here the evidence is not specified as certain facts but is only referred to as not further specified knowledge of the speaker. This demonstrates that the

182 The four evidential constructions in present-day German situations introduced by scheinen & zu-infinitive may be inferred from pieces of the speaker’s (personal) knowledge (as it is the case with the conceptual inferential construction werden & infinitive) as well as from information learned from others. The spectrum of types of evidence which scheinen & zu-infinitive may refer to in particular contexts reaches thus from irrefutable facts to unspecified pieces of the speaker’s subjective knowledge. Moreover, scheinen is very often used in situations where direct (visual) information serves as the basis for the statement (see Section 6.2.4). For this reason we classify scheinen & zu-infinitive as an inferential evidential construction that combines perceptual and conceptual inferential values. This characteristic of scheinen & zu-infinitive reflects its diachronic development (see Chapter 8.2 for further details). A brief note on epistemic modal nuances of scheinen is in place here. In our synchronic corpus material we found a number of examples in which scheinen co-occurs with expressions indicating the strong speaker’s commitment to the truth of the proposition, cf.: (57)

(58)

(59)

Also scheinen unsere Behauptungen, die wir ja auch beweisen können, unter allen Umständen zu stimmen. (DWDS, Gesprochene Sprache) ‘Our statements then, which we also can prove, seem to be right whichever way you look at it.’ Denn in der Tat scheint es so zu sein, dass es ein schweres Trauma gegeben hat… (DWDS, Gesprochene Sprache) ‘For indeed it seems to be that there was a severe trauma.’ Ich sage das noch einmal; man kann es nämlich nicht oft genug sagen; denn es scheint offensichtlich noch nicht angekommen zu sein. (DWDS, Gesprochene Sprache) ‘I will say it one more time; one cannot say it often enough; for, obviously, it seems not to have been registered properly.’

Remarkable in this connection is that we have not found any examples in which scheinen co-occurs with epistemic modal adverbials like vielleicht ‘possibly’, vermutlich ‘supposably, assumingly’, sicher ‘surely’ etc. Bearing in mind that scheinen has been often described as a verb expressing epistemic modality (see Chapter 3) one would expect to detect such combinations. The German epistemically used modal verbs, on the contrary, often allow such combinations:

Scheinen & zu-infinitive 183

(60)

(61)

Das könnte vielleicht klappen. (www.artechock.de/film/text/interview/h/hader_2005.htm, accessed 18.08.2007) ‘This might possibly work out.’ Dich müsste ich sicher schon gesehen haben. (www.dsdsnews.de/27-andreas-eitel, accessed 18.08.2007) ‘I surely must have seen you before.’

This fact supports our claim that scheinen does not have any inherent epistemic modal semantic components and that such components have the status of (generalized) conversational implicatures (see Chapter 3 for more details). In our corpus material we detected a number of instances of the construction scheinen & zu-infinitive which cannot be considered evidential. Leaving aside cases where the construction can be interpreted as direct – visual – evidential marker (for this see Section 6.2.4), we focus here on some uses which, as we think, may be best explained as a persistence of certain meaning components of the original – lexical – semantics of scheinen. The phenomenon of persistence in grammaticalization was described by Hopper 1990, 1991: When a form undergoes grammaticization from a lexical to a grammatical function, so long as it is grammatically viable, some traces of its original lexical meanings tend to adhere to it, and details of its lexical history may be reflected in constraints on its grammatical distribution. (Hopper 1991: 22)

In short, we are talking about persistence when a grammaticalized linguistic unit displays distributional, semantic and structural restrictions which are due to some preserved feature of its original lexical source item. For scheinen, most of the ambiguous interpretations of the infinitive construction we have found in our corpus may be explained in terms of persistence effects. Consider the following examples: (62)

Der Teppichboden gibt hier den neutralen Untergrund ab, auf dem sich die Rot-Töne entfalten können. Selbst etwas schwerere Möbel scheinen bei dieser Zusammenstellung zu “schweben“. (DWDS, Kernkorpus) ‘The wall-to-wall carpeting serves here as a neutral background against which the shades of red can unfold. Even the slightly heavier pieces of furniture seem to “hover” in this configuration.’

184 The four evidential constructions in present-day German Syntactically, the verb scheinen occurs within infinitive construction which we described above mainly as an evidential construction. Semantically, however, the meaning of scheinen & zu-infinitive in the example above is not evidential: scheinen denotes the imagined, non-real status of the described situation. It introduces situations which directly contradict reality, and this is obvious from the surrounding context. In such cases scheinen shows the preservation of the meaning component of its lexical source: the highlighting of the non-consistence between the appearance of an object/event and the real nature of this object/event. This meaning component, which we call “make-believe” meaning, may be formulated in short by the following formula: scheinen ‘seem, appear’  sein ‘be’; it is mostly present in the lexical constructions of type 2 and 3 described in Section 6.2.1. Thus, scheinen highlights the opposition between the appearance of an entity and its real nature. There is a second type of infinitive constructions with scheinen which is frequently found in our corpus and which cannot be easily subsumed under the evidential meaning of scheinen. Scheinen cannot be interpreted as an evidential marker in such situations: it denotes the perspective (often the visual perspective) of an observer who is part of the described scene. This interpretation of scheinen & zu-infinitive may be exemplified by the following examples from our corpus: (63)

(64)

Die letzten Hoffnungen scheinen in den Augen der englischen Politiker … auf neuen Komplikationen zu beruhen (DWDS, Gesprochene Sprache) ‘Seen through the eyes of the English politicians the last hopes seem to rest on new complications.’ Dahinter aber scheint mir eine ideologische Strategie zu stehen, die mir einerseits zu erklären scheint, warum dieses Buch im Moment so erfolgreich ist… (DWDS, Gesprochene Sprache) ‘An ideological strategy seems (to me) to be behind it, which on the one hand seems (to me) to explain why this book is so successful at the moment…’

For the explanation of such uses of the construction scheinen & zuinfinitive we apply the notion of persistence again. The meaning of personal attitude can be traced back to the lexical meaning of personal impression which is again very strong in the copula and copula-like uses of scheinen:

Scheinen & zu-infinitive 185

(65)

(66)

DER BARON - erhebt sich, der Moment für den ersten Kuß scheint ihm gekommen. (DWDS, Kernkorpus) ‘The baron stands up, the moment for the first kiss seems to him to have come.’ er schien ihr sehr gut bekannt (DWDS, Kernkorpus) ‘She seemed to know him very well.’

In the examples above scheinen renders the impression the situationinternal observer has of an entity, which usually is based on perceptual (often visual) clues. The situation-internal observer is often encoded by a free dative, a structural possibility, which, too, has persisted through all the history of scheinen (see Section 8.2). Furthermore, it should be noted that this “subjective, observer-based” meaning and the evidential meaning of scheinen are examples of the semantic relationship of gradience or polysemy respectively (see Chapter 5), i.e., they belong into the domain of ambiguity within lexical meaning. It should be noted however that both meaning components described above are mostly realized in certain contexts, i.e., they arise in certain contextual configurations which trigger a particular interpretation of the infinitive construction diverging from the evidential one. Thus, the “makebelieve” interpretation of the proposition with scheinen arises typically in contexts where the opposition to the actual, or the “real”, state of affairs is obvious. The function of the construction scheinen & zu-infinitive, which we have described as “subjective” and observer-based, on the other hand, is mostly limited to such cases where the personal involvement of an observer is explicitly referred to by the free dative (mir ‘(to) me’, ihnen ‘(to) them’, ihm ‘(to) him’ etc.). Hence both of the interpretations of scheinen & zu-infinitive that question its evidential status in PDG can be easily revealed as cases of the persistence mechanism typical of every grammaticalization process.

2.3. Lexical versus evidential scheinen: some (more) crucial differences In this section some more relevant points revealing the categorical divergence between lexical and grammatical scheinen will be addressed. The following observations will show that there is a clear difference between scheinen as a lexical verb and scheinen as a grammatical evidential marker.

186 The four evidential constructions in present-day German As we have pointed out in Chapters 1 and 2, German evidential verbal constructions qualify as secondary predications. This means that they convey additional information to the proposition, which is the primary predication. Together with the proposition they build a complex predication consisting of two predications: the proposition centered around the infinitive verb, on the one hand, and the evidential qualification of the proposition expressed by the evidential auxiliary, on the other. So, every utterance entailing an evidential marker may be split up in two separate parts: the proposition or the factual claim (cf. Anderson 1986) and the evidential status of this proposition. This may be explicated in the following way: (67)

Dieser Roman scheint doch deutlich autobiographische Züge des Autors zu haben. (DWDS, Gesprochene Sprache) ‘This novel seems to have clear autobiograpfical traits of the author.’

proposition (P): Dieser Roman hat deutlich autobiographische Züge des Autors. evidential status/secondary predication: Origo (i.e., the speaker) infers P from known facts (Q) Similar analysis may be conducted for every instance of the evidential construction scheinen & zu-infinitive, as well as for the other constructions with werden, drohen and versprechen. In Section 6.1 we used the werden-construction to exemplify that the basic semantic template of evidential constructions is a relational structure linking a source to an endpoint. This, too, applies to all four constructions treated here, which, however, for reasons of space is explicated in detail only for werden. As long as scheinen functions as a main verb, including its copula and a number of its copula-like uses, it forms a main (and only one) predication, i.e., scheinen is the essential part of the proposition: (68)

Heute scheint auch die Sonne mal wieder. (DWDS, Kernkorpus) ‘Today the sun is shining once again.’

proposition: paraphrase:

Heute scheint auch die Sonne mal wieder. Heute strahlt auch die Sonne mal wieder. ‘Today the sun is gleaming once again.’

Scheinen & zu-infinitive 187

For copula uses of scheinen the same situation holds. For copula-like uses with the added zu-infinitive of the verb sein, the situation is more complicated since in most cases the ambiguous interpretation between more lexical copula use and more grammaticalized evidential use is possible. The semantic ambiguity originates from the structural ambiguity of such constructions: on the one hand, they resemble copula uses of scheinen in that they can be “reduced” to them: Sie scheint krank zu sein > Sie scheint krank; on the other hand, they resemble evidential uses of scheinen insofar as they are infinitival construction of the form scheinen & zu-infinitive. For this reason copula-like constructions with scheinen are often ambiguous between more lexical – copula – readings of scheinen and more grammatical – evidential – readings of this verb, cf.: (69)

… allerdings schien diesmal das Material noch besser zu sein als sonst. (DWDS, Kernkorpus) ‘… this time the material indeed seemed to be better than usual.’

lexical reading: proposition: paraphrase:

Das Material scheint diesmal noch besser als sonst. Das Material sieht diesmal besser aus als sonst. ‘This time the material looks better than usual.’

grammatical reading: proposition (P): Das Material ist diesmal besser als sonst. evidential status/secondary predication: Origo (i.e. the speaker) infers P is from some accessible evidences (Q) Often, the context conveys additional information which supports one or another interpretation, cf.: (70)

Gaigern scheint hier draußen Stammgast zu sein, jedermann grüßt und kennt ihn. (DWDS, Kernkorpus) ‘Gaigern seems to be a regular guest here, everyone greets and knows him.’

In this example the evidential interpretation of scheinen is most appropriate one, since the context gives exact information on which the statement is based, i.e., it is an observation of the speaker that everyone greets Gaigern.

188 The four evidential constructions in present-day German Thus scheinen serves here as an inferential evidential marker and should be seen as a secondary predication, cf.: proposition (P): Gaigern ist hier draußen Stammgast. ‘Gaigern is a regular guest here.’ evidential status/secondary predication: Origo (i.e. the speaker) infers P from Q [jedermann grüßt und kennt ihn] Some formal characteristics of lexical vs. grammatical uses of scheinen may be helpful for the interpretation of ambiguous readings. Scheinen as a copula may in turn occur as an infinitive complement of other auxiliaries. Evidential scheinen, on the contrary, does not occur as infinitival complement of other auxiliaries. Furthermore, in copula-like constructions, lexical scheinen is often accompanied by adverbial complements. In such cases it will be obvious that, if the verbal character of scheinen is specified, scheinen cannot be considered to be an auxiliary, but only a main verb forming the main proposition of the sentence. The most frequent adverbial complements are heute ‘today’, jetzt ‘now’, damals ‘at that time’, auf den ersten Blick ‘at first glance’, nun ‘now, then’ etc. For evidential scheinen, however, it is impossible to add such adverbial complements: scheinen as an evidential auxiliary serves itself to characterize the verbal event which is introduced by the infinitival complement. Since scheinen has auxiliary status, it cannot be further specified. One important observation on lexical versus evidential scheinen deserves further attention here. We have found out that lexical scheinen does not normally occur with first and second person subjects. This restriction is implied in the core lexical semantics of this verb: a visual effect typically comes from an entity which is distinct from the observer (and from the second person which is a part of the communicative situation). Lexical uses with the second person subjects are not fully absent, but they are very rare. First person subjects are totally absent in lexical readings of scheinen. Lexical scheinen does not have intentional semantics, i.e., it refers to the situations where an effect or an impression is not intentionally caused by the entity under consideration. Thus the uses of scheinen with first person are blocked. In this point scheinen differs significantly from such verbs like vorgeben ‘pretend’, täuschen ‘delude’ and others which do have inten-

Scheinen & zu-infinitive 189

tional semantics. In its evidential uses, however, scheinen does occur with first person subjects, cf.: (71)

Ich scheine mich in der Tür geirrt zu haben. (DWDS, Gesprochene Sprache) ‘I seem to have taken the wrong door.’

That scheinen does not have any syntactic restrictions with regard to the selection of its subjects anymore, is a convincing evidence for the high degree of auxiliarization of evidential scheinen. It allows for combinations with the first person subjects marking explicitly that the speaker’s statement about himself/herself is based upon inference and is not made directly. In sum, in this section we have listed some further points which provide strong evidence for scheinen to be treated as a grammaticalized evidential marker in PDG. Relying on these observations we claim that scheinen has two general readings: the lexical reading with the variety of semantic interpretations in specific constructions and the grammatical – evidential – reading which is restricted to infinitive constructions which serve as secondary predications.

2.4. Two types of evidential scheinen and their conceptual relationship As we have mentioned in the previous sections scheinen & zu-infinitive may be attributed two distinct evidential readings in PDG. On the one hand, this construction serves as a perceptual/conceptual inferential evidential marker. On the other hand, it is used to express direct – mostly visual – evidential value. The two meanings may be paraphrased as follows: perceptual/conceptual inferential value of scheinen & zu-infinitive: P (proposition), because Q1: Origo (i.e. the speaker) can observe some signs/traces of P OR because Q2: Origo (i.e. the speaker) knows Q, and Origo knows that Q entails P

direct evidential value of scheinen & zu-infinitive: P (proposition) and Q: Origo (i.e. the speaker) sees P

190 The four evidential constructions in present-day German These two readings usually arise in particular contexts, so they can be easily distinguished from each other. Since a number of examples already have been given for the former reading of the construction (see previous sections) the following examples from our corpus illustrate only the direct evidential reading of scheinen & zu-infinitive: (72)

(73)

Jetzt will der Arm über den Kopf hinaus, um etwas zu verscheuchen ... jetzt krümmt er sich und scheint ein Gefäß umklammern zu wollen. (DWDS, Kernkorpus) ‘Now the arm is trying to move overhead to shoo something away … now it is bending and seems to clasp a vessel.’ Die blassen Wangen rötheten sich, auf der kahlen Stirn schwollen die Adern, die Zähne schienen sich in einander zu beißen, als wollten sie einen Schmerzlaut zurückhalten. (DWDS, Kernkorpus) ‘The pale cheeks turned red, the veins swelled on the bald forehead, the teeth seemed to clench as if they were trying to hold off a sound of pain.’

In these examples scheinen can only be given direct evidential interpretation. It indicates that the speaker observes directly – visually – the described event. The described situation P coincides with the time of perceiving the situation Q: scheinen & zu-infinitive marks that the ongoing situation is directly observed by the speaker. The information that the situation is perceived directly at the time of speaking is added to the proposition describing the occurring event. Thus this information expressed by scheinen & zu-infinitive can be conceived of as a secondary predication. Note that the described situation P and the fact of seeing (perceiving) this situation Q always coincide, i.e. P = Q. We have already mentioned in Chapter 2 that this is the defining property of direct evidentials (understood as deictic expressions). We can see that this property is present in the above mentioned uses of scheinen. One important empirical observation proves this fact: scheinen as a direct evidential occurs in past tense (as in (73)). Due to the past occurrence of the described event the evidential specification thereof is also given in the corresponding tense. The coincidence property has its formal representation on the linguistic level, in that it is made transparent by the tense of the evidential construction. Scheinen & zu-infinitive in its uses as indirect inferential construction, on the other hand, shows the deictic ‘separation’ between the evidential source Q and the described situation P. As we have found out, the inferen-

Drohen & zu-infinitive 191

tial scheinen-construction does not usually occur in the past tense. The particular location of the described situation in time is thus marked by the infinitive itself. Therefore, a clear distinction of two different – direct and indirect – evidential values of the construction scheinen & zu-infinitive may be retraced not only semantically but morphosyntactically as well. As for the relationship between two described evidential readings it can be stated that the direct evidential value of scheinen is much more closely connected to the lexical reading of the main verb scheinen. The component of visual perception is still preserved in the meaning of the direct evidential construction scheinen & zu-infinitive. The inferential evidential construction, on the other hand, is semantically more bleached. Therefore we can speak of two different readings of evidential scheinen which are distinguished semantically as well as formally. The differences between two types of evidential constructions point to the differences in their degrees of grammaticalization. Direct evidential scheinen shows a lesser degree of grammaticalization and can be seen as an earlier stage in the grammaticalization of scheinen towards indirect inferential evidential marker.

3. Drohen & zu-infinitive The infinitive constructions with drohen and versprechen share the same evidential value in PDG; they both serve as perceptual inferential constructions. In this function they indicate that the described situation is inferred by the speaker on the basis of some presently accessible information. The information source is thus specified. On the one hand, the proposition is marked as inferred, i.e., acquired through the inference process, and this is the primary – indirect – evidential meaning of both drohen and versprechen & zu-infinitive. On the other hand, the inference process is represented as resulting from some presently accessible information. Thus the pieces of information on which the inference process is based are normally perceived at the time of speaking and are closely linked (or even identified) with the described situation/object and, furthermore, they can be easily recognized by the hearer. In this respect the infinitive constructions with drohen and versprechen differ from the constructions with scheinen and werden described in previous sections. In this section we will concentrate on the uses of the construction drohen & zu-infinitive in PDG. More fine-grained semantic descriptions of the construction readings in specific syntactic and contextual structures will be

192 The four evidential constructions in present-day German discussed in Section 6.3.3. The synchronic analyses of this construction (e.g. Askedal 1997, Gunkel 2000, Kokutani 2004, Reis 2005) concentrate to a large extent on finding differences, not similarities between various meanings and uses of the verb drohen. We are however much more interested in discovering semantic features all uses of drohen have in common. By revealing conceptual commonalities between various readings of drohen we attempt to identify the functional status of the construction drohen & zu-infinitive within the evolving evidential paradigm of German.

3.1. Relevant construction types of drohen in present-day German In PDG the verb drohen shows several different readings. The majority of German linguists dealing with this topic posit two distinct readings of the verb drohen: the lexical reading on the one hand and the so called modal, or epistemic, reading on the other (cf. Askedal 1997, Zifonun et al. 1997, Eisenberg 2004, Gunkel 2000, Kokutani 2000 et al.). As has been suggested by Heine and Miyashita 2004 and 2008, the verb drohen is associated with the following four relevant constructions (the examples (74) to (77) are taken from Heine and Miyashita 2008): (74) (75) (76) (77)

Karl droht seinem Chef, ihn zu verklagen. ‘Karl threatens his boss with a lawsuit.’ Uns droht nun eine Katastrophe. ‘We are now threatened by a disaster.’ Das Hochwasser droht die Altstadt zu überschwemmen. ‘The high water threatens to flood the old town.’ Mein Mann droht krank zu werden. ‘My husband threatens to fall ill.’

According to Heine and Miyashita 2008 the verb drohen in (74) is a lexical main verb, while (76) “tends to be described as a raising construction, more specifically a subject-to-subject raising construction” (Heine and Miyashita 2008: 56). After having classified the (syntactic) constructions of drohen into a four types the authors introduce a twofold (functional) classification of the drohen-variants. These are said to be the lexical main verb drohen (as in (74)) and the so-called functional verb drohen that is usually referred to as the epistemic, subjective, modal, semi-modal, or grammatical verb (exemplified in (75) to (77)). The following semantic

Drohen & zu-infinitive 193

paraphrases are given for the two types of drohen, respectively (Heine and Miyashita 2008: 56): lexical drohen: ‘Someone points out that s/he intends to do something that is undesirable to someone else.’ functional drohen: ‘Something undesirable is about to happen.’ In a similar way Reis 2005 distinguishes between two variants of drohen in PDG. She terms them ‘agentive’ versus ‘non-agentive’ readings of the verb and illustrates them by the following examples (examples together with paraphrases are taken from Reis 2005: 125): (78)

Paul droht den Hund umzubringen. a. ‚Paul verpflichtet sich (gegenüber jemandem), den Hund umzubringen (falls nicht…).‘ ‘Paul is committing himself to kill the dog (if not…).’ b. ‚Paul bringt, so wie es jetzt aussieht, den Hund um.‘ ‘Paul is killing the dog, as it seems now.’

The first variant of drohen is said to be the lexical verb. The second variant is classified as an obligatorily coherent raising-verb. Reis 2005 rejects the notion of an epistemic or otherwise modal-like status of the second variant of drohen and argues instead for treating it as a temporal-aspectual verb. The fact that drohen occurs in two distinct readings in PDG is not controversial. What is controversial, however, is the status of the second variant of drohen. Should it be classified semantically as a modal, semimodal, epistemic, subjective, aspectual, or temporal verb? Should it be seen syntactically as an auxiliary, semi-auxiliary or raising verb? We will deal with such questions in the next sections and will put forward the hypothesis that the second variant of drohen is best described as a relatively highly grammaticalized auxiliary within the functional domain of (inferential) evidentiality. It should be noted, first, that we absolutely agree with the commonly acknowledged classification of drohen that distinguishes between its two variants. On the basis of our own corpus analysis we suggest that there are (at least) three different readings of the verb drohen:

194 The four evidential constructions in present-day German 1. 2. 3.

drohen as a (lexical) verb with agentive semantics; drohen as a (lexical) verb with non-agentive semantics; drohen as a grammatical (auxiliary or semi-auxiliary) verb with evidential semantics.

From the three types listed above the first two correspond to what have been called “lexical” drohen, and only the last type represents the grammatical, or functional, variant of the verb. Note that we qualify nonagentive uses of drohen like in (75) lexical, whereas Heine and Miyashita 2008 identify them as belonging to the functional variant of this verb. Although the meaning of drohen in (75) may be paraphrased as ‘Something undesirable is likely to happen soon’, there are some relevant formal characteristics which do not allow for functional – or grammatical – classification of this drohen-type. For example, this variant of drohen does not combine with zu-infinitives (which would be most expectable for an auxiliarylike verb), it does exhibit a valency relation to the subject of the sentence, it does allow for (free) dative complements, it does allow for adverbial modification, it may be itself a gerundival attribute, etc. Since it shares these relevant properties with the first – lexical – drohen, it cannot be considered to be functional. Rather, it is a variant of lexical drohen with distinct semantics. We will come to this in more detail below. Drohen as a main verb with agentive semantics shows the highest variability with regard to its combinatorial possibilities. Its main distinctive feature is that it normally occurs with animate subjects (and is not restricted to the 3rd person referents). It occurs with infinitives, with complement clauses and with direct speech complements, it can further take a dative participant and can be modified by adverbial complements or dative complements with instrumental semantics. Drohen has the basic meaning ‘to threaten’ and has a strong intentional component which explains the fact that this variant of drohen – with rare exceptions – is restricted to human subjects. Furthermore, it is a speech act verb without inherent performative semantics. The second variant of drohen with non-agentive semantics is usually restricted to non-animate subjects (with negative semantics). The most frequent subject is Gefahr ‘danger’, the combination of drohen and Gefahr has nearly acquired the idiomatic status in PDG. Other subjects with negative semantics like Krieg ‘war’, Verlust ‘loss’, Untergang ‘fall’, Ausrottung ‘extinction’, Strafverfahren ‘criminal procedure’, Katastrophe ‘catastrophe’ etc. are usually found in combination with this variant of drohen. The crucial semantic difference to the first type of drohen is that an inten-

Drohen & zu-infinitive 195

tional component can hardly be attributed to the non-agentive drohen. This variant of drohen also takes dative participants (as mir ‘to me’, uns ‘to us’), it can be modified by adverbial complements and it shows the whole range of inflectional forms, including periphrastic constructions. Note also that this type of drohen is mostly restricted to the 3rd person (inanimate) subjects. Drohen as a (relatively highly grammaticalized) evidential auxiliary is restricted to the infinitive constructions, cf.: (79)

(80)

(81)

Wenn man also merkt, daß eine Diskussion (etwa über Politik) in einen Streit auszuarten droht… (DWDS, Kernkorpus) ‘When one realizes that a discussion (for example about politics) threatens to degenerate into a conflict…’ Wenn ein Haus aus den Nähten zu platzen droht. (DWDS, Kernkorpus) ‘When a house threatens to burst at the seams.’ Sie bestehe aus Mitarbeitern JELZINs, einer mächtigen Korporation unberechenbarer Leute aus seiner Umgebung, die das Land in die Katastrophe zu führen drohten. (DWDS, Gesprochene Sprache) ‘It consists of Jelzin’s collaborators, a powerful corporation of unpredictable people around him which threatens to lead the country into a catastrophe.’

As these examples from our corpus illustrate, the grammatical variant of drohen shows no restrictions in regard to the subject of the sentence: it can be either a human/animate or an inanimate subject. The meaning of this drohen-variant can be paraphrased as follows: The speaker has (had) perceptual access to certain pieces of information which s/he interprets as pointing towards the described event (which is evaluated negatively from the speaker’s point of view). P (proposition), because Q: Origo (i.e. the speaker) can observe some signs/traces of P; P is undesirable

The semantic difference between the lexical and the grammatical variants of drohen is made clear in this paraphrase:

196 The four evidential constructions in present-day German 1. 2. 3.

while the lexical reading does not imply any connection to the speaker, the grammatical reading is always speaker-oriented, i.e., it is deictic; while lexical drohen refers to an (undesirable) situation which is represented by the subject of the sentence, grammatical drohen serves as a secondary predication; while lexical drohen does not entail any indication about the information source (instead, it only denotes that something negative is going to happen), grammatical drohen always relates the proposition to the information pieces from which the proposition is inferred by the speaker.

3.2. Lexical versus evidential drohen In this section we are primarily concerned with the infinitive constructions with drohen. From what we have described so far the occurrence of drohen within infinitive constructions cannot be regarded as a defining property for the evidential, i.e. grammatical variant of this verb. In contrast to werden, which appears with infinitival complements only in its grammatical reading, drohen (as well as versprechen and scheinen) within infinitive constructions may serve both as lexical verb (82) and as grammatical auxiliary (83), cf.: (82)

(83)

Die Behörden drohten jedoch, Nadeshda Krupskaja in einem solchen Fall in das Gouvernement Ufa auszuweisen. (DWDS, Kernkorpus) ‘The authorities threatened however to expel Nadeshda Krupskaja to the Governorate of Ufa, should this be the case.’ Wenn ein Haus aus den Nähten zu platzen droht. (DWDS, Kernkorpus) ‘When a house threatens to burst at the seams.’

Nevertheless, in most instances, these two readings of drohen can be differentiated. In the following we are dealing with those factors that usually serve as disambiguating devices. From these observations we hope to be able to draw some important conclusions, which concern the relevant properties of the contexts triggering one of the two distinct readings and the relevant properties of grammatical (as opposed to lexical) drohen. From there we hope to discover changes occurred to drohen on the way from lexical verb to evolving evidential auxiliary.

Drohen & zu-infinitive 197

The characteristics of drohen and its immediate constituents have been extensively investigated in the last few years (e.g. Askedal 1997, Gunkel 2000, Heine and Miyashita 2004, 2008, Kokutani 2004, Diewald 2004, Reis 2005 et al.). We first summarize the relevant factors focusing on the characteristics of evidential drohen as opposed to lexical drohen within infinitive constructions: 1. 2. 3. 4. 5. 6. 7. 8.

there are no restrictions with regard to the animacy and intentionality of the subject referent (subject may be impersonal, and a complement clause); there is no semantic relation to the subject, instead, drohen modifies the whole proposition (i.e., it has the status of the secondary predication); drohen has a limited inflectional paradigm (no periphrastic forms, no infinitives, no imperatives, no passive forms); extraposition of the infinitive complement is not possible; pronominalization of the infinitive complement is not possible; adverbial modification of drohen is not possible; clausal complements are not possible (dass-, damit-clauses, indirect and direct speech complement clauses); drohen cannot be topicalized and nominalized.

As we have described in Chapter 4, all these parameters reveal high degree of grammaticalization of the functional variant of drohen in contrast to its lexical counterpart(s). In addition to the factors listed above there are some further points which have been mentioned in the literature and are mainly concerned with the semantics of infinitive verbs. The nature of the verbal process introduced by the infinitival complement has been considered very important for the disambiguating between two distinct readings of drohen. Gunkel (2000: 116) formulates the following restriction for the functional variant (which he calls ‘modal’): “Modales drohen/versprechen kann keine Verben einbetten, die eine Handlung bezeichnen” [‘The modal variant of drohen/versprechen cannot combine with infinitive verbs which denote an action’]. He admits however that drohen in its ‘modal’ reading occurs with infinitive verbs designating intentional events (examples taken from Gunkel 2000: 117): (84)

Die feindlichen Truppen drohen die Stadt zu erobern. ‘The enemy troops threaten to conquer the town.’

198 The four evidential constructions in present-day German (85)

Die Schwiegermutter drohte uns zu besuchen. ‘The mother-in-law threatened to visit us.’

Gunkel 2000 rightly points out that the subject role is not conceptualized as an agens in such cases. Instead the holistic view of the event is taken: the situation is described as a whole (we will return to this further below). Therefore the situation is not seen as a process which is controlled and fulfilled by an agens but rather as an event which happens due to its own dynamics. Gunkel (2000: 118) proposes a subdivision of events into intentional versus non-intentional events. Thereby the mechanism of the socalled ‘sort shift’ is proposed, due to which verbal semantics may be ‘shifted’ between intentional and non-intentional. This attempt is rather problematic as it starts from the assumption that there is only one variant of drohen and shows how the infinitives trigger different readings of this verb. It fails however to explain under which circumstances and due to which factors an ‘event’-verb has the ‘intentional’ or ‘non-intentional’ interpretation. We argue here for another view on the ambiguity of drohen & zuinfinitive between lexical and grammatical readings. In place of searching for factors determining one or another reading of the construction in particular communicative contexts, which are undoubtedly very important for the particular interpretation of the construction in use, we concentrate on the grammatical – as arising from the grammaticalization process – nature of the evidential construction drohen & zu-infinitive. In our view the ability of evidential drohen to combine with animate subjects and infinitive verbs denoting intentional events/processes and even actions is due to its high degree of grammaticalization. In other words, the occurrence of drohen & zu-infinitive with infinitive verbs denoting actions and intentional events should be seen as a consequence of the increasing degree of grammaticalization of this construction. That is, drohen as a lexical verb usually combines with verbs denoting an action. The verbal process of the infinitive verb is seen as intended by the subject of the sentence. In the course of its grammaticalization process a reinterpretation of the core semantics of the lexical verb drohen takes place: drohen comes to be used in contexts with infinitive verbs designating states and non-intentional events (and with non-intentional subjects). In such contexts the core lexical meaning of drohen is ruled out. What happens is the reinterpretation of the lexical semantics in the sense of ‘something undesirable is about/going to happen’. The process of grammaticalization, including the mechanisms of semantic bleaching and generalization, continues: drohen gradually loses semanti-

Drohen & zu-infinitive 199

cally motivated restrictions with regard to the choice of its infinitive complements. Due to this process other verbs come to serve as infinitives of grammaticalized drohen. As a consequence of the increasing degree of grammaticalization drohen gains the ability to combine with any infinitive verb and starts to function as a typical auxiliary (for the details of the historical development of evidential drohen in German, see Chapter 8.3). Taking this into account, the fact that ambiguous interpretations of drohen & zu-infinitive arise in PDG if drohen combines with action verbs is not surprising. This is only the natural result of the grammaticalization process via which drohen as evidential auxiliary develops. We believe that, instead of searching for restrictions with regard to the infinitive verbs of the grammaticalized verb drohen, one should rather look for those restrictions for lexical drohen. Gunkel 2000 has proposed that if we are dealing with the grammatical variant of drohen, the holistic perspective on the event is taken, which is not the case for the lexical drohen. Taking into account principles of grammaticalization, this change of perspective may be easily explained. Drohen as a lexical verb serves as a primary predication of the sentence, i.e., it says something about the subject of the sentence. Namely lexical drohen expresses that the subject is going to do something (to perform an action). The infinitive complement explicates what exact the subject is going to do. (86)

Diesmal drohen die Abonnenten, die ‘Neue Freie Presse’ zu bestrafen... (DWDS, Kernkorpus) ‘This time the subscribers are threatening to punish the ‘Neue Freie Presse’.’

Drohen as (a relatively highly grammaticalized) evidential auxiliary, on the other hand, serves as a secondary predication. In this function drohen provides additional information to the proposition, i.e., information about the evidential status of the proposition. As such, drohen is not closely linked to the subject of the sentence. Instead, it predicates something about the (evidential) status of the whole proposition including the subject of the sentence and the infinitive verb (together with its possible complements). (87)

Berichte, die recht interessant zu lesen sind, der Konferenz aber ihren Charakter als eines Meinungsaustausches über wichtige Fragen zu rauben drohten. (DWDS, Kernkorpus)

200 The four evidential constructions in present-day German ‘Reports, which are very interesting to read but threaten to steal from the conference its character as an exchange of views.’ From this it is not surprising that the perspective from which the situation is seen changes from a narrow perspective, which concentrates on the action introduced by the infinitive complement, to a wider holistic perspective, which embraces the situation represented by the whole proposition. Furthermore, there are some more observations that favor the account of drohen & zu-infinitive proposed here. Due to the increasing degree of grammaticalization drohen gains the capacity to take infinitives of different types. So, grammatical drohen occurs also with passive infinitives: (88)

Und nun steigt der Rhein … zu einst so seltener Höhe an, droht der Wasserstand des Jahres 1920 wieder erreicht zu werden. (DWDS, Kernkorpus) ‘Now the Rhein is rising to a level seldom previously reached and threatening to reach the level of the year 1920.’

This combinatorial possibility is not possible for lexical drohen. Since a verb in passive voice cannot refer to an action it cannot combine with lexical drohen. As grammatical drohen, on the other hand, does not show any restrictions with regard to the semantic nature (particularly, to the process type expressed by the verb) of its infinitive complement, this combination is not ruled out. By way of conclusion, we can add a further point to the list of properties that serve to distinguish grammatical drohen from its lexical variant: 9.

there are no restrictions with regard to the semantic nature of the infinitive, whereas lexical drohen is restricted to infinitive complements denoting actions.

3.3. The perceptual inferential value of evidential drohen In the previous sections we have shown that there are two distinct readings of the construction drohen & zu-infinitive in PDG. In the following we will argue for treating drohen & zu-infinitive as a perceptual evidential construction providing the results of our synchronic corpus investigation. In the literature the semantics of the construction drohen & zu-infinitive is mostly described as pertaining to the domain of epistemic modality (cf.

Drohen & zu-infinitive 201

Askedal 1997, Heine and Miyashita 2004, Zifonun et al. 1997, Engel 2004). In this view drohen & zu-infinitive is seen to express a degree of the speaker’s commitment to the truth value of the proposition. Some authors (e.g. Reis 2005: 128) do not agree with this position and consider drohen to be a temporal-aspectual verb that locates the described event in the immediate future (with a note that this is done by the speaker on the basis of some available evidences). We argue that drohen & zu-infinitive is an evidential construction (for a similar view see Diewald 2004). The distinct semantic feature of any evidential marker is the indication of the information source, i.e., the indication of the way in which the speaker acquired the information contained in the proposition. The information source indicated by drohen & zu-infinitive is inference: any proposition introduced by this construction is based on the inference (as information source). Consider the following examples: (89)

(90)

Die Hintersten drängten rücksichtlos nach und die Vordersten wurden so unwiderstehlich gegen uns geschoben, daß das gesamte Bureau mit seinem Tisch über die Rampe in den Saal hinabzufallen drohte. (DWDS, Kernkorpus) ‘The backmost were pushing ruthlessly and the front line was being irresistibly pushed against us, so that the whole bureau with its desk threatened to fall over the ramp into the hall.’ Der Stand ist folgender: Spitzengruppe Lorenz - Saldow, Poulain Miquel, de Mara - Stabe, Rudel - Pawke, Arend - Hübner; eine Runde zurück Techmer - Großmann, zwei Runden zurück Marx Finn, vier Runden zurück Neumer - Rosellen. Während Lorenz Saldow sich am frischesten fühlen, drohen Arend - Hübner jeden Augenblick aus der Spitzengruppe zurückzufallen. (DWDS, Kernkorpus) ‘The actual state is as follows: the top flight …, one lap behind …, two laps behind…, four laps behind… While Lorenz and Saldow feel fresh, Arend and Hübner are threatening to fall out of the top flight any moment.’

The situations here are represented as developing towards a certain end point which is introduced by drohen. From the context it is obvious that the situations are directly perceived by the speaker and the portions of the texts containing drohen describe situations which are not realized yet but are the likely end point of the ongoing development. Thus the events introduced

202 The four evidential constructions in present-day German by drohen are definitely inferred by the speaker on the basis of presently accessible information. In such cases the perceptual inferential evidential value of drohen is out of question. There are however a number of contexts in which the perceptual inferential evidential value of drohen is not as obvious as in the above examples: (91)

Schon der wachsende Einfluß fundamentalistischer Christen und ihre lärmende Forderung nach einem Verbot der Abtreibung und der Rückkehr zu traditionell-reaktionären Werten droht die Partei zu spalten. (DWDS, Kernkorpus) ‘The increasing influence of fundamentalist Christians and their noisy demands for the prohibition of abortion and a return to traditionally reactionary values atr threatening to split the party.’

In cases like these, there is hardly any presently accessible piece of information to be detected. However, it is clear that the situation introduced by drohen is seen as a consequence from what is manifest in some way at the time of speaking. In each case the inference process starts from the speaker’s present experience and results in the conclusion expressed by drohen. Some further examples from our corpus show that this interpretation of the construction drohen & zu-infinitive can be applied easily: (92)

(93)

Der erhöhte Posttarif droht vielen Geschäften zum Verhängnis zu werden. (DWDS, Kernkorpus) ‘The raised postal rate threatens to be the downfall of many stores.’ Da wird mit Petroleumöfen geheizt, dann droht das Flugzeug zu brennen. (DWDS, Gesprochene Sprache) ‘They use kerosine stoves for heating, then the plane threatens to burn.’

The interpretation of the construction as a marker of epistemic modality is, as we think, not appropriate. Relating the proposition to (the information source of) the speaker, drohen does not attach a particular degree of speaker’s confidence to the proposition: it only indicates that the proposition is the result of the inference process. The same holds also for the other evidential construction concerned here. They all are semantically neutral with respect to their degree of epistemic modal evaluation because they do not entail such evaluations in their inherent semantics (see Chapter 3 for fur-

Drohen & zu-infinitive 203

ther details). In contrast to Reis 2005, we do not regard the temporalaspectual interpretation of the construction as basic, but as a case of persistence of more lexical components of the item (note that a speech act of threatening per definition refers to some imminent action). Let us now turn to the lexical readings of the verb drohen. Taking for granted that evidential drohen has evolved through the process of grammaticalization out of its lexical reading(s) we look for similarities and possible motivations for the present – evidential inferential – value of the construction drohen & zu-infinitive. For the details of the diachronic development and overlapping meanings see Chapter 8.3. We propose that the agentive variant of lexical drohen served as a structural template enabling the grammaticalization of drohen & zu-infinitive. The infinitive complement of lexical drohen introduces an action which is the “content” of the threat. Thus drohen installs the threat, which is the announcement of a (future, possible, avoidable, negative, undesirable, etc.) action, and the infinitive complement describes that action. A similar – though highly schematized – structure is preserved in the inferential semantics of drohen & zu-infinitive. The evidential construction drohen & zu-infinitive installs an analogous relationship between the inference on which the statement is based – as information source – and the proposition. The infinitive verb describes the situation: in the case of grammatical drohen the situation does not follow the threat but results from the inference process of the speaker. Hence the abstract semantic structure of drohen has been preserved in the grammatical reading of this verb. As concerns the second lexical reading of the verb drohen, it also shows relevant characteristics resembling the evidential construction drohen & zu-infinitive. As has been rightly pointed out by Heine and Miyashita 2004, 2008 this lexical reading diachronically precedes the grammatical variant of drohen. Moreover it gave rise to the development of the grammatical reading of drohen, since it already was able to be combined with inanimate subjects. The assumption made by Heine and Miyashita 2004, 2008, that the development of the infinitive construction occurred due to the extension from nominal to non-finite complements, will however be criticized and refined (see Chapter 8.3). We suggest that the second lexical reading “paved the way” for the grammaticalization of the evidential construction drohen & zu-infinitive not only because it allowed for inanimate subjects to be combined with drohen. In our view the holistic perspective on the situation has been established by such syntactic combinations. In the structures like Gefahr droht, Katastrophe droht, Krieg droht etc., the noun refers to the whole situation, i.e., it comprises the whole situation in one

204 The four evidential constructions in present-day German nominal expression. On the one hand, the whole situation is introduced by the single verb drohen now. On the other hand, the static view of the situation is represented, i.e., the situation is seen as a state which is going to be the fact soon, cf.: (94)

Hin und wieder drohte Regen... (DWDS, Kernkorpus) ‘Every now and then rain threatened.’

It can generally be said that the holistic and static view of the situation is represented by the second lexical reading of drohen. In order to express the situation from the dynamic perspective, a verb is needed. We think that not the extension but the reformulation of the single noun referring to the negative situation by means of an infinitive was the crucial step towards the evidential construction drohen & zu-infinitive. The syntactic structure was already available: the first lexical reading of drohen already allowed for combinations with infinitives. This structure was analogically recruited for the new structure. As a result, the present evidential construction drohen & zu-infinitive developed. Importantly, the non-agentive variant of drohen already bears some components of evidential semantics (cf. also Heine and Miyashita 2004, 2008 who explicitly paraphrase this reading with the same formula as they use for their ‘functional’ drohen). It introduces a situation which is not realized yet, but is likely to occur soon. Namely, the “threatening” situation is likely to occur, because something in the present situation points towards it. That is, already at this – still lexical – stage drohen develops evidential semantics. To conclude, we have tried to show that the evidential characterization of the relatively highly grammaticalized construction drohen & zuinfinitive is the most appropriate description of the present function of this construction. We have proposed that semantically the grammatical variant of drohen is closely connected to both lexical readings of this verb. Moreover we have suggested that the core evidential meaning goes back to the second lexical reading of drohen. In this reading drohen already relates the described situation to some present factors which point towards the “threatening” situation.

Versprechen & zu-infinitive 205

4. Versprechen & zu-infinitive In this section we are concerned with the construction versprechen & zuinfinitive. It should be noted, first, that this construction is very closely connected with the construction drohen & zu-infinitive described in the previous section. First, they show similarities in regard to their compositional structure (cf. Askedal 1982, Reis 2005). Second, they show semantic affinities (cf. Askedal 1982, Heine and Miyashita 2004, 2008, Diewald 2004, Reis 2005) in that they predicate something about the ontological status of the proposition and function on the propositional level. For this reason the verbs drohen and versprechen (and their infinitive constructions) have often been investigated jointly by several authors (see e.g. Askedal 1997, Diewald 2004, Heine and Miyashita 2004, Reis 2005 for German; cf. further Cornillie 2004, 2005 for Spanish; Traugott 1997 for English; Verhagen 2000, Vliegen 2006 for Dutch). The observed similarities between the infinitive constructions with drohen and versprechen are due to the fact that they are (gradually becoming) members of the same grammatical paradigm, i.e., the evolving paradigm of inferential evidentiality of German. They represent the same distinct evidential value within the paradigm, i.e., the perceptual inferential evidential value, and stay in opposition to the other paradigm members. Note that they differ from one another in that drohen expresses a negative evaluation, whereas versprechen indicates a positive evaluation of the described event. However, the story becomes more complicated if one looks closely at the details. Infinitive constructions with evidential drohen are much more frequent in PDG than those with evidential versprechen (cf. also Cornillie 2004, 2005 for Spanish, Vliegen 2006 for German and Dutch for similar observations). Furthermore evidential versprechen & zu-infinitive still shows restrictions with regard to subjects and infinitive verbs, i.e., the selection mechanisms of evidential versprechen & zu-infinitive are more restricted and thus less entrenched than those of drohen & zu-infinitive. On the basis of these and other observations (which will be presented below) we suggest that versprechen has not acquired the same degree of grammaticalization yet. This section is structured in the following fashion: we first give a brief outline of the relevant construction types of versprechen in PDG (relying primarily on our own corpus analysis, but also considering the relevant research literature). In the next step we concentrate on the construction versprechen & zu-infinitive and uncover its semantic and structural proper-

206 The four evidential constructions in present-day German ties in more detail. In so doing we mainly deal with the above mentioned similarities and differences between the infinitive constructions with drohen and versprechen. Finally we propose an explanation for the constrained or even missing grammaticalization of versprechen towards an evidential auxiliary, from the synchronic point of view.

4.1. Relevant construction types of versprechen in present-day German Usually two distinct readings of the verb versprechen are distinguished in the literature: the lexical reading and the so-called ‘modal’, ‘functional’ or ‘epistemic’ reading (cf. Askedal 1982, 1997; Zifonun et al. 1997; Engel 2004; Eisenberg 2004; Heine and Miyashita 2004; Reis 2005 et al.). These two distinct readings are exemplified by the following examples (taken from Askedal 1997: 12): (95) (96)

Er hatte seiner Tochter versprochen, sie nach Paris mitzunehmen. ‘He had promised his daughter to take her to Paris.’ … die großen Erfindungen, die dieser Streit hervorzurufen versprach. ‘… the great inventions which this dispute promised to bring forth.’

In general the situation is very similar to drohen. In view of the very high functional variability of the verb versprechen (the list of possible meanings and structures employing the verb versprechen presented in the DWB, for example, lasts for several pages), we concentrate in the following only on the occurrences of versprechen that we have found in our corpus material. It should be noted here that we excluded the intransitive reflexive verb sich versprechen meaning ‘to misspeak’ from our analysis, since it no longer has any considerable semantic (but just etymological) connections to the contemporary meaning of the verb versprechen we are dealing with here. On the basis of our corpus findings, we basically agree with the above mentioned twofold classification. These are indeed the most common uses of the verb in the PDG. However we would like to add two further variants of versprechen to the list: the non-agentive versprechen and the construction sich etwas (von etwas) versprechen ‘to expect something of something’ where versprechen is a transitive reflexive verb. The reasons why we include this construction in our analysis will be discussed further be-

Versprechen & zu-infinitive 207

low. Our classification of the relevant readings of the verb versprechen consists thus of the four following entries: 1. 2. 3. 4.

versprechen as a speech act verb with agentive semantics; versprechen as a lexical verb with non-agentive semantics; sich etwas von etwas versprechen as a lexical(ized) construction with non-agentive semantics, expressing the personal view of the subject on the situation; versprechen as an auxiliary (or: semi-auxiliary) verb with evidential semantics.

For these readings the following semantic paraphrases can be formulated: 1. 2. 3. 4.

‘Someone points out that s/he intends to do something (that is desirable to someone else or to him/herself)’. ‘Something positive (desirable) is about to happen’. ‘Someone expects something positive (desirable) from something else’. ‘Something points towards something else, which is positive (desirable) and is about to happen’.

The last paraphrase will be modified further below. It is given here in this form for reasons of simplification and better comparability with other readings. The proposed four variants of versprechen usually correspond to the certain structures: 1. 2. 3. 4.

human subject referent & versprechen & (facultative) dative object & accusative object (NP)/(in)direct speech sentence/complement clause with dass/zu-infinitive complement/main clause; inanimate subject & versprechen & accusative object (NP); versprechen & sich & PP with von & accusative object (NP); versprechen & zu-infinitive.

The first variant of versprechen has the basic meaning ‘to promise’ and exhibits a strong intentional component which explains the fact that its use is mostly restricted to human subjects: (97)

Ich hatte ihnen versprochen, wenn sie artig wären, etwas vorzulesen. (DWDS, Kernkorpus) ‘I promised them to read something to them if they behaved well.’

208 The four evidential constructions in present-day German The second type of versprechen, i.e. the non-agentive lexical verb, is usually restricted to non-animate subjects and occurs with accusative objects with positive evaluative semantics. The most frequent noun thereby is Erfolg ‘success’. Further nouns with positive evaluative semantics are also found in combination with this variant of versprechen, e.g. Genuss ‘pleasure’, Erlösung ‘salvation’, Vergnügen ‘amusement’, Nutzen ‘advantage’, Spaß ‘fun’, Sicherheit ‘safeness’, etc. (98)

Das verspricht aber keinen großen Erfolg. (DWDS, Kernkorpus). ‘This is however not very promising.’

The crucial semantic difference to the first type of versprechen is that an intentional component can hardly be attributed to the non-agentive versprechen. Rather, it describes a whole situation which is seen as being likely to occur in the future. This type of versprechen is mostly restricted to the 3rd person (inanimate) subjects. The third lexical reading of versprechen, which is associated with the fixed syntactic structure sich NPacc von NPdat versprechen has specific semantics and expresses someone’s personal expectation. In this reading versprechen introduces someone’s personal view of the described situation, particularly expressing the subject’s positive evaluation of the ongoing development. (99)

Ich verspreche mir gerade von dieser Kombination gute Erfolge. (DWDS, Kernkorpus) ‘I hope [lit.: promise to me] to achieve success with this combination.’

The personal attitude is made clear by the obligatory dative reflexive pronoun. We will show in the following sections that the development of this particular variant of versprechen might have served as one of the constraining factors in the evolution of the evidential construction versprechen & zu-infinitive. Since this particular structure has been lexicalized and has received the status of a conventional lexical unit, it has established a particular meaning which is one of the possible interpretations of the evidential meaning of the grammaticalizing construction versprechen & zuinfinitive. In this way it “reserved” this semantics and restricted the new grammatical construction in its “attempts” to include this meaning as well.

Versprechen & zu-infinitive 209

As the development of this lexical reading has influenced the development of the evidential construction versprechen & zu-infinitive to a considerable extent, we have decided to include it in our analysis. The evidential variant of versprechen is mostly associated with zuinfinitive constructions and is used only with inanimate subjects: (100)

…und nach längerem Suchen und nachdem er ihn mit den Zähnen getestet hatte, fand Kürenberg einen Reis, der körnig zu kochen versprach. (IDS-Corpora, W) ‘… and after a lengthy search and after testing it with his teeth, Kürenberg found a type of rice which promised to boil grainy.’

At this point we would like to refine the semantic paraphrase for the evidential construction versprechen & zu-infinitive given above. We suggest that this construction has the same evidential value as the construction drohen & zu-infinitive and differs from it only in regard to the additional evaluative semantic component. Therefore the paraphrase is largely identical to what has been given for drohen & zu-infinitive in the previous chapter: The speaker has perceptual access to certain pieces of information which s/he interprets as pointing towards the described event (which is evaluated positively from the speaker’s point of view). P (proposition), because Q: Origo (i.e. the speaker) can observe some signs/traces of P; P is desirable

We should add some words about the perceptual nature of the inferential evidential value of versprechen & zu-infinitive. Similarly to drohen & zuinfinitive, versprechen & zu-infinitive mostly indicates that the inference process of the speaker is based upon some pieces of information which are somehow relevant and obvious at the time of speaking. The factor of obviousness is given if the information is either directly perceived or observed at the time of speaking, or it is made directly accessible at the time of speaking. In each case the inference process starts from the speaker’s presently accessible experience and results in the conclusion expressed by drohen/versprechen & zu-infinitive.

210 The four evidential constructions in present-day German 4.2. Lexical versus evidential versprechen As noted in Section 6.3 the construction drohen & zu-infinitive has both – lexical as well as evidential – readings in PDG. A similar situation can also be observed for the construction versprechen & zu-infinitive (see examples (95) and (96)). These two readings can however always be distinguished as they occur in different contextual environments. Hence we can speak of a clear division of labor between lexical and grammatical versprechen, insofar as each interpretation is restricted to specific contextual configurations. This also means that the factors characteristic for one reading are exclusively characteristic for it. The features of versprechen and its immediate constituents have been described earlier (e.g. Askedal 1997, Gunkel 2000, Heine and Miyashita 2004, Reis 2005). A list of the relevant characterisitcs given in Table 1 is based on the findings found in the literature as well as on our own observations. Table 1. Lexical versus grammatical versprechen lexical versprechen subject referent

versprechen bears semantic relation to… inflectional paradigm extraposition of the infinitive complement possible? pronominalization of the infinitive complement possible? adverbial modification of versprechen possible? clausal complements possible? topicalization/ pronominalization/ passivization of versprechen possible? dative participant infinitive verb denotes…

animate, agentive 1st, 2nd, 3rd person referents the subject

grammatical versprechen inanimate, nonagentive only 3rd person referents the whole proposition

full yes

limited no

yes

no

yes

no

yes

no

yes

no

yes an action

no a non-action

Versprechen & zu-infinitive 211

We concentrate here only on two properties, the subject referent and the infinitive verb. Other characteristics have already been extensively discussed in earlier works on this topic (see especially Askedal 1997 and Heine and Miyashita 2004). Only inanimate referents can take the subject position in a sentence with the evidential versprechen & zu-infinitive. If a human subject referent is used, only lexical reading is possible. Collective referents such as Regierung ‘government’, Staat ‘country’, Partei ‘party’, Unternehmen ‘company’, etc. also fall into the category of human subjects. They are able to perform an illocutionary act (even if the act itself is normally performed by a person who is a member of a collective) and therefore are compatible with the prototypical lexical reading of versprechen. It should also be noted that inanimate subject referents may sometimes occur with lexical versprechen & zu-infinitive if they allow for a metonymic interpretation, cf.: (101)

Der letzte der dreizehn Punkte versprach, die Integrität Norwegens zu achten und uns nach dem Krieg unsere Unabhängigkeit zurückzugeben. (DWDS, Kernkorpus) ‘The last of the thirteen articles promised to respect the integrity of Norway and to give us our independence back after the war.’

The subject referent Punkt ‘article’ is used metonymically and it represents an oral statement that has been written down in the form of a document containing several articles. Thus the subject refers literally to a promise given by some human referent which is available in the written form now. Other subject referents used metaphorically with lexical versprechen are e.g. Buch ‘book’, Titel ‘title’, Kapitel ‘chapter’ etc. Hence even inanimate subjects used with lexical versprechen are seen as referring to the illocutionary act of promising. To conclude: as far as the aspect of animacy and agentivity of the subject referent is concerned, two readings of versprechen (within infinitive constructions) seem to be mutually exclusive. The lexical reading is restricted to animate and agentive subject referents, whereas the grammatical reading is restricted to inanimate and non-agentive subject referents. Moreover it should be noted that the grammatical reading of versprechen has even more restrictions with regard to their subjects: it combines only with 3rd person subject referents. In its lexical reading, on the other hand, versprechen occurs with all personal forms. This fact can be seen as resulting from what we have stated above: inanimate subjects can hardly be understood as being speaker or hearer, the roles to which the 1st and 2nd person inflexions correlate. From this, 1st and 2nd person subjects

212 The four evidential constructions in present-day German are excluded from the evidential uses of the construction versprechen & zu-infinitive.3 The next relevant point we would like to address here is the nature of the infinitive verb combining with versprechen. As has been noted by Gunkel (2000: 116), grammatical versprechen does not occur with infinitive verbs denoting an action. We have shown in the preceding chapter that this claim does not hold for evidential drohen. For versprechen, however, this restriction does hold in PDG. The most frequent infinitive verb of evidential versprechen is werden ‘become’ that unambiguously denotes a process which is not intended by the subject of the utterance, cf.: (102)

London steht in den nächsten Wochen vor einer Versteigerung von Autographen und Büchern, die bei Sotheby - zu einer großen Sensation zu werden verspricht. (DWDS, Kernkorpus) ‘In the next weeks London will experience an auction of autographs and books which promises to be a big sensation at Sotheby’s.’

Further infinitive verbs found with evidential versprechen are enthalten ‚contain‘, weichen ‚lose ground‘, Aufschwung nehmen ‚boom‘, bieten ‘offer’: their semantics can be best described as non-agentive, non-intended and non-controlled by the subject referent. Rather they denote a state or an unintended event, in short: non-actions. In its lexical reading, on the other hand, versprechen most frequently combines with infinitive verbs referring to intended and controlled actions (which are carried out by intentional subjects), e.g. unterstützen ‚support‘, helfen ‚help‘, nehmen ‚take‘, Hilfe leisten ‚give aid‘, Maßnahmen ergreifen ‚take action‘, geben ‚give‘, schicken ‚send‘, schenken ‚make a gift‘, erzählen ‚tell‘, tun ‚do‘ etc. Hence we again observe that two readings of versprechen within infinitive constructions are mutually exclusive: the lexical reading is most often restricted to action verbs whereas the grammatical reading is associated with verbs denoting non-actions.

4.3. Drohen & zu-infinitive versus versprechen & zu-infinitive In this section we compare the evidential constructions with drohen and versprechen and focus primarily on their differences. Let us first summarize the relevant properties these two constructions have in common:

Versprechen & zu-infinitive 213

1.

2. 3.

4.

5.

they both function as evidential constructions in PDG, sharing the same evidential value: they mark the described event as inferred on the basis of some (perceptually) founded facts (see Chapters 6.3.3 and 6.4.1); they both show the high degree of grammaticalization (see Chapter 4); drohen and versprechen still show persistence of some semantic aspects from their lexical meanings: they introduce events which are not realized, but rather expected; they express a particular evaluation of the described event, i.e., while drohen refers to negatively evaluated situations and provides those with the label ‘undesirable’, versprechen introduces positively evaluated situations, which can be labeled ‘desirable’ (from the speaker’s point of view)’; both drohen and versprechen with infinitives are similar in their lexical readings with respect to the following restrictions: they mostly occur with animate (individual or collective) subjects which take control over the event represented in the infinitive complement; their infinitives express an action which is controlled or fulfilled by the subject; there are some features which are typical for evidential readings of both verbs: they mostly take an inanimate subject which can not be interpreted as having control over the event presented in the infinitive complement; the infinitives describe an event rather than an action.

Apart from these similarities there are, however, considerable differences between the evidential infinitive constructions with drohen and versprechen in PDG. First, the evidential construction drohen & zu-infinitive is much more frequent in PDG than the construction versprechen & zuinfinitive. This fact has been already observed in other studies (e.g. Askedal 1997 and Vliegen 2006) and is confirmed by the results of our own corpus investigation. Therefore this fact cannot be considered to be a mere coincidence. Given the fact that both verbs undergo a parallel process of grammaticalization, as has been shown in the previous chapters and will be discussed in more detail in the diachronic part of this volume, one might assume that the frequency of both constructions would show equal or at least to a large extent similar value. However, given massive differences in frequency between drohen and versprechen, this fact requires explanation. For the synchronic part of our study we looked through a random selection of 2000 instances of drohen and versprechen respectively, covering the 20th and 21st centuries. The following table shows the figures for the distribution of lexical and grammatical readings of the two constructions found in our corpus data.

214 The four evidential constructions in present-day German Table 2. The distribution of lexical and grammatical readings of the infinitive constructions with drohen and versprechen in PDG

drohen versprechen

infinitive construction 603 265

lexical reading

evidential reading

% lexical

% evidential

133 237

470 28

22,0 89,4

80,0 10,6

Second, evidential drohen sometimes occurs with animate subjects, as we have mentioned above (see Section 6.3.2). Evidential versprechen, on the other hand, never occurs with animate subjects. Third, evidential drohen takes infinitives which encode an action (e.g. bereiten ‘prepare’, schaffen ‘create’, niederdrücken ‘press down’, vernichten ‘destroy’, brechen ‘break’, sprengen ‘detonate’ etc.), combining with inanimate as well as animate subjects (see Section 6.3.2). Evidential versprechen, on the other hand, never occurs with action verbs. As noted in Section 6.3.2, evidential drohen also takes passive infinitives, which is not possible for evidential versprechen. By way of conclusion we can state at this point that versprechen is far less grammaticalized as an evidential auxiliary than drohen. Now the question arises why versprechen has not been grammaticalized as an evidential auxiliary to the same extent as drohen. In what follows we will propose a tentative explanation bringing together some observations about synchronic status of versprechen. The first possible explanation for the constrained grammaticalization process of versprechen & zu-infinitive is the strong speech act semantics of this verb. This leads to a heavy dominance of animate, mostly human subjects associated with versprechen. Furthermore this makes it difficult for non-human subjects to be associated with versprechen, particularly when versprechen occurs within infinitive constructions where the infinitive verb describes an action promised by the subject. Vliegen (2006: 87) proposes a similar explanation for the more frequent use and much wider variation of infinitive complements of drohen as compared to versprechen: “the dominant illocutionary character of the verbs versprechen/beloven places an appreciable limitation on the variation of the infinitives associated with them in their epistemic reading, while the verbs drohen/dreigen show a much wider variation here”. Askedal 1997 also suggests that the specific speech act semantics of versprechen as opposed to drohen is responsible for the differences between grammatical variants of these verbs. So, only versprechen can be used both to perform a commissive illocutionary act and to refer to this kind of illocutionary act.

Versprechen & zu-infinitive 215

Drohen, on the other hand, is restricted to the second function, i.e. it can only refer to a speech act. Consider the difference between the following sentences (examples taken from Askedal 1997: 18): (103) (104)

Ich verspreche Ihnen (hiermit), Sie sofort einzustellen. ‘I (hereby) promise you to engage you immediately.’ *Ich drohe Ihnen (hiermit), Sie sofort zu entlassen. ‘*I (hereby) threaten to dismiss you immediately.’

When combined with human subjects versprechen conveys its commissive meaning. The same holds also when versprechen combines with nominal accusative objects. Drohen, on the contrary, when it combines with prepositional complements with mit ‘with’, may refer either to a speech act, as in (105)

Sie drohte mit Polizei. ‘She threatened to call the police.’

or to a non-verbal action, as in (106)

Sie drohte mit der Faust. ‘She threatened with a fist.’

Therefore drohen is semantically less restricted to the speech act semantics than versprechen. This difference in semantics has most probably been the reason why drohen combined much more easily first with inanimate subjects and then with infinitives which do not necessarily introduce an action to which the subject of the sentence is committing him/herself. Most important, this is not the speech act semantics per se, but rather the commissive semantics of the speech act verb versprechen that serves as a constraining factor in the semantic change from lexical to evidential meaning of this verb. The next possible explanation, which we would like to address here, has also been proposed by Askedal 1997 and concerns pragmatic characteristics of versprechen. Given that drohen introduces negatively evaluated events, and versprechen introduces positively evaluated events, it is likely (according to the maxim of quantity by Grice 1975) that the speaker will more frequently use infinitive constructions with drohen. Since versprechen refers to expected, planned, i.e. normal developments of situations, the speaker will not necessarily use the construction versprechen & zu-

216 The four evidential constructions in present-day German infinitive to highlight the “normal” way the things develop. With drohen, on the other hand, unexpected, unplanned situations, i.e. developments deviating from the “normal” status, are introduced. So, the speaker will more frequently use the construction drohen & zu-infinitive just to express that same value. Vliegen (2006: 92) suggests that the nature of the described event – either its realization or non-realization – has an impact on the differences in frequency between grammatical variants of drohen and versprechen. “The increased likelihood of non-realization of the proposition expressed in the infinitive construction associated with the verb pair dreigen/drohen, as compared to the cases where there is no evidence whether this proposition is realized or not”. We do not, however, agree with Vliegen at this point. We think that the realization or non-realization of the described situation is not relevant at all because at the time, when a sentence with versprechen or drohen is uttered, there can be no difference between realization and nonrealization status of the described situation. The only difference versprechen and drohen express is the difference between positive and negative evaluation of the introduced event, respectively. Whether the situation will be realized or not, both infinitive constructions do not convey any information about this; instead, the situation is seen as most likely to develop in the described way and not in another one. We would like to add one more explanation to the three mentioned above which, as we think, could have been responsible for the constrained grammaticalization of versprechen, too. This concerns the paradigmatic changes within the lexical readings of versprechen. By this we mean the rising of the third lexical reading of versprechen which corresponds to the structure jemand verspricht sich etwas von etwas ‘someone expects something from something else’ described in Section 4.1. According to the DWB this reading has been attested from the beginning of the 18th century and was spread very rapidly in the course of the 18th century, i.e. in the period proximately preceding the evolution of the evidential construction versprechen & zu-infinitive. We suggest that the development of this specific lexical form of versprechen, which is undoubtedly very close to the evidential reading of the construction versprechen & zu-infinitive, hampered the development of the evidential construction versprechen & zuinfinitive to a large extent. The lexical verb versprechen refers to the personal attitude the (human) subject of the sentence has towards the described situation, i.e. it refers explicitly to the epistemic sphere of the subject person. The evidential construction versprechen & zu-infinitive, similarly, links the described situation (conceived of as a particular devel-

Versprechen & zu-infinitive 217

opment of the actual situation) to the mental world (i.e. information status) of the speaker. Thus the meaning of this variant of lexical versprechen and the evidential meaning of the construction versprechen & zu-infinitive are in some way competitive. Since the lexical sich etwas von etwas versprechen has established itself as a lexical element in the German language, this particular function could not be adapted by the construction versprechen & zu-infinitive. This suggestion will be further examined in the diachronic part of this study.

Chapter 7 Intermediate summary

Having laid out our theoretical background as well as the empirical data for the evidential constructions in PDG, this chapter is intended as a survey and intermediate summary of our findings on the evidential system of PDG and on its locus in a typological perspective, i.e. its relation to known evidential systems. This synopsis of the current results of the longstanding development of the four evidential constructions will serve as a starting point for a concise look back into their diachronic development which will be the objective of Chapter 8. Section 7.1 concentrates on those properties of the four constructions which allow us to integrate them into a functional paradigm. Section 7.2 compares well established typological classifications of evidential values to our classification of the German evidential values.

1. The inferential evidential paradigm in present-day German As suggested in Chapter 2, we classify the four infinitive constructions scheinen/drohen/versprechen & zu-infinitive and werden & infinitive as inferential evidential constructions. Inferential evidentiality is a subcategory of indirect evidentiality. Irrespective of more fine-grained subdivisions, the common core of inferential evidential expressions is that they mark the proposition as inferred by the speaker (cf. Section 2.2.2 for quotes and discussion). As explained in Chapter 2, this core meaning of inferential evidential expressions may be summarized as follows: (i) The speaker has no direct access to the described event, i.e. s/he has not perceived/does not perceive the situation directly, e.g. by means of perception. >PQ (ii) The speaker infers the described situation from some pieces of information, which may be either observable/observed facts or reported information or generally known facts etc. > P, because Q

The inferential evidential paradigm in present-day German 219

Taking these two criteria as pivotal for defining inferential evidentiality and referring to the facts and results presented so far, we contend that these two properties are present in the semantics of the German infinitive constructions with scheinen, werden, drohen and versprechen, which thus make up a system – a paradigm – of inferential evidential markers the details of which are spelled out in the rest of this section.

1.1. Formal characteristics of a paradigm Before we present our view on the conceptual configuration of the evidential paradigm constituted by the four German infinitive constructions, a short overview of the relevant formal characteristics of these four constructions and the paradigm as a whole will be given. Though we will deal with these largely diachronically motivated issues in Chapters 8 and 9 in greater detail, there are some points which do deserve attention at this stage. To begin with, the four infinitive constructions, which are in the focus of this study, form a relatively homogenous paradigm, as concerns their formal characteristics. Most important is the periphrastic status of these forms: they are all infinitive constructions consisting of an auxiliary verb and the infinitive of the main verb. Apart from this, there are some formal differences between the constructions. The most obvious formal difference between werden & infinitive on one hand and scheinen/drohen/versprechen & zu-infinitive on the other is the absence of the infinitive particle zu ‘to’ in the construction with werden. As to this difference, we would like to mention two points. First, werden has never been used with infinitives and the particle zu ‘to’ in the history of German. Moreover werden has been attested with infinitives since the 13th century. As will be discussed in greater detail later, werden & infinitive has a specific status within the evidential system and in the German language in general.1 Werden & infinitive is the oldest and the most grammaticalized construction among the four constructions considered here. Therefore it exhibits some properties which still cannot be found in other constructions. As is known in grammaticalization theory and typology, grammatical categories may be formally inhomogeneous and derive from disparate sources like, for example, German tense grams which derive from phonotactically induced vowel changes (the system of Ablaut) as well as from independent lexical sources presently displaying diverging states of grammaticalization (periphrastic and inflectional items). The same holds true

220 Intermediate summary for evidential markers, as the following typological observation illustrates. Aikhenvald (2004) points out that evidential markers of one language do not necessarily form a formally coherent paradigm: The morphemes then occur in different slots of the verbal word, enter in different paradigmatic relationships with non-evidentials, and have different restrictions on co-occurrence with other categories. The expression of evidentiality may itself be obligatory – but different evidentiality specifications ‘scattered’ throughout the verbal system by no means make up a unitary category. They still, however, qualify as grammatical evidentials. (Aikhenvald 2004: 80)

Moreover it has been detected that evidential markers, like other grammatical categories, may originate from different unrelated sources, which, however, does not prevent them from forming a coherent category in a language: The few available studies of the genesis of large systems show that, by and large, they are etymologically heterogeneous. That is, different terms come from different sources. Of the evidentials in Wintu, three came from independent verbs, and one from a particle. […] [T]he non-visual sensory in Nganasan, Nenets, and Enets probably derives from a noun. Both inferred and reported evidentials in these languages contain a formative -bi- which is most probably related to a past tense or a past participle marker. (Aikhenvald 2004: 285)

Against the background of such observations the German evidential paradigm appears to be a very homogeneous system consisting only of verbal infinitive constructions and arising only from main verbs. Thus, even if the German evidential paradigm seems to be incoherent from the narrower – language-specific – perspective, a wider – typological – view helps to relativize this. Other signs of formal inhomogeneity of the German evidential system, such as e.g. the impossibility of drohen and versprechen to combine with perfect infinitive forms, or their restrictions with regard to their subject referents and infinitive verbs, are due to the different degrees of grammaticalization of these constructions. Werden & infinitive is the oldest and the most grammaticalized one, followed by scheinen & zu-infinitive, which is highly grammaticalized as well but to a lesser extent than werden & infinitive. Scheinen & zu-infinitive again is followed by the “youngest” members of the evidential system – the infinitive constructions with drohen and versprechen which are less grammaticalized than the other two. Adopting the perspective of grammaticalization theory, variability and differences be-

The inferential evidential paradigm in present-day German 221

tween the constructions can be easily explained without denying their membership in a coherent grammatical paradigm. As discussed before (see Chapter 4), the crucial property of a grammatical paradigm is the close – formal as well as conceptual – relations between its members. Lehmann (2002) proposes a similar definition of a grammatical paradigm, cf.: The size of a paradigm is, to repeat it, a superficial and not always reliable aspect of paradigmaticity. Over and above it, the integration of the paradigm has more intrinsic and less easily quantifiable aspects. It also comprises the formal and functional homogeneity of the paradigm, i.e. a certain amount of similarity among its members and of regularity in their differences. […] On the semantic side, the members of a paradigm have a common semantic basis with varying differentiae specificae. (Lehmann 2002: 120)

This section has shown that the formal homogeneity of the German evidential paradigm can be said to fulfill this condition, insofar as it consists of – formally analogous – constructions displaying the same pattern: [auxiliary] + [infinitive]. The next section will show that the criterion of semantic homogeneity applies as well, as all members of the German evidential paradigm share “a common semantic basis” (their inferential evidential meaning) and contrast with each other by specific distinctive values.

1.2. Conceptual structure of the paradigm Having already proved the high degree of grammaticalization of each construction and having delineated each of their semantic and lexical peculiarities in previous chapters, our central concern here is to show that the four constructions are semantically very close to each other, i.e. that they all pertain to one conceptual sub-domain of evidentiality. In PDG there are no overt grammatical elements expressing direct evidentiality. Direct evidentiality – if needed – may be expressed either by means of lexical signs (e.g. the adverbs sichtlich, sichtbar ‘visible’ etc.) or by certain syntactic constructions (e.g. complex sentences with matrix verbs of perception, as illustrated in (3), AcI constructions as in (2), or parenthetical uses of verbs of perception as in (1), etc.), cf.: (1)

Sie ist – wie ich sehe/höre – im Garten. ‘She is – as I see/hear – in the garden.’

222 Intermediate summary (2) (3)

Ich sehe/höre sie im Garten arbeiten. ‘I see/hear her working in the garden.’ Ich sehe/höre, wie sie im Garten arbeitet. ‘I see/hear that she is working in the garden.’

Recently it has been suggested (cf. de Haan 2001, Mortelmans 1999; see also Fabricius-Hansen 2002, Letnes 1997, DUDEN 2005: 535, where, however, the explicit reference to the evidential nature of these verbs is lacking) that the German modal verbs wollen and sollen serve as (grammaticalized) markers of mediated evidentiality as – in specific constructions – they serve to mark reported speech. Wollen is said to indicate reports of a particular speaker who is identical with the subject of the sentence. Sollen on the other hand is said to introduce reports of an unspecified person’s speech, cf.: (4) (5)

Sie will nicht am Tatort gewesen sein. ‘She claims that she wasn’t at the crime scene.’ Sie soll nicht am Tatort gewesen sein. ‘It is said that she wasn’t at the crime scene.’

Although both verbs without doubt pertain to a wider domain of “linguistic utterance of alter ego” (see Section 2.2.3), their status as evidential markers, i.e. markers of mediated information, is still a matter of discussion. Bearing in mind the controversial status of quotatives within the evidential domain (cf. Plungian 2001, Aikhenvald 2004: 132-135, see also Chapter 2 for the discussion of “quotatives” vs. “mediated information”), the question is yet to be answered whether the German modals sollen and wollen should be considered as purely evidential markers. As this question is beyond the scope of this study and calls for separate treatment, the grammatical evidential system of German presented here does not include the direct or the mediated branch of the general classification of evidentiality (see Figure 20 in Chapter 2). As regards the four infinitive constructions with werden, scheinen, drohen, and versprechen, they constitute the (newly evolving) grammatical paradigm of inferential evidentiality. Each of the four constructions meets the definition of inferential evidentials given in Chapter 2: they pertain to the same sub-domain of evidentiality. Moreover, each construction specifies the inferential evidential value in a different way. This makes for the distinctive feature – or oppositive values – among the four constructions in

The inferential evidential paradigm in present-day German 223

the paradigm. As is typical of grammatical categories, the whole paradigm is structured hierarchically and asymmetrically (cf. Jakobson [1939] 1974). There is one member of the paradigm which is unmarked in contrast to all other members, which is the construction werden & infinitive. It marks the (conceptual) inferential evidential value without specifying the type of information on which inference is based. Inference expressed by werden & infinitive may be indiscriminately based on experienced evidence (i.e., on pieces of information which have been in some way acquired by the speaker) as well as on personal and general knowledge (i.e. on internalized pieces of information). This is to say that the basic meaning of werden & infinitive is to indicate the (neutral) inferential value (see Section 6.1.4 for a detailed description).2 The other three evidential constructions represent marked inferential evidential values. Scheinen & zu-infinitive is opposed to werden & infinitive as it normally means that the inference process is based on some specific pieces of information (see Section 6.2.4 for a detailed account). For the sake of simplification, this type of (pieces of) information may be called “experienced information” which means that the pieces of information are marked as in some way acquired (i.e. as observed, as heard from others etc.) by the speaker. The constructions drohen und versprechen & zu-infinitive are in opposition to the construction with scheinen (as well as with werden) in that they specify the basis of inference as present experience (see Sections 6.3.3 and 6.4.1 for details). This means that the speaker bases his/her inference on some specific, presently accessible (i.e. observed) pieces of information. Thus, the difference between werden & infinitive, which is positioned on the lowest (first, see Figure 1) level of the hierarchy, and the constructions drohen/versprechen & zu-infinitive, which represent the highest (third, see Figure 1) level, consists in two distinctive features, one of which they share with scheinen while the other one is specific for drohen and versprechen. The grammatical paradigm of German evidential constructions in PDG which is hierarchically formed upon binary oppositions is represented in Figure 1.

224 Intermediate summary First level of inferential distinctions: features: [+speaker’s inference] paraphrase: P, because Q [Origo has information pointing towards P] = P, because Q [Origo knows Q, and Origo knows that Q entails P] typological label: (conceptual) inferential represented by: werden & infinitive Second level of inferential distinctions: features: [+speaker’s inference] [+experienced information as a basis for inference] paraphrase: P, because Q [Origo knows Q, and Origo knows that Q entails P] + P, because Q [Origo can observe some signs/traces of P] typological label: perceptual inferential represented by: scheinen & zu-infinitive Third level of inferential distinctions: features: [+speaker’s inference] [+experienced information as a basis for inference] [+presently accessible information as a basis for inference] paraphrase: P, because Q [Origo knows Q, and Origo knows that Q entails P] + P, because Q [Origo can observe some signs/traces of P] + P, because Q [Origo has direct access to observed sign/traces of P] typological label: perceptual inferential represented by: drohen/versprechen & zu-infinitive Fourth level of inferential distinctions: as level three plus features: [non-desired] [desired] drohen & zu-infinitive versprechen & zu-infinitive Figure 1: The conceptual structure of the grammatical paradigm of inferential evidentiality in PDG

The four constructions form four levels of conceptual organization; each lower level is defined by the presence of an additional distinctive feature.

German evidential system in a typological perspective 225

The neighboring levels form binary oppositions, differing from each other only with regard to the presence or absence of one distinctive feature. The first three levels reflect genuinely evidential distinctions. The fourth level, on the other hand, contains the features [desired] and [non-desired] which originate from the lexical semantics of the verbs drohen and versprechen. Being structured in this manner, i.e., displaying close formal and conceptual relations and oppositions, the four evidential constructions form a grammatical paradigm as discussed in Section 7.1. Bearing in mind (i) that the grammatical category of evidentiality has been completely newly evolving in the German language and (ii) that the paradigm of inferential evidentiality is still involved in the process of paradigmatic integration (see Section 4.3), it is not surprising that the distinctive conceptual features of the individual members of the German paradigm do not yet exactly correspond to the acknowledged typological distinctions (see Chapter 2). However, at this stage of grammaticalization a strong tendency towards establishing a tightly organized paradigm with internal distinctions which are genuinely evidential can be observed. This tendency is represented in Figure 1 above whereby a typological label is attached to each German infinitive construction.

2. German evidential system in a typological perspective The German evidential system, as described in the preceding section, may also be seen in the broader context of typologically discovered evidentiality systems. In order to do so, we have to take into account not only the closed system of the four inferential evidential constructions, but also its “outside” neighbors. By this we mean the placement of the grammatical category of inferential evidentiality with regard to other possible evidential distinctions. In Section 7.1.2 we pointed out that there are no grammatical markers of direct and mediated evidentiality in PDG. Thus the overall system of evidentiality marking in PDG consists of the marking indirect evidentiality, whereby direct evidentiality is not overtly marked. This situation is typical all over the world, cf.: In most types of evidentiality systems, the visual (or direct, or eyewitness) evidential may acquire a zero-realization, i.e. be formally unmarked. […] (Aikhenvald 2003: 7) The fact that only one language in the present sample does not contrastively mark indirect evidence, as opposed to seven languages that do not contras-

226 Intermediate summary tively mark direct evidence is a further indication that languages treat these categories differently. Perhaps this is because, in many languages, unless otherwise indicated, an assertion is to be interpreted as based on direct evidence rather than indirect. (Willett 1988: 58-9) [original emphases]

Thus we can assume for PDG that the lack of overt evidential markers of direct evidentiality does not necessarily mean that there is no such category at all. Rather, direct evidentiality is not explicitly marked. In other words: as soon as a (sub-)category is established in a language, it is reasonable to suggest that the whole category is present in this language. In a system like that zero-marked elements are interpreted as having a particular – normally the unmarked – value of the category under consideration. Such a situation can be assumed for the domain of evidentiality in PDG. The category of inferential – or more generally: indirect – evidentiality, which is marked in contrast to the sub-category of direct evidentiality, has arisen, thus legitimating the whole category of evidentiality to be present in German. This has been followed by the re-interpretation of all other forms (e.g. the indicative): they are either considered neutral with respect to this category, or unmarked. In this way, an opposition between marked and neutral/unmarked forms comes into being, where the latter can be equally interpreted as an unmarked (or least marked) member of the category, i.e. not longer being localized outside the category. So far, we can conclude that the overall German evidential system consists of two opposed values: zero-marked direct evidentiality (which can be equaled with the value “non marked with respect to evidentiality” due to the privative character of the opposition) versus overtly marked indirect evidentiality, namely inferential evidentiality.3 Although German does not exactly match one of the systems of evidential markers identified in Aikhenvald (2004), it shows striking resemblance to two typologically relevant systems. On the one hand, the German evidential system can be classified as an “A2”-system, following the classification suggested by Aikhenvald (2003, 2004), which is characterized by the values ‘nonfirsthand’ versus ‘everything else’: “The nonfirsthand term typically covers inference based on visible traces and reported information” (Aikhenvald 2003: 4). On the other hand, as the German evidential system does not only distinguish between two – inference and ‘everything else’ – terms, but also differentiate values within the domain of inferentiality it also shares features with the “C1”-system consisting of the following terms: visual, in-

German evidential system in a typological perspective 227

ferred 1 (i.e. inference from direct physical evidence), inferred 2 (i.e. inference from general knowledge), and reported, cf.: In Tsafiki the visual (used for ‘directly witnessed events’) term is formally unmarked; there is one suffix marking information inferred from direct physical evidence, another for inference from general knowledge, and an additional one for reported, or hearsay. A similar system is found in Pawnee. (Aikhenvald 2003: 5)

In German there are no overt grammatical markers of direct evidentiality; there are three constructions with scheinen, drohen and versprechen which approximately resemble the above mentioned category “inference from direct physical evidence”, and the construction werden & infinitive which covers both “inference from direct physical evidence” and “inference from general knowledge”. And, there are (still controversial) grammatical constructions marking reported, or hearsay, evidentiality, i.e., the constructions with modal verbs wollen and sollen (for the discussion see Chapter 2.2.2.3). Hence we can assume that the main distinctions grammatically present in the German category of evidentiality may be summarized as follows: 1.

direct evidentiality (= unmarked with respect to evidentiality status) – formally unmarked (= zero-marked) 2. indirect evidentiality – expressed by periphrastic infinitive constructions; subdivided in: 2.1. indirect inferential evidentiality, consisting of: 2.1.1. werden & infinitive, indicating the (neutral) inferential value 2.1.2. scheinen/drohen/versprechen & zu-infinitive, indicating inference from experienced information 2.2. mediated information – expressed by infinitive constructions with sollen and wollen (to be confirmed yet, see Chapter 2.2.2.3). By way of conclusion we can say the German evidential system, though not fully corresponding to one of the systems attested typologically so far,4 follows the general distinctions which are attested cross-linguistically. Therefore German infinitive constructions may be treated as members of the grammatical category of evidentiality. One specific trait of German inferential evidentiality is noteworthy here: while in most well investigated evidential systems (e.g. Turkish, Bulgarian) inferred evidentials usually introduce past events,5 German inferential markers may introduce past, present, and future events (except for dro-

228 Intermediate summary hen and versprechen, which still show some restrictions of their lexical semantics). The German inferential evidential constructions only indicate the way in which the described situation is acquired by the speaker, which is inference. The temporal status of the described situation is mostly specified by other grammatical or lexical means, cf.: (5)

Das scheint mir ein Spiel mit Worten zu sein/gewesen zu sein… ‘That seems (to me) to be just a game with words/to have been…‘

In PDG the domain of evidentiality can be treated independently of other grammatical domains, such as verbal mood and tense (see Chapter 3). This shows that evidentiality in German deserves its status as a grammatical category of its own.

Chapter 8 Diachronic corpus study of the four constructions

This chapter presents the results of our diachronic corpus research on infinitive constructions with werden, scheinen, drohen and versprechen. The constructions will be described in their chronology from OHG period onward using the theoretical concepts introduced in Chapter 5. As we are dealing with verbs of (relatively) high frequency and high combinatorial potential, a full treatment of all diachronic ramifications is beyond the scope of this study. We will therefore focus on those periods and constructions that are most decisive for the development of the evidential marker in each verb. This means that, depending on the verb, particular diachronic periods as well as particular lexical and constructional variants are treated differently as to weight and explicitness of exposition depending on their relevance for the grammaticalization process under investigation. For werden & infinitive, which is the oldest infinitive construction in the set, the development into an evidential marker is accomplished around 1800, i.e., the period from the 18th century onward will not be looked at in this chapter. As werden & infinitive can be taken to be an instance of “polygrammaticalization”, the status of its function as a future marker and its interrelation with the evidential function is discussed in some detail. The construction scheinen & zu-infinitive, too, will be traced back to its ancestors in OHG to the 18th century, when – as we will show – it was already grammaticalized to a very high degree and functioned in a similar way as it does in PDG. However, as this is a new result of this study, in contrast to werden, a control study with corpus data from the 19th century was conducted for scheinen. This study confirmed that all relevant changes leading to the establishment of the (grammaticalized) evidential construction scheinen & zu-infinitive had occurred before the 19th century. The verbs drohen and versprechen were much later in developing the relevant constructions and evidential meanings. Therefore, for these two verbs, the older diachronic stages up to ENHG are treated with much less detail than it is done for werden and scheinen. Instead, the 18th and 19th centuries, which are the formative period for the grammaticalization of the evidential construction in these verbs, are treated extensively with a large corpus of additional data. As versprechen has a particularly complex ety-

230 Diachronic corpus study of the four constructions mology, we have imported research from literature (DWB and Paul 1992). Our own diachronic analysis starts with the situation at the beginning of the 18th century and looks for those changes which most probably led to the establishment of the construction versprechen & zu-infinitive as an evidential construction. Thus, the development of the evidential construction of each of the four verbs can be tracked in this chapter, designing a diachronic scenario of the grammaticalization of these constructions that shows the individual path of each verb and at the same time reveals significant parallels to the development of other evidential constructions. A survey focussing on the overall development and the formation of an evidential paradigm including the four infinitive constructions will be given in Chapter 9. Before starting the diachronic study proper, some terminological clarifications would be in order. The term “construction” is used as a nontechnical term of everyday language in many places in this study when this does not lead to confusion. However, as linguistic change takes place in specific linguistic surroundings, the term construction in its technical sense has been employed here to describe these phenomena. Therefore, this chapter calls for some additional terminological arrangements: As constructions in the technical sense are always defined as signs, i.e. as inseparable units of form and meaning/function, we need a technical term to talk about those linguistic forms and structures that travel through history taking on different meanings and functions during this journey, i.e. forming distinct constructions although their formal make-up may remain the same. Therefore, in our basic list/classification of relevant linguistic realizations of the items in question, the term “construction” can only have pretheoretical meaning. Alternatively (more precisely but more clumsily) we use the term “syntagmatic pattern”, which just refers to a specific pattern of syntagmatic co-occurrence which is retrievable through time and which does not contain any ascriptions of its semantic and functional status. In other words, these syntagmatic patterns (also “construction types” in nontechnical terminology) serve as a starting point in our diachronic investigation, whereby their particular meaning(s) and their interrelations and interactions will be identified separately for each period of time. In any construction (seen in the technical sense as a form-meaning association) changes may affect the formal side as well as the content side independently, thereby leading to new constructions. Thus, the view taken here obviously is but one of several possible approaches and indeed the

Werden & infinitive 231

mainstream way of looking at change. It is applied in most diachronic and all grammaticalization studies. In traditional terms, it is a semasiological approach as it starts by taking the formal side for granted. We pay attention to changes and shifts in the interpretations of formally unchanged (or only slightly changed) constructions and relate them to specific contexts of usage, some of which vary over time. On the other hand, grammaticalization per definition is the complex change of semantic/functional and formal/structural aspects, i.e. we are confronted with drastic formal changes in the linguistic system. This aspect is accommodated by the concept of a regular succession of distinct construction/context types, which are defined via relevant constellations of form-meaning-usage parameters (bundles of features of form, meaning, and function). These are the “context types” identified in Chapter 5. Beyond this, whenever there is need to speak of some recurrent textual or contextual characteristics that are relevant but do not lend themselves to integration into a constructional scheme (as they are too variable in their formal expression, too far dispersed in the textual and situation surroundings, too inhomogeneous etc.), we use the term “contextual configuration”.

1. Werden & infinitive We first give a brief survey of the diachronic development of the construction werden & infinitive as it has been described in the literature so far. As there is a large number of studies dealing with the history of werden we start with a brief research review which, far from being exhaustive, concentrates on those hypotheses that are relevant for the topic treated here. As noted in Section 4.1, we do not fully agree with the currently dominant opinion that werden & infinitive had first developed temporal function which then gave rise to other – amongst them evidential – interpretations of this construction. Instead, we argue that the evidential function of the construction developed before or in parallel to the temporal meaning of werden & infinitive, so that its evidential meaning cannot be derived from its temporal meaning. In what follows we will present diachronic empirical evidence underpinning our hypothesis.

232 Diachronic corpus study of the four constructions 1.1. Previous work on the development of werden & infinitive Today, the construction werden & infinitive is mostly considered a (grammaticalized) future tense form (see Section 4.1). In accordance with this view, most diachronic studies dealing with werden & infinitive more or less explicitly describe its development as a grammaticalization process of a future tense marker, whereby the temporal meaning is usually seen as preceding other (e.g. modal and evidential) meanings of the construction. In addition to treating the temporal meaning of this construction as chronologically prior to its other meanings, it is often seen as its main or default function and its further meanings as derived. In our view, however, the future temporal meaning of the construction werden & infinitive has always been only one of several alternative interpretations of its core semantics. Further possible interpretations, namely the aspectual, the evidential and some other textually triggered readings, should be best seen as competing with the temporal one rather than having evolved from it. We argue that in the course of the historical development the evidential reading of the construction has prevailed over all other interpretations, including the temporal one. Before showing this, a rough sketch of the development of werden & infinitive is in order: Combinations of werden with bare infinitives first occurred in the 13th century (cf. e.g. Saltveit 1962, Westvik 2000, Diewald and Habermann 2005), i.e. in the late MHG period. At that time the construction had ingressive aspectual semantics and marked the onset of the event denoted in the infinitive verb. In the course of ENHG the construction is said to have developed future tense readings. In this respect it was in competition with the infinitive constructions with modal verbs sollen ‘shall’, wollen ‘will’ and müssen ‘must’. By the end of the ENHG period, however, werden & infinitive finally prevailed over the constructions with modal verbs (cf. Bogner 1989). At about the same time werden within infinitive constructions gradually lost its (indicative) past tense form ward/wurde. Thus the construction werden & infinitive was relatively highly grammaticalized. From that time on, it is supposed to function as a grammatical future tense marker in German, cf. e.g.: In der Verbindung werden mit Infinitiv hat die deutsche Sprache ein Mittel zum Ausdruck des Futurums erhalten, das von jeder modalen Nebenbedeutung befreit ist. [‘In the combination of werden plus infinitive the German language has received a device to express future tense that is free of any modal nuances.’] (Dal 1962: 132)

Werden & infinitive 233 Werden mit dem Infinitiv bezeichnet die reine abstrakte Zukunft. [‘Werden with infinitive designates a purely abstract future.’] (Kleiner 1925: 1) So hatte also die Sprache in ihrer Entwickelung vom Mhd. zum Nhd. ein Mittel gewonnen, das ein reinerer Ausdruck des Futurums war als die Verba sollen und wollen. [‘Hence during the development from the MHG to NHG the language received a device that was a purer expression of the future tense in contrast to the verbs sollen and wollen.’] (Wilmanns 1906: 178)

As mentioned above, other readings, including different (epistemic) modal and evidential interpretations, are usually assumed to be derived form that future reading as a later development. The path of derivation leading from future to modal meaning is typically described as in the following quotes (cf. also Fritz 2000): In derartigen bifunktionalen Verwendungsweisen, die dadurch zu erklären sind, dass Aussagen über Zukünftiges immer mehr oder minder hypothetischen Charakter haben müssen, kann die Brücke von der Temporalität zur Modalität gesehen werden. Plausibel ist ein Kontinuum mit folgenden Phasen: 1. Gewähr durch den Sprecher > 2. sichere Annahme > 3. Annahme > 4. bloße Vermutung. [‘Such bi-functional usages, which are due to the fact that statements about the future are always more or less hypothetical, may be seen as a bridge between temporality and modality. It is plausible to assume a continuum with the following phases: 1. speaker’s assurance > 2. sure assumption > 3. assumption > 4. bare guess.’] (Schmid 2000: 20) With the conventionalization of the future sense, the ground is cleared for new grammaticalizations, and German has in fact followed universal paths of development in this respect. One of these paths involves the reinterpretation of the prediction sense as a modal one in contexts involving durative verbs like sein ‘be’ […] This reinterpretation is likely to have been triggered by an invited inference […] according to which future events are conceptually treated like unbounded states. […] The second path of expansion involved a highly restricted context, namely utterances having second person subject referents, whereby an inference […] was invited, leading to the emergence of a kind of imperative sense, or, more precisely, a form that Bybee et al. (1991) refer to as speaker-oriented modality. […] To summarize, the future sense of the werden-construction has given rise to two modal senses, one of them expressing epistemic and the other speaker-oriented modality. (Heine 1995: 126-127)

As we will show in the remainder of this section, this view is problematic. First, werden & infinitive manifests evidential readings as far as back in ENHG, i.e. the evidential reading co-exists with the temporal reading and

234 Diachronic corpus study of the four constructions does not follow it chronologically. Second, other readings of werden & infinitive such as the text-connective one are already present at that time. Furthermore, there is no agreement concerning the details of the complex development of werden & infinitive. First, there is dispute on the question of whether the infinitive complement of werden is an independent development or whether it should be seen as a result of the phonetic reduction of the present participle, which had combined with werden from the OHG period onward. The second opinion may be called ‘traditional’. It has been adhered to in numerous studies and grammar books (cf. e.g. Weinhold 1883, Behaghel 1924, Ebert 1978, Betten 1987, von Polenz 1991). However, this opinion has been increasingly criticized in the last few decades (cf. e.g. Saltveit 1962, Leiss 1985, Schmid 2000, Harm 2001). The proponents of the latter view see the infinitive construction with werden as an independent development that has little to do with combinations of werden with present participles. The exact details of this development are, however, still a matter of dispute (see Schmid 2000 and Harm 2001 for further details). We follow the suggestions put forward by Wilmanns 1906, which were taken up by Schmid 2000 and slightly modified by Diewald and Habermann 2005. The hypothesis offered there puts major weight on the process of analogical change with semantically close verb-plus-infinitive constructions serving as a model. It rests on the observation that in OHG verbs like beginnan ‘to begin’, gistanten ‘to be situated, to begin’ etc. were frequently combined with an infinitive to express ingressive notions (cf. also Westvik 2000: 237). The meaning of these verbs is maximally close to the ingressive meaning of werden. Therefore we assume that werden & infinitive was analogically formed after the model of the beginnan & infinitive construction and became a member of a relatively closed class of aspectualizing ingressive expressions. The next unanswered question is why werden & infinitive prevailed over competing infinitive constructions with the modal verbs sollen and wollen in ENHG. As the development of future tenses from modal sources is far more frequent in the languages of the world, this question appears to be relevant here. The following characteristics of werden have been considered in the literature as its advantages over the modal verbs: a) The lexical source semantics of the ingressive (inchoative) verb werden shows close affinity with the future temporal meaning (cf. Harm 2001, Diewald and Habermann 2005). b) In contrast to wollen, there is no intentional semantic component in the lexical source semantics of werden (cf. Kotin 2003, who also speaks of

Werden & infinitive 235

the feature “anthropocentric” in this context, Diewald and Habermann 2005; a different view is taken by Heine 1995, see below). c) The absence of modal nuances in the lexical source semantics of werden allows us to postulate a future reference without further overlap with modal meanings (cf. Leiss 1985, Fritz 1997, Dal 1962, Wilmanns 1906). d) A wide range of constructions (lexical as well as grammatical) in which werden participated is assumed to have allowed werden a faster auxiliarization towards a future tense marker (cf. Kotin 2003, Diewald and Habermann 2005). We agree with the opinion that the specific features of werden mentioned above testify to a very different character of this verb in contrast to modal verbs. We do not agree, however, that these characteristics necessarily explain the prevalence of werden over the modal verbs sollen and wollen as far as their functional status as future tenses is concerned. As mentioned above, we regard the temporal meaning of werden & infinitive as one of several alternative readings of this construction, in its synchrony as well as in its diachrony. Within the framework of grammaticalization studies, future tense markers are usually considered to arise via four main channels (cf. Bybee, Perkins and Pagliuca 1994, Ultan 1978). The relevant source structures are the following: 1) aspectual forms, 2) verbs denoting agent-oriented modalities, 3) verbs denoting movement towards a goal, and 4) temporal adverbs. German werden seems to represent the first channel as its lexical source is a verb with aspectual – ingressive/inchoative – semantics. However, it stands quite alone among other languages of the world: From a typological perspective, the German werden future is rather exotic; among the 100-plus languages world-wide we have looked at so far, there is not a single one that has the same conceptual source. (Heine 1995: 125)

In Heine 1995 it is proposed that the development of the construction werden & infinitive is nevertheless in line with universal developmental tendencies mentioned above. Heine (1995: 125-126) reconstructs the evolution of the German construction werden & infinitive, assuming an

236 Diachronic corpus study of the four constructions intermediate stage of an intention sense for it, which he locates chronologically between the ingressive meaning and the future meaning as shown in examples (1) - (4), which are taken from Heine (1995: 126-127): (1) (2) (3) (4)

Ich werde Arzt, ich bin fest entschlossen. ‘I am going to be a doctor, I have made up my mind’. Ich werde kommen, ich bin fest entschlossen. ‘I am going to come, I have made up my mind’. Es wird runterfallen. ‘It is going to fall down’. Es wird wahr sein. ‘That will be true’.

Thus the following developmental path is assumed for werden: ingressive > intention > future > [epistemic/speaker-oriented modality] As many diachronic studies have shown, however, the scenario proposed by Heine 1995 is rather problematic. First of all, the analysis is based upon exclusively synchronic observations. Furthermore the establishment of the intention sense of werden & infinitive at the intermediate stage of its grammaticalization is considered doubtful: Heines Erklärungsversuch ist aber vor allem deshalb abwegig, weil die Grammatikalisierung des Futurs eine Rückstellung der „intention“ neben Rückstellung der aktionalen Markiertheit erfordert […] Es kann kaum angenommen werden, dass ein von vorn herein rückstellungsbedürftiges Merkmal die Rolle eines Bindegliedes auf dem Weg der Entität zur Grammatikalisierung übernimmt und das andere, ebenfalls (aber nicht stärker!) „rückstellungsbedürftige“ Merkmal (Mutativität) eine Zeit lang einfach mechanistisch ersetzt. Werden ersetzt wollen und sollen in dem prospektiven Bereich gerade deshalb, weil letztere ihre Deontik bzw. Voluntarität und somit ihren Anthropozentrismus beibehalten. [‘Heine’s explanation is fallacious in particular because the grammaticalization process of the future tense requires the cancellation of the “intention” sense in addition to the cancellation of the aspectual meaning. […] It can hardly be assumed that a feature, which from the beginning was also bound to be deleted, served as a link in the grammaticalization process and for a period of time merely mechanically replaced another property (i.e. mutativity), which likewise was bound to be cancelled. Werden replaces wollen and sollen in the prospective

Werden & infinitive 237 domain exactly the latter ones preserved their deonticity or volitionality and, thus, their anthropocentricity.’] (Kotin 2003: 174)

Diewald and Habermann (2005: 235) state that werden never fully established the semantic component of intentionality. In a similar way, Hilpert 2007 states: Intentional meanings of werden, which are present in biblical translations from the 16th century, become only typical of more casual uses of werden in later periods of German. (Hilpert 2007: 246)

Our claim – that werden & infinitive has never been a pure future tense marker in the history of German – can be substantiated by the typologically informed observation that aspectual forms as conceptual sources of future tense categories differ in their development from other types of sources (cf. Bybee, Pagliuca and Perkins 1991, Bybee, Perkins and Pagliuca 1994). They are considered to be capable of developing future meanings as “just one interpretation of a much broader semantics”. Aspectual futures develop as imperfectives or perfectives, which happen to accommodate future readings, but they do not develop explicit future semantics. (Bybee, Pagliuca and Perkins 1991: 32)

In the following we will show that this situation holds for the German construction werden & infinitive as well. In the next sections we will reconstruct the development of the construction werden & infinitive as we have retraced it on the basis of an extensive corpus study.

1.2. Werden in the history of German In general, the following five syntagmatic patterns can be distinguished in the diachrony of German: 1) werden as a one-place predicate, functioning as an intransitive verb; 2) werden & nouns and adjectives, functioning as a copula verb; 3) werden & perfect participle; 4) werden & present participle; 5) werden & infinitive. The syntagmatic patterns (1) and (2) have been present all along in the history of German since OHG.1 In its function as an intransitive verb wer-

238 Diachronic corpus study of the four constructions den means ‘to happen, to rise, to occur, to come about, to come into being’, etc. (5)

Fóne demo îo álle geskáfte sínd . únde geríchte uuérdent. (Notker 12971, KALI-Corpus) ‘From this all creatures are and laws come into being.’

Werden as a copula verb usually combines with nouns and adjectives serving its predicative complements. (6)

wir wurden aber sô vrô dâ vôn… (KvHeimUrst 2034, MHDWB) ‘but we became glad thereof.’

The third syntagmatic pattern, i.e. werden & perfect participle, is attested from the OHG period onwards and developed towards a grammaticalized passive construction in the course of MHD. At the end of the 13th century this construction was already highly grammaticalized. (7)

die wurden âne zagen/alle meisteil erslagen/und die andern gevangen (Iwein 6, 3745-3747, taken from Kotin 2003: 189) ‚most of them were struck dead without hesitation and the other [were] arrested’

We will not deal with this syntagmatic pattern in the following (for detailed descriptions of its development see Oubouzar 1974, Valentin 1987, Abraham 1991, Eroms 1990, 1992, Leiss 1992, Kotin 2003). As the grammaticalization process of werden & perfect participle towards a passive construction had been completed roughly at the end of the 13th century, its development could hardly have had direct impact on the development of the construction werden & infinitive which first appeared at about the same time. Werden as a passive auxiliary has completely lost its lexical semantics and thus no longer has a direct semantic connection to the other syntagmatic patterns with werden. It came to function as an auxiliary and can be considered fully independent of its lexical source. This parallel and diverging development of werden into different grammatical categories (passive, evidential marking) is a clear case of “polygrammaticalization”, where “a single form develops different grammatical functions in different constructions” (Hopper and Traugott [1993] 2003: 112). A split occurred, whereby werden & perfect participle, after a short time of being integrated in the paradigm of werden-constructions, left this paradigm and followed

Werden & infinitive 239

its own developmental path. In the following we will refer to the construction werden & perfect participle and its uses only in cases when they are relevant for our primary subject – the construction werden & infinitive. The fourth syntagmatic pattern, i.e. werden & present participle, was present in the German language from the OHD period till the 17th century. The meaning of this construction was ingressive aspectual, i.e. it indicated the initial stage of the process expressed in the present participle. (8)

Tho ward múnd siner/sar spréchanter (Otfrid 9376, KALI-Corpus) ‘His mouth began to speak [lit.: became speaking]’

The construction werden & infinitive has been first attested in the 13th century. Single attestations of this construction have been found as early as in the late OHG period, namely in the texts by Notker, e.g.: (9)

Íh uuárd iu fórn guar chad er . in nâh philologia chélen (Notker, taken from Kotin 2003: 135) ‘A while ago I began – he said – to tell you about philology.’

The initial meaning of this construction was ingressive aspectual, i.e. the construction marked the described event in its onset. In this respect it was conceptually very close to verbal constructions with aspectually marked verbs, e.g. biginnan ‘to begin’, anefahen ‘to start’, gistantan ‘to be situated, to begin’ etc. (10)

Bigonta ez der rîcho man vilo harto zurnen… (Georgslied 248, KALI-Corpus) ‘The rich man began to be very cross about this.’

As we have noted in Section 8.1.1, the construction werden & infinitive was analogically built after the model of such verbal constructions. In PDG the construction werden & infinitive has a wide range of alternative readings (see Section 4.1) and has the inferential evidential reading as its main meaning. Taking the existence and the relevance of the described werden-patterns as a starting point, we analyze their typical contextual surroundings at different historical periods of German in order to better understand shifts in their semantic content. In doing so, we reconstruct the semantic development of the construction werden & infinitive, (i) representing it as a devel-

240 Diachronic corpus study of the four constructions opment which is related to the development of other werden-constructions, (ii) retracing it exclusively on the basis of corpus data, and (iii) adopting a usage-based view on language and grammar.

1.3. Werden in Old High German In OHG only four of the five constructions described above are present.2 It should be noted right at the beginning that all werden-constructions in OHG can be subsumed under one meaning. It is the ingressive aspectual semantics of werden, i.e. indicating the initial stage of an event, situation or state. The only exception is werden as a main verb with the meaning ‘to happen, to occur’, i.e., displaying durative aspectual semantic components as well. This meaning was the original meaning of the etymon werden in the former historical stages of Germanic and can be seen as a case of preservation of the source meaning of this etymon.3 Usages of werden as an intransitive main verb are quite frequent in OHG. The verb is used in the present as well as in the past tense and in all personal forms. This syntagmatic pattern of werden has two interpretations: the ingressive (i.e., ‘to come into being’) and the non-ingressive (i.e., ‘to happen, to take place’), exemplified in (11) and (12), respectively: (11)

(12)

Táz-tir uuírdet . táz zegât óuh . únde uuéhselôt síh (Notker 33514, KALI-Corpus) ‘That which comes into being, passes by and changes.’ Ofto wírdit, oba gúat/thes mannes júngoro giduat… (Otfrid 2955, KALI-Corpus) ‘It often happens that …’

Werden as an intransitive verb occurs normally as a one-place predicate and combines with nouns/noun phrases in the nominative case. Nouns and noun phrases in the dative case are often found in sentences with the intransitive werden, functioning to introduce the benefactive role. This possibility is still available in PDG, though to a far lesser extent. Furthermore werden as an intransitive verb is not restricted to any specific contexts of usage. It occurs in all temporal and personal forms and allows for syntactic variations (e.g. through a facultative benefactive role). Werden as a copula verb combines with nouns and pronouns in the nominative and genitive, with nouns and noun phrases with the preposition zu

Werden & infinitive 241

‘to’, and with predicative adjectives and adverbs. It appears in all inflectional forms. As this variant of werden is still present in PDG, only the no longer extant use with a genitive predicative noun will be exemplified here: (13)

daz sie schebedîg uuurden unde tôdes (Notker 2 319, 9 Piper, taken from DWB 29, 221-276) ‘that they became leprous and moribund’

Copula constructions with werden indicate a change in the status of the described entity: they introduce a new status or a new quality of an entity which had not been attached to it before. In this respect the copula construction of werden is fully integrated in the paradigm of other werdenconstructions. Werden as a copula is not restricted to any specific contexts: it is used in all temporal and personal forms, and allows for syntactic variations (e.g. the adjective can be either inflected or non-inflected). The construction werden & perfect participle is very frequent in OHG. At this time it is already involved in the initial stages of the grammaticalization process towards a passive voice marker. Kotin (2003: 85-96) points out that the change in the syntactic status of the perfect participle – from an adjective-like predicative into an integral part of an analytic grammatical construction – begins relatively late in OHG (i.e., in the texts by Notker). In other OHG texts werden & perfect participle usually allows for a separation into two relatively autonomous parts: the copula-like verb werden on the one hand and the adjective-like perfect participle on the other. Furthermore many perfect participles in combination with werden are inflected, as shown in (14). This fact provides evidence for treating many of OHG instances of werden & prefect participles as a copula-like construction. (14)

(15)

Tho ward ther fáter aler/gótes wihi irfúlter (Otfrid 9526, KALICorpus) ‘The father was filled with all God’s blessing.’ Chindh uuirdit uns chiboran, sunu uuirdit uns chigeban, endi uuirdit siin herduom oba sinem sculdrom… (Isidor 3380, KALICorpus) ‘A child will be born (to us), a son will be given (to us), and his reign will be over his evil…’

242 Diachronic corpus study of the four constructions Such instances of werden & perfect participle are of particular interest. The most relevant observation is that werden & perfect participle initially displayed ingressive semantics. Irrespective of the temporal perspective (either present or past) the construction referred to initial stages of events or situations. In other words: new situations, which have not been present(ed) before, were introduced by the construction werden & perfect participle. A specific semantic feature of this construction is its resultative interpretation which is due to the resultative semantics of the perfect participle. As the grammaticalization process proceeded, werden gradually lost its ingressive component, and the resultative component also disappeared. First signs of this development can already be found in Notker: (16)

tíu uuírt ófto fóne uuíllen ferduuénget (Notker 10875, KALICorpus) ‘she is often prevented by the will’

The use of the temporal adverb ófto ‘often’ in (16) makes an ingressive (and also a resultative) interpretation of the construction rather unlikely. Instead, a habitual (iterative, durative, imperfective) reading is favored here. To summarize, the development of the construction werden & perfect participle towards a grammatical(ized) passive construction is already at an advanced stage at the end of OHG: there are contexts in which the original – ingressive – interpretation is no longer possible. Nevertheless some instances of its use still display ingressive semantics (this holds for at least the early MHD period as well). There are singular attestations of the construction werden & present participle in OHG. In this construction werden behaves like a copula verb, as the construction usually allows for a separation into two relatively autonomous parts: the copula-like werden and the adjective-like present participle. Note that present participles are often inflected. (17)

Thô sliumo uuard thar mit themo engile menigi himilisches heres got lobontiu inti quedentiu… (Tatian 4831, KALI-Corpus) ‘Suddenly the crowd was there with the angel praising God and saying …’

The meaning of this construction is unambiguously ingressive: it introduces a change from some former situation into a new situation or state. Moreover, the situation not only starts but also lasts for some time. The

Werden & infinitive 243

durative meaning component comes from the inherent meaning of the present participle (whereas the perfect participle form marks resultativity, see above). Werden & present participle is used in the present as well in the past temporal perspective. An important contextual restriction for this construction is that – in our data – it combines only with animate subjects. In all other readings werden & present participle seems to be used synonymously to the copula constructions with adjectives. It should be noted that werden is often used in “presentational” constructions in OHG; this usage disappears however in the later periods. The combination of werden with the adverbs tho/dô ‘there, then’ and thane ‘then’ appears mostly in a sentence initial position, cf.: (18)

Uuard thô thaz arfuorun fon in thie engila in himil (Tatian 4915, KALI-Corpus) ‘It happened there that the angels in heaven learned about them.’

The presentational construction indicates either that a new situation emerges, i.e., werden has ingressive meaning, or that a situation holds, i.e., werden has non-ingressive meaning. Importantly, these contexts often serve as a narrative means of text-construction, whereby werden marks a new episode in the narration, introducing a new development or a new turn in the story. In short, in this usage werden may be considered as an explicit marker of new textual information. Interestingly, contextual configurations of werden and the adverbs tho/dô and thanne are typical not only of werden as a main verb, but also of werden as a copula and of werden & perfect participle. Thus we can sate that the introduction of a new situation, i.e., marking of new information in a story, is in line with the common aspectual ingressive semantics of the verb werden.

1.4. Werden in Middle High German In MHG some relevant changes in the overall system of the werdenconstructions may be recorded. In the following we will only discuss those cases where a considerable change (in the structure of a construction, in the contextual uses of a construction etc.) could be detected on the basis of our corpus investigations. First, the construction ward & tho/dô/thanne disappears. This has been seen in connection with the gradual decrease of werden as a main verb (cf. Kotin 2003: 139). As the intransitive variant of werden slowly fades away,

244 Diachronic corpus study of the four constructions the possibility to use it in a variety of syntactic constructions also decreases. We suggest however that it was primarily the slowly vanishing noningressive (i.e. durative) semantic component of the verb werden that has had a considerable impact on this development. Moreover we postulate that the meaning of the OHG ward & tho/dô/thane construction was partly taken over by the MHG constructions with the adverb sô. The first indications for this process can already be detected in the texts by Notker, cf.: (19)

Sô uuírdet sâr uuúnterônnes énde. (Notker 30959, KALI-Corpus) ‘So the amazement comes to its end.’

It should be also noted that we detected a considerable gain in formulaic uses of the main verb werden in phrases with rat ‘advice’, teil ‘part’, buoz ‘substitute’ etc., cf.: (20)

wie möhte des immer werden rât… (KvHeimHinv 771, MHDWB) ‘how one might ever find a remedy …’

The spread of such formulaic lexical uses of werden and the decrease in its absolute uses as a main verb occur at about the same time. This fact provides evidence for the statement that werden as a main verb slowly faded away during the MHG period. While absolute uses of werden gradually leave the “scene”, there is an increasing amount of copula uses of werden in MHG. The formal structure and the meaning of this construction remain roughly the same as compared to OHG. (21)

So er alt wird so wahsit sin oberer snabel. (JPhys 19,16, MHDWB) ‘As he gets elder his upper beak grows.’

In combination with (perfect and present) participles werden often has the status of a copula (or copula-like) verb. However a strong tendency towards the reanalysis of these combinations in terms of verbal periphrases can be observed. Werden & perfect participle continues to establish its grammatical status as a passive construction (for details see e.g. Kotin 1998, 2003, Eroms 1990, Öberg 1907, Schröder 1955). The syntagmatic pattern werden & present participle is much more frequent in MHG than in OHG. Still, it shows the preservation of the OHG structure, being more likely a combination of a copula-like werden and a

Werden & infinitive 245

predicative participle. Whereas werden & perfect participle was reanalyzed into a verbal periphrasis, werden & present participle remained unchanged. In MHG contextual restrictions of this construction are very much the same as in the OHG period. It always requires a grammatical subject, in most cases an animate one. The participles mostly originate from action verbs. Thus the construction usually indicates the beginning of an action carried out by an animate actor. In MHG the construction werden & infinitive comes into existence. Apart from two occurrences found in Notker (see previous section), it was absolutely absent in OHG (as far as we can state on the basis of the data available from that period). In MHG this construction is semantically very close to the construction werden & present participle, so all we have said about werden & present participle equally applies to the construction werden & infinitive. The following examples have been often cited in the literature on werden & infinitive, in order to demonstrate the closeness of these two constructions (cf. Wiesinger 2001: 178, Kotin 2003: 156). Both sentences are taken from the MHG text Iwein by Hartmann von Aue, line 5891. (22) appears in the older version (manuscript) of the text, whereas (23) is taken from the younger one: (22) (23)

Dô sî si vrâgende wart… ‘There she began to ask them [pres. part.].’ Dô sî vrâgen wart… ‘There she began to ask [infinitive].’

The verb werden, both in constructions with present participle and with infinitive, is semantically very close to ingressive verbs such as biginnan, gistantan, and others. These verbs often combine with infinitives (mostly without particle zu) in OHG and MHG. As already mentioned, we propose that werden & infinitive was formed analogically after the model of infinitive constructions with such ingressive verbs. We would like to emphasize that werden & infinitive does not display an inherent future meaning in MHG, though some signs of incipient grammaticalization towards a future tense marker can be observed already in this period (cf. e.g. Kotin 2003: 173, Saltveit 1962: 191, Wiesinger 2001: 182). Indeed, future interpretations of this construction are possible in many local contexts; they nevertheless remain contextually induced interpretations of this construction. We can thus summarize that, apart from the construction werden & perfect participle, which has established its gram-

246 Diachronic corpus study of the four constructions matical status as a passive marker, all other werden-constructions still form a semantically homogeneous system in MHG. They all have ingressive semantics and differ from each other with regard to formal and functional characteristics of the elements combining with werden. At this point, we would like to mention a specific contextual use of werden which we have detected in our diachronic data. We consider this use to be a new and unusual one at that time, and want to emphasize its role in the further development of werden & infinitive. Let us first give some examples for each syntagmatic pattern of werden: (24)

(25)

(26)

(27)

intransitive werden sô ein bâbst gelac tôt/und ein ander wider wart,/der trat zehant an die vart. (Ottokar 258-260, taken from Kotin 2003: 140) ‘as a pope lay dead and another came, he immediately set off’ copula werden und welheu schâf auz dem andern trinkent, die werdent weiz. (BdN 484,15, MHDWB) ‘and those sheep which drink from the other (they will) become white’ werden & perfect participle/present participle dat tout er drîstunt. so wirt er geiunget unte gesehente (JPhys 19,8, MHDWB) ‘he does this for three hours, thus he becomes younger [perf.part.] and able to see [pres.part.; lit.: seeing]’ werden & infinitive iǕt aber daz manz niht vz Ǖchepfet, Ǖo werdent die vnden oben hin in Ǖlahen. (PrBerthKl 2,63, MHDWB) ‘but if it is so that one does not scoop it, (thus) the lower (will) begin to break in at the top’

At first sight the contexts in (24) to (27) seem to have little in common: werden occurs in main clauses as well as in embedded clauses and in relative clauses; it occasionally appears with adverbs such as do and so; it is used in present and past tenses, etc. However, we can observe that these contexts have one important commonality: the meaning of werden cannot be described exclusively in terms of ingressive aspectual semantics, but displays an additional semantic component. Werden not only indicates the beginning of a situation (which is its ingressive meaning), but also marks the situation as resulting from something else. This interpretation of wer-

Werden & infinitive 247

den in MHG (i) is compatible with the ingressive semantics, i.e. it co-exists with and accompanies the ingressive meaning of werden without overriding it; (ii) it arises in local contexts supported by further contextually present linguistic elements; and (iii) it allows for a wide range of semantic specifications of the abstract meaning of sequential relation, which may be a temporal relation as in (24), a causal effect as in (26), or a conditional effect as in (25) and (27). These usages, which share the feature leading to the (contextually triggered) meaning of a sequential relation are instances of untypical contexts for werden, and thus may be seen as pre-conditions for its further grammaticalization. Contexts expressing this kind of sequential relation become even more prominent in ENHG, as we will show in the next section.

1.5. Werden in Early New High German Approximately at the end of ENHG, the development of werden & infinitive towards a future tense marker slowed down, and the inferential evidential meaning of werden & infinitive established itself. Thus, as far as the (potential) future marker is concerned, we may speak of an instance of ‘halted’ or ‘arrested grammaticalization’ (Janda 2001; see also Andersen 1973, Fischer 2000, Haspelmath 2004).4 The relevant instances of werden & infinitive and present participle in ENHG display the following tendency: while their past tense forms with ward/wurde continue to express aspectual ingressive semantics, their present tense forms develop a future temporal reading in some contexts. The following example from our corpus data illustrates the future tense reading of the present tense form of the construction werden & infinitive, cf.: (28)

daß wir nicht anders bishero bekennet haben/nicht anders annoch bekennen/nicht anders hinfuoran bekennen werden/als daß uns erloeset habe die Allerheiligste Dreyfaltigkeit (B 117) ‘that we have confessed in this way so far, we still confess in this way, we will continue to confess in this way that we have been redeemed by the Holy Trinity’

Here werden & infinitive is contrasted with the present and perfect tense forms of the infinitive verb bekennen ‘confess’.

248 Diachronic corpus study of the four constructions On the other hand, werden & infinitive/present participle is very often used in specific contextual configurations, namely after embedding expressions containing verbs of cognition, perception and emotion. This use of werden is listed in the DWB as a particular semantic variant of werden & infinitive: oft findet sich die umschreibung auch beim ausdruck von befürchtungen, hoffnungen und mutmaszenden überlegungen, die sich auf künftiges richten, wobei der sicherheitsgrad der aussage vielfach durch einen entsprechenden ‘vorspann’ eingeschränkt ist. [‘The periphrasis is often found in expressions of fears, hopes and conjectures which are oriented towards future events, whereby the degree of certainty in many cases is restricted by a pertinent opening.’] (DWB, Vol. 29, column 254)

Here is an example from our corpus: (29)

Ach Argenis, Argenis, ich besorge/dieser Schiffbruch wird mehr Todes--Faelle nach sich ziehen. (B 147) ‘Oh Agrenis, Argenis, I am worried that this shipwreck will involve more deaths.’

Importantly, not only the present indicative form of the construction is found in such contexts but also its present subjunctive form in (30) as well as its past tense indicative (31) and subjunctive (32) forms: (30)

(31)

(32)

Das erste/das der Oberste glaubet/der HERR Jesus werde sein Tochter wider lebendig machen/wenn er sie nur mit der hand anruere. (B 135) ‘The first thing is that the Supreme Being believes the Lord Jesus will bring his daughter back to live by touching her with his hand.’ Der Consul war hierüber ubel zufriden/dieweil er sahe/das solche uberlauff den seinigen beschwerlich sein wurden… (B 125) ‘The consul was not satisfied with it while he saw that such a crowd would be exhausting for his people.’ etliche andere aber zuckten jhre Sebel uber uns/das wir anderst nicht gedachten/es wurde stuck und trim-er geben (B 125) ‘quite a few others drew their swords on us (so) that we thought there would be bloodshed and destruction [lit. ‘pieces and debris’]’

The future temporal interpretation of werden & infinitive is indeed possible in all these contexts. However in each case it is overlaid with additional

Werden & infinitive 249

meaning components: most often they concern some aspects of the speaker’s inner world of perception, cognition or emotion. Although this meaning to a considerable extent is motivated by matrix verbs such as believe, see, think, worry, etc., its realization seems to bear a close syntagmatic relationship to werden & infinitive. Hence we interpret the contextual configurations with embedding verbs of cognition, perception and emotion and werden & infinitive in embedded clause as the critical contexts, as in our view, they are the first ambiguous contexts to trigger evidential interpretation. Note that they do not follow temporal readings of this construction in time, but rather exist in parallel to them. In ENHG there are also other patterns of werden & infinitive/present participle which represent non-temporal meaning. First, the ingressive aspectual interpretation of this construction is still prominent, cf.: (33)

do kom ain veintlicher nebel, und bedäkt mit ainer finstriu den luft, da wurdend die wind vast wäen, und von donerslegen ain grozz krachen, do ward der regen giezzen mit den donern und mit feintlichen blitzen. (B 221) ‘there came a hostile fog and covered the air with a darkness, the wind began to blow strongly and the thunderclaps to crash, the rain began to pour with thunders and with hostile flashes of lightning’

The ingressive reading is found only when past (indicative) forms of werden & infinitive are used. It disappears for good in the course of the ENHG time together with this inflectional form. What we can observe in ENHG is that this reading is gradually reinterpreted towards marking a sequential relationship between described situations (see Section 8.1.4). The frequency of the contexts in which werden & infinitive exhibits this new reading, i.e. marking a sequential relation, is relatively high in the ENHG period (for further details see Smirnova 2006: 242-259), cf.: (34) (35)

Was Gott befilcht/das wirdt on dein bitt beschehen. (B 115) ‘What God orders that will happen without your request.’ Weil dann ohne das in deß Menschen Namen und Herz das Sigill der Allerheiligsten Dreyfaltigkeit scheinet eingedruckt zu sein/also wird uns dieses noch ein groessere Anmahnung seyn zur Verehrung dieses Goettlichen Drey. (B 117) ‘While then the signet of the Holy Trinity seems to be pressed in

250 Diachronic corpus study of the four constructions the men’s name and heart, thus it will be a greater reminder to deify the Holy Trinity.’ Such uses of werden & infinitive may have future temporal meaning, indeed, but often they have an atemporal interpretation as in (34) instead. Thus, the marking of a sequential relation between two textually neighboring situations is a firmly entrenched meaning of the construction werden & infinitive in ENHG. We also observe a further development and reinterpretation of the sequential relation reading of werden & infinitive in ENHG. The sequential relationship between two events is gradually reinterpreted as a sequential relationship between premises and conclusions, i.e. as a logical relation between two situations (which is installed by the speaker). As the following examples show, werden & infinitive cannot be given a future temporal interpretation. Werden & infinitive is rather indifferent with regard to its temporal value. Instead, it has inferential evidential reading, indicating that the described situation is based upon speaker’s inference. The main information conveyed by this construction is that the statement made by the speaker is the result of his/her inference process. The inference process itself rests upon certain premises, which in ENHG are still often mentioned in the preceding text. (36)

(37)

Weil nicht zuhoffen steht… daß die jenige so in den Romanen bewandert/und ihre beste Zeit under disen Blaettern verscharret haben/die schaedlichkeit derselben mercken oder erklaeren werden (B 217) ‘While it cannot be expected … that those who are well versed in novels and have spent their best time among these pages will be able to notice and explain their harmfulness.’ So werden andere gelehrte vnd vngelehrte ein jeder gern wissen wöllen/was er bedeute (G Kep) ‘Thus each of other scholars and non-scholars is bound to want to know what he represents.’

A comparison of these contexts to contexts with embedding verbs of perception, cognition and emotion mentioned above reveals some remarkable similarities. First, they both exhibit a strong link to the (inner world of the) speaker insofar as the former refer to the process of reasoning and the latter explicitly mention the speaker’s cognitive or emotional stance. Second,

Scheinen & zu-infinitive 251

they are not restricted to the future temporal perspective: they allow for different temporal locations of the described situations. Third, they are not restricted to the present tense form of the construction werden & infinitive, but concern the whole inflectional paradigm (excluding the disappearing indicative past tense form) of the construction. It should also be noted that we have found combinations of werden with perfect infinitives in our ENHG data. As is well known, such combinations never had pure future temporal interpretations. The fact that such combinations are already present in ENHG underpins our hypothesis about a parallel (and not successive) development of temporal and inferential meaning of the construction werden & infinitive. (38)

Dan wie es die Astronomische rechnung bezeugt/so ist in der nacht nach dem sechsten Aprilis ein conjunctio Jovis Septentrionalis et Lunae dividuae australis gewest in Sagittario, das also ditz khein newer stern/sondern der alte Jupiter gewest sein würt (G Kep) ‘As the astronomical calculation shows there was a conjunctio Jovis Septentrionalis et Lunae dividuae australis in Sagittario in the night after April 6th, so that this would not have been a new star but the old Jupiter’

The inferential interpretation of werden & infinitive is most appropriate here. The evidence from which the speaker (here: the writer) draws his conclusion is explicitly mentioned in the text. The fact that the described event occurred in the past motivates the use of the perfect infinitive of the infinitive verb sein ‘be’. From this empirical evidence we suggest that the establishment of the inferential reading of werden & infinitive was triggered in the following critical context in ENHG: there is an embedding verb of cognition, perception and emotion, which indicates the existence of premises on which an inference process may be based.

2. Scheinen & zu-infinitive In this section we deal with the diachrony of the construction scheinen & zu-infinitive from OHG up to the 18th century with an outlook on the 19th century data for control.

252 Diachronic corpus study of the four constructions 2.1. Scheinen in the history of German In general, we could distinguish between the following syntagmatic patterns with the verb scheinen in the history of German (for an overview of the present syntagmatic patterns of scheinen see Section 6.2): 1) scheinen as a main intransitive verb; 2) scheinen as a copula verb; 3) scheinen with zu-infinitive; 4) scheinen followed by a complement clause (with als, als ob, dass); 5) the parenthetical scheinen. The first syntagmatic pattern, the main verb scheinen, is the oldest one of this verb. It has been present all along in the history of German, counting from the OHG period onward. In its original function as an intransitive verb scheinen means ‘to shine’ and is mostly used with subjects referring to something being able to give off light, cf.: (39)

Lioht thaz thar scínit/inti alla worolt rinit (Otfrid 11869, KALICorpus) ‘The light that is shining and lighting up the whole world.’

As this syntagmatic pattern has had different meanings in the history of German (ranging from ‘to shine, to give off light’ to ‘to be clear, to prove’) it will receive more attention in the next sections. Scheinen as a copula verb usually combines with all types of predicative expressions. It does not refer to a visual effect. Instead its meaning can be best paraphrased with ‘to appear, to seem to be’. This syntagmatic pattern of scheinen has been attested from the MHG period onward: (40)

er was schön und wol gestalt an synen gliedern, er schein starck und kúne ... (Lanc 376, 34, MHDWB) ‘he was beautiful and his limbs were well formed, he seemed strong and bold …’

Sentence (40) exemplifies the gradient transition from the intransitive use of the verb (stark scheinen where stark – used in isolation – could be an adverb) to the copula use (kühn scheinen where kühn favours an interpretation as a predicative element reducing scheinen to a copula); in (40) the copula use is the most plausible one for both complements.

Scheinen & zu-infinitive 253

Often, the copula scheinen is used with (free) dative participants referring to the observer to whom something or somebody appears in this or that way, cf.: (41)

Wanne daz selb dinch scheinet uns greozzer, so ez nahent ist, danne so ez verre ist. (KvMSph 13, 10, MHDWB) ‘The same thing seems to us bigger, when it is nearby than when it is far away.’

The third syntagmatic pattern, scheinen & zu-infinitive, appears first in ENHG and at the beginning is restricted to the infinitives of sein ‘to be’. Later on, further infinitive verbs come to be used with scheinen. As this construction will be discussed in greater detail in the following sections, the following example may suffice to exemplify this use: (42)

Was da scheinte mein unglück zu seyn. (Simpl. 3 (1684), 669, taken from DWB 14, 2442) ‘What seemed to be my bad luck.’

In the fourth syntagmatic pattern, scheinen usually embeds a complement clause with dass ‘that’ or als ob ‘as if’. This syntagmatic pattern also displays different meanings in the history of German. It has been pointed out that it has been attested from the 18th century onward (cf. Paul and Henne 1992, DWB). As our corpus data reveal, however, it already existed in the MHG and OHG period (see 48). (43)

Nun scheinet daß M. G. Herr zugebe daß im nachdencken und erforschen sich schöhne vortheil finden (BA Leib 4) ‘Now it seems that M. G. Herr admits that there are lovely advantages in thinking and investigating.’

The parenthetical scheinen in constructions like (mir) scheint ‘it seems (to me)’ or (so) scheints ‘(as) it seems’ has the status of an adverbial and does not show any syntactic connection to the clause it is used in (44). In our corpus data this syntagmatic pattern, which probably developed out of syntagmatic pattern 4, occurs first in ENHG. It will not be treated further here. (44)

Es war, wie es schien, niemand schoener, niemand geschickter und vollkommner, als ich. (BA Gel)

254 Diachronic corpus study of the four constructions ‘There was, as it seemed, nobody more beautiful, smarter and more perfect than me.’

2.2. Scheinen in Old High German In OHG the first syntagmatic pattern of scheinen represents the most common use of this verb. Neither scheinen as a copula nor scheinen with infinitival complements are found in this period. In our OHG data we registered 27 occurrences of scheinen all told. 15 (i.e., 55,6%) of them are clear cases of intransitive scheinen with the meaning of visual effect. Scheinen is either literally used as a verb of visual appearance, as in (45), or it is used metaphorically to describe a (visual) effect, as in (46): (45)

(46)

daz kemeine ring ist. allero dero anderro sternon. die in_heiterero naht skînent. (Notker 18496, KALI-Corpus) ‘there is the common ring of the other stars which shine during bright nights’ Bluot skein in uuangon (Ludwigslied 492, KALI-Corpus) ‘the blood shone in cheeks’

Moreover, scheinen is used (3 out of 27 occurrences) with the preposition an ‘at, on’, cf.: (47)

An allen dien rationibus. skînet tero guoton maht. unde dero ubelon guissiu unchraft (Notker 22687, KALI-Corpus) ‘In all these reflections the power of good and the sure powerlessness of evil shows up’

Here, scheinen can be interpreted as a verb of appearance, meaning ‘to appear’, as it refers to something which shows itself (clearly). On the other hand, scheinen can be read as meaning ‘to be recognized’. Importantly, scheinen does not have any nuances of non- or counter-factual status of the described situation. Rather the verb is used to emphasize the clearly factual status of what is being said. Further occurrences of scheinen may be interpreted in this manner as well. In 9 out of 27, i.e. in one-third of all instances scheinen is followed by a complement clause with daz ‘that’, uuio ‘as’, cf.:

Scheinen & zu-infinitive 255

(48)

Unde skînet. taz platonis reda uuariu ist. (Notker 22716, KALICorpus) ‘And it is evident [lit. ‘shines’] that Plato’s words are true.’

These findings obviously contradict the observation mentioned in the last section that scheinen with complement clauses appears only from the 18th century onwards. As in Notker, these instance translate the Latin unpersonal predicate liquet ‘it becomes obvious’ and thus are ‘loan translations’, we will not discuss this construction any further, although they obviously have been handed down to MHG (see next section).

2.3. Scheinen in Middle High German Some changes may be detected in MHG, although the overall situation is fairly similar to OHG. Scheinen is still very common as a main intransitive verb denoting a visual effect emanating from entities like the sun, the morning, jewels, swords etc. Scheinen is also used with a more abstract meaning, expressing the obvious status of factual situations. In this reading, scheinen combines with subjects referring to abstract entities, e.g.: (49)

Dô got daz gebôt, sô der chôr wart zestôret. dô scein der gotes gwalt (Gen 35, MHDWB) ‘As God had demanded so the chore was destroyed. There shone God’s power’

A relevant innovation in MHG is that scheinen starts to be used very frequently as a copula. The number of such uses is relatively high (ca. one third) in our MHG corpus data. Scheinen as a copula combines with various types of predicative expressions from the very beginning: with noun phrases, adjectives, prepositional phrases, perfect and present participles. The copula scheinen is sometimes accompanied by a free dative object which refers to the person seeing or in some other way perceiving the situation. Furthermore the copula scheinen often occurs in contexts where the appearance of an entity (the subject of the sentence) is compared to another entity that usually has the ability to shine (which is normally gold), as in (50). Scheinen as a copula occurs also in other contexts of comparison where the way in which an entity or person appears (to some observer) is explicitly compared – or confronted – to the way it (really) is, cf. (51).

256 Diachronic corpus study of the four constructions (50) (51)

Sin schilt schein als ein golt. (Herb 9977, MHDWB) ‘His sword shone as gold.’ So schinent si uswendig wise und sint doch alle leider inwendig toren. (Mechth 5: 8, 35, MHDWB) ‘Thus they seem wise outside and are however silly inside.’

In such contexts the original lexical meaning of scheinen of emanating a visual effect is still present, however the comparative structure allows a conversational implicature focussing on the difference between real existence and visual impression. Thus, this use of the copula scheinen may be seen as an untypical context. The next section will show that it gained in importance and became very frequent in ENHG. It is particularly relevant that in MHG specific structures appear and are spreading which can be best described as a constructional fusion of the old structure as represented in (48) with the (newly developing) copula verb scheinen. In such contexts scheinen partly has the formal characteristics which are typical for a copula, i.e. it combines with a subject predicative. In the vast majority of the examples found in our corpus data the predicative element is the adverb wol ‘well, certainly, indeed’. At the same time the formal similarity to OHG constructions with the complement clauses is obvious. (52)

(53)

(54)

ez schînet wol… daz mîn vrouwe ein wîp ist… (Iw 3127, MHDWB) ‘It is indeed obvious that my wife is a woman’ Das schinet uch an uwerm libe wol das ir darumb nit enwißent. (Lanc 283, 12, MHDWB) ‘It is also obvious from your body that you do not know about it.’ wie wol daz an im schein,/daz er an freuden was verzaget! (Eracl 3140, MHDWB) ‘how well it seemed from him that he gave up hope of being happy’

As the above examples also show, scheinen co-occurs with prepositional phrases with an (see (53) and (54)), similar to the OHG structures described in the previous section. The prepositional phrase refers to something which serves as a kind of proof for the statement made in the complement clause. We suggest that these constructions – together with its context sensitive interpretation – can be seen as the first predecessors of

Scheinen & zu-infinitive 257

the present evidential meaning of the construction scheinen & zu-infinitive. In these specific contextual configurations scheinen shows first signs of a development towards a secondary predication, as it qualifies the statement of the complement clause as obvious and without doubt. It would be premature to say that scheinen already exhibits an evidential meaning, however. At this point we can speak of a new untypical context of scheinen with its unique formal as well as semantic characteristics. This construction has the form [scheinen & (anPP) & (wol) & daz] and the meaning ‘(from this) it is obvious that’ or ‘(from this) one can clearly see that’. Further evidence for the existence of this specific construction type of the lexical verb scheinen in the MHG period is provided by examples, where scheinen appears as an infinitive complement of the verb lassen ‘to let’. The combination with lassen may be paraphrased with ‘to make sth. clear’, ‘to show’: (55)

dô lie aber schînen/der grüene ritter, wer er was. (UvZLanz 2960, MHBWB) ‘The green cavalier made it look obvious who he was.’

We can conclude that scheinen undergoes two relevant changes in MHG: (i) It develops a copula verb use, whereby it combines with all types of predicative expressions. (ii) It comes to be used in a new construction [scheinen & (anPP) & (wol) & daz] with a specific meaning which can be seen as a first predecessor for the present evidential meaning.

2.4. Scheinen in Early New High German The most relevant changes towards the establishment of the evidential construction scheinen & zu-infinitive happened in ENHG. As has been pointed out by several authors (cf. Ebert 1976: 41f., Behaghel 1924: 319, Askedal 1998, Diewald 2000, 2001), at this time scheinen first comes to be used with infinitive complements. Before we deal with this development in more detail, let us give a brief survey of the constructional variation of scheinen found in our ENHG data. Scheinen is still common in its original lexical variant as a main intransitive verb, though with decreasing frequency. The copula scheinen is the most frequent syntagmatic pattern in that period of time. From this we can conclude that scheinen had established its copula function among its stan-

258 Diachronic corpus study of the four constructions dard uses. It should be noted that scheinen as a copula exhibits a large semantic diversity when it is used in different contexts. Its meaning ranges from the expression of the appearance of something/someone, as in (56), to the indication of a perceptional effect of something/someone to someone, as in (57), and further to the expression of a deceptive or delusive appearance of something/someone that is explicitly contrasted to its actual being, as in (58). These manifold interpretations of scheinen as a copula seem to be new for this verb. Neither in MHG nor in OHG did scheinen have such a variety of readings. (56)

(57) (58)

er schein ein starker svfer man…, vnd schein wol vffe sine fvnfe vnd fierzig iare. (B 231) ‘he seemed a strong neat man … and seemed well in his forty-fifth year‘ Die Zeitung scheinet mir zu unglaublich. (B 147) ‘The message seems (to) me too unbelievable.’ es sind noch einige kleinigkeiten übrig, welche so nüzlich seyn als sie gering scheinen. (BA Leib 1) ‘there are still some small things left, which are as useful as they seem trivial.’

The construction [scheinen & (anPP) & (wol) & daz] described in the last section continues to exist in ENHG where it displays a much lower frequency and slight shift in form and meaning. First, it takes embedded complement clauses with als wenn and als ob ‘as if’: (59)

Es scheinet/als wenn diese Tapfferkeit mit keinem geringen Lohne werde zufrieden seyn. (B 147) ‘It seems as if this bravery will not be satisfied with a lesser wage.’

Second, scheinen combines with complement clauses with dass ‘that’ with the subordinate verb used either in its subjunctive or in its indicative form. (60)

Es ist auch kein wunder, wie mans damit macht, denn offt scheint daß mans nicht brauchen wolle hinter die wahrheit zu kommen... (BA Leib 4) ‘This is no wonder how one does it, because (it) often seems that one does not want to use it to get to the truth of the matter …’

Scheinen & zu-infinitive 259

Furthermore the old constructional pattern can be recognized in sentences with complement clauses without complementizer: (61)

Diese leüte sind ohne erregung und feuer, es scheint, sie seyen zwar aus der Adamischen erde gemacht, allein der Geist des lebens sey ihnen nicht eingeblasen worden. (BA Leib 1) ‘These people are without any excitement and fire, it seems, although they are made from Adam’s earth, the spirit of life has not been blown into them.’

Finally, scheinen is used parenthetically as in the following example: (62)

Dan obe woll der Herr daselbst im anfang seins Sermons (wie es scheint) allein von der Geistlicher/vnd nit Sacramentalischer Niessung redet (B 155) ‘Then although at the beginning of his sermon the Lord himself speaks (as it seems) only about the spiritual and not about the sacramental pleasure.’

In all these structures, which first evolve in ENHG, scheinen has a slightly different semantics compared to the meaning of the MHG construction, from which we suppose it to have arisen. Namely, their meaning is very similar to the meaning of the copula verb scheinen and from the assertion of the obviousness (factuality) of a situation to the expression of its doubtfulness, i.e. its mere appearance (to someone). An additional note is in place here concerning the semantic changes of scheinen. As has been noted in the previous section, the prepositional phrase with an in the MHG construction [scheinen & (anPP) & (wol) & daz] could be potentially interpreted as indicating evidence for the statement formulated in the complement clause. Though the construction itself has been formally disintegrated, i.e. it was replaced by several new constructions of ENHG, its semantic content has remained roughly the same. In (63), for example, it is the parenthetical prepositional phrase with nach ‘according to’ that refers to visually perceived evidence. Thus the relational meaning of scheinen, i.e. the expression of a logical connection between two textually mentioned situations, is still preserved in the new developed constructions in the ENHG period, too. (63)

so ist gewiß, daß sie, eusserlichem ansehen nach, scheinet, ob ihr gantzes Hertz nichts als Andacht, Traurigkeit vnd Kyrieleyson wä-

260 Diachronic corpus study of the four constructions re. (B 237) ‘thus it is certain that she, according to her outward appearance, seems as if her heart consists of nothing but devotion, sadness and Kyrie eleison.’ Now, let us turn to the infinitive construction with scheinen which first appear in the 16th century. The only infinitive verb scheinen first combined with was the verb sein ‘to be’ with or without the particle zu ‘to’ (see (64) and (65) respectively, which are taken from one and the same text). (64)

(65)

Wiewol es nun gar frembdt vnd schier vnglaubbar scheint seyn/ist es doch mehr denn warhafftig (B 245) ‘Although it seems to be very strange and almost unbelievable, still it is more than true’ aber ich habs erfahren/daß es keine Marcasita seynd/sondern scheinen Goldtreiche Mineralien zu seyn/vnnd seyndt nichts anders (B 245) ‘but I learned that they aren’t Marcasita but seem to be golden minerals and are nothing else’

The appearance of scheinen with infinitives can possibly be traced back to the copula uses of this verb and analogical extensions using werden and its constructional spectrum as a model (also cf. Diewald 2000, 2001; Dal 1966: 102ff). Thus, a new syntagmatic pattern of scheinen is established, namely the construction [scheinen & (zu) & sein]. We suggest that [scheinen & (zu) & sein] is the critical context for the development of the inferential construction, which means that it served as a kind of catalyst triggering the development of a more general constructional pattern scheinen & zu-infinitive with all types of infinitives. That is, the construction [scheinen & (zu) & sein] opened a door for another development, i.e., for the grammaticalization of the inferential evidential construction. Let us first quickly describe the changes on the formal side of this construction. Already in the 17th century scheinen shows combinations with infinitives other than sein: (66)

so scheinets doch im ubrigen nicht ganz ungefahr zu kommen… (B 257) ‘Though for the rest it seems to be [lit.: come] not without reason’

Scheinen & zu-infinitive 261

(67)

Und weiln die Teutsche Sprache vor vielen andern dem Ursprung sich zu nähern scheinet, so sind auch die Grund-Wurtzeln in derselben desto besser zu erkennen (G Leib 5) ‘And while the German language seems to approach the origin prior to all other [languages], the better the basic sources of the others can be recognized.’

The following example from our corpus provides strong evidence that the construction [scheinen & zu-infinitive] originates from the copula construction. In (68) scheinen combines simultaneously with an adjective heilig, i.e. it is used as a copula, and with a zu-infinitive zuverehren, i.e. represents an instance of the new infinitive construction: (68)

Und diese Ehre begleitete von außen ein vortrefflicher Ruhm bei Fremden/eine Verwunderung von Abgelegenen/so daß HORST mit Ruhm und Preiß wie uberschuttet lebte/HORSTENS Nahme allen Musen heilig und zuverehren schiene. (B 257) ‘And this honor was accompanied outside by admirable glory from strangers and by astonishment from remote people, so that Horst lived as if he were overwhelmed with glory and awards, Horst’s name seemed to all muses holy and to be adored.’

Not only zu-infinitives of verbs other than sein come to be used with scheinen. In (69) scheinen combines with a passive infinitive, in (70) it is used with a perfect infinitive, in (71) a complex infinitive phrase with a modal verb wollen serves as an infinitive complement of scheinen. (69)

(70)

Jn Metaphysicis und Physicis, welche doch zu seiner Profession so genau erfordert zu werden nicht schienen/wuste er die subtilesten Streitigkeiten zuerzahlen/ja was erzahlen? mit scharfem Urtheil zu entscheiden. (B 257) ‘In metaphysics and physics, which seemed not to be required for his profession that exactly, he knew…’ alle Lieblichkeiten schienen ihre Hande in seiner verewigten Dinte abgewaschen zu haben. (B 257) ‘all the loveliness seemed to have washed its hands in his eternalized ink’

262 Diachronic corpus study of the four constructions (71)

Der Sterbende schien sich noch einmal aufrichten zu wollen… (BA Gel) ‘The dying man seemed to try and raise himself up once again.’

Such a rapid spread of the constructional pattern scheinen & zu-infinitive is indeed very curious and has been neither satisfactorily understood nor explained so far. One possible explanation for this has been offered in Diewald 2000. It is suggested that the grammaticalization process of the construction scheinen & zu-infinitive was to a large extent influenced by the grammaticalization process of the deictic variants of the German modal verbs. Since the constructional pattern MV & infinitive was already grammaticalized to a high degree in ENHG, scheinen could easily form similar structures due to analogical transfer. Here we would like to offer a slightly revised explanation. The grammaticalization process of the German modal verbs towards (modal deictic) factuality markers was without any doubt an important factor in the development of infinitive constructions with scheinen, as they represented the same formal constructional pattern. However scheinen has not developed towards a grammaticalized factuality marker but towards an evidential marker. From this point of view the development of the construction werden & infinitive most probably served as a model for analogical transfer. Since werden & infinitive is formally very similar to modal verbs, the above presented hypothesis is still plausible. As, however, werden has evidential semantics and has developed towards an inferential evidential marker, it is more reasonable to suggest that scheinen was formed analogically to werden. To conclude, the development of the construction scheinen & zuinfinitive in ENHG can be described as a series of analogical extensions, leading from the copula uses to the development of the infinitive construction. After the construction [scheinen & (zu) & sein] was built analogically to other copula verbs, the next analogical transfer was made possible. In analogical to the modal verbs and especially to the construction werden & infinitive scheinen came to be used with infinitives of other verbs as well. As for the semantics of the verb scheinen in ENHG, it can be said that it still remains rather inhomogenous. Apart from the main intransitive verb, where scheinen preserved the meaning of visual effect, in all other syntagmatic patterns it developed a more complex meaning, including various nuances (and therefore permitting variable implicatures depending on the context of usage) from indication of the factual status of the described event to the expression of the mere appearance (which may reveal either the real or the ‘pretended’ status of the described event). This – to some

Scheinen & zu-infinitive 263

extent underspecified – meaning of scheinen is shared by (i) the copula verb, (ii) scheinen with complement clauses, (iii) the parenthetical scheinen, and (iv) scheinen & infinitive. 2.5. Scheinen in the 18th century In the 18th century each of the syntagmatic patterns described in the previous sections – though remaining stable formally – establishes its own, i.e. specialized, meaning. We begin with scheinen with complement clauses. The formal realization of this syntagmatic pattern becomes relatively stable and remains so till today. Scheinen typically takes the formal subject es ‘it’ and is followed by a complement clause with the verb in the subjunctive (either subjunctive I or subjunctive II). The complement clause is introduced by the conjunction dass ‘that’, als ‘as’, or als ob/als wenn ‘as if’, or it appears without any conjunction at all. The same formal distribution of the pattern can be observed in PDG (see Section 6.2). The meaning of this syntagmatic pattern underwent specialization from the ENHG period onwards. From the 18th century onwards it has referred to an impression (of an observer) which is produced by some entity or situation. The impression may be more or less specific, it may also be more or less plausible; it can further coincide with reality or may contradict it. Scheinen is normally chosen to mark the perspective of the witness of the situation (who may be the speaker, too). That is, the situation is represented from the point of view of an observer. In this construction scheinen can be used to introduce visually (or in some other way directly) perceived situations as well as to describe opinions and personal beliefs. This meaning may be paraphrased as follows: “from my/someone’s point of view, p.” Compared to the earlier historical periods, scheinen lost its affirmative meaning, i.e., it cannot be used to emphasize the obvious and unchallenged status of events anymore. It furthermore gave up the meaning of universal validity of facts. Now it expresses the (point of) view of an observer. The copula verb scheinen, including its uses as a compound copula with added infinitive sein and some of its uses with other infinitive verbs,5 on the other hand, has preserved the ambiguous meaning. It can be used in its neutral meaning referring to the fact that something appears (to somebody) in a particular way, i.e., without further evaluation of this appearance. In

264 Diachronic corpus study of the four constructions other words, it describes an impression that is caused by something (which is the subject of the sentence) by means of its visual or other kinds of appearance. Usually, there is an observer who has this particular impression, but s/he does not have to be mentioned. This core meaning of the copula scheinen may be paraphrased as follows: “something/someone is perceived (by an pbserver) as X.” Three different interpretations of this core meaning of the copula scheinen may be distinguished. First, the observer may be explicitly mentioned by means of a (free) dative object. In this case, the personal (subjective) impression of the observer is expressed. This reading corresponds to the present reading of the construction which we named “subjective” or observer-based scheinen in Section 6.2.2. (72)

Nun schien es aber dem Herrn von Mehlfeld Zeit zu seyn, seinen Kopf gänzlich aus der Schlinge zu ziehn und zugleich den Neid der übrigen Dienerschaft seinem Gegner auf den Hals zu laden. (Knigge, DE-Gutenberg) ‘Now it seemed to Herr von Mehlfeld to be the right time to pull his head out of the noose…’

Second, the copula scheinen may be used to express the illusory appearance of something/someone. In this case scheinen is often used in contexts where the real state of existance of something is explicitly confronted with its illusory appearance (this meaning corresponds to the present meaning of scheinen which we described as “make-believe” scheinen, see Section 6.2.2). Another possible interpretation is that the illusory appearance is intended by the described person. (73)

Du kannst zur rechten Zeit zerstreuet scheinen, wenn Du etwas nicht hören, auf etwas nicht antworten willst. (Knigge, DEGutenberg) ‘You can seem distracted at the right time, if you don’t want to hear something or to answer a question.’

Third, the copula scheinen may introduce situations about whose factual status there is no doubt. In such cases scheinen has evidential meaning

Scheinen & zu-infinitive 265

comparable to the meaning of the present evidential construction scheinen & zu-infinitive. The construction scheinen & zu-infinitive, which, as we have shown in the previous section, originated from the copula scheinen, manifests a similar semantic ambiguity, i.e. persistence (cf. Section 6.2 for persistence in the evidential construction scheinen & zu-infinitive in PDG). It is reasonable to assume that in the 18th century the persistence in the infinitive construction was even stronger than it is in PDG and that the construction was grammaticalized to a lesser degree as compared to PDG. For example, in the 18th century, scheinen & zu-infinitive still occurs in analytical tenses and moods and sometimes functions as an infinitival complement. Moreover it still allows for the extrapolation of the infinitive, which is no longer possible in PDG (see Section 6.2): (74)

(75)

… das alles kontrastierte so sonderbar mit der Absicht, derentwegen er sie besuchte und welche sie durch die höfliche Annahme seines Briefes und die darauf ertheilte mündliche Antwort zu begünstigen geschienen hatte (Knigge, DE-Gutenberg) ‘All this contrasted so curiously with the intention with which he visited her and which she, having politely accepted his letter and given him an oral answer, had seemed to abet.’ Leider bin Auch ich ein Ding von vielen Seiten, die Oft nicht so recht zu passen scheinen mögen. (Lessing, DE-Gutenberg) ‘Unfortunately I am also a thing of many sides which may often seem not to match properly.’

As early as in the 18th century scheinen & zu-infinitive shows clear indications of its semantic development towards an inferential evidential marker. The lexical meaning of scheinen is already bleached to a considerable degree. As exemplified in (76), scheinen may combine with infinitives which are very close to the original lexical meaning of scheinen (here, it is the verb hervorleuchten ‘to shine’): (76)

Das Kind auf ihren Armen ist ein Kind über gemeine Kinder erhaben, durch ein Gesichte, aus welchem ein Strahl der Gottheit durch die Unschuld der Kindheit hervorzuleuchten scheinet. (Winckelmann, DE-Gutenberg) ‘The child in her arms is a child that is above all usual children,

266 Diachronic corpus study of the four constructions due to a face out of which a divine ray seems to shine through the innocence of childhood.’ Furthermore scheinen comes to be used with first person subjects, whereby the subject does not control the verbal action. The inferential evidential meaning of the construction is evident here: (77)

Unterdessen bat er mich doch, einige Zeit bey ihm zu bleiben, wurde nach und nach vertraulicher gegen mich, ich schien ihm zu gefallen, und endlich both er mir sogar an, mich als Gesellschafter bey sich zu behalten. (Knigge, DE-Gutenberg) ‘Meanwhile he asked me to stay with him for a while, he got more and more intimate towards me, I seemed to please him…’

There are many examples in our corpus in which scheinen & zu-infinitive has direct evidential meaning, i.e. it indicates that the speaker makes his/her statement on the basis of some direct (perceptual, visual) input. (78)

Sie schien etwas Leidendes und Furchtsames in ihren Gesichtszügen zu verrathen. (Knigge, DE-Gutenberg) ‘Her countenance seemed to reveal suffering and timorousness.’

Infinitive verbs with abstract semantics, i.e. verbs which do not denote any particular action or state, come to be used with scheinen. This was not the case in the earlier periods: in ENHG scheinen combined only with verbs with specific semantics of action such as gehen ‘to go’, kommen ‘to come’, verändern ‘to change’ etc. In the 18th century verbs of different kinds are found in combination with scheinen. Such infinitive verbs as deuten ‘to suggest’, andeuten ‘to hint at’, anzeigen ‘to indicate’ in particular point to the fact that scheinen is involved in the process of semantic reinterpretation, whereby its lexical semantics is bleached. The unambiguously (inferential) evidential reading of scheinen & zu-infinitive is illustrated by the following examples: (79)

Alles, wonach du strebst, ist neu und schön und groß, aber du scheinst es nicht zu wissen... (Wackenroder, DE-Gutenberg) ‘Everything you strive for is new and beautiful and big, but you seem not to know this.’

Scheinen & zu-infinitive 267

(80)

(81)

Er schien mir vollkommen zu gehorchen; es war ihm unmoeglich mir etwas abzuschlagen; er hielt alles fuer genehm, was ich verlangte. (BA Gel) ‘He seemed to entirely obey me; it was not possible for him to refuse me something; he regarded everything I asked for as pleasant.’ Die schönern Arten von Tieren unter andern Himmelstrichen scheinen ihnen nicht bekannt gewesen zu sein; wenn man aus einzelnen Fällen, von dem Pferde des Marcus Aurelius, von den beiden Pferden in Monte Cavallo, ja von den vorgegebenen Lysippischen Pferden über dem Portal der St. Markuskirche in Venedig, von dem Farnesischen Ochsen und den übrigen Tieren dieses Gruppo, schließen darf. (Winckelmann, DE-Gutenberg) ‘They seem to have been unfamiliar with the finer species of animals under other skies; if one may infer from individual cases, from the horse by Marc Aurel, from both horses in Monte Cavallo…’

While in (80) and (81) the premises, i.e. pieces of information used by the speaker as a basis for his/her statement, are explicitly mentioned, in (79), there is no textual explication of information available to the speaker. As can be seen from the examples above the meaning of scheinen & zuinfinitive is in many ways different from that of the copula verb scheinen. The infinitive construction does not show a great diversity in its readings, having only evidential meaning as its basic semantics. It indicates that there is some evidence which is used by the speaker as a basis for his/her statements. Thus, we may state that the evidential meaning – direct as well as inferential – developed and established itself already in the 18th century. This meaning of scheinen & zu-infinitive in the 18th century may then be described as follows: “due to some pieces of information available to the speaker, p.” To summarize, in the 18th century scheinen & zu-infinitive had already reached a critical stage in its grammaticalization towards an evidential marker. With regard to its formal characteristics it is still very close to the copula construction with scheinen. However, obvious signs of grammaticalization may already be detected. As regards the semantics of scheinen within infinitive construction, it is already bleached to a considerable de-

268 Diachronic corpus study of the four constructions gree. Furthermore, in addition to the direct evidential value, which scheinen & zu-infinitive inherited from the copula construction, it establishes the inferential evidential meaning.

3. Drohen & zu-infinitive 3.1. The development of drohen & zu-infinitive: a short survey As we generally agree with the scenario offered in Heine and Miyashita 2008 for the diachrony of the German construction drohen & zu-infinitive we would like to give a brief review of this study before presenting the results of our own reconstruction work for this verb. Heine and Miyashita treat the developmental process of drohen & zu-infinitive towards a “functional category” as an instance of grammaticalization, with the lexical variant of drohen preceding functional drohen in time, and the latter originating from the former. The authors propose four successive developmental stages for grammaticalized drohen each of them is connected with a particular construction type, which have been already expounded in Chapter 6.3.1 and are briefly recapitulated here: (82) C1: Karl droht seinem Chef, ihn zu verklagen. ‘Karl threatens his boss with a lawsuit.’ C2: Uns droht eine Katastrophe. ‘We are now threatened by a disaster.’ C3: Das Hochwasser droht die Altstadt zu überschwemmen. ‘The high water threatens to flood the town.’ C4: Mein Mann droht krank zu werden. ‘My husband threatens to fall ill.’ The overall developmental scenario hypothesized is the following (cf. Heine and Miyashita 2008: 78): C1 > C2 > C3/C4 The historical development of drohen is thus divided into four successive stages which are associated with the constructions introduced above. Let us briefly survey these four stages of grammaticalization. At the first stage,

Drohen & zu-infinitive 269

covering the OHG and MHG periods, only C1 is attested, i.e. drohen is used as a lexical verb. There are already some occurrences of drohen with inanimate subjects in Middle High German, but they are not suggestive of a pattern; rather, they appear to be individual instances of metaphorical transfer, where human agents serve as metaphorical vehicles for certain undesirable forces. (Heine and Miyashita 2008: 79)

At the second stage, in ENHG, the new construction C2 begins to emerge: “Rather than standing for a human or animate concept, the construction begins to be extended productively to contexts with inanimate subject referents” (Heine and Miyashita 2008: 79). In parallel to this, after 1600 AD, the drohen-construction, which during the 16th century continued “to be associated to a large extent with events referring to temporally unrestricted situations not implying any change of state” (Heine and Miyashita 2008: 80), becomes increasingly restricted to contexts referring to changes. This important innovation was accompanied by a second one, i.e. the loss of the speech act semantics of the verb drohen. At the third stage, namely in the 17th century, the construction C3 emerges. In the 18th century it establishes itself as a “distinct category”. At this period, drohen already shows considerable changes which are described by the authors as evidence of decategorialization. Drohen in C3 loses relevant morpho-syntactic properties as it can no longer be passivized, topicalized, used as an infinitival complement or undergo extraposition etc. (for a detailed description see Section 6.3). In short, drohen in C3 is grammaticalized to a very high degree. Finally, the construction type C4 develops roughly from 1750 onwards. At this stage functional drohen comes to be used with human subjects. In such contexts the subject cannot be interpreted as controlling the action or acting intentionally. While agreeing with this general outline of the developmental process of drohen sketched by Heine and Miyashita 2008, our analysis also led to some diverging conclusions – not least because we have analyzed the development of drohen not in isolation but in connection with other constructions.6 One of the main differences between our analysis and the analysis presented in Heine and Miyashita 2008 concerns the status of the construction C4, which is treated by them as a distinct category. For us, however, C3 and C4 are functionally equivalent, as they represent the grammaticalized variant of drohen as an evidential marker. The differences between

270 Diachronic corpus study of the four constructions them are a matter of further grammaticalization and integration in the verbal grammatical system of German, as will be discussed below. Heine and Miyashita 2008 have demonstrated that the development of drohen is not a language specific phenomenon (this observation holds also for the cognates of the verb promise). Comparative data from other European languages (French, Spanish, Dutch and English) point to parallel developments in these languages and provide strong evidence for the existence of a language independent grammaticalization channel, which in German is instantiated by the change of drohen from a lexical towards a grammatical item. The next sections sketch our reconstruction of the developmental history of German drohen drawing primarily on the extensive corpus analysis.

3.2. Drohen in Old and Middle High German The following three syntagmatic patterns can be distinguished for the verb drohen in the diachrony of German: 1) drohen as a lexical verb with agentive semantics; 2) drohen as a lexical verb with non-agentive semantics; 3) drohen as an auxiliary verb with evidential semantics. This classification differs from the construction types proposed by Heine and Miyashita (for details see Section 6.3.1). First, it does not classify C2 (Uns droht eine Katastrophe ‘We are now threatened by a disaster’) as a functional variant of drohen. Despite some signs of decategorialization7 observed for it in contrast to the first syntagmatic pattern of drohen, most of its morphosyntactic properties are identical with the lexical variant of drohen (i.e. C1). Nevertheless, the very frequent presence of accusative complements suggests that we are dealing with a specific construction type of lexical drohen that undoubtedly has served as an intermediate stage in the semantic development of the verb towards a grammaticalized usage. Second, we integrate C3 and C4 into a single syntagmatic pattern (3), as in both cases drohen manifests the same – evidential inferential – meaning, whereby it is insubstantial whether the subject of the proposition can or cannot in principle control the situation. What Heine and Miyashita classify as C4 is nothing but the expectable extention of the new grammatical meaning/function to contexts where it had not been used earlier. Table 1

Drohen & zu-infinitive 271

gives a summarizing view of the three construction types of drohen we are dealing with in the following. Table 1. Construction types of drohen construction type

formal characteristics

meaning

1 drohen as a lexical agentive verb

subject is not restricted to 3rd person, drohen takes dative complements, complements with instrumental mit, damitcomplements (instrumental, consecutive, final) etc. subject referent is restricted to 3rd person (inanimate), no clauses as complements, source/cause complement (with von), accusative object, etc. no restrictions with regard to the subject referent, only infinitives as complements, extraposition not possible, etc.

lexical meaning: someone points out that s/he intends to do something that is undesirable for someone else

2 drohen as a lexical nonagentive verb 3 drohen as an evidential auxiliary

lexical meaning: something undesirable is going to happen

grammatical meaning: inferential evidential = marking the proposition as inferred

Construction Type 1 has been present throughout the history of German. (83)

Daz ist: unekihafteem indi unstilleem scal hartor drauuen… (Benediktinerregel 3848, KALI-Corpus) ‘That is: to the disobedient and anxious ones there shall be severe threatening…’

The second construction type of drohen has been attested from ENHG onwards. As noticed by Heine and Miyashita 2008, individual appearances of this type can already be found in MHG. (84)

dergleichen auch ohn-zahlbar sorgen, troen im abendt und den morgen (Hans Sachs; taken from Heine and Miyashita 2008: 80) ‘Countless sorrows of this kind are threatening him evening and morning.’

While Heine and Miyashita 2008 claim that first attestations of the third construction type of drohen have been found in the 17th century texts, in

272 Diachronic corpus study of the four constructions our diachronic corpus, the first unambiguous instances of evidential drohen date from the 18th century, cf.: (85)

Nach einem so langen Glücke, nach einer so glänzenden Reihe von Siegen… sieht sich der österreichische Monarch zum zweitenmal an denselben Abgrund geführt, in den er beim Antritt seiner Regierung zu stürzen drohte. (Schiller, Geschichte des 30jährigen Krieges, DE-Gutenberg) ‘After such a long period of luck, after such a glorious row of victories … for the second time the Austrian monarch finds himself on the edge of the same abyss in which he threatened to fall when he first entered the cabinet.’

In the following, we briefly summarize the findings on drohen in OHG and MHG. In OHG only the agentive variant of drohen was present. In our OHG data we registered 5 occurrences of drohen altogether. All of them are used to introduce a speech act: (86)

Threuuita ín ther heilant quedanti… (Tatian 29403, KALI-Corpus) ‘The Redeemer threatened them saying…’

In two of five instances, drohen co-occurs with the adverb harto ‘strong, hard, very much’. In three of five instances drohen takes a dative complement introducing the addressee of the threat. From this rather small sample we can nevertheless conclude that in OHG drohen is restricted to the first construction type: it occurs only with human subjects, it can be modified by adverb(s), and it takes an optional dative participant. Drohen is clearly a speech act verb, meaning that the threat is performed by a human agent. In MHG like in OHG, agentive drohen is still the only existing variant. It combines with dative complements and is often modified by adverbs (most frequently by harte, but also by sere ‘very, a lot’), cf.: (87)

Ich bin sô harte niht verzaget,/daz er mir sô harte solte dröun. (MF: Reinm 21: 3, 2, MHDWB) ‘I am not as much hopeless that he should threaten me that strongly.’

However, as compared to OHG, drohen im MHG displays a much higher variability of its complements. For example, drohen is often followed by complement clauses, introducing direct or indirect speech, cf. e.g.:

Drohen & zu-infinitive 273

(88)

Als er daz gehôrte ze stet er uns drôte/wir nechômen ûz deme lante ê wir den minnisten besanten. (Gen 2213, MHDWB) ‘As he had heard it he immediately threatened: we would not be able to leave [would not come out of] the country before we sent the Minister.’

These instances suggest that the speech act semantics of drohen is still very strong in MHG. Another piece of evidence in favor of the dominance of the speech act semantics of drohen is its co-occurrence with such verbs as klagen ‘complain’ or schɺlten ‘scold’. In MHG the first attestations of drohen with zu-infinitive are found, cf.: (89) (90)

daz man ime trôte ze slahen. (Alexanderlied 8164, KALI-Corpus) ‘that they threatened to defeat him’ Er brach im den helm abe und trauwet im das heubt abzuschlagen. (Lanc 350, 17, MHDWB) ‘He broke his helmet off and threatened him to behead him.’

Another important observation is the appearance of drohen with accusative objects and with prepositional phrases, most often with the prepositions an ‘on’, in ‘in’, and zu ‘to’ and with nouns such as lîp ‘body’ and leben ‘life’. In such contexts drohen cannot be interpreted as a pure speech act verb. Rather drohen refers to the action or behavior of someone which in some way is threating and which need not be communicated verbally, cf.: (91)

Daz ir mir dreuwet an daz leben? (Herb 8260, MHDWB) ‘That you threaten my life?’

Drohen also comes to be used with the prepositional complements with mit which specifies the kind of the threatening action, cf.: (92)

uns dunket daz er uns dreu/mit sînem zornigen site. (Iw 6694, MHDWB) ‘It seems that he threatens us with his angry manner/habit.’

The variability of drohen with regard to its complements suggests a relevant shift in the semantics of this verb: the meaning of drohen gradually widens from referring to a speech act towards denoting a more abstract notion of a threat. This shift in meaning of drohen, as we think, was crucial

274 Diachronic corpus study of the four constructions for metaphorical extensions, whereby drohen began to be used with nonhuman subject referents, cf.: (93)

Set, wie des meien hant dem winter drouwet. (Mügeln 399,1, MHDWB) ‘See how the hand of May is threatening the winter.’

To conclude this section, we can state the following: (i) The agentive variant of drohen is predominant up to the MHG period. However, it undergoes significant meaning changes from OHG to MHG: it no longer refers to speech acts performed by human subjects. Instead, it refers to every action or behavior which is characterized by a negative impact on the undergoer. (ii) This shift in the semantics of drohen is manifested by the high variability of its complements which show considerable increase in contrast to the OHG period. Most importantly, drohen starts to take infinitival complements. (iii) The metaphorical transfer, or personification, which allows inanimate subjects to combine with drohen helps to form an untypical context in the development of a new drohen-construction.

3.3. Drohen in Early New High German For the ENHG period, some significant tendencies in the development of drohen may be recorded. First, the non-agentive usages of drohen become established as an independent lexical variant of this verb (see also Heine and Miyashita 2008: 80f. for further details and figures). Having evolved via metaphorical extension from the agentive drohen, this variant of the verb becomes independent. It manifests itself most often in intransitive sentences with an optional dative object. (94)

dergleichen auch ohn-zahlbar sorgen, troen im abendt und den morgen. (Hans Sachs, taken from Heine and Miyashita 2008: 80) ‘Countless sorrows of this kind are threatening him evening and morning.’

Drohen & zu-infinitive 275

Heine and Miyashita (2008: 80) point out: “the popularity of C2 rises dramatically, reaching its peak with authors born between 1650 and 1669, where it occurs with the same rate of frequency (48%) as C1”. Moreover the construction is mostly associated with temporally unrestricted situations. That is, the non-agentive drohen introduces situations which are seen as not being limited to a particular time interval. Rather, the mere fact of the existence of an observable threatening situation is being asserted, without specifying its localization in time. Second, the agentive construction type of drohen shows an increasing tendency to integrate infinitives within its structural scope. On the one hand, there is an increase in token frequency of infinitives combining with drohen (from zero in OHG, to 10% in MHG and further to 15% in ENHG). On the other hand, while in MHG only three infinitive verbs schlagen ‘hit’, abschlagen ‘behead’ and brennen ‘burn’ (in the sense of ‘set fire’) combine with drohen, in ENHG further infinitives of clearly agentive verbs come to be used with drohen (e.g. tun ‘do’, erstechen ‘stab’, umbringen ‘kill’, etc.), cf.: (95)

sie tröuwend euch mit dem schwerdt auszureuten und ze verbrennen (Maaler 409c, DWB) ‘they threaten to destroy you with sword and to burn you up.’

Hence we can state that the agentive variant of drohen increasingly tends to take infinitives of agentive verbs which specify the threat: the threat concerns such actions as killing, beating someone, which all cover negative actions being dangerous and undesirable. Note that no infinitives are found within the non-agentive construction type. This fact suggests that (i) we are dealing with two relatively independent constructions of drohen, whereby (ii) infinitives are restricted only to those which describe an intended action. We propose that the development of a new construction type of drohen, i.e., the non-agentive drohen-construction, on the one hand and the establishment of infinitives as standard complements of the agentive drohen, both of which occur at roughly the same time, can be seen as relevant in the development of the evidential construction drohen & zu-infinitive, for the following reasons. First, the evolving and strengthening ability of agentive drohen to take infinitives yielded a conceptual split between the act of threatening itself and the content of the threat which itself is conceived of as an action. Second, the possibility to distinguish between two situations is not given in the non-agentive uses of drohen. Instead, the non-agentive

276 Diachronic corpus study of the four constructions construction type of drohen is restricted to introducing immanent undesirable, unpleasant or dangerous situations (as a whole). It is necessary to note that drohen in this function represents situations which are seen from the speaker’s point of view – they are observed and judged by the speaker as unpleasant and dangerous. This means that drohen is used to describe situations which are in some way perceived by an observer. As the subject of the sentence is always inanimate, it cannot be given the ability to have intention and to be able to carry out some (unpleasant) action. Thus the interpretation of drohen in the sense of a declaration of intention to do something is blocked. Therefore a conceptual split into two distinct situations is not available for this variant of drohen. For this reason the combinatorial variability of this construction type of drohen is limited to a considerable extent. 3.4. Drohen in the 18th and 19th centuries In the 18th century, the above described variants of drohen – agentive and non-agentive drohen – continue to exist. We have found 57 instances of drohen in our corpus data from the 18th century altogether. The agentive construction type counting 30 instances is the most frequent type in our data, whereby the number of infinitive complements in this type is remarkably high: it amounts to 12 instances, i.e. ca 40%. In comparison to ENHG there is thus a considerable increase. Moreover, further infinitive verbs come to be used as infinitive complements of agentive drohen. Hence the infinitive complement of the agentive drohen became customary within the spectrum of its possible complements, together with dative and accusative objects, complement clauses, and prepositional phrases with mit ‘with’. Agentive drohen is rarely used in its old speech act meaning, serving as a one-state predicate, cf.: (96)

(97)

Er nannte uns Landläufer, drohete uns festsetzen und ausliefern zu lassen (Knigge, DE-Gutenberg) ‘He called us hoboes, threatened to arrest us and to hand us over.’ Man schrie, raste, tobte, drohte, und nur die Mittagsglocke konnte die Streitenden auseinanderbringen. (Klinger, DE-Gutenberg) ‘They shouted, rushed, rioted, threatened, and only the midday bell could separate the fighting people from each other.’

Drohen & zu-infinitive 277

With the exception of the structure with the accusative object, which is almost absent in the language of today, the situation in the 18th and 19th centuries is very similar to the situation observable in PDG (see Section 6.3). The next frequent construction type in our corpus data is non-agentive drohen with 13 instances. Note that 5 of them represent a new structure with an accusative complement,8 cf. e.g.: (98)

Eine gefährliche Trennung drohte dem ganzen protestantischen Bunde den Untergang. (Schiller, Geschichte des 30jährigen Krieges, DE-Gutenberg) ‘The whole Protestant union was doomed by a dangerous schism.’

It should be noted at this point that this (new) valency pattern of drohen is parallel with the common valency pattern of versprechen (for versprechen see Chapter 8.4). We suggest that this combinatorial structure was formed analogically to the parallel structure of the agentive drohen. In connection with the structural analogy, semantic properties are transferred to the new structure of the non-agentive drohen. In these contexts the non-agentive construction type of drohen manifests the conceptual split into two separable situations which we have observed for the agentive variant of drohen in the ENHG period. On the one hand, there is a dangerous situation, i.e. the subject of the sentence, which is conceived of as a source of a threat.9 On the other hand, there is another situation which can be interpreted as a concrete effect of this threat, i.e. the accusative complement. In other words: the non-agentive drohen comes to occur in contexts where an explication of a threat can be added. In this way, drohen in its non-agentive uses receives now a temporal-aspectual reinterpretation. The second event, i.e., the content of the threat, is seen as non-realized at the time of speaking. At the same time, the (inanimate) subject, which cannot be interpreted as an agent, is now reinterpreted as a source, or a cause of a dangerous or unpleasant situation. The first attestations of the infinitive constructions with non-agentive semantics are present as early as in the 18th century. These are the critical contexts for the development of drohen, which is treated in Section 5.3 in detail. We found 12 instances of this pattern, which obviously came into being via analogical transfer from the agentive drohen with an infinitive complement.10 It thus represents the typical case of distributional extension of emergent constructions. Remarkably, 7 of the 12 infinitive constructions

278 Diachronic corpus study of the four constructions with non-agentive drohen found in the 18th century corpus data share relevant properties with non-agentive drohen plus accusative object: (i) their subjects have the feature ‘inanimate’; (ii) they are interpreted as a cause of the situation, (iii) which, in turn, is introduced in the infinitival complement or in the accusative complement respectively. (99)

schon loderte die Flamme, und drohte das herrliche Werk in die Asche zu legen. (Klinger, DE-Gutenberg) ‘the flame blazed and threatened to burn up the lovely work’

The affinity between (98) and (99) is obvious. The only difference between them is that the situation effected by the subject is exemplified in the form of an infinitive complement and not in form of an accusative object, which typically is a nominalization. Note that the infinitive verbs occurring in such structures are all action verbs with negative semantics (zerstören ‘destruct’, entreißen ‘snatch’, zerreißen ‘disrupt’, etc.). Heine and Miyashita (2008: 84) propose a similar view: “The infinitival complements of C3 and C4 can not be traced back to C1; rather, they are the result of a development from nominal accusative complement in C2 to infinitival complement in C3 and C4” [original emphasis]. In the light of our data this statement must be modified in the following way. Although the agentive construction type of drohen (the type C1 in the terminology of Heine and Miyashita) is not the direct precursor of the grammaticalizing construction, it still served as the structural model that was analogically transferred into the nonagentive construction type of drohen. This analogical transfer was ‘mediated’ semantically via the structure with nominalized accusative complement, which are semantically close to infinitives and which thus was a catalyst for the conceptual shift in the semantics of the non-agentive variant of drohen described above. The first, still rare, evidential readings of drohen & zu-infinitive are attested in our text data from the 18th century, cf. e.g.: (100)

Unser Schiff bedurfte einer völlig neunen Verzimmerung, die sich bis in den November 1755 zu verzögern drohte. (Nettelbeck, DEGutenberg) ‘Our ship needed a completely new superstructure which threatened to be delayed to November 1755.’

The distinct feature of such constructions is that the subject of drohen cannot be interpreted as a causing force of the situation introduced in the infi-

Drohen & zu-infinitive 279

nitive complement. Rather, a situation is represented in its entirety, whereby the subject together with the infinitive complement forms a core predication. Here, drohen exhibits a true evidential semantics for the first time in its history: it serves as a secondary predication and modifies the whole proposition. The situation modified by drohen is conceived of as (i) nonrealized yet and (ii) as being likely to occur. In the following example, the evidential semantics of the infinitive construction with drohen is obvious: (101)

Walther Fürst schwankt und droht zu sinken, Berta hält ihn. (Schiller, Wilhelm Tell, DE-Gutenberg) ‘Walther Fürst is staggering and is threatening to fall, Bertha holds him.’

The subject is a human referent, and the event described in the infinitive complement is a non-action. The subject cannot be interpreted as an agent. It functions instead as an experiencer of the event formulated by the infinitive verb. Thus we are dealing here with a reanalyzed structure, whereby drohen functions as an auxiliary verb, which modifies the (primary) proposition formed by the subject and the infinitive verb (including its own valence relations). Thus drohen loses all other valence relationships exept for the ability to combine with an infinitive complement. There is also another important characteristic of the contextual configurations in which the evidential reading of the infinitive construction arises. The preceding contexts contain descriptions of some observable pieces of information which are interpreted as resulting in the situation introduced with drohen. In other words: the proposition modified by drohen is semantically integrated into a larger text segment. These contextual configurations are – as we think – crucial for the establishment of the evidential reading of the infinitive construction with drohen. What happens here is a semantic reinterpretation of the infinitive construction with drohen occurring in specific contexts, whereby the whole proposition is interpreted as resulting from some other situation. Neither does the subject referent intend (literally or metaphorically) to do something described in the infinitive complement, nor does it function as the cause of the described situation. Instead, it is the pieces of information available to the speaker that are the source of the described situation now. While such contextual configurations are only rarely observed in the data from the 18th century, they come to be very frequent in the 19th century. It is these contexts which form the isolating contexts for the evidential reading of drohen & infinitive.

280 Diachronic corpus study of the four constructions The following conclusions can be made at this point about the relevant developments within the overall paradigm of drohen constructions in the 18th century: 1. The agentive use of drohen continues to exist with the whole range of its combinatorial possibilities. Importantly, its use with infinitive complements establishes itself as a standard structure. 2. The non-agentive construction type of drohen expands its combinatorial potential, as it shows new structures, such as combination with nominalizations as accusative objects and – via this stage – with zuinfinitive complements. These new structures are built analogically to the existing structures of the first – agentive – construction type of drohen. Most of these uses may be interpreted as cases of metaphorical transfer, whereby the inanimate subject of the sentence is given intentional semantics. We consider such structures here as untypical contexts, whereby an inanimate subject occurs in combination with nonagentive drohen and an infinitive verb with (mostly) agentive semantics. 3. The first attestations of the evidential construction type may be detected. Evidential interpretations arise in specific contextual configurations and are mostly triggered by the presence of a non-action verb and/or an inanimate subject. Frequently, these structures are ambiguous between a non-agentive reading described in (2) and a truly evidential reading. These contexts may be thus classified as critical contexts, whereby two different interpretations implying two distinct structural analyses are possible. On the one hand, it is the metaphorical interpretation implying that a threat comes from an entity, and the infinitive complement explicates this threat. In this case, the infinitive verb is seen as a complement of the lexical (main) verb drohen, whereby drohen modifies the subject. On the other hand, it is the evidential reading implying that drohen introduces a situation which results from some other facts (introduces earlier in the text). In this case, the subject together with the infinitive verb forms a proposition which – as a whole – is modified by drohen as a secondary predication. As instances of the evidential construction type are still very rarely found in the 18th century, let us briefly sketch the situation of the 19th century. From now on we will concentrate only on the occurrences of infinitive constructions, which are polysemous as to the three types of meaning illustrated above. The evidential reading is restricted to this structure only so it

Drohen & zu-infinitive 281

is most important to look at the whole class of infinitive constructions more closely. The number of infinitive constructions increases from 18th to the 19th century (see Table 2), and the number of evidential constructions increases rapidly (see Table 3). Therefore it can be suggested that the most relevant changes must have occurred in these structures. Table 2. The frequency of drohen with infinitives in the corpus data verb ¦ drohen

57

18th century with % infinitive 24 42,1

¦ 196

19th century with infinitive 97

% 49,5

Table 3. Lexical and evidential readings of drohen & zu-infinitive lexical reading

18th century evidential (or ambiguous) reading

¦

lexical reading

12 (50%)

12 (50%)

24 (100%)

19 (19,6%)

verb

drohen

19th century evidential (or ambiguous) reading 78 (80,4%)

¦

97 (100%)

The unambiguous lexical reading of the construction drohen & zuinfinitive is found in 19 instances out of 97. The subject is usually a human referent, and the infinitive verb refers to an action. The threat is not restricted to a verbal act. The amount of untypical contexts (32 instances) is remarkably high, where the non-agentive drohen occurs with an inanimate subject and an infinitive verb denoting an (undesirable) action. The subjects found in such contexts are, for instance, Toben ‘uproar’, Unheil ‘harm’, Verwirrung ‘confusion’, Krampf ‘convulsion’, das Ungeheuere ‘tremendousness’, Revolution ‘revolution’, Hindernis ‘barrier’, etc. The infinitive verbs occuring in these contexts are e.g. (zer)reißen ‘break’, rauben ‘rob’, fortschleppen ‘drag off’, fesseln ‘tie’, hereinbrechen ‘break in’, vernichten ‘destroy’, ersticken ‘stifle’, überwältigen ‘overpower’, etc. Most of these expressions allow for a metaphorical interpretation, meaning that an unpleasant entity causes an undesirable action. Most of the remaining 46 instances of the non-agentive drohen with infinitive may be described as cases of critical contexts. As such, they are – at least potentially – ambiguous, and further contextually present information is often required for one or the other interpretation. For us, cases

282 Diachronic corpus study of the four constructions where the evidential reading is most likely are of particular interest. Consider the following examples: (102)

(103)

(104)

(105)

Irland erhob sich so rasch, daß seine Bewohner die Engländer zu überflügeln drohten. (mew/WAG.03667, IDS-Corpus) ‘Ireland revolted so quickly that its inhabitants threatened to outdistance the English.’ Der Abend kam herbei, die Nacht drohte hereinzubrechen. (GOE/AGM.00000, IDS-Corpus) ‘The evening came, the night threatened to break in.’ Mit seiner Frau, Maria von Tübing, hatte er nur zwei Töchter und keinen Sohn erzeugt, und sein gräfischer Stamm drohte zu erlöschen. (GRI/SAG.00543, IDS-Corpus) ‘With his wife, Maria von Tübing, he had fathered two daughters and no son, and his noble ancestral line threatened to disappear.’ aber die Sonnenhitze brennt gar zu stark, daß mir das zarte Kraut zu welken droht… (GRI/KHM.00122, IDS-Corpus) ‘But the sun heat burns so strongly that the tender herb threatens (to me) to wither.’

There are no particular characteristics, which could be singled out for the infinitive verbs or for the subject referents of this – evidential – drohen. As can be seen from the examples above, the subject may be either animate or inanimate. As regards the infinitive complement, it is either an action verb or a non-action verb. Without denying the fact that a particular combination, especially [animate subject + non-action infinitive verb], could indeed have helped to trigger the evidential reading of the construction, we assume that another factor had much more impact on the rise of an unambiguous evidential meaning. In all of the examples represented above, there is some explicitly mentioned situation which precedes the situation described with drohen. We propose that this contextual configuration is crucial for the reinterpretation of the semantics of the whole construction. There is some entity (i.e. situation, observation) which is distinct from the subject referent and which has the situation described with drohen as its result. The link between some (textually present) situation and the described situation is established by the speaker: it is the speaker who recognizes that there is a relationship between two situations. In other words, the speaker draws on available (and in some way obvious) information to make a statement about the situation. Drohen is employed to mark this link be-

Drohen & zu-infinitive 283

tween some situation and the proposition. This reinterpretation process, whereby a sequential-relationship is established between the proposition itself and the preceding texts, is very similar to the process reconstructed for werden (see Section 8.1). It should be noted that contexts like those illustrated in (102)-(105) are still frequent in PDG, though their numbers have fallen as compared to the 19th century (in our corpus of PDG, such structures amount to less than 1% of the overall occurrences of drohen, whereas their percentage in our data from the 19th century is ca. 10%). As concerns the situation in the 19th century, one important observation may be added to what we have already said. There is also another kind of contextual configurations of drohen where only the evidential reading is possible, cf.: (106)

Schon drohte der Zug abermals um eine Felswand zu verschwinden. (GOE/AGM.07859, IDS-Corpus) ‘Now the train once again threatened to disappear behind a rock face.’

Here, there are no explicitly mentioned facts in the text which may be interpreted as preceding or causing the situation introduced by drohen. Drohen indicates that the speaker relies on some observations in making his/her statement. In such contexts the situation, i.e. pieces of information, which functions as a source for the described situation, is not textually present. Instead, it is implicitly present in the speech situation, which is made clear by the use of drohen & zu-infinitive. Summarizing this section we state that, in the course of the 18th and 19th centuries, there occurs a shift in the meaning of drohen within infinitive constructions, which can be best described as a relocation of the source entity. As regards the agentive drohen, the source of a threat lies always in the subject entity, which is usually a human referent. In the case of the non-agentive drohen, the source of the threat lies the subject entity (or its appearance, its characteristics, etc.), too. As regards the evidential drohen, however, the source lies outside the subject entity: this is some information available to the speaker which s/he interprets as pointing towards the described event. This information is either textually present or contextually given at the time of speaking. This particular (re-)location of the source entity is the distinct characteristic of the evidential construction type of drohen.

284 Diachronic corpus study of the four constructions 4. Versprechen & zu-infinitive As the etymon versprechen has a highly complex, multifacetted history and accordingly, a wide-spread, diversified web of meanings, the following sketch renders only a small segment of this field, namely those semantic components which are relevant for the development of the evidential usage of this verb. Nevertheless, in order to at least hint at the inhomogeneity and complexity of the etymology of the verb versprechen in German, a – rather lengthy – quote from Grimm’s dictionary is in place here (the whole entry extends over 38 columns): Versprechen hat sich in seiner bedeutung nach den verschiedensten richtungen hin verzweigt; die übersicht zu geben ist nicht ganz leicht. […] Besonders fällt das nebeneinander von positivem und negativem sinne in diesem worte auf […] diese verschiedenheit könnte vielleicht auf verschiedene grundformen des präfixes zurückgehen […] das verhältnisz der verschiedenen bedeutungen zueinander und ihr vorkommen überhaupt ist verschieden in den verschiedenen sprachperioden und gegenden […] Versprechen, das die heutige sprache nur in positivem sinne kennt, wird von LUTHER vorwiegend abweisend oder gehässig gebraucht; in den älteren sprachperioden sind beide richtungen nebeneinander vertreten […] Mit dem 18. jahrh. schwindet die negative verwendung von versprechen. [‘The meaning of versprechen has developed in various directions; it is not that easy to give an overview. […] Particularly remarkable is the coexistence of negative and positive semantics in this word. […] This diversity might probably have been influenced by different basic forms of the prefix. […] The relation of different meanings to each other and their overall usages are different depending on different historical periods and areas. […] Versprechen, which in the present-day language has only positive meaning, is largely used negatively by LUTHER; in the elder periods of German both readings coexist […] From the 18th century onwards the negative reading of versprechen disappears.’] (DWB, vol. 25, 1448-1501)

The following meanings and uses of versprechen are noted in the DWB (we have shortened the list to the most central meanings and uses): I. Versprechen with positive semantics, including the following variants: A1-3: ‘to promise that something will happen’. Versprechen is a clear speech act verb which can be used performatively. The subject is usually able to fulfill the promise, i.e. it takes on the responsibility to bring about the promised situation.

Versprechen & zu-infinitive 285

A4: A sub-type of versprechen with the positive semantics and a distinct feature meaning that the subject does not perform a speech act of promising. Instead, an ‘expectation’, a ‘hope’ is expressed by versprechen, whereby there is something in which this particular assessment originates. B: ‘to advocate’, ‘to act on someone’s behalf’. This meaning type of versprechen was present in the elder periods of German and is absent now. Its use was largely restricted to legal contexts. C: ‘to claim’, ‘to confiscate’. This variant, like the preceding one, is absent in PDG and, in the earlier historical periods, was restricted to legal contexts. II. Versprechen with negative semantics, including: A 1-4: ‘to deny’, ‘to negate’, ‘to refuse’, ‘to reject’, ‘to give up’, etc. This negative meaning of versprechen has disappeared in the course of the history of German. Interestingly, in the texts of Luther versprechen is used only with this meaning. B: ‘to blame’, ‘to condemn’, ‘to asperse’, etc. This meaning is also no longer available for versprechen in PDG. C: reflexive, ‘to make a mistake while speaking’. In PDG only one nuance of this meaning is still present, meaning ‘to misspeak’. Since the second group has been completely lost (except the reflexive form sich versprechen ‘to misspeak’), we focus on the first group. As the quotations show, the modern variant of the verb versprechen with positive semantics – as we know it in PDG – had been but one of several alternative interpretations of the etymon for a long time. It developed towards the central lexical variant of the verb, while other meanings either disappeared or became very rare (and are now realized only in specific contexts). This process begins in the 18th century. At about the same time, the first evidential interpretations of the construction versprechen & zu-infinitive have been attested. According to Paul (1992, s.v. versprechen), the basis for the evidential meaning developed in the 18th century. A similar view is presented in the DWB: “dieser gebrauch hat sich erst in der jüngeren sprache herausgebildet; er erscheint um 1700, verbreitet sich dann aber sehr schnell in der schriftsprache und hält sich in dieser lebendig” [‘this usage has not developed until the younger periods of the language; it first appears around 1700 AD, spreads very rapidly in the written language where it continues to exist’]. As our investigations also prove, the evidential meaning of versprechen originates in the positive lexical semantics of the verb. Without dealing with the complex questions concerning the development of this verb (e.g.why the positive variant of versprechen could estab-

286 Diachronic corpus study of the four constructions lish itself as the central lexical variant) our diachronic analysis starts from the situation at the beginning of the 18th century and looks for relevant changes which most probably led to the establishment of the construction versprechen & zu-infinitive as an evidential construction.

4.1. Versprechen in the history of German and its development up to Early New High German Basically, the following syntagmatic patterns can be distinguished for the verb versprechen (with positive semantics as delimited in the previous section) in the diachrony of German: 1) versprechen (as a lexical verb) with commissive speech act semantics; 2) versprechen (as a lexical verb) without reference to speech act; 3) versprechen as a grammatical (semi-)auxiliary with evidential semantics. To each of these three types, specific characteristics may be assigned: 1) The subject is usually a human referent, versprechen indicates the verbal act of promising; the subject is (in principle) able to carry out the promised action or – in some way – to bring about the promised situation; versprechen can be used as a one-place predicate, more common are, however, its combinations with direct/indirect speech clause complements, with infinitival complements, and with nominal accusative objects. This syntagmatic pattern corresponds to the sub-types A.1-3 described in the DWB (see previous section). 2) The subject is mostly inanimate11 (and cannot be interpreted as performing a speech act), versprechen indicates an expectation and has strong future-oriented semantics; most often, versprechen combines with accusative objects. This syntagmatic pattern basically corresponds to the sub-type A4 described in the DWB (see previous section). 3) The subject is mostly inanimate and cannot be interpreted as performing a speech act; versprechen modifies the whole proposition, assigning an inferential evidential value to it; versprechen combines with infinitive complements denoting non-actions; all other complements depend on the infinitive verb.

Versprechen & zu-infinitive 287

As for the first syntagmatic pattern, it has been attested in all historical periods of German. The OHG example in (107) illustrates versprechen followed by a complement clause introducing direct speech, in the MHG sentence (108) versprechen combines with an infinitive complement with zu. (107)

(108)

sô medicus infirm, ságet, mit uuíu er genésen sól … únde ióh fúresprichet: daz ist signum recuperandae sanitatis (Notker 1, 238, 6, taken from DWB) ‘so the doctor tells the sick one how he should convalesce … and also promises: this is signum recuperandae sanitatis.’ daz der kunic versprechen het ze leisten sîne bet. (Ottokar, reminchronik 85573, taken from DWB) ‘that the King has promised to satisfy his appeals.’

The second syntagmatic pattern occurs first in the 18th century, cf.: (109)

Gleich der Anfang der Fahrt versprach wenig Gutes. (Nettelbeck, DE-Gutenberg) ‘The beginning of the journey promised little good.’

From then on, its frequency increased rapidly, and this variant established itself as an independent lexical variant of versprechen. As for the third syntagmatic pattern, we found its first instances in our corpus data from the 19th century, cf.: (110)

die Vorarbeiten zur “Farbenlehre”, mit denen ich mich seit zwölf Jahren ohne Unterbrechung beschäftigte, waren so weit gediehen, daß sich die Teile immer mehr zu runden anfingen und das Ganze bald selbst eine Konsistenz zu gewinnen versprach. (GOE/AGA.02616, IDS-Corpus) ‘the preparations for my teory of colour, which have kept me busy for the last twelve years, were already advanced to a degree that the parts began to take shape and the whole promised to gain in consistency.’

Before looking at the details of this evidential construction of versprechen, we briefly summarize the results of our corpus analysis for the earlier periods. In our corpus data from OHG we recorded 5 instances of the verb

288 Diachronic corpus study of the four constructions firsprehhan altogether. All of them can be interpreted as expressing the negative semantics of versprechen, meaning generally ‘to disavow, to deny’, cf.: (111)

échert ér dero fórderûn zálo iéhe. des ûzlâzes nemág ér nîeht ferspréchen (Notker, 16966, KALI-Corpus) ‘if he confesses to the previous claim he cannot deny the consequence.’

In the MHG data (consisting of 96 instances of versprechen altogether), most of the instances, too, carry the negative meanings mentioned above, cf.: (112)

ich wolt uch ere thun, und ir habt sie versprochen. (Lanc 388, 12, MHDWB) ‘I wanted to venerate you and you have refused it.’

However, there are also some examples of positive versprechen (ca. 2025). We should mention here, however, that an unambiguous interpretation of individual cases is often very difficult, as both – negative as well as positive – interpretations are often possible. Further contextual information and extra-linguistic knowledge is often required to interpret individual uses of versprechen. (113)

(W)er den andern vor gericht versprechen sol… (StRAugsb 4, 12, MHDWB) ‘Who should advocate the other in court…’

As regards our corpus data from the ENHG period, the situation is slightly different. First, the number of versprechen uses with positive semantics, namely versprechen as a speech act verb representing a promise in a modern sense, is relatively high (14 out of 24 instances), as opposed to the earlier historical periods. (114)

Als wult er sagen/Ich hab euch hieruor zuogesagt vnd versprechen. (B 155) ‘As if he wanted to say: I have agreed to this and promised it to you.’

Versprechen & zu-infinitive 289

Second, more than a half of these uses (8 of 14) manifest a combination of versprechen with an infinitive. The infinitive may precede or follow the verb, cf.: (115) (116)

…wie ich jetz zuo thuon versprochen hab. (B 155) ‘… as I have promised you to do.’ des was der guot mann gar wol zuo friden/versprach im auch seinen willen und bitt zuo volziehen. (B 235) ‘the good man was very satisfied and promised him to fulfill his will and request.’

Summarizing the central characteristics of versprechen, which we consider relevant for the development of the evidential meaning of the construction versprechen & zu-infinitive, we note that the the polarity between a negative and a positive meaning of this verb (which is most probably due to different origins of its prefix, see above), got resolved in favour of the positive one. While most of the OHG and MHG instances of versprechen display its negative semantics, in ENHG the positive variant already dominates. Another relevant innovation of ENHG is that versprechen comes to be combined with infinitive complements, which may either precede or follow versprechen and thus, in the more recent history, represents a standard complement of (positive) versprechen. At this point, we can hypothesize that versprechen, starting as a very ambiguous item with different, sometimes even mutually exclusive meanings, had an inherent semantic component of speaking from the very beginning. Up to ENHG this meaning narrowed towards strong commissive speech act meaning. In the course of this development versprechen came to be used as an illocutionary verb in the strict sense. Thus at the end of the ENHG period the first lexical variant of this verb was established. We suggest that this development (i.e. specification of speech act semantics), which in contrast to drohen occurred relatively late, may have worked as a retarding factor for the grammaticalization of the evidential construction versprechen & zu-infinitive. 4.2. Versprechen in the 18th century From the 18th century onward versprechen with the positive semantics was the central variant of this verb. At the same time the negative uses of the

290 Diachronic corpus study of the four constructions verb almost entirely disappeared. Moreover, again from the 18th century onward, versprechen gradually develops evidential semantics, whereby the denoting of a speech act slowly fades away. For this reason, it is worth looking into the data from the 18th century in greater detail. We detected 152 occurrences of versprechen in our data from the 18th century. The first important observation is that most of them (122 instances) represent the positive variant of this verb: versprechen most often denotes a commissive speech act. The verbal act is mostly performed by a human referent. The content of a promise is always an action or a situation, which is controlled by the subject of the sentence. That is, versprechen expresses that someone commits him/herself to do something that s/he is able to do. This is the prototypical meaning of versprechen which is still very prominent in PDG. This variant of versprechen shows different realization structures: versprechen combines with nominal accusative objects, with infinitive complements and with complement clauses which mostly explicate the content of the promise. In more than half of its occurrences (i.e. 62 instances), versprechen combines with infinitive complements. The second syntagmatic pattern of versprechen is represented in our corpus data with 10 instances, i.e. ca. 6% of the overall amount. The main characteristic of this syntagmatic pattern is that versprechen does not denote a speech act, but expresses an expectation with regard to the subject entity, instead. The subject is mostly inanimate (118), though human subjects are not excluded (117), cf.: (117)

(118)

Er ist jetzt acht Jahre alt und verspricht ein redlicher teutscher, grader, gesunder Kerl an Leib und Seele zu werden. (Knigge, DEGutenberg) ‘He is now eight years old and promises to become an honest German, an upright citizen, sound in body and soul.’ Der eine Theil versprach Ersatz, der andere Vergebung. (Schiller, Geschichte des 30jährigen Krieges, DE-Gutenberg) ‘The one part promised replacement, the other forgiveness.’

Here versprechen means ‘to portend, to presage’. Note that the subject still controls the predicate. In other words: the subject is the source of the promise. It has features which promise something, so that one might speak of metonymical transfer (pars pro toto) here. What versprechen indicates here is that something is expected from the subject of the sentence. It may

Versprechen & zu-infinitive 291

be suggested that such structures arose via metaphorical transfer. Indeed, we found some cases of a clear metaphorical transfer in our corpus data: (119)

Mein Herz sagt mirs, und es verspricht mir die Wollust… (BA Gel) ‘My heart is saying it to me, and it promises me lust…’

Thus we can assume that, similarly to the development of the verb drohen described in Section 8.3, the inanimate subjects came to be used with versprechen via metaphorical transfer, i.e. personification. The subject cannot be interpreted as performing a verbal act of promising. Instead, there is something about the unintentional subject which can be interpreted as metaphorically ‘promising’ something else. The content of the promise is expressed either by a nominal accusative complement or by an infinitive complement. Note, however, that both, the first lexical variant of versprechen, denoting a commissive speech act, and the second variant, lacking the reference to a speech act, appeared at about the same time. Before the 18th century none of them was common. This fact suggests that there must be something else apart from metaphorical transfer that could have influenced such a rapid development of the second construction type. We argue that an analogical transfer could have served as an additional motive for this development. For drohen (see Section 8.3) we have detected that its first nonagentive uses, which – semantically as well as formally – very much correspond to the second construction type of versprechen, appeared first in MHG and were already quite common in ENHG. Versprechen, on the other hand, was much ‘slower’ in its development towards a specific speech act verb and had not established itself as such until ENHG. Since the latter resembled drohen in many respects, i.e. semantically (both are speech act verbs by means of which something is introduced as intended by a human subject) as well as formally (both combine with animate subjects, nominal accusative complements, infinitive complements, and direct/indirect speech clause complements), it is not implausible to assume that drohen could have served as an analogical model for versprechen. For this reason metaphorically reinterpreted structures of the second construction type of versprechen could have appeared so quickly after its first construction type had been established. We propose that the second construction type of versprechen manifests the untypical context, where versprechen lacks its prototypical interpretation as a commissive speech act verb. This stage in

292 Diachronic corpus study of the four constructions the development can be seen as the initial stage in the grammaticalization process leading to versprechen as an evidential marker. Another specific structure of versprechen deserves particular attention here. In 16 instances versprechen occurs in a specific structural configuration which is semantically very close to the structures just described. However, there is one important difference between the two structures, i.e. versprechen always combines with a dative (reflexive) pronoun, cf.: (120)

(121)

Und sie versprachen sich von diesem Richter nichts als Gerechtigkeit. (BA Gel) ‘And they expected [lit.: promised themselves] of this judge nothing but justice.’ Ich versprach mir von Ihrem Einfalle so viel! (Lessing, Emilia Galotti, DE-Gutenberg) ‘I expected [lit.: promised myself] so much of your idea!’

As the above examples show, there are some relevant constituents which form this structure: (i) a subject which is usually a human referent; (ii) versprechen, meaning ‘to expect something of something/someone’; (iii) a dative reflexive pronoun referring back to the subject; (iv) a nominative accusative complement, representing the content of expectation; (v) information about the thing/person as a source of the expectation which is mostly introduced by a prepositional phrase with von ‘of’ or durch ‘via, by’. The distinct property of such structures is that they always entail the reflexive pronoun referring back to the subject. A particular perspective on the situation is expressed: it is the subject’s personal assessment of the situation. This personal assessment of the situation is anchored in some entity or situation which is mostly explicated in the text by means of a prepositional phrase with von or durch. This is what crucially distinguishes these structures from the second construction type of versprechen. For this reason we consider these structures as a relatively independent construction type of versprechen that develops in the course of the 18th century (see also Section 6.4, where we have argued for treating this type as an independent lexical variant of versprechen). We suggest that this variant originates partly in the first construction type of the verb and partly in the former structures which were existent earlier in the history of German. The combination of versprechen with reflexive pronouns was already quite common in MHG and ENHG, whereby versprechen could be interpreted in both its positive as well as its negative readings. We assume that the new

Versprechen & zu-infinitive 293

lexical variant evolves via integration of the existing formal structures and a newly established semantics. To summarize the relevant findings concerning the syntagmatic patterns of versprechen in the 18th century, we can state the following: 1) Versprechen establishes itself as a commissive speech act verb, whereas older uses and meanings of this verb disappear. 2) The (lexical) syntagmatic pattern 2 – versprechen without speech act semantics – develops first in the 18th century as a result of two interrelated processes. We hypothesize that, on the one hand, the new syntagmatic pattern evolves via metaphorical transfer, whereby a nonintentional and non-committing subject is reinterpreted as a source for a future situation. On the other hand, this syntagmatic pattern is formed analogically after the model of the already existing non-agentive construction type of drohen. As we have noted, this construction type may be classified as untypical context in the grammaticalization process of the evidential construction versprechen & zu-infinitive. 3) A new lexical variant arises – versprechen with dative reflexive pronoun – meaning ‘to expect something of something/someone’ whose development is (only) partly motivated by the establishment of the first and the second construction types of versprechen. As we will show in the next section, this construction type develops towards an independent lexical variant of versprechen in PDG. 4.3. Versprechen in the 19th century In the 19th century the first lexical reading of versprechen represents the central use of the verb. As shown in Table 4, it accounts for 79,5% of the overall amount of versprechen instances in our corpus material. Table 4. The overall distribution of versprechen in the 19th century construction type

amount

%

versprechen with commissive speech act semantics

427

79,5

versprechen without speech act semantics

54

10,1

versprechen with a dative reflexive pronoun

56

10,4

537

100%

total

294 Diachronic corpus study of the four constructions The construction type 1 has remained the same until now. Versprechen has commissive speech act semantics, denoting a verbally performed communicative act of promise. This construction type has typical formal realizations which have been preserved till PDG (see Section 6.4 for more details). Versprechen in (obligatory) combination with a dative reflexive pronoun, i.e. the reflexive construction of versprechen, is quite common in our data (10,4%). As presented in Table 5, in our corpus data we found the most frequent uses of this construction in the 18th and 19th centuries. Table 5. Reflexive versprechen in the 18th, 19th and 20th centuries verb versprechen

18th century

19th century

20th century

reflexive

¦

reflexive

¦

reflexive

¦

16 (10,5%)

152 (100%)

56 (10,4%)

537 (100%)

49 (7,6%)

644 (100%)

Thus we can speak of a newly established lexical variant of versprechen that manifests itself in particular formal structures and has become largely independent from other semantic variants of versprechen, most obviously from its central lexical variant denoting commissive speech acts. The distinct formal properties of this construction are: (i) a dative reflexive pronoun referring back to the subject; (ii) a nominal accusative complement, indicating the content of the expectation; and (iii) a facultative prepositional complement with von or durch introducing the entity which is the source of the expectation. Yet another development of versprechen deserves attention here. As we have mentioned in the previous section, in the 18th century versprechen starts to be used in contexts without reference to a speech act meaning ‘to portend, to presage’. This syntagmatic pattern continues to develop. In the 19th century this usage is expanded and accounts for ca. 10% of all uses of versprechen at that time. (122)

Mit solchen angenehmen Pflichten beschäftigt, sah ich die Sonne untergehen, die einen folgenden heitern Tag verkündigte und unserm Fest ihre frohe glänzende Gegenwart versprach. (GOE/AGD.06345, IDS-Corpus)

Versprechen & zu-infinitive 295

(123)

(124)

‘While attending to such agreeable duties, I watched the sun going down, and it heralded sunshine for the morrow, promising to grace our celebrations with the warm glow of its presence.’ nicht von großer aber gewandter Gestalt, versprach sein Gesicht und sein ganzes Wesen eine anmutige rasche Entschlossenheit (GOE/AGD.06345, IDS-Corpus) ‘Not big but agile, his face and his whole constitution conveyed a sense of [lit: promised] a graceful quick decisiveness.’ indem sie saß, versprach ihre Figur mehr Länge (GOE/AGI.00000, IDS-Corpus) ‘When she sat her figure seemed [lit: promised] to increase in stature.’

Such structures with nominal accusative complements (and with an optional dative complement as in (122)) are very close to structures with infinitives, they are sometimes even equivalent. Consider, for instance, the following pair of examples: (125)

a. die auf zahlreiche Nelkenstöcke verwendete Sorgfalt versprach den mannigfaltigsten Flor (GOE/AGM.07859, IDS-Corpus) ‘The care that had been taken to cultivate numerous carnation bushes promised to produce the most abudand display.’ b. Und wie manche Laube versprach nicht zunächst mit Geißblatt, Jasmin, rebenartigen und rankenartigen Gewächsen zu prangen und zu schatten. (GOE/AGM.07859, IDS-Corpus) ‘And how many a garden-house did not immediately promise to give a resplendent display of honeysuckle, and jasmine, with vines and tendrils that would provide shade.’

In each sentence in (125) versprechen conveys the same meaning. It indicates that the subject of the sentence, which cannot be interpreted as (verbally) committing itself to something, in some way senses the (future) situation expressed either in the form of a nominal accusative complement (a) or in the form of an infinitive complement (b). This meaning can be also described in Traugott’s (1997: 188) words, who argued for a similar view on the development of the English verbs promise and threaten: “there is something about the subject that leads to an expectation of the proposition coming into being”. This interpretation of versprechen arises only in specific – untypical contexts – where the subject is usually non-animate and

296 Diachronic corpus study of the four constructions cannot be interpreted as an intentional agent. Note that similar suggestions were made about the verb drohen described in the previous section. Importantly, human subjects are also found in structures where an intentional interpretation of the subject is ruled out by further contextually present information: the subject is not really committing himself/herself to something. Note that the only infinitive verb occurring in such structures is werden ‘to become’, cf.: (126)

sie hatte vieles von der Mutter, versprach schon frühe sehr zierlich und reizend zu werden. (GOE/AGM.00000, IDS-Corpus) ‘She had a lot from her mother and promised to become very delicate and charming.’

In most cases such structures with infinitive complements allow for an evidential interpretation as we know it from PDG. We consider these structures the critical context for the triggering of the evidential construction of versprechen. This is due to the following observations: (i) all of them have the same meaning, i.e. some person is described as showing a potential to become something in the future, (ii) the only infinitive verb used in such structures is werden, and (iii) the source of promise is always the subject, i.e. the person itself, be it its outward appearance, its known talents, qualities or something else. From today’s point of view, given that we have an evidential construction versprechen & zu-infinitive in PDG, such structures may be interpreted as having evidential meaning. Seen in the system of versprechen uses existing in the 18th and 19th centuries, where this development has not yet reached this advanced stage, an evidential interpretation of these structures is rather problematic and should be better seen as a bridging structure which is ambiguous between two alternative interpretations. On the one hand, it shares the relevant properties with the existing variant of versprechen with the meaning ‘to portend’. On the other hand, the evidential interpretation is possible, because the personal involvement of the speaker is prominent – it is the speaker who makes the statement on the basis of either his/her knowledge about the person or the person’s outward appearance. We suggest that these structures played a relevant role in the grammaticalization of the evidential construction versprechen & zu-infinitive, as they triggered the appearance of contexts, where other verbs than werden come to be used as infinitive complements of versprechen and the subject could not be interpreted as having control over the verbal process anymore.

Versprechen & zu-infinitive 297

(127)

Offenbach am Main zeigte schon damals bedeutende Anfänge einer Stadt, die sich in der Folge zu bilden versprach. (GOE/AGD.06345, IDS-Corpus) ‘Even then, Offenbach am Main displayed significant features of the town into which it promised to develop in the future.’

The distinct property of such contexts is that the subject is not (only) the source entity that leads to the expectation expressed in the proposition. Rather, versprechen links the situation described in the proposition to some other situation, which is now seen as a source pointing towards the described situation. Thus a relevant shift in the meaning of the construction versprechen & zu-infinitive occurs (importantly, this shift concerns only infinitive constructions, no other structures participate at this development), whereby the source entity is re-located from the subject entity to some other entity. This shift is the same change that we have observed for drohen (see Section 8.3.6). This can be best illustrated if the following structures are compared to each other: (128)

(129)

Geht die Sonne des Morgens auf und verspricht einen feinen Tag. (GOE/AGW.00000, IDS-Corpus) ‘The sun of the morning is rising and promises a nice day.’ „Gehen wir aber zur Ruh … um den morgenden Tag, der heiter zu werden verspricht, von früh auf zu nutzen“. (GOE/AGM.07859, IDS-Corpus) ‘Let us rest now … to use the following day right from the early morning, which promises to be clear.’

In (128) it is the sun, i.e. the subject of the sentence, which is the source of the nice day. In (129), on the other hand, the subject ‘day’ cannot be interpreted as a source of itself becoming clear. Instead, something else serves here as a source: most probably, there are some observable pieces of information (e.g. clear sky, warm weather in the evening, etc.) which point towards this particular expectation. Thus we can already speak here of the initial stage in the development of a new construction type with evidential semantics, represented by versprechen & zu-infinitive. Taking into account our analysis of the synchronic status of this construction (see Section 6.4), we can state that the construction versprechen & zu-infinitive has been grammaticalized towards an evidential auxiliary and that this process was crucially influenced by the above described contextual configurations at-

298 Diachronic corpus study of the four constructions tested first in the 19th century. This particular (re-)location process of the source entity from the subject referent to some entity outside the proposition is the distinct characteristics of the evidential construction type of versprechen. In our data from the 19th century we found clear examples of evidential versprechen that manifest the evidential function of the verb, cf.: (130)

(131)

zu schlafen, um alsdann mit frischem Blick die dem Aufgang der Sonne voreilende Venus, welche eben heute in ihrem vollendeten Glanze zu erscheinen verspräche, zu schauen und zu begrüßen. (GOE/AGM.07859, IDS-Corpus) ‘to sleep in order to see and to greet [...] Venus who promised to come today in her perfect splendor.’ was ausserdem Arnolds Verhältniß zu Mazzini zu befestigen versprach… (meg/GAN.01007, IDS-Corpus) ‘what additionally promised to strengthen Arnold’s relationship to Mazzini…’

As has been shown in Section 6.4, versprechen has not reached the same degree of grammaticalization as drohen, which otherwise displays many parallel features and is its immediately opposed partner among the evidential constructions. In particular, versprechen is restricted to specific contexts, e.g. it does not occur with animate subjects and action verbs at the same time, whereas drohen does.

Chapter 9 Summary: the diachrony of the four constructions

This chapter summarizes the overall diachronic development of the four German evidential constructions. It integrates the results of the corpus investigation with the theoretical considerations discussed in Chapter 5. As a result, the development of the four constructions of German will be represented as a unified grammaticalization process which starts from individual lexical sources and leads to a consolidation of their (more) grammaticalized variants into a relatively tightly organized paradigm of inferential evidential markers as we observe it in PDG. The first three sections of this chapter are concerned with the relevant changes in the semantics of the verbs werden, scheinen, drohen and versprechen in the course of their developmental processes towards evidential (semi-)auxiliaries. Semantic shifts, which in the previous chapter were described at length, will be integrated into the dynamic grammaticalization concepts introduced in Chapter 5: metaphorization and metonymization scales (Section 9.1.1), scheme retention (Section 9.1.2), and successive semantic-functional stages (Section 9.1.3). In Section 9.2 the model of the successive stages in grammaticalization associated with specific context types or constructions (see Section 5.3) will be applied to the diachronic development of the German evidential constructions.

1. Relevant semantic and functional changes 1.1. Metaphorization and metonymization scales The notion of metaphorization and metonymization was introduced in Section 5.1.1. As has been shown in numerous studies dealing with grammaticalization, a large percentage of meaning changes accompanying grammaticalization can be described in terms of successive procedures whereby meanings evolve from concrete/less abstract to more abstract. In the following we aim at reconstructing such scales of semantic development for the four German evidential constructions treated here. The empirical proof for these paths and a detailed description of each individual semantic shift and of contextual factors having promoted them – which necessarily bear

300 Summary: the diachrony of the four constructions specific peculiarities for each individual construction – have been dealt with extensively in the previous chapter. An abstraction scale represents a gradual and continuous semantic development whereby there are no individual stages with clearly defined boundaries. It is a simplification when we speak of individual steps in the semantic development of a verb in the following; the boundaries between individual steps are by necessity fuzzy. The relevant semantic changes occurred to the construction werden & infinitive on its way towards an inferential evidential marker may be sketched in the following manner: Table 1. Relevant changes in the semantics of werden concrete process > (ingressive) ‘designation of the beginning of an event’

relation between situations (sequential relation) ‘indication of a relation between two events’

>

mental state (inf. evidential) ‘marking of inf. evidentiality’

As shown in Table 1, in the course of the grammaticalization process the meaning of werden becomes more abstract: it gradually changes from denoting the initial state of a situation towards marking a very abstract (grammatical) notion of inferential evidentiality. This shift in meaning happens in two steps and occurs in specific contexts of use (for further details see Section 9.2). First, starting from the aspectual ingressive lexical meaning, in the MHG and in the early ENHG periods werden acquires the meaning of a (textually supported) sequential relationship. In a second step, the meaning of werden becomes still more abstract: in ENHG werden & infinitive develops towards an inferential evidential construction. In this function it indicates the mere fact that the speaker infers the proposition from some information available to him/her. The inferential evidential meaning can always be described as the expression of a particular mental state of the speaker, as it implies an obligatory link to the speaker, who infers the asserted proposition from some information pieces. The evidential variant of werden means that (i) the speaker has access to some information, which (ii) s/he interprets as pointing towards the described event. Thus the indication of the speaker’s mental operation of inference serves as the primary meaning of the grammaticalized (evidential) verb werden. The relevant semantic changes associated with the grammaticalization process of the construction scheinen & zu-infinitive may be represented in terms of an abstraction scale as well:

Relevant semantic and functional changes 301 Table 2. Relevant changes in the semantics of scheinen concrete process (visual effect) ‘designation of a visual effect’

>

psychophysical observation > (observed situation) ‘designation of an observed situation’

mental state (inf. evidential) ‘marking of inf. evidentiality’

The developmental path represented in Table 2 indicates the relevant semantic changes that happened to the verb scheinen during its grammaticalization towards an inferential evidential marker. Starting from its prototypical uses as an intransitive verb with the meaning ‘to produce a visual effect’, thus referring to a dynamic process, scheinen develops a more abstract meaning and comes to refer to states which are perceived in some way by an observer. This semantic development occurs in the MHG period (and is contextually marked by the presence of a facultative dative participant introducing the perceiving person) and continues during the ENHG period. In the course of the 18th century the construction scheinen & zuinfinitive semanticizes the inferential evidential meaning, which is still more abstract than the designation of an observed situation. The inferential evidential meaning of scheinen & zu-infinitive is the expression of a particular mental state of the speaker (see above for werden). The semantic development of the verb drohen from denoting a specific speech act (performed by the subject of the sentence) to referring to a particular mental state/operation (of the speaker) constitutes a particular realization of the abstraction scale, too, and may be summarized as follows: Table 3. Relevant changes in the semantics of drohen concrete action > (speech act) ‘declaration of the intention to do something undesirable’

psychophysical observation > (observed situation) ‘designation of an observable undesirable situation’

mental state (inf. evidential) ‘marking of inferential evidentiality’

The source of this development is provided by drohen as a speech act verb. The original semantics of lexical drohen gradually changes into a designation of an (observed/observable) and negatively evaluated situation which emanates from some entity. This process starts in the ENHG period and continues till the 18th century, where the non-agentive variant of drohen establishes itself as an independent lexical variant of drohen (see Section 8.3 for details). In the course of the 18th and 19th centuries the meaning of drohen becomes even more abstract: the inferential evidential meaning of

302 Summary: the diachrony of the four constructions (the grammaticalized variant of) drohen evolves, i.e. drohen comes to indicate a particular mental state/operation of the speaker. As described in the previous chapter, the general developmental processes of drohen and versprechen are parallel to a great extent, and so are the steps in the semantic change of the two verbs. The abstraction scale of versprechen on its way towards a grammaticalized evidential verb is represented in Table 4 below: Table 4. Relevant changes in the semantics of versprechen concrete action > (speech act) ‘declaration of the intention to do something’

psychophysical observation > (observed situation) ‘designation of an observable situation’

mental state (inf. evidential) ‘marking of inferential evidentiality’

Early in the development the lexical verb versprechen with commissive speech act semantics gradually changes into a verb which refers to an observed/observable situation, which is usually evaluated positively. This shift occurs in the 18th century and is closely linked to the analogical development of drohen. In the 19th century this meaning variant of versprechen is already quite common, and in the course of the 19th century, the construction versprechen & zu-infinitive develops towards an inferential evidential grammatical construction. However, as noted in Section 8.4, versprechen & zu-infinitive has not reached the same degree of grammaticalization as drohen & zu-infinitive and, accordingly, the grammaticalized variant of versprechen shows a lesser degree of semantic abstraction as compared to drohen. This fact is not explicitly represented in Table 4 as it focuses on the most basic changes in the semantics of versprechen. It should be mentioned here that, although the semantic changes of the four verbs are represented in a unified way, i.e. as two basic shifts in the meaning of the respective verb, the degree of semantic abstraction of each verb is not the same. The (diagrammed) correspondence between the stages in the semantic development processes of the four verbs neither implies a chronological correspondence between the individual stages (see Section 9.2) nor does it mean that the degree of semantic abstraction (in the intermediate stages as well as at the outcome of the grammaticalization process) is identical for each verb.

Relevant semantic and functional changes 303

1.2. Scheme retention and reinterpretation In semantic change accompanying grammaticalization, there is a common denominator, i.e. an aspect of meaning that remains stable over time and links different stages of the development to each other. It can be described as a semantic relational template that is preserved from the source (lexical) entity through all stages up to the grammaticalized functions, whereby its fundamental relational aspects get successively transferred to different semantic domains (see Section 5.1.2). The domains are characterized, on the one hand, by increasing abstraction, but, on the other hand, they also provide new semantic features that had not been present before. In the following the relational templates of the verbs werden, scheinen, drohen and versprechen and of their grammaticalized variants will be reconstructed with the help of the basic schema consisting of three positions – source, path, and goal – introduced in Section 5.1.2. It will be shown that the basic semantic template of the lexical sources has been preserved notwithstanding significant reinterpretations and transformations, the most crucial of which concern the respective reinterpretations of the source position in the scheme. The original lexical semantics of the verb werden is aspectual ingressive. In this meaning werden indicates a situation in its onset. Today, this ingressive meaning is clearly manifested in the copula verb werden, cf.: (1)

Sie wird krank. ‘She is getting ill.’

This basic semantics may be visualized by postulating the following abstract relational template for werden: Table 5. Relational template of the lexical verb werden source > subject of the sentence

path > werden

goal (new) situation

The subject of the sentence functions as a source entity which either performs or undergoes the process/the situation introduced by the verb werden. The “new situation”, the beginning of which is indicated by werden, takes over the position of the goal. Werden itself manifests the turning point, i.e. the change in the subject entity from some old into some new

304 Summary: the diachrony of the four constructions situation (which may be conceived of as a quality, position, status, entity, etc.). As for the highly grammaticalized evidential construction werden & infinitive, as it is exemplified in (2) below, the same abstract relational scheme may be used as a base: Table 6. Relational template of evidential werden source > speaker’s information

path [werden

>

goal proposition]

The source, i.e. the entity to which the construction werden & infinitive bears a close relation and to which the described event (the proposition) is anchored, is the speaker’s information source. The goal, i.e. the event introduced by werden & infinitive (the proposition) is the result of the speaker’s inference process. The third element in this abstract semantic structure is the relational component which holds between the source and the goal: this pure relational component can be defined as the inference process itself, i.e. the mental operation of the speaker. (2)

Sie wird davon gehört haben. ‘She is bound to have heard of it.’

As can be seen from this analysis, the basic semantic template of the lexical variant of werden has been preserved. The individual positions of the scheme, however, underwent changes whereby they have been transferred to other domains, e.g. ‘situation’ > ‘proposition’ and ‘subject of the sentence’ > ‘speaking subject (i.e. the speaker)’. In the following, we will show that the other verbs show the preservation of the semantic template of their original lexical verbs as well. While the individual positions have been transferred and reinterpreted, the basic schema has remained stable. The lexical verb scheinen is an intransitive verb denoting the emission of a visual effect (by an entity). It usually requires a subject referring to the source of the ‘shining’ process; it is always possible to further specify the verbal process of producing visual effect by adding e.g. modifying elements, which – under suitable conditions – may be interpreted as adverbial or predicative elements, thus producing ambiguous utterances like in (3): (3)

Sie scheint hell. ‘She shines brightly’, i.e. ‘the shining is bright.’

Relevant semantic and functional changes 305

OR ‘She seems bright’, i.e. ‘she is bright.’ This potential is mostly exploited in the copula construction of scheinen, where the combination of scheinen with a predicative expression is obligatory. The prototypical semantics of scheinen may be represented as follows: Table 7. Relational template of the lexical verb scheinen source > subject of the sentence

path scheinen

> goal property of the subject entity/or the whole process

The relational structure of scheinen displays three basic elements: the source of the verbal process, which is usually an entity producing a visual effect, the path, and the goal, i.e. the manner in which the verbal process is realized. The source position is occupied by the subject of the sentence, the goal is realized either in the form of a predicative expression (as concerns the copula scheinen) or in the form of a facultative (mostly adverbial) complement of the main intransitive verb scheinen, see (3). As for the evidential construction scheinen & zu-infinitive, the same abstract relational scheme can be used as a basic description of its semantic content. The positions in the scheme are realized as follows: Table 8. Relational template of the evidential verb scheinen source > speaker’s information

path [scheinen

>

goal proposition]

The basic relational structure has thus remained the same, while the status of each position has changed in the course of the grammaticalization process. The verb scheinen serves to mark a specific evidential value, i.e. the inferential evidential value. The goal position is represented by the event introduced by scheinen & zu-infinitive (the proposition), which is the result of the speaker’s inference process: (4)

Sie scheint dich zu mögen. ‘She seems to like you.’

306 Summary: the diachrony of the four constructions Drohen as a speech act verb indicates a communicative act which is normally performed by a human subject. (5)

Sie droht mich zu verlassen. ‘She threatens to leave me.’

Usually drohen requires complements (infinitive complement, direct or indirect speech clauses as complements, prepositional complement with mit, rarely accusative object) which further explicate the content of the threat (for details see Section 6.3). This prototypical semantics of drohen can be visualized in the following manner: Table 9. Relational template of the lexical verb drohen source > subject of the sentence

path drohen

>

goal content of the threat

This core meaning of drohen is clearly manifested in its every use as a lexical verb. The source position is always taken over by the subject of the sentence, the goal is realized in the form of infinitive complement, (in)direct speech clause complement, prepositional complement or accusative object. The evidential construction drohen & zu-infinitive displays the same basic relational scheme as the semantic templates of evidential werden and scheinen introduced above. The individual positions in the scheme have been reinterpreted in the course of grammaticalization and are now realized as follows: Table 10. Relational template of the evidential verb drohen source > speaker’s information source

path [drohen

>

goal proposition]

Thus, for the inferential evidential construction drohen & zu-infinitive, the abstract relational scheme of the lexical verb drohen has been preserved, together with a change in the concrete linguistic realization and semantic content of the position. (6)

Sie droht krank zu werden. ‘She threatens to fall ill.’

Relevant semantic and functional changes 307

Versprechen as a lexical verb prototypically indicates a commissive speech act preformed by a human subject. Usually it requires further complements (infinitive complements, nominal accusative complements, direct or indirect speech clauses as complements) which explicate the content of the promise. (7)

Sie verspricht (mir), mein Geheimnis zu bewahren. ‘She promises (me) to keep my secret.’

The following abstract relational template may be taken as a basis of the meaning of lexical versprechen: Table 11. Relational template of the lexical verb versprechen source > subject of the sentence

path > versprechen

goal content of the promise

The scheme exhibits three obligatory elements: the source of the promise, which is a human subject performing the commissive speech act, the path, i.e. the specification of the speech act, and the goal, i.e. the content of the speech act. Versprechen itself designates the verbal announcement of some action or situation. This core meaning of versprechen is clearly manifested in its every normal use, in PDG as well as in the earlier periods of German (for details see Sections 6.4. and 8.4). As regards the relational semantic structure of the evidential inferential construction versprechen & zu-infinitive, it can be visualized in the following fashion: Table 12. Relational template of the evidential verb versprechen source > speaker’s information source

path > [versprechen

goal proposition]

The abstract semantic template, containing three basic positions, has remained the same as that of the lexical verb. A relevant shift has taken place: the position of the source, being inherent in the semantics of the evidential versprechen, is now the speaker’s information source, and it is now implicit, i.e. it does not require an explicit realization in form of the subject of the sentence. Evidential versprechen modifies the whole proposition.

308 Summary: the diachrony of the four constructions (8)

Es verspricht sonnig zu bleiben. ‘It promises to stay sunny.’

Summarizing this section, it can be said that each of the four grammaticalized verbs treated here illustrates a clear instance of the preservation of the basic semantic template of the respective source item (scheme retention). Moreover, all four verbs show very similar developments concerning the reinterpretation of the basic positions in the scheme. The most relevant change in the semantic structure of werden, scheinen, drohen and versprechen consists in the transformation of the source position. Namely, it shifts from a subject referent, which is responsible for the process introduced by the respective lexical verb, to the deictic position of the speaker, i.e. his/her information source. In their function as evidential markers, the grammaticalized verbs deictically anchor the propositional content in the speaker’s perspective. The position of the source is permanently occupied by the speaker’s information source. That means that we can speak here of the externalization or relocation of the source position and its identification with the speaker’s (information status) position.

1.3. Successive semantic-functional stages As we have argued in Section 5.1.3, in addition to the semantic change accompanying grammaticalization there is a functional change concerning the semiotic status of the item in question. The development runs from mainly referential to mainly textual and connective functions, and further evolves to indexical grammatical functions. To visualize this development, we have used Traugott’s (1989, 1995) model of semantic change that is formulated as three tendencies of semantic change. The development, as we understand it, can be sketched as follows: (i) (self-contained) referential function > (ii) text-integrative/connective function > (iii) indexical-grammatical function

This model has been applied to the development of the German evidential constructions. The common feature of the four constructions, i.e. their indexical-grammatical function is taken as granted and will not be treated further (for details see Chapters 2 and 3). The four constructions are indexical, i.e. deictic, signs, as they relate the proposition they modify to a

Relevant semantic and functional changes 309

relevant aspect of the speech situation, namely to the sphere of knowledge of the speaker. The relatively high degree of grammaticalization of these constructions proves their (formal) status as grammatical elements. Werden as a lexical verb with aspectual semantics serves to introduce events which are bounded in the onset. It has referential function as it refers to processes in the world “out there”. In the MHG period werden & infinitive comes to be used in specific contextual configurations which trigger its reinterpretation as a marker of a (textually present) sequential relation. Here, werden has a clearly text-connective function: it serves to link two textually neighboring clauses/propositions to each other. Finally, werden as an evidential marker has indexical-grammatical function. It incorporates a strong link to the speaker, i.e. to the deictic sphere of the speaker. In other words, it relates the proposition to a relevant reference point, which is the speaker’s information source. This change in the functional status of werden can be illustrated in the following manner: Table 13. Functional changes in the development of werden referential aspectual

>

text-connective > successive relation between two textually expressed events

indexical-grammatical inferential evidential status of the proposition

The lexical verb scheinen codes a visual effect. From the late 16th century onward this verb in combination with adjectives, adverbs and infinitives comes to indicate a relation between two situations which are compared to each other (on the basis of their appearance). In such cases, scheinen very often occurs in textual configurations in which two different entities or events are explicitly mentioned in the text. Together with comparative particles such as als ‘as, than’ and wie ‘as’, scheinen contributes to the establishment of such textually cohesive structures. From the beginning of the 18th century, in construction with infinitives, scheinen develops towards an inferential evidential, indicating that the described event is inferred from some information pieces available to the speaker and is thus anchored in the deictic position of the speaker. Table 14. Functional changes in the development of scheinen referential > visual effect

text-connective > comparative relation between two textually expressed events/entities

indexical-grammatical inferential evidential status of the proposition

310 Summary: the diachrony of the four constructions The verbs drohen and versprechen display a very similar functional development. For this reason we will treat the changes in functional status of these verbs in conjunction. The original lexical verbs drohen and versprechen serve to (verbally) announce a situation which is intended by a person. In this meaning they refer to a particular communicative event, which however may or may not coincide with the speech event. That is, they have a referential function. From the ENHG period onward, especially in specific contextual configurations, e.g. in combination with accusative objects or with infinitive complements, both verbs come to mark a successive relationship between two situations. The speech act semantics is largely bleached and only the relational connection between textually represented situations is promoted, whereby the event introduced by drohen or versprechen is conceptualized as following others. Finally the textually anchored successive relationship between the described situations gradually changes into the inferential evidential meaning. Evidential variants of drohen and versprechen now indicate that there is a logical connection between two events, one of which pertains to the speaker’s deictic sphere and is conceived of as information source for the other. Table 15. Functional changes in the development of drohen and versprechen referential > speech act

text-connective > successive relation between two textually expressed events

indexical-grammatical inferential evidential status of the proposition

By way of conclusion we can state that, first of all, the four German evidential periphrases arise via grammaticalization processes developing from main verbs towards auxiliary verbs. Their overall semantic development is an exemplary case of the successive reinterpretation of an abstract relational template (see Section 9.1.2). Furthermore, in their grammaticalization, they exhibit an intermediate stage at which they are integrated in specific textual structures, where they serve a particular text-connective or text-integrative function. In this function, they contribute to the forming of cohesive textual segments whereby they operate between propositions. That is, they participate in creating and establishing links across clause boundaries. As grammaticalization proceeds, they gradually become grammatical markers, i.e. they acquire indexical-grammatical functional status. They modify propositions by relating them to a particular deictic reference point. At the same time, their integration into particular text

Stages, contexts and constructions 311

structures becomes looser till they are no longer identified with specific contextual configurations.

2. Stages, contexts and constructions This section summarizes the development of the four German evidential constructions with respect to the model of the relevant context types introduced in Section 5.3. The model integrates semantic, morphological and structural aspects into the definition of context types, and emphasizes the role of paradigmatic relations among constructions at a certain historical stage in a particular language as well as the influence of paradigmatic oppositions in the target category of a grammaticalization process. Below, the model is reproduced from Section 5.3: Table 16. Context types in grammaticalization Context

Meaning/Function

Construction types

Stage I preconditions of grammaticalization Stage II triggering of grammaticalization

untypical contexts

conversational implicature

no particular construction type

critical context

multiple opacity

“extragrammatical idioms”

Stage III reorganization & differentiation

isolating contexts

polysemous / heterosemous items

“formal or lexically open idioms”

Stage IV paradigmatic integration

paradigmatic context

paradigmatic oppositions (choices) with reduced relational meanings, i.e. grammatical meanings

paradigmatic choices within an abstract constructional schema

In applying the model to the diachronic development of the four verbs treated here, we concentrated primarily on the reconstruction of critical and isolating contexts. As untypical contexts make use of existing constructions and can be easily interpreted due to their compositional structure, their role in the grammaticalization processes is less relevant than those of critical and isolating contexts. Furthermore, describing the structure of

312 Summary: the diachrony of the four constructions isolating contexts for each case, only the isolating context for the newly developed – i.e. grammaticalized – variants of the verbs is stated explicitly. This does not mean, however, that there are no isolating contexts for the still pertaining – original – lexical sources which continue to exist. Usually, isolating contexts for the grammaticalized and for the non/less grammaticalized variants show complementary distribution with regard to their relevant semantic and structural properties. Hence isolating contexts for non/less grammaticalized variants of the elements in question may be defined negatively with respect to the isolating contexts for the grammaticalized variants. Let us start with the diachronic development of werden. As we have shown in Section 8.1, there are two relevant context types which considerably triggered the development of the inferential evidential construction werden & infinitive. First, in the MHG period, werden & infinitive occurs in specific contextual configurations such as temporal, conditional, causal, consecutive syntactic structures with correlative conjunctions, adverbs etc. In these untypical contexts werden receives an additional interpretation, i.e. it comes to indicate a sequential relationship between two textually explicated situations. Werden marks that there is a close link between the described situations, and the specific interpretation of this sequential relation as conditional, temporal, consecutive etc. depends on contextually present cues. Second, in the ENHG period, werden & infinitive progressively occurs in another kind of specific contextual configurations, i.e. it is embedded under verbs/expressions of cognition, perception and emotion. This is the critical context which triggers a further reinterpretation of the construction, whereby werden & infinitive incorporates a strong link to the speaker, i.e. to the speaker’s inner world. The meaning of the successive relationship gradually changes into a logical relationship between situations, whereby the first one serves as a premise for the conclusion about the second one. The co-operation of these two relevant contexts, which follow each other in time, as we propose, leads to the establishment of the inferential evidential semantics of the construction werden & infinitive. The meaning of a sequential relationship is “put” in the contexts with verbs of cognition, perception and emotion and is so reinterpreted as a specific sequential relationship, i.e. as a logical relationship between premises and conclusions. Note that both contexts show the tendency to disappear: whereas werden & infinitive is only rarely found in adverbial clauses (with the meaning of sequential relationship) in PDG, structures with embedding cognitive verbs are still common (see Section 6.1.2 for a detailed description). However, in PDG inferential evidential werden &

Stages, contexts and constructions 313

infinitive is used not only within such syntactic structures but it has broadened its use also beyond those (which is captured in the description of the isolating context in Table 17). An important structural property of the critical context in the development of werden is that the functional status of werden is highly ambiguous: it can be interpreted either as a lexical verb (more particularly, as a copula verb) in a combination with a verbal (i.e. infinitival) or adjectival (i.e. participial) predicative complement or as an auxiliary in combination with a main verb (i.e. its infinitive or present participle). That is, the status of werden is highly opaque. Table 17. Context types in the grammaticalization of werden Context

Meaning/Function

Construction types

Stage I preconditions of grammaticalization Stage II triggering of grammaticalization

untypical contexts

ingressive + conversational implicature of a sequential relation ingressive OR sequential relation OR inferential evidential

werden in adverbial clauses

Stage III reorganization & differentiation

isolating context

critical context

werden & (present) infinitive/present participle in contextual structures with embedding expressions of cognition, perception, and emotion inferential evidential werden & infinitive (present and perfect) in indicative sentences, in present and past temporal perspective1

The same observation can be made about the critical contexts in the development of the other three verbs: being used in the critical contexts, they equally allow for both – more lexical and more grammaticalized – interpretations. On the one hand, the structure of the infinitive constructions may be conceived of as a complex structure whereby the verb and the infinitive complement are seen as forming two different predications. On the other hand, the same structure may be interpreted as one complex predication, whereby the verb and its infinitive complement form a structural unity. This means that the critical context is the locus of reanalysis: it is the context type in which a reanalysis in terms of an analytical grammatical form (i.e. auxiliary & main verb) takes place or at least is made possible. Most

314 Summary: the diachrony of the four constructions importantly, it is the context that triggers the ambiguous interpretation of the item under consideration, and not the item itself. That is, the grammaticalizing element does not contain the possible alternative interpretations by itself, but only in this particular type of contextual configurations. The fourth stage in the grammaticalization process of the verbs will be treated in the next section, as it is the paradigmatic relationships between constructions which are relevant for this stage of the development. Here, we concentrate on the relevant context types which have played a crucial role in the grammaticalization processes of the verbs werden, scheinen, drohen and versprechen respectively. For the development of the evidential construction scheinen & zuinfinitive, we have detected a relevant – critical – context type, to which crucial changes in the semantics of scheinen are directly linked. First, in the ENHG period, namely in the 16th century, a new construction type develops: scheinen & zu & sein. Most probably, it happens via analogical transfer, whereby other copula verbs serve as a model for this extension. The relevant pre-conditions for this development construction was created in the untypical contexts, where the original lexical meaning of (the copula) scheinen of emanating a visual effect was still present, however the comparative structure allowed a conversational implicatures focusing on the difference between real existence and visual impression. The possibility to compound the copula with the infinitive of the verb sein creates the possibility to explicitly compare or confront two situations: on the one hand, the situation of scheinen, i.e. the situation as it appears to someone; on the other hand, the situation of sein, i.e. the situation as it really is. It should be noted, however, that these two situations, albeit conceptualized separately, may well coincide, i.e. the situation may be well the same way it appears. That is, the description of an event by means of scheinen does not have to contradict reality. In this way, the conceptual split in the semantics of scheinen occurs, i.e. the distinction between two situations: the one expressed by scheinen and the other implied by it. It is at this stage that the direct evidential meaning of scheinen most probably develops. The central characteristics of this contextual configuration are (i) the presence of a 3rd person subject referent (animate as well as inanimate), and (ii) the presence of the infinitive of the verb sein. Second, starting from the 16th century and further continuing in the 17th and 18th centuries, the construction scheinen & zu & sein is again analogically extended to other types of infinitives. Most relevant in this connection is the fact that scheinen comes to be combined with perfect and passive infinitives (like werden does). In such – isolating – contexts, the

Stages, contexts and constructions 315

meaning of simultaneity between two compared situations is weakened in favor of the new inferential evidential interpretation. That is, the situation of acquiring knowledge about some facts on the on hand and the described situation on the other cannot be brought into a direct (simultaneous) relation any more. Instead, the first one is reinterpreted in terms of pieces of evidence on which the statement about the described situation is based. Several contextual factors may be named here which characterize such contexts: (i) scheinen combines with all (morphological) types of infinitives and with infinitives of all verbs, including verbs referring to nonobservable situations, e.g. cognitive verbs; (ii) it combines with 1st and 2nd person subject referents; (iii) it is often used in contexts where the pieces of evidence are textually mentioned. Table 18. Context types in the grammaticalization of scheinen Context

Meaning/Function

Construction types

Stage I preconditions of grammaticalization Stage II triggering of grammaticalization

untypical contexts

scheinen in contexts of comparison

Stage III reorganization & differentiation

isolating context

visual effect + conversational implicature of a comparative relation comparison of an entity to another one due to its visual appearance OR due to its impression on an observer (direct evidential) OR due to some other pieces of information (inferential evidential) inferential evidential

critical context

scheinen & zu & (present) infinitive of the verb sein with 3rd person subject referents

scheinen & zu & present and perfect infinitives of any verb with 1st and 2nd person subject referents

The interplay of these contexts in the history of German has led to the establishment of the inferential evidential function of scheinen & zuinfinitive. The meaning of a simultaneous relationship between two (comparable) situations undergoes further shift towards inferential evidential semantics, i.e. a logical (inference) relationship between some information

316 Summary: the diachrony of the four constructions pieces as premises for a conclusion (proposition). The relevant stages in the grammaticalization of scheinen are summarized in Table 18. For the development of the inferential evidential function of drohen & zu-infinitive, again two relevant contextual configurations played a crucial role (see Table 19). First, in the ENHG period contextual configurations evolve, where drohen starts to be frequently used with inanimate subjects. In such – critical – contexts, the source of a threat is no longer restricted to human subjects being able to communicate their intentions. Instead, it is extended to other subject referents. The relevant preconditions for the critical context to evolve were created in the untypical contexts where inanimate subjects came to be combined with drohen via metaphorical transfer, or personification. Additionally, in critical contexts drohen receives a kind of temporal-aspectual interpretation, meaning that the situation (the content of the threat) has not yet occurred at the time of speaking. As far as the complementation structure is concerned, the verb drohen in this context is combined either with an accusative object or with an infinitive complement. Table 19. Context types in the grammaticalization of drohen Context

Meaning/Function

Construction types

Stage I preconditions of grammaticalization

untypical contexts

metaphorical usage of drohen with inanimate subjects

Stage II triggering of grammaticalization

critical context

Stage III reorganization & differentiation

isolating context

speech act of threat + conversational implicature of undesirable situation emanating from the subject entity undesirable situation emanating from some entity OR temporal-aspectual OR inferential evidential inferential evidential

drohen & zuinfinitive/drohen & nominal accusative complement with inanimate subject referents drohen & zu & (present) infinitive of verbs denoting a nonaction; animate and inanimate 3rd person subject referents

Second, in the course of the 18th and 19th centuries, the construction drohen & zu-infinitive comes to be used in contexts, where the subject of the sen-

Stages, contexts and constructions 317

tence cannot be interpreted as a cause or an initiator of the situation introduced in the infinitive complement. These – isolating – contexts are represented either by contextual configurations, where an explicit link to another situation, from which the situation described with drohen results, is textually present; or by contextual configurations, where there is an implicit link to the speech situation, in which some facts (pointing towards the described situation) can be reconstructed. A further crucial semantic change of drohen occurs at this stage: drohen loses the ability to bear a close semantic connection to the grammatical subject of the sentence. Instead, it has the ability to indicate the information source for the proposition. An important observation which holds for every grammaticalization process including the diachronic development of the four evidential verbs described in this book is the following. The chronological succession of the two relevant contexts is such that the second – isolating – type of context crucially depends on the existence of the first – critical – one. The critical context necessarily creates the possibility for other contexts (among them isolating context) for the newly developing grammatical functions to develop. The isolating context could have not arisen without the prior existence of the critical context. As concerns the development of drohen, the meaning of an (observable) causal/sequential relationship between the subject referent and the (infinitive) complement of drohen, which had been already established in the critical context, was realized in the second – isolating – contexts and was in this way reinterpreted as a specific sequential relationship, i.e. as a relationship between premises and conclusions. Note that the critical contexts show the tendency to disappear. The use of non-agentive drohen with inanimate subjects in combination with accusative complements is largely absent in PDG. Thus we can assume that the structure with nominal accusative objects served as a kind of catalyst in the grammaticalization of the evidential construction drohen & zu-infinitive. As the infinitive construction got used with more and more types of infinitive verbs, this construction was completely replaced by the later. The development of the evidential construction versprechen & zu-infinitive is in many ways similar to the development of drohen & zu-infinitive. In the 18th century, contextual configurations evolve, most probably via metaphorical and analogical extension, whereby versprechen is combined with inanimate subjects (i.e. untypical contexts, see Section 8.4.2). Due to this, the verb receives a new interpretation: it no longer denotes a speech act performed by a human speaker, but is able to refer to devel-

318 Summary: the diachrony of the four constructions opments occurring without an active participation of any intentional subject. The untypical contexts create the pre-condition for the critical contexts to arise, which are characterized by the presence (i) of a 3rd person (usually) inanimate subject that cannot be interpreted as committing itself to something; and (ii) of a nominal accusative complement or an infinitive complement describing the expected situation. A crucial semantic change occurs to the verb at that time: it receives a kind of aspectual-temporal interpretation, meaning that the situation has not yet occurred, but the first indications of the ongoing development are already perceivable. Starting from the 19th century and continuing until now, the construction versprechen & zu-infinitive shows the tendency to be used in contextual configurations where the subject of the sentence can no longer be understood as an entity initializing the development (or the situation) described in the proposition. Instead, the subject itself is the entity which undergoes the development introduced in the proposition. In such isolating contexts, another entity has to be assumed – which are the information pieces available to the speaker – as a source of the proposition expressed. The subject does not control the verbal process of versprechen. Table 20. Context types in the grammaticalization of versprechen Context Stage I preconditions of grammaticalization Stage II triggering of grammaticalization

Stage III reorganization & differentiation

Meaning/Function

untypical speech act of promise contexts + conversational implicature of a development emanating from the subject entity critical (desirable) development context emanating from some entity OR temporal-aspectual OR inferential evidential isolating inferential evidential context

Construction types metaphorical usage of versprechen with inanimate subjects

versprechen & zu & infinitive/versprechen & nominal accusative complement with inanimate subject referents versprechen & zu & (present) infinitive of verbs denoting a nonaction; inanimate 3rd person subject referents

Common grammaticalization path 319

Thus, versprechen turns into a marker that modifies the whole proposition, assigning it an inferential evidential value. It should be noted that only in the isolating contexts versprechen & zu-infinitive starts to function as a grammatical auxiliary indicating inferential evidentiality. The relevant stages of the grammaticalization of versprechen are summarized in Table 20. As we have proposed in the previous chapter, the grammaticalization process of the inferential evidential construction versprechen & zuinfinitive was hampered due to the following facts: (i) the relatively late establishment of the commissive variant of lexical versprechen as the central lexical variant in the paradigm of its construction types; (ii) the establishment of a specific lexical meaning of the construction versprechen & sich & accusative object & von/durch-pp. Therefore, the isolating contexts, in which the inferential evidential construction versprechen & zu-infinitive manifests itself, are more restricted than those of the construction drohen & zu-infinitive described above. In this respect the two constructions differ from each other. As we have repeatedly emphasized, versprechen & zu-infinitive is less grammaticalized than drohen & zu-infinitive, even though they show the same (chronological, distributional, semantic, and structural) regularities in their diachronic developments.

3. Common grammaticalization path In the remainder of this chapter we will argue that the individual developmental paths of the four German constructions scheinen, drohen, versprechen & zu-infinitive and werden & infinitive may be integrated into a unified grammaticalization scenario that represents the establishment of a new grammatical paradigm of inferential evidentiality in German. As suggested in Chapter 7, the four constructions constitute a paradigm of inferential evidentials and represent different semantic values within the domain of inferential evidentiality. As for the diachronic development of these constructions, we may formulate the following conclusions. 1. In the course of the diachronic development, each verb underwent a process of semantic abstraction, which can be represented in the form of metaphorization and metonymization scales (see Section 9.1.1). Irrespective of the source lexical semantics of the verbs werden, schei-

320 Summary: the diachrony of the four constructions

2.

3.

4.

5.

6.

nen, drohen and versprechen, at the final stage the inferential evidential meaning is found, i.e. the indication of a particular mental state/operation of the speaker. Each grammaticalized verb has preserved the basic semantic template of the respective source element. All four constructions have in common that, on the way towards grammaticalized evidential markers, each verb undergoes a process of the externalization of the source entity (see Section 9.1.2). At the final stage of this development, the position of the source is identified with the speaker’s information source. In the process of grammaticalization, each verb passes through three successive semantic-functional changes, whereby its functional status gradually changes from referential function > textual-connective function > indexical-grammatical function (see Section 9.1.3). At the final stage of the development, there are four deictic elements which relate the propositional content to the sphere of knowledge of the speaker. In the process of grammaticalization, each verb passes through four successive stages which are associated with specific construction/context types. At the third stage there occurs a lexical split into two coexisting items: a lexical sign and a grammatical sign (see Section 9.2). In PDG, the grammatical variants of each verb exist independently from their lexical sources,which means the verbs are polysemous. The construction werden & infinitive accomplished its development towards evidential marker already in ENHG. Scheinen & zu-infinitive developed later and was established as an inferential evidential construction in the 18th century. Drohen and versprechen & zu-infinitive developed their evidential readings only in the course of the 18th and 19th centuries. Already in the 19th century, drohen shows a higher degree of grammaticalization as an inferential evidential than versprechen. In fact, versprechen is much less grammaticalized in this function in PDG, too (see Section 6.4).

To summarize our diachronic investigation of the four constructions, we integrate the individual developmental paths of each verb into a unified grammaticalization path. In doing so, we propose a schema which shows the relevant semantics of the source constructions for German inferential evidential constructions and sketch their common developmental path. Table 21 demonstrates the four successive diachronic stages in the grammaticalization of the German evidential verbal periphrases from their lexical source semantics via the intermediate stage of a construction with textually relational meaning towards the expression of inferential eviden-

Common grammaticalization path 321

tiality and further towards the establishment of a grammatical paradigm of inferential evidentiality. Table 21. Successive stages in the grammaticalization of the German evidential constructions source item I: original meaning critical context

II: constructional meaning isolating context

target meaning of grammatical item paradigmatic context IV: paradigmatic oppositions target item

werden

scheinen

ingressive aspectual werden & inf./ pres. part. in contextual structures with embedding expressions of cognition, perception, and emotion successive relation between two events werden & inf. (pres./perf.) in indicative sentences, in present and past temporal perspective

visual effect

drohen

versprechen speech act

scheinen & zu & inf. of the verb sein with 3rd person subject referents

drohen/versprechen & zu & inf. and drohen/versprechen & nom. acc. complement with inanimate subject referents

simultaneous relation between two events scheinen & zu & inf. (pres./perf.) with 1st and 2nd person subject referents

successive relation between two events

drohen/versprechen & zu & (pres.) inf. of verbs denoting a nonaction 3rd person inanimate subject 3rd person referents subject referents inferential evidential meaning

(grammatical) marking of inferential evidentiality [+ speaker’s inference] werden & infinitive

[+ based on experienced information] scheinen & zu & infinitive

[+ based on presently accessible information] drohen & versprezu & chen & zu infinitive & infinitive

322 Summary: the diachrony of the four constructions It should be noted that Table 21 does not claim a simultaneous development for all four verbs. It only demonstrates the existence of a common grammaticalization path with analogous stages for the German evidential constructions. As represented, we distinguish between two basic source construction types from which the German inferential evidential markers have evolved. The first lexical source construction type manifests the successive/consecutive relation between two events. On the one hand, it is represented by the aspectual source semantics of the verb werden which indicates a strongly related continuum between the beginning of the event and the event itself. At the second stage of the development, i.e. in the constructions with werden & infinitive or present participle, its lexical semantics is reinterpreted as an indication of the successive relation between two (different) situations which are connected to each other. Finally, at the third stage, where there is only the infinitive construction left, inferential evidential meaning develops, marking primarily a logical connection between two situations, whereby one of them pertains to the sphere of knowledge of the speaker and is regarded as the information source of the proposition. On the other hand, drohen and versprechen follow a similar developmental path. First, as speech act verbs, they serve to announce an event which is intended by the (grammatical) subject and thus introduce a strong connection between two situations. At the second stage, in constructions with accusative objects or with infinitives, these verbs come to mark a successive relationship between two situations. Finally, the relational connection between the described events turns into the inferential evidential meaning, presenting a logical link between situations one of which pertains to the sphere of knowledge of the speaker and is seen as the information source of the proposition. The second lexical source construction type manifests the simultaneous relationship between two situations. Scheinen, a verb expressing visual effect, implies that some entity triggers a particular effect by means of its appearance. In combination with adjectives, adverbs and infinitives, this verb comes to indicate a relation between two situations which are compared to each other on the basis of their appearance. At the third stage, in construction with infinitives, scheinen develops towards an inferential evidential, indicating that the described event is inferred from some pieces of evidence. To put these two developmental paths in one, we can say that the first source construction type represents a relation between two situations projected to a syntagmatic axis (succession) whereas the second source con-

Common grammaticalization path 323

struction type manifests a relation mapped onto the paradigmatic axis (simultaneity). Table 22 shows the same semantic development focusing on the building of a uniform semantic category: starting from different lexical semantics, the four verbs develop to mark inferential evidentiality passing through the stage of establishing a relational constructional meaning. Table 22. Chronological path of the semantic development of the German evidential constructions beginning (Q) of P ingressive aspectual event Q ĺ event P successive relation announcement (Q) of P speech act

appearance (Q) of P visual effect

event Q § event P simultaneous relation

evidence Q ĺ event P inferential evidential meaning: „Q points to P“

Chapter 10 Summary and outlook

This book has set out to discover a new grammatical category of German: the grammatical category of evidentiality as it is expressed in verbal periphrastic constructions with infinitives – namely werden & infinitive, scheinen/drohen/versprechen & zu-infinitive. Its aim has been to delineate the make-up of this category together with its diachronic development since OHG. The study is founded on a combination of theoretical principles of grammaticalization theory, construction grammar and theories of language change. Its data analysis is based on an in-depth diachronic corpus study guided by the methodological principles to continually adjust empirical and theoretical aspects to each other. The results prove that these procedural decisions have been the right choice. They have led to a complete picture of the present-day structure and of the diachronic development of the category of evidentiality expressed by these four constructions. Moreover, they have led to further insights into several typological issues in the domains of evidentiality and epistemic modality. In the following the major results of this study will be summed up, proceeding from language-specific to more general issues, presenting the condensed results of the investigation into the German system, its implications for the typological description of evidentials and general theoretical issues. The chapter will end with some open questions that still need to be addressed, and with an outlook on some further projects. The central hypothesis, as it is rendered in Section 1.3, is that the German language has been about to develop a system of evidential markers by the grammaticalization of four constructions with infinitives and the verbs scheinen, versprechen, drohen and werden. This means, thus, that the German language has acquired a new grammatical category, i.e., the category of evidentiality, which is integrated into the growing system of verbal periphrastic paradigms in German. This hypothesis has been verified, whereby arguments on all relevant levels of linguistic structure and grammaticalization theory have been taken into account. In the synchronic, present-day part of our study, the following specific assumptions on the synchronic evidential system of present-day German have been confirmed (for a detailed account see Chapter 6 and 7):

Summary and outlook 325

1. The constructions werden & infinitive, and scheinen & zuinfinitive, drohen & zu-infinitive, and versprechen & zu-infinitive are evidential markers in PDG. 2. The four evidential constructions are grammaticalized to a relatively high degree. 3. They build a paradigm and represent distinctions within the domain of indirect inferential evidentiality. These assumptions have been underlined with the help of Lehmann’s (1985) grammaticalization parameters. While the diachronic processes working on Parameter 1a (‘morphological attrition’) and Parameter 6 (‘fixation’) are not distinctive between less grammaticalized and more grammaticalized forms in any of the four items, all other parameters (i.e., Parameters 1b: ‘semantic attrition’, 2: ‘condensation’, 3: ‘paradigmaticization’, 4: ‘coalescence’, and 5: ‘obligatorification’) are highly distinctive. In all four items there is a remarkable difference between less grammaticalized and more grammaticalized usages. The latter ones show a loss of intentional or ingressive semantics of the central verb, they require obligatory complementation by infinitives (auxiliarization), they display close intraparadigmatic (inferential evidential) distinctions, they allow only coherent infinitive constructions, and they have acquired a certain degree of obligatorification. Thus, although the four items differ in the degree of grammaticalization among each other (werden being the most progressed, versprechen the most retarded), they form a relatively highly grammaticalized system of evidential markers in PDG. By measuring the data of PDG against typologically relevant distinctions and undertaking thorough lexeme specific semantic investigations, it has been possible to produce succinct derivations of semantic-functional paraphrases of each item, elucidating the internal, i.e. intraparadigmatic distinctions. Most importantly, we were able to show that the evidential distinctions in German cluster in the indirect evidential domain and form a highly differentiated system of internal distinctions, which may be formulated as a system of four hierarchical levels incorporating binary distinctions. As this is one of the most important findings of this study, the table summarizing these distinctions is repeated here (see Section 7.2 for a detailed account):

326 Summary and outlook Table 1: The conceptual structure of the grammatical paradigm of inferential evidentiality in PDG First level of inferential distinctions: features: [+speaker’s inference] paraphrase: P, because Q [Origo has information pointing towards P] = P, because Q [Origo knows Q, and Origo knows that Q entails P] typological label: (conceptual) inferential represented by: werden & infinitive Second level of inferential distinctions: features: [+speaker’s inference] [+experienced information as a basis for inference] paraphrase: P, because Q [Origo knows Q, and Origo knows that Q entails P] + P, because Q [Origo can observe some signs/traces of P] typological label: perceptual inferential represented by: scheinen & zu-infinitive Third level of inferential distinctions: features: [+speaker’s inference] [+experienced information as a basis for inference] [+presently accessible information as a basis for inference] paraphrase: P, because Q [Origo knows Q, and Origo knows that Q entails P] + P, because Q [Origo can observe some signs/traces of P] + P, because Q [Origo has direct access to observed sign/traces of P] typological label: perceptual inferential represented by: drohen/versprechen & zu-infinitive Fourth level of inferential distinctions: as level three plus features: [non-desired] [desired] drohen & zu-infinitive versprechen & zu-infinitive

It has been shown that the first three levels form genuinely evidential oppositions, while the fourth level with the features “non-desired” versus

Summary and outlook 327

“desired” incorporates a type of meaning that is not specific to the domain of evidentiality. Instead, this meaning belongs to a lexically important domain which may be roughly labeled “emotive-affective”/“evaluative”. Its application as a distinctive criterion in the evidential paradigm is triggered by the reinterpretation of the semantic content of the lexical verbs drohen and versprechen (see below). These results, i.e. the preponderance of the indirect branch of evidential distinctions in the German system, and the integration of “non-evidential” “emotional-affective” features together with the decline of the distinctive relevance of aspectual values like “ingressive”, are unexpected and shed new light on possible structural layouts of the evidential domain in general as well as on the particularities and language specific tendencies of German. The demonstrated importance of the indirect branch as opposed to the branch of direct evidentiality, which – in many typological studies – is emphasized greatly, in our view, calls for enhanced attention to the indirect section in future studies in the field. Furthermore, if this tendency is not but an idiosyncrasy of the German language (which can be characterized as a language with a very strong tendency towards a highly differentiated system of markers of speakerbased knowledge processing in the evidential and epistemic domains), this observation may lead the way to a better understanding of the conceptual relations between evidentiality and epistemic modality, and the developmental paths of linguistic items expressing these values. Finally, by looking at the four verbal constructions in PDG together, we were able to motivate the classification of the verb werden as a verb which is primarily an evidential, but nevertheless has ample ramifications into other categories. It is the evidential domain where werden displays the most obvious paradigmatic oppositions, and it is its abstract indirect evidential meaning which is the most natural starting point for reinterpretations in the direction of other semantic values (e.g. future meaning, epistemic modality meaning, habitual meaning etc.). As far as the diachronic development of the four constructions into evidential markers is concerned, we have been able to show that the diachronic development of the four constructions since OHG can be described as a process of grammaticalization, and that the grammaticalization paths of these constructions can be traced back with reference to known paths for the development of evidential markers. As the details of the diachronic study are summarized in Chapter 9, it may suffice here to resume only some general results. While werden & infinitive – as the oldest infinitive construction in the set – and scheinen &

328 Summary and outlook zu-infinitive had completed their development into evidential markers as early as about 1800, the verbs drohen and versprechen were much later in developing the relevant constructions and evidential meanings, so that important developmental stages for these verbs can be observed from the 18th century until today. Most importantly, our diachronic analysis has shown that, although the four items stem from very different lexical sources and display very idiosyncratic individual histories (see, e.g., the complexities of versprechen), they converge – step by step – into a common grammaticalization path thereby shedding their particularities and at the same time gaining common features relevant for grammatical markers of indirect (inferential) evidential distinctions. Furthermore, we have found that there are two basic source constructions from which these inferential evidential meanings arise diachronically in German, one representing a successive relationship between two events (in the originally ingressive werden as well as in the verbs drohen and versprechen through their original speech act semantics) and the other coding a simultaneous relationship between two events (scheinen, which started out as a verb denoting visual appearance or effect). It could be shown that the general parameters of these diachronic processes are the ones identified in many other grammaticalization studies before under the labels of metaphorization (and/or metonymization), scheme retention, and the three stages of semantic-functional change starting from (self-contained) referential function to text-integrative/connective function, and proceeding further to indexical-grammatical function. A further essential finding is the applicability of the model of successive contexts and construction types, which had been developed and applied to other instances of grammaticalization before by the authors of this book. While earlier versions of the context model assumed three chronologically ordered stages in the diachronic rise of grammatical functions, the latest version contains four stages each of which is associated with a particular type of context. These four stages are associated with specific processes and features and are assumed to describe the general developmental stages of grammaticalization.Table 2 summarizes the central aspects of each of the four stages (see Section 5.3 for further details). As the addition of the fourth and final stage is an important novelty of the model and is claimed to be generalizable for other instances of grammaticalization, it is sensible to repeat some of the major arguments that led to the revision here. It is the fourth stage which ultimately defines a given developmental process as grammaticalization: the paradigmatic

Summary and outlook 329

(re)integration or ‘paradigmaticization’ of a new grammaticalized element is essential for the completion of the full process of grammaticalization. Furthermore, it is the criterion to distinguish grammaticalization from other types of language change as paradigmaticity is the very property that distinguishes grammatical items from lexical ones, which may conjoin into word fields (i.e. loosely organized, open classes) but not into paradigms. Table 2. Context types in grammaticalization Context

Meaning/Function

Construction types

Stage I preconditions of grammaticalization Stage II triggering of grammaticalization

untypical contexts

conversational implicature

no particular construction type

critical context

multiple opacity

“extragrammatical idioms”

Stage III reorganization & differentiation

isolating contexts

polysemous / heterosemous items

“formal or lexically open idioms”

Stage IV paradigmatic integration

paradigmatic context

paradigmatic oppositions (choices) with reduced relational meanings, i.e. grammatical meanings

paradigmatic choices within an abstract constructional schema

Thus, beyond morpho-syntactic and semantic changes, the most important factor of grammaticalization is the increase in paradigmaticity. In Stage IV the new grammatical meaning becomes subject to the meaning of the whole paradigm and at the same time dependent on the meanings of other members of the paradigm. To render this fact, we propose that a paradigm may be described as an abstract constructional schema consisting of the common semantic basis of a paradigm (e.g. evidential marker) and the abstract structural properties of its participating constructions (e.g. verbal inflection or periphrasis), thus representing the type of non-reducible association of form and meaning crucial for classifying a given linguistic entity as a construction. If, like in the case of the German evidential markers, a new category arises in a language, the members of the growing category build a paradigm among themselves, whereby the older members are the blueprint and give the pattern for analogous developments etc. The para-

330 Summary and outlook digmatic integration of the evidential constructions in present-day German is given in the following diagram, which is taken from Section 5.3 and slightly modified: Table 3. Paradigmatic oppositions of inferential evidentials in PDG CAT. LABEL: evidential marker F. CHAR.: modifies the main verb CAT. MEANING: information source for the proposition

formal realizations

unmarked form

auxiliary & infinitive

semantic oppositions within the paradigm (distinctive values of individual members)

unmarked (with respect to) evidentiality

inferential evidentiality (= speaker’s inference) + experienced information as a basis for inference

+ presently accessible information as a basis for inference

scheinen

drohen versprechen

indicative (zeromarking) werden

Thus, the fourth stage of grammaticalization – the paradigmatic integration – refers to a whole set of constructions, i.e. the constructional network building the paradigm, on the one hand, and every single construction within it, on the other. On the whole, the grammaticalization scenario laid out here includes the perspective of tracing back a change through time, as well as the process of segmenting a continuum into discernible stages. As has been shown, both aspects – the dynamic one, i.e. factors steering the process of change itself, as well as the more static, classificatory and comparative one – must be taken into account and must be combined in order to model a complete diachronic grammaticalization scenario in an adequate way. The applica-

Summary and outlook 331

tion of this complex model of grammaticalization to the four evidential constructions has resulted in a succinct and comprehensive description of the history of these forms and the evolution of an evidential system in German. Above all, it has shown the validity and applicability of the model as such. As to more general findings on evidentiality, modality and evidential distinctions, it may be noted that it has been possible to argue for some partial revisions of the generally accepted assumptions, and to provide some additions to typological evidential distinctions. The three major points of general interest are: 1. the indexicality of evidential (and modal) markers, 2. the principle conceptual distinction and factual overlap between the evidential and epistemic-modal domain, and 3. the special role of quotatives and reportives. These three aspects may be briefly resumed here. First, the fact that evidentials, like epistemics, are deictic grammatical markers, is a most important outcome of this study. The deictic relational character of the evidential category is the basis of many further theoretical distinctions and elaborations, and the grammaticalization of evidential markers in German can be described as a change in the semiotic status of the lexical non-deictic source items to grammatical deictic markers. The relational diagram of evidential markers illustrating its deictic nature is repeated here (see Section 2.2): primary event

secondary event functioning as evidence

Origo

evidential dimension

Figure 1. Evidentiality as a deictic category

In contrast to other deictic dimensions, the evidential dimension includes a secondary event (the shaded figure in the middle of the large arrow), which is logically situated between the Origo and the primary event and functions as a third point in the vectored relation. Furthermore, there is a “reflexive”

332 Summary and outlook relation pointing from the evidence to the Origo and providing the Origo with the “pair of glasses” to perceive the primary event. The deictic component can be related to the generally accepted distinction between direct and indirect evidentiality in the following way. In the case of direct evidentiality, all relevant points of the relation are treated as co-present in the same deictic field: the secondary event, functioning as evidence for the primary event, and the primary event are co-present with the Origo, i.e. the marker encodes an “origo-inclusive” value. In the indirect branch, the primary event is not co-present with the Origo (and the secondary event): the marker expresses an “origo-exclusive” value. We could constitute the following general classification of evidential values, which takes into account their deictic foundation (see Table 4 below). The second general issue, the relation between evidentiality and epistemic modality, had to be discussed at some length as this issue has been hotly debated in the literature for quite some time now. We would like to offer the solution represented in Table 4. Table 4. Overview of general classification of evidential value evidential values Level I

direct

indirect

deictic value

‘origo-inclusive’

‘origo-exclusive’

P is co-present with Q

P is non co-present with Q

visual, auditory, etc. P and Q [Origo can see/hear etc. P]

inferential

description Level II paraphrase

P, because Q [Origo has information pointing towards P]

Level III

perceptual

conceptual

paraphrase

P, because Q [Origo can observe some signs/traces of P]

P, because Q [Origo knows Q, and Origo knows that Q entails P]

mediated information P, because Q [Origo was told that P]

We argue that both categories are in principle independent of each other, though closely related. This means that there is a clear conceptual distinction between evidential markers and epistemic modality markers (i.e. modal verbs) and that their distinctive features can be clearly delineated. On the other hand, evidentiality overlaps with epistemic modality due to their common denominator. The potential areas of overlap are particularly de-

Summary and outlook 333

termined by general cognitive principles. The similarities of both domains are due to a common abstract relational template, which, too, accounts for their functions as deictic categories. Variations in the “filling” of the positions of the template account for semantic and functional differences of the forms and categories in question, whereby the specification of the source features of a relational structure play a decisive part. It is only on the basis of a sharp distinction that overlaps and gradation between the two domains can be properly integrated into a linguistic description. Thus, we postulate two distinct paradigms for PDG, one modalepistemic, one evidential. Both systems are semantically coherent, formally homogeneous, derived from their own specific grammaticalization paths respectively, and compatible with typological classifications. The epistemic system is closely integrated into the paradigm of verbal moods, i.e. it encodes deictic values concerning speaker-based factuality judgments. The evidential system, on the other hand, encodes speaker-based information source and is in principle independent of the system of factuality markers. The third important general issue is the relation between quotatives and hearsay evidentials (= mediated evidentials). Although this complicated sub-domain of evidentiality has not been in the focus of this study, it was necessary to make some basic decisions in order to be precise about the extent of the evidential (and modal epistemic) distinctions in the grammaticalized verbal periphrases of German. In short, we suggest that there is a super-ordinate distinction between linguistic means expressing any type of “origo shift” – these are the quotatives in the genuine sense of the term – and evidential meanings, which do not contain an origo-shift but instead indicate that the speaker has access to mediated evidence. While the former ones are NOT evidentials (though quotative expressions may often serve as diachronic sources for evidentials), the latter are. In German, some uses of the modals sollen and wollen probably qualify as mediated evidentials, as is exemplified by sollen in the following sentence, where sollen indicates that the speaker learned about the described event from someone else. (1)

Was ich über 2003er aus der Bourgogne lese, weckt wenig Erwartung. Es soll aber, und das kann man sich gut vorstellen, ein gutes Jahr für Beaujolais gewesen sein. (DWDS, Kernkorpus) ‘What I have read about the 2003 Bourgogne vintages raises few expectations. However, they say that it has been a good year for Beaujolais, and it is easy to imagine that to be the case.’

334 Summary and outlook However, as this was not the focus of this book, more research is necessary to investigate this issue, which – as the reader is sure to have noticed – touches the question of the connection between evidential constructions and modal verbs. Now, as an outlook and topics for further research, we would like to take up the issue of the further development of the four levels of distinctive markers in the evidential constructions (see Table 1). As we have shown, the three levels of distinctive features draw on genuinely evidential semantics, while the fourth level draws on non-evidential meanings, namely from the domain of emotional and affective distinctions. Both layers can be put down to the developmental potential of the original source items: Both distinctions qualify as potentially relevant distinctions in grammatical systems. For the third level the typological literature on evidentiality quoted in this study (e.g. Aikhenvald 2004) gives ample evidence. The emotive-affective distinction of “desired” vs. “non-desired”, on the other hand, has received less attention so far as it is not a prominent distinction in grammatical systems. However, in many languages it does figure as a relevant distinction in lexical subfields which are close to grammatical distinctions (see e.g. the Latin “modal” verbs velle ‘want’ – nolle ‘not want’ – malle ‘rather want’). Furthermore, an evaluative component distinguishing between a negative and a positive value is often present in linguistic comparison, i.e. in the expression and differentiation of a number of speech act types, and those modal meanings that are usually classified as deontic or volitional. Therefore, although this distinction at first sight seems to be alien to a grammatical paradigm, there are chances that the fourth layer may persist in the further development. As to the evidential distinctions of the third level, they, too, might persist and be further developed, but there is also the possibility that this level of distinction will get lost as the paradigm is subject to tighter integration. There are cases of usage where the distinction between level two and three blurs, as the distinctive feature of “presently accessible information as a basis for inference”, which is relevant for separating drohen and versprechen (which show this feature) from scheinen (which does not show this feature any longer). As soon as this feature gets lost in drohen and versprechen, these two levels will be collapsed and the former level four will gain weight as its distinction becomes decisive in order to differential between the three items. As to the history of the level three feature (“presently accessible information as a basis for inference”), it is sensible to assume that this is a reflex of the still persisting residual ingressive component of drohen and ver-

Summary and outlook 335

sprechen, which has often been noted in literature and which is responsible for the restriction on drohen and versprechen to take a perfect infinitive complement as in (2): (2)

*Sie drohen/versprechen am Morgen abgereist zu sein. ‘They threaten/promise to have left in the morning.’

It may be assumed that – if and as soon this component gets bleached (which may or may not happen) – the evidential distinction of level three will become weak and get lost. However, as these are speculative reflections on possible further developments of the evidential paradigm, we leave this issue here and come to a final conclusion. During the last couple of years interest in research into evidentiality in language has taken an enormous upsurge. In this situation, we hope that our book which is the first comprehensive study on evidentiality and its grammaticalization in German will fulfil its task as a kind of landmark for future research into this field. As far as the German language is concerned, we claim to have shown that it is time to include evidentiality as a grammatical category into the linguistic descriptions of present-day German in reference grammars and linguistic text books.

Notes Notes Chapter 1 1.

We would like to thank an anonymous reviewer whose insightful comments were of great help in making the book as good as we could. This book is the final outcome of a research project “Evidentialitätsmarker im Deutschen”, which was funded by the Fritz Thyssen Stiftung between 2006 and 2009 (Az. 10.06.1.081). We would like to express our sincere gratitude to the Fritz Thyssen Stiftung. Thanks are also due to the Belgian Science (Interuniversity Attraction Poles programme project GRAMIS P6/44) for partial funding of work on this topic. 2. The description of scheinen as a visual evidential is preliminary here and will be modified in the course of the book. 3. Anderson (1986: 274): “Evidentials are not themselves the main predication of the clause, but are rather a specification added to a factual claim about something else.” 4. The examples are taken from the IDS-corpora and from the ZEIT-sub-corpus of the DWDS-corpora. Since every example can be traced back easily, we only indicated the sub-corpora they were taken from: W = “Archiv der geschriebenen Korpora”, DSAv = “Deutsches Spracharchiv”. 5. The fact that ‘indexicality’ is a defining feature of grammatical elements was first mentioned by Jespersen (1921), and Jakobson ([1957] 1971), who dubbed indexical grammatical signs “shifters”. 6. Bühler [1934] 1990 defines the origo as the source or zero point of the “coordinate system of ‘subjective orientation’, in which all partners in communication are and remain caught up” ([1934] 1990: 118). 7. See also Leiss 1992 and Diewald 1997. 8. For the distinction between characterisation and deictic pointing see Bühler [1934] 1990. 9. An example of the temporal function of the unmarked category “present” is: (Wo kann ich sie jetzt antreffen?) – Sie ist in der Bibliothek. ‘Where can I find her, right now? – She is in the library.’ 10. Engel (1996: 868) has an entry for the lemma eventuell as an example of what he calls “verifikative Angabe”, which belongs to modality; the function of these adjuncts is described as “[sie] modifizieren den Realitätsgrad eines Sachverhalts”; among the items grouped together here are also adverbs with evidential meaning like angeblich, anscheinend, offenbar, offensichtlich (Engel 1996: 230). 11. By drohen2 and versprechen2 Reis refers to what we call the more grammaticalized usages of these verbs.

Notes 337 12. Cf. also De Haan (1999: 210-219) for a discussion on this issue. 13. The Leipzig Glossing rules are available under the following address: http://www.eva.mpg.de/lingua/resources/glossing-rules.php (10.01.2009) 14. The corpus and further information about it are available under the following address: http://titus.fkidg1.uni-frankfurt.de (22.09.2008) 15. Please consult the following page for more information: http://www.hsaugsburg.de/~harsch/augustana.html (29.01.2009) 16. Detailed information about the corpus is given at: http://mhdbdb.sbg.ac.at (15.11.2008) 17. For technical details and principles of the architecture and procedures of KALI see Lehmberg, Diewald, and Smirnova 2007. 18. For further information concerning the texts, e.g., year, edition, page in the edited version, etc., please consult the corpus webpage. The corpus is available online under the following addresses: http://www.kali.uni-hannover.de/ and http://www.gabrielediewald.de/index.php/kali-korpus (15.03.2009) 19. http://www.mhdwb-online.de (14.02.2009) 20. http://virt052.zim.uni-duisburg-essen.de/Fnhd/ (25.11.2008) 21. http://www.hs-augsburg.de/~harsch/augustana.html (21.11.2008) 22. http://www.staff.uni-marburg.de/~Egloning/etexte.htm (13.10.2008) 23. The project is an initiative of the online version of the German illustrated magazine Spiegel-Online. Since 1994, numerous German books from different periods of time have been made accessible via the World Wide Web. Meanwhile, the project has become the largest collection of classical texts in the German language. The online library with more than 4,000 books is available at: http://gutenberg.spiegel.de/ (09.01.2009) 24. More information is available at: www.ids-mannheim.de (19.07.2008) 25. It should be mentioned here that the corpora from the 18th and 19th centuries contain predominantly narrative texts. This means that this text type clearly dominates over the others. Nevertheless, we analyzed the data available, since, to our knowledge, there are no other corpora covering data of other text types from these periods of German. 26. The corpus is available at http://www.dwds.de (04.02.2008)

Notes Chapter 2 1. 2.

For a good account of terminological problems in the evidential domain see Jacobson 1986. See also the following quote by Anderson, who – as early as in 1986 – made the importance of the distinction perfectly clear: “It is important to distinguish true evidential categories from other forms which seem evidential, but are not. The noun form of the term “evidentials” or “an evidential” does not simply in-

338 Notes

3.

4.

5.

6.

7.

8.

9.

clude anything one might consider to have an evidential function, that is, to express evidence for something else. Rather, evidentials are a special grammatical phenomenon” (Anderson 1986: 274). Cf. also Giacalone Ramat and Topadze (2007: 7-8), who, referring to Aikhenvald 2004, also emphasize this distinction. A word of caution here concerning the term “periphrasis”: verbal periphrases, or periphrastic verbal forms are morphologically complex expressions of grammatical categories in verbs, e.g. the German perfect etc. Another expression for this is analytic verb forms. This is not to be confused with what we call “evidential periphrasis”, the latter term applying to any type of nongrammaticalized, non-stereotypical construction having evidential meaning. The term “source” in diachronic contexts is used to refer to the lexical predecessors of grammaticalizing items; it is not to be confused with “source” as a relational concept used in the description of the semantic template of evidentials and modals (see below). The English equivalents seem to have the same type of ambiguity (cf. Whitt 2008). For evidential and ambiguous uses of aussehen cf. Whitt (2008: 114), for klingen and sich anhören also cf. Whitt (2008: 158): “Klingen is German’s key carrier of evidential meaning among object-oriented verbs of auditory perception, in contrast to the much lesser used (sich) anhören.” Aikhenvald (2004: 10): “These expressions are not obligatory and do not constitute a grammatical category; consequently, they are only tangential to the present discussion. Saying that English has ‘evidentiality’ (cf. Fox 2001) is misleading: this implies a confusion between what is grammaticalized and what is lexical in a language. Lexical expressions may, of course, provide historical sources for evidential systems (see Chapter 9). Lexical ways of indicating source of information may reinforce grammatical evidentials”. Aikhenvald (2004: 6): “Linguistic evidentiality is a grammatical system (and often one morphological paradigm). In languages with grammatical evidentiality, marking how one knows something is a must. Leaving this out results in a grammatically awkward ‘incomplete’ sentence […].Those who cannot get their evidential right are in trouble: they are considered linguistically incompetent and generally not worth talking to. Only in some systems can an evidential be omitted if recoverable from the context. This is very much unlike languages where saying explicitly how you know things is a matter of choice for the speaker”. In other places, Aikhenvald clearly sees the difference, cf. the following quote: “Saying that English parentheticals are ‘evidentials’ is akin to saying that time words like ‘yesterday’ or ‘today’ are tense markers” (Aikhenvald 2004: 10). For further descriptions of the structure of an evidential setting in terms of the relevant deictic relations (i.e. the relational structure in a deictic process) see also Chapter 3.

Notes 339 10. If necessary it is possible to draw a further distinction between “speech sounds” and sounds in general (“acoustic”). 11. We should bear in mind that the origo and his/her information source may be disconnected from each other (and thus not co-present) by means of e.g. tense markers. So, if the statement is made about an event in the past, the speaker normally expresses the fact that s/he has directly witnessed the situation. Though s/he is now detached from the event – as well as from his/her own perceptual input of it –, this fact is indicated not by using an(other) evidential form, but is achieved only by using particular tense markers. In such cases, it becomes clear that the information source is still conceived of as co-present with the described event. That means that the coincidence of the speaker’s information source (e.g. speaker’s perception of an event) and the perceived event itself is most relevant for attributing a particular evidential value to a proposition. 12. See also Giacalone Ramat and Topadze (2007: 38). 13. There are evidential systems that do not display any further specifications within the subgroup of indirect evidentiality. That means that inference (reasoning, assumption, etc.) and mediated information are expressed by one and the same form (cf. the two term systems A1 described in Aikhenvald 2003, 2004). This fact provides further evidence for treating the distinction direct vs. indirect evidence as a basic opposition within the domain of evidentiality. 14. Cf. research into intertextuality in cultural studies, role play in psychological programs, legal issues around plagiarism and copyright etc.

Notes Chapter 3 1.

As explicated in Diewald 1999, the internal distinctive values of this grammatical category can be described with reference to Bühler’s concept of deixis. Bühler points out that the relational structure of the deictic process can be applied to different deictic fields, i.e., the origo can be abstracted from the present speaker and transferred to some other point. These modifications create different “modes of deixis”, i.e. deixis in the actual deictic field of the utterance (“demonstratio ad oculos”, Bühler [1934] 1990: 94), “imaginationoriented deixis”, and thirdly “the anaphoric use of deictic words”. In Diewald 1999, 2001 it has been shown that theses modes of deixis play a central role in the internal structure of the grammatical category of mood, whereby the indicative and the modals können and müssen represent deixis proper, the subjunctive I and the modals sollen and wollen “imagination-oriented deixis”, whereby the two modals qualify as “quotatives”, and the subjunctive II and the modals mag and dürfte the anaphoric mode, as they make use of an additional phoric relation. Thus, the application of the concept of modes of deixis reveals

340 Notes

2. 3.

4.

5.

6.

7. 8.

the high degree of functional specialization in the category of mood, and makes it possible to describe the distinctive value of each modal verb as it has become integrated into this paradigm as a factuality marker. As these distinctions do not affect the classification of evidentials, they will not be discussed further here. It should pass without notice that the parenthesis in (5) is of course not an evidential (not a grammatical marker) but an evidential expression. The German modals show two distinct uses: a typically narrow scope reading in which the modals are used as lexical verbs to predicate a complex state on the subject (1) and a wide scope reading comprising the objective epistemic and the so-called “subjective epistemic” use (2) (cf. Lyons 1977, Traugott 1989, Diewald 1993, 1997); only the latter one is of importance here. As the adverbials used in the paraphrases are not grammatical but lexical means, it should be noted that they are not equivalent to the grammaticalized modals or the verbal moods as far as their semiotic status is concerned (cf. Diewald 1991, 1999). Diewald 1999 shows in detail how the epistemic modal source is derived from the non-epistemic modal source, and how the resultative notion is preserved and reinterpreted as a speaker-based “judgment”. In the full-verb use with a nominal complement (e.g. Er kann Englisch ‘He can speak English’), the modal is a stative verb with a primary experiencer subject because there is only one relational structure. As soon as there is an infinitive complement, there are two verbal relations and thus a complex and condensed relational structure with a secondary experiencer occupying the subject position (e.g. Er kann/muss/darf/... Englisch verwenden ‘He can/must/may/...use English’). The meaning of the modals in this use retains aspectual and passivelike components that reflect their origin in preterito-presents. This meaning is: the state of having been given a directive. The experiencer role can be fully analyzed as follows: the goal feature derives from the fact that the subject is seen as the recipient of an issued directive; the source feature derives from the intended action, which the subject is meant to carry out and which is represented in the infinitive construction. All relational positions of the underlying directive speech act are semantically present and recoverable from the modal meaning, but the two relations (the issuing of the directive and the carrying out of the directed activity) have been condensed into a complex modal structure that is held together in the experiencer subject. See also Plungian (2001: 354): “[...] not all evidential markers are modal in that they do not all necessarily imply an epistemic judgment.” An instance of this type of implicatures may be the case of Italian, where, according to Giacalone Ramat and Topadze 2007, epistemic modals are used as (inferential) evidentials.

Notes 341

Notes Chapter 4 1. 2. 3.

4. 5.

6.

7.

See also Kuryłowicz (1965: 52), Traugott and König (1991: 189), Croft (2000: 156), Heine (2003: 584). See also Bybee 1985, Haspelmath 1999, Heine (2003: 579ff.), Himmelmann (2004: 31). See also: “A sign is grammaticalized to the extent that it is devoid of concrete lexical meaning and takes part in obligatory grammatical rules.” (Lehmann 1995 [1982]: viii) This is the essence of many definitions of grammar, see for example Jakobson (1959: 489), Mel’þuk (1976: 84), Hagège (1987: 250), Radtke (1998: 10). See also Radtke (1998: 10), who with reference to the verbal categories of German states: „Zwar hat der Sprecher keinerlei Freiheit bezüglich der Frage, ob eine Verbalkategorie gewählt werden soll oder nicht. Er muß sich hier für jeweils eine Verbalkategorie entscheiden, und zwar für genau eine. Bezüglich der Frage, welche Verbalkategorie er dabei realisieren möchte, besteht jedoch Wahlfreiheit.“ [The speaker may not have any freedom of choice concerning the question whether a verbal category has to be chosen or not. But in each case, he must choose a verbal category, and what is more, he must choose one and only one. Concerning the question, which verbal category he wants to realize, however, there is freedom of choice.] In the approach taken here the notion of communicative obligatoriness is not restricted to verbal categories but is used to cover any grammatical category displaying the combination of obligatory realization and freedom of choice between several paradigmatic options according to intention. There has been a lively discussion of the question of the degree of grammaticalization of the dative-passive in the past decades, which however cannot be taken up here (for an overview cf. e.g. Diewald 1997, Leirbukt 1997, Askedal 2005). The fact that passives are not possible with all verbs is disregarded here. Restrictions of this type are not sufficient for deciding about the status as a grammatical category, as the applicability to all relevant category members (host class extension in the diction of Himmelmann (2005: 89)), too, is a matter of degree. It affects many grammatical categories (there are, for example, nouns without plurals, inclinable pronouns etc.). This is quite clearly stated in Himmelmann (2005: 89): “Host classes of individual grams differ significantly in size. Plural markers may be restricted to nouns denoting animate beings, passive markers to transitive verbs, applicatives to motion and transfer verbs, etc. Only certain types of grams are associated with a host class which is coextensive with a major lexical category. Well-known examples are tenseaspect-mood auxiliaries, (clitic) articles, (some) simple adpositions (or clitic case markers) and negators. These gram types can be used to define highly

342 Notes

8.

9.

10.

11.

12.

13.

general syntactic slots where practically all the members of a given major lexical category may occur.” It should be noted that this kind of rule can be set up for individual category members (as illustrated for the werden-passive) as well as for groups of marked values of a paradigm (as illustrated for the evidential constructions). The latter option is chosen here, because the exact distinctive values of each of the four evidential constructions have yet to be explored in later chapters. For the classification of the construction with würde see e.g. Zifonun et al. 1997, Eisenberg 1999, 2004, Fabricius-Hansen 1998, 2000, Thieroff 1992, Smirnova 2006. The attribute “structural” was added to the label in Lehmann (2002: 110), as the notion of “scope”, which in the first papers stood alone, allowed misinterpretations in the sense of “semantic scope”, (like in negation or adverbials) which is not what is meant here (e.g. Traugott 1997 criticized this parameter on the basis of this misinterpretation). In order to avoid misunderstandings, we prefer the term “syntagmatic weight” to the term “scope” here. The first two parameters can be equated to what Hopper (1991: 22) calls “decategorialisation”, which is described as a loss of the “morphological markers and syntactic privileges characteristic of the full categories Noun and Verb”. In this process verbs lose their capability to denote events, i.e. their “predicativity” (Hopper 1991: 30f., Lehmann 1985: 309), or “referentiality” (Dasher 1995). There is a number of different sets of parameters that have been proposed in grammaticalization theory so far beyond Lehmann's (see e.g. Heine, Claudi and Hünnemeyer 1991; Hopper and Traugott 1993, Bybee, Perkins and Pagliuca 1994); a brief overview of parameters existing within the grammaticalization theory is given in Askedal 2005). They differ with regard to their degree of specificity, i.e. whether they concern changes of a particular item or changes in the whole paradigms and systems, with regard to the favored – semantic or formal – factors, etc. Moreover, they differ from each other as some approaches consider particular parameters to be definitional while declaring others as deriving from the basic ones: for instance, the parameter of obligatoriness, or obligatorification, is seen by Lehmann [1982] 1995 as one of the definitional criteria of the process of grammaticalization, whereas in Heine and Kuteva 2002 the same parameter is considered to be “only” a by-product of another – as they claim more important – property of grammaticalization process, namely decategorialization. For less developed, incipient instances the parameters are usefully complemented by the criteria set up by Hopper (1991: 22ff.), which he calls “the five principles of grammaticization”; these are: layering, divergence, specialization, persistence, de-categorialization. As has been already convincingly demonstrated by Askedal (1997), Heine and Miyashita (2004: 20-27) and Diewald (2004: 250-253), versprechen within infinitive constructions – similarly to drohen – can be seen as a canonical case

Notes 343 of grammaticalization (and auxiliarization). In this section, we combine the results of those former studies with our own corpus data, which will be dealt with later.

Notes Chapter 5 1.

2. 3.

4.

5.

6. 7.

Cf. Heine, Claudi and Hünnemeyer (1991: 48), Matlock 1989, for category specific regularities concerning the preferred source domains for grammaticalization in the area of modality and evidentiality also see van der Auwera and Plungian 1998, Anderson 1986, Willett 1988. For a discussion of the notions of indexicality and subjectivity see Diewald (1999b: 14-16). Coates herself uses the term “ambiguity” instead of “polysemy” as opposed to conversational implicature. As already explained, we do not follow this practice. Bybee, Perkins and Pagliuca (1994: 11), for example, state that “… it is the entire construction, and not simply the lexical meaning of the stem, which is the precursor, and hence the source, of the grammatical meaning.” In a similar vein, Bisang (1998: 20) concludes that “constructions provide the framework within which combinations of syntactic units and semantic components can be analysed in a new way which may lead to language change if it is propagated from a linguistic individual to a language community.” Finally, under the heading “Constructions in Grammaticalization”, Traugott 2003 explicitly takes up this problem and suggests the following definition of grammaticalization which includes the notion of “construction”. Grammaticalization is seen as “the process whereby lexical material in highly constrained pragmatic and morphosyntactic contexts is assigned grammatical function, and, once grammatical, is assigned increasingly grammatical, operator-like function” (Traugott 2003: 645). Cf. Goldberg 1995, 2006, Diewald 2006, Bergs and Diewald 2008, Stefanowitsch and Fischer 2008. As argued for in Diewald 2006, there are at least four areas in which the assumptions of construction grammar converge with basic concepts of studies in language change in general and grammaticalization in particular. These are 1. the definition of the basic unit, 2. the range of phenomena covered, 3. the flexibility and openness of the formalism, 4. the dynamic potential of the concept. Of course, contexts like this one have persisted through all the history of German and still can be found in PDG. Examples proving that the critical context is attested with each of the six modals can be found in Diewald (1999: 361-431); cf. also Deeg 1948.

344 Notes 8.

9.

Iconicity is understood here in the broad sense of the term as a semiotic mechanism representing the parallelism between different layers of structure (see e.g. Croft 2001: 108). In terms of the grammaticalization parameters developed by Lehmann 1985, this is an instance of the loss of “morphological integrity”, i.e. the loss of inflectional distinctions pertaining to the main word class the item originally belonged to.

Notes Chapter 6 1. 2.

3.

At this point, we provisionally note that the construction is polysemous. In the following, this view will be revised. The exact description of the aspectual semantics of werden has been a matter of intense debate in recent years. So, for example, Leiss (1992: 167-170) attributes two distinct semantics to werden: the additive and the nonadditive one. Fritz (2000: 37) distinguishes between the “slow” and the “punctual” werden. Amrhein (1996: 121) considers werden as a prototypical mutative verb which is indifferent in regard to the punctuality component. In a similar vein, Kotin (2003: 31) classifies it as a mutative verb: “Die Bedeutung der Zustandsänderung ist stets aktional markiert und kann als mutativ eingeordnet werden, wobei das Verb ‚werden‘ den Endpunkt einer Phase bezeichnet“ [‚The meaning of change of state is always aspectually marked and can be classified as mutative, whereby the verb ‘werden’ expresses the endpoint of a phase’]. Assuming the mutative component to be the core aspectual semantics of werden, Kotin 2003 explains its two different readings – the static one meaning ‘to happen, to occur’ and the dynamic one meaning ‘to emerge, to come into being’ – as two possible interpretations of this core meaning. Smirnova (2006: 114-120) adopts Kotin’s view because the label mutative covers best the aspectual semantics of werden and allows us to explain its integration into the different grammatical domains of the German verbal system through grammaticalization. As the question of the exact aspectual identification of werden is not our central concern here, we will use the most established term “ingressive” for the aspectual characterization of werden. It has been pointed out by Heine and Miyashita (2004: 22) that grammatical versprechen is also found in contexts with animate subjects which cannot be thought of as referents controlling or intending the action introduced by the infinitival complement, cf. Er versprach bald aus der Ohnmacht zu erwachen – ‘He promised to wake up soon after fainting’. Though we accept that such combinations are potentially possible (and acceptable) and that they would strengthen the claim about the high degree of grammaticalization of versprechen, we did not however find any such occurrences in our corpus data. In this

Notes 345 connection, we can still draw the generalization that combinations of grammatical versprechen and animate subject are not possible in PDG (though not fully inacceptable).

Notes Chapter 7 1.

2.

3.

4.

5.

Werden & infinitive started its grammaticalization process as an ingressive construction in MHG. The construction was analogically formed after the model of the (bi-)ginnan/ (bi-)stan(tan) & infinitive constructions, and, in the course of MHG and ENHG, competed with infinitive constructions with modal verbs (wellen ‘will’, suln ‘shall’ and muozan ‘must’) as a candidate for a future tense marker. It should be said here that werden & infinitive – being the unmarked member of the paradigm of inferential evidentiality – can also be interpreted as marking the conceptual inferential status of the proposition exclusively, i.e. it can be contrasted with the perceptual inferential value (which all other members of the paradigm share). In this case, werden & infinitive realizes the second meaning of an unmarked term whereby the absence of a feature (which is the perceptual value) is explicitly marked. This is to say that we regard the opposition conceptual vs. perceptual within the domain of inferential evidentiality as a binary opposition where the conceptual value is the unmarked one and is thus tantamount to the (neutral) inferential evidential value (see also Note 3). For this cf. Jakobson (1939 [1974]: 48) where he shows that zero-elements do have a double function in grammatical systems: they can be first “wo A and Nicht-A nicht unterschieden werden” [‘where A and non-A are not distinguished’] that is where the category in question is neutralized, not relevant, not expressed; and secondly, “wo Nicht-A bezeichnet werden soll” [‘where non-A should be expressed’], i.e. the unmarked value inside the paradigm as opposed to the marked values of it. Here, it can be noted that the classification of the world’s evidential systems which have been presented in Aikhenvald 2003, 2004 does not imply that there are clear-cut boundaries between systems; the proposed system types do not necessarily cover all existing evidential systems. Moreover Aikhenvald herself admits that “evidentiality systems may vary as to the relative markedness of terms” (2003: 7), and that “languages can have more that one evidentiality system, or more that one paradigmatic set of forms with evidentiality meaning” (2003: 8). Furthermore, languages may combine grammatical marker of evidentiality with so-called “evidentiality strategies”. Moreover, this characteristic is not restricted to inferred evidentials but applies to all evidential distinctions; cf. Aikhenvald, who states that “most frequently, evidentiality distinctions are found in past tenses” (2003: 15).

346 Notes Notes Chapter 8 It should be noted that, in PDG, the use of werden as an intransitive verb has become quite rare (see Section 6.1 for further details). 2. Several instances of the construction werden & infinitive have been attested in the texts by Notker. They will be taken up in the next section. 3. The meaning of the IE stem *wert(h)-, which is usually translated as ‘to turn, to rotate’, allows for both – ingressive and durative – interpretations. In the course of the history of German werden has favored the meaning ‘to come into being, to rise’, i.e. the ingressive interpretation of the etymon. This development can be detected in OHG, where a gradual decrease of the amount of durative uses of werden is attested. 4. This may be based on observations like the following: the future temporal meaning of werden & infinitive did not develop before the inferential one, but in parallel with it (see Section 8.1.5); the future temporal meaning of werden & infinitive has not developed further since the ENHG period (e.g. with regard to its frequency, its distribution, its meaning components, etc.). 5. It should be noted that, in PDG the situation is fairly similar: as we have mentioned in Section 6.2, scheinen & zu-infinitive has readings which are not clearly evidential. These readings can be traced back to the less grammaticalized source construction, i.e. to the copula construction of scheinen, which permitted various readings, among them the expression of “delusive appearance”, i.e. the mere appearance of something as opposed to its real being. 6. The possible impact of this factor is conceded by the authors themselves, as the following quote shows: “We have discussed the item drohen as an isolated phenomenon of German grammar; as a matter of fact, however, it is not: what we observed about drohen applies in much the same way to a range of other German items, most notably to versprechen ‘to promise’, and in a number of studies, drohen and versprechen are treated as exhibiting essentially the same structure and/ or development” (Heine and Miyashita 2008: 98). 7. Heine and Miyashita (2008: 81) note the lack of certain complement types, such as infinitive complement, finite complement clause and direct speech proposition, which are characteristic for C2. 8. Heine and Miyashita (2008: 81) point out that first instances of this innovation emerged in the course of the 16th century. 9. Heine and Miyashita (2008: 81) notice another important innovation: “The construction acquires the ability to take a source complement introduced by von, which can be interpreted either as an agent or as a locative source. This innovation appears to have taken place long after C2 had established itself, as its first attestations in the second half of the 18th century suggest.” 10. As Plank 1985 argued, it is natural for marked constructions to be structured in analogy to unmarked ones to the highest possible degree. In our case this argument applies as follows: since agentive and non-agentive drohen1.

Notes 347 constructions are formed with the same lexical verb, and the non-agentive construction type represents a marked construction, it is not surprising that it forms its infinitive according to the model of the agentive variant. 11. The subject may sometimes be a human referent, though. However it cannot be interpreted as having verbally communicated a promise: the component of speaking is totally absent here.

Notes Chapter 9 1.

It should be noted here that it is very difficult to formulate the exact structural properties of the isolating context for werden. This is due to the fact that, presently, werden is very highly grammaticalized as an auxiliary and potentially serves as an evidential marker in every context. Therefore, in contrast to other three verbs treated here, whose developmental peculiarities are still (at least partly) observable in their present uses, werden does not show a clear complementary distribution as compared to its less grammaticalized variants. An additional reason for this situation may be found in the special status werden has in the German language. As is well known, werden has served as a source for several grammatical markers in German: apart from its evidential variant, it has been grammaticalized as a passive auxiliary (werden & perfect participle) and as a preterit subjunctive auxiliary (würde & infinitive). This special diachronic behavior of werden makes the task of formulating the exact nature of particular – isolating – contexts, which clearly favor a grammatical reading of the verb, if not impossible, though very difficult.

References

Aikhenvald, Alexandra Y. 2003 Evidentiality in typological perspective. In Studies in Evdentiality. Alexandra Y. Aikhenvald and Robert M. W. Dixon (eds.), 1-31. Amsterdam/Philadelphia: John Benjamins. 2004 Evidentiality. Oxford: Oxford University Press. 2006 Evidentiality in grammar. In Encyclopedia of Language and Linguistics, 2nd ed. Keith Brown (ed.) Vol. 4 (article 0252), 320-325. Oxford: Elsevier. Aikhenvald, Aleksandra Y. and Robert M. W. Dixon (eds.) 2003 Studies in Evidentiality. Amsterdam/Philadelphia: John Benjamins. Amrhein, Jürgen 1996 Die Semantik von werden. Grammatische Polysemie und die Verbalkategorien Diathese, Aspekt und Modus. Trier: Wissenschaftlicher Verlag Trier. Anderson, Lloyd B. 1986 Evidentials, paths of change, and mental maps: typologically regular asymmetries. In Evidentiality. Wallace L. Chafe and Johanna Nichols (eds.), 273-312. Norwood/New Jersey: Ablex. Askedal, John O. 1982 Zur semantischen Analyse deutscher Auxiliarkonstruktionen unter dem Aspekt der Unterscheidung von summativer und nichtsummativer Bedeutung. Deutsch als Fremdsprache 19: 159-167. 1989 Über den Infinitiv ohne bzw. mit zu im heutigen Deutsch: Klassenbildung regierender Lexeme und Hauptzüge der Distribution (I). Deutsch als Fremdsprache 26: 2-7. 1997 Drohen und versprechen als sogenannte ‚Modalitätsverben‘ in der deutschen Gegenwartssprache. Deutsch als Fremdsprache 34: 12-19. 1998 Zur Syntax infiniter Verbalformen in den Berthold von Regensburg zugeschriebenen deutschen Predigten. Vorstufe der topologischen Kohärenz-Inkohärenz-Opposition. In Historische germanische und deutsche Syntax: Akten des Internationalen Symposiums anläßlich des 100. Geburtstages von Ingerid Dal. John O. Askedal (ed.), 231259. Frankfurt am Main: Peter Lang. 1999 Zur Frage der Auxiliarisierung einiger deutscher Verben. In Linguistik und Deutsch als Fremdsprache. Festschrift für Gerhard Helbig

References 349 zum 70. Geburtstag. Bernd Skibitzki and Barbara Wotjak (eds.), 113. Tübingen: Niemeyer. 2005 Grammatikalisierung und Persistenz im deutschen „RezipientenPassiv“ mit bekommen/kriegen/erhalten. In Grammatikalisierung im Deutschen. Torsten Leuschner, Tanja Mortelmans and Sarah de Groodt (eds.), 211-228. Berlin/New York: de Gruyter. Autenrieth, Tanja 2002 Heterosemie und Grammatikalisierung bei Modalpartikeln: eine synchrone und diachrone Studie anhand von „eben“, „halt“, „e(cher)t“, “einfach“, „schlicht“ und „glatt“. Tübigen: Niemeyer. Auwera, Johan van der and Vladimir A. Plungian 1998 Modality’s semantic map. Linguistic Typology 2 (1): 79-124. Bech, Gunnar 1949 Das semantische System der deutschen Modalverba. Copenhague: Einar Munksgaard. 1951 Studien über das deutsche Verbum Infinitum. Copenhague: Einar Munksgaard. Bergs, Alex and Gabriele Diewald (eds.) 2008 Constructions and Language Change. Berlin/New York: Mouton de Gruyter. Birkmann, Thomas 1987 Präteritopräsentia: morphologische Entwicklungen einer Sonderklasse in den altgermanischen Sprachen. Tübingen: Niemeyer. Bisang, Walter 1998 Grammaticalization and language contact, constructions and positions. In The Limits of Grammaticalization. Anna Giacalone Ramat and Paul J. Hopper (eds.), 13-58. Amsterdam/Philadelphia: John Benjamins. Brinkmann, Hennig 1971 Die deutsche Sprache: Gestalt und Leistung. Düsseldorf: Schwann. Bühler, Karl 1932 Das Ganze der Sprachtheorie, ihr Aufbau und ihre Teile. In Bericht über den XII. Kongreß der Deutschen Gesellschaft für Psychologie in Hamburg vom 12.-16. April 1931. Gustav Kafka (ed.), 95-122. Jena: Fischer. 1989 Reprint. Sprachtheorie: Die Darstellungsfunktion der Sprache. 3rd ed. Stuttgart: UTB. Original edition, Jena: Fischer, 1934. Bybee, Joan L. 1985 Morphology. Amsterdam/Philadelphia: John Benjamins. Bybee, Joan L., William Pagliuca and Revere Perkins 1991 Back to the future. In Approaches to Grammaticalization. Traugott, Elizabeth Closs and Bernd Heine (ed.), 17-58. Amsterdam/ Philadelphia: John Benjamins.

350 References Bybee, Joan L., Revere Perkins and William Pagliuca 1994 The Evolution of Grammar. Chicago: University of Chicago Press. Calbert, Joseph P. and Heinz Vater 1975 Aspekte der Modalität. Tübingen: Narr. Chafe, Wallace L. and Johanna Nichols (eds.) 1986 Evidentiality. Norwood/New Jersey: Ablex. Coates, Jennifer 1983 The semantics of the modal auxiliaries. London/Sydney: Croom Helm. Cornillie, Bert 2004 The shift from lexical to subjective readings in Spanish prometer ‘promise’ and amenazar ‘threaten’. A corpus-based account. Pragmatics 14 (1): 1-30. 2005 A paradigmatic view of Spanish amenazar ‘to threaten’ and prometer ‘to promise’. Folia Linguistica 39 (3-4): 385-415. 2007 Evidentiality and epistemic modality in Spanish (semi)auxiliaries: a cognitive-functional approach. Berlin/New York: Mouton de Gruyter. Croft, William 2000 Explaining language change: an evolutionary approach. Harlow: Longman. 2001 Radical construction grammar. Oxford: Oxford University Press. Dasher, Richard B. 1995 Grammaticalization in the System of Japanese Predicate Honorifics. Ph.D. diss., Stanford University. Deeg, Käthe 1948 Der Infinitiv Perfekt im Frühmittelhochdeutschen. Ph.D. diss., University of Munich. Dendale, Patrick and Liliane Tasmowski 2001 Introduction: evidentiality and related notions. Journal of Pragmatics 33 (3): 339-348. Dieling, Klaus 1982 Das Hilfsverb werden als Zeit- und Hypothesenfunktor. Zeitschrift für Germanistik 3: 325-331. Diewald, Gabriele 1991 Deixis und Textsorten im Deutschen. Tübingen: Niemeyer. 1993 Zur Grammatikalisierung der Modalverben im Deutschen. Zeitschrift für Sprachwissenschaft 12 (2): 218-234. 1997 Grammatikalisierung: Eine Einführung in Sein und Werden grammatischer Formen. Tübingen: Niemeyer. 1999a Die Modalverben im Deutschen. Tübingen: Niemeyer. 1999b The integration of the German modals into the paradigm of verbal mood. In Langue and Parole in Synchronic and Diachronic Perspec-

References 351 tive. Selected Proceedings of the XXXIst Annual Meeting of the Societas Linguistica Europaea, St Andrews 1998. Beedham, Christopher (ed.), 119-129. Oxford: Pergamon. 2000 Scheinen als Faktizitätsmarker. In Wortschatz und Orthographie in Geschichte und Gegenwart. Festschrift für Horst Haider Munske zum 65. Geburtstag. Mechthild Habermann, Peter O. Müller and Bernd Naumann (eds.), 333-355. Tübingen: Niemeyer. 2001 Scheinen-Probleme: Analogie, Konstruktionsmischung und die Sogwirkung aktiver Grammatikalisierungskanäle. In Modalität und Modalverben im Deutschen. Reimar Müller and Marga Reis (eds.), 87110. Hamburg: Buske. 2002 A model for relevant types of contexts in grammaticalization. In New Reflections on Grammaticalization II. Ilse Wischer and Gabriele Diewald (eds.), 103-120. Amsterdam/Philadelphia: John Benjamins. 2004 Faktizität und Evidentialität: Semantische Differenzierungen bei den Modal- und Modalitätsverben im Deutschen. In Tempus/ Temporalität und Modus/ Modalität im Sprachenvergleich. Oddleif Leirbukt (eds.), 231-258. Tübingen: Stauffenburg. 2005 Werden & Infinitiv - Versuch einer Zwischenbilanz nebst Ausblick. Deutsch als Fremdsprache 42 (1): 23-32. 2006a Context types in grammaticalization as constructions. In Constructions. Special Volume 1: Constructions all over - case studies and theoretical implications. Schönefeld, Doris (ed.). http://www.constructions-online.de/articles/specvol1/ 2006 Discourse particles and modal particles as grammatical elements. In Approaches to discourse particles. Fischer, Kerstin (ed.), 403-425. Amsterdam: Elsevier. 2008 Grammatikalisierung, Grammatik und grammatische Kategorien: Überlegungen zur Entwicklung eines grammatikalisierungsaffinen Grammatikbegriffs. In Grammatikalisierung und grammatische Kategorien. Stolz, Thomas (ed.), 1-32. Brockmeyer: Bochum. to appear On some problem areas in grammaticalization theory. In What’s New in Grammaticalization. König, Ekkehard, Elke Gehweiler and Katerina Stathi (eds.). Amsterdam/Philadelphia: John Benjamins. Diewald, Gabriele and Kerstin Fischer 1998 Zur diskursiven und modalen Funktion der Partikeln aber, auch, doch und ja in Instruktionsdialogen. Linguistica 38 (1): 75-99. Diewald, Gabriele, Timm Lehmberg and Elena Smirnova 2007 KALI - A Diachronic Corpus for the Investigation of Grammaticalization and Semantic Change. Data Structures for Linguistic Resources and Applications. Proceedings of the Biennial GLDV Conference 2007. Georg Rehm, Andreas Witt and Lothar Lemnitzer (eds.), 103-112. Tübingen: Narr.

352 References Diewald, Gabriele and Elena Smirnova in print The German evidential constructions and their origins. A corpus based analysis. In Papers presented at the International conference on tense, aspect, mood, and modality, 18-20 september 2006, University of Antwerp, Belgium. Brisard, Frank, Tanja Mortelmans and Walter de Mulder (eds.) Diewald, Gabriele and Ilse Wischer 2005 Reconstructing grammaticalization and the definition of grammar, New Reflections on Grammaticalization 3, Santiago de Compostela, Spain. 19th-23rd July 2005 [unpublished manuscript]. DUDEN (ed.) 2005 Grammatik. Vol. 4. Mannheim: Dudenverlag. Eastman, Robert and Elizabeth Eastman 1963 Iquito syntax. In Studies in Peruvian Indian Languages. Vol. 1. 14592. Norman/Oklahoma. Ebert, Robert P. 1976 Infinitival Complement Constructions in Early New High German. Tübingen: Niemeyer. Eisenberg, Peter 2004 Reprint. Grundriß der deutschen Grammatik. Vol. 1: Das Wort. 2nd ed. Stuttgart/Weimar: Metzler. Original edition, Stuttgart/Weimar: Metzler, 1998. Engel, Ulrich 2002 Deutsche Grammatik. München: Iudicium. Fabricius-Hansen, Cathrine 1986 Tempus fugit: über die Interpretation temporaler Strukturen im Deutschen. Düsseldorf: Schwann. 1998 Über den Platz des würde-Gefüges im Tempus-Modus-System des gegenwärtigen Deutsch. In Sprachgermanistik in Skandinavien III: Akten des IV. Nordischen Germanistentreffens auf Schloss Sandbjerg, 5. bis 8. Juli 1996. Pors, Harald, Lisbeth Falster Jakobsen und Flemming Talbo Stubkjær (eds.), 135-158. Århus: Det Erhvervssproglige Fak., Handelshøjskolen. 2000 Die Geheimnisse der deutschen würde-Konstruktion. In Deutsche Grammatik in Theorie und Praxis. Rolf Thieroff (ed.), 83-96. Tübingen: Niemeyer. 2002 Nicht-direktes Referat im Deutschen. Typologie und Abgrenzungsprobleme. In Modus, Modalverben, Modalipartikeln. Cathrine Fabricius-Hansen, Oddleof Leirbukt and Ole Letnes (eds.), 3-27. Trier: Wissenschaftlicher Verlag Trier. Faller, Martina 2002 Semantics and Pragmatics of Evidentials in Cuzco Quechua. Ph.D. diss., Stanford University.

References 353 Fillmore, Charles and Paul Kay 1995 Construction Grammar. Ph.D. diss., University of California, Berkeley. Fillmore, Charles, Paul Kay and Mary C. O’Connor 1988 Regularity and idiomaticity in grammatical constructions: The case of let alone. Language 64 (3): 501–538. Fischer, Kerstin and Anatol Stefanowitsch (eds.) 2006 Konstruktionsgrammatik: Von der Anwendung zur Theorie. Tübingen: Stauffenburg. 2008 Konstruktionsgrammatik und grammatische Konstruktionen. Tübingen: Stauffenburg. Flämig, Walter 1991 Grammatik des Deutschen: Einführung in Struktur- und Wirkungszusammenhänge. erarbeitet auf der theoretischen Grundlage der „Grundzüge einer deutschen Grammatik“. Berlin: Akademie Verlag. Floyd, Rick 1999 The Structure of Evidential Categories in Wanka Quechua. Dallas, Texas: Summer Institute of Linguistics. (Summer Institute of Linguistics and the University of Texas at Arlington publications in linguistics). Fox, Barbara A. 2001 Evidentiality: Authority, Responsibility, and Entitlement in English Conversation. Journal of Linguistic Anthropology 11 (2): 167-192. Frachtenberg, Leo J. 1922 Coos: An illustrative sketch. In Handbook of American Indian languages. Vol.2. Franz Boas (ed.), 297-299, Washington: Government Print Office (Smithsonian Institution, Bureau of American Ethnology). Friedman, Victor A. 2000 Confirmative/nonconfirmative in Balkan Slavic, Balkan Romance, and Albanian with additional observations on Turkish, Romani, Georgian, and Lak. In Evidentials. Lars Johanson and Bo Utas (eds.), 329-366. Berlin/New York: Mouton de Gruyter. Fritz, Thomas 2000 Wahr-Sagen: Futur, Modalität und Sprecherbezug im Deutschen. Hamburg: Buske. Gamon, David 1993 On the development of epistemicity in the German modal verbs mögen and müssen. Folia linguistica historica 14(1-2): 125-176. Gelhaus, Hermann 1975 Das Futur in ausgewählten Texten der geschriebenen deutschen Sprache der Gegenwart: Studien zum Tempussystem. München: Hueber.

354 References Giacalone Ramat, Anna and Manana Topadze 2007 The Coding of Evidentiality: A Comparative Look at Georgian and Italian. Rivista di Linguistica Italiana 19 (1): 7-38. Goldberg, Adele E. 1995 Constructions: a construction grammar approach to argument structure. Chicago: University of Chicago Press. 2006 Constructions at Work. The Nature of Generalization in Language. Oxford: Oxford University Press. Grice, Herbert P. 1989 Reprint. Logic and conversation. In Studies in the way of words. Herbert P. Grice (ed.), 22-40. Cambridge/Massachusetts: Harvard University Press. Original edition Syntax and semantics. Vol.3: Speech acts. Peter Cole and Jerry L. Morgan (eds.), 41-58. New York: Academic, 1975. Guentchéva, Zlatka 1996 Le Médiatif en Bulgare. In L' énonciation médiatisée. Zlatka Guentchéva (ed.), 47-70. Louvain/Paris: Peeters. Gunkel, Lutz 2000 Selektion verbaler Komplemente. Zur Syntax der Halbmodal- und Phasenverben. In Deutsche Grammatik in Theorie und Praxis. Thieroff, Rolf (ed.), 111-121. Tübingen: Niemeyer. Hagège, Claude 1987 Le Français et les siècles. Paris: Jacob. de Haan, Ferdinand 1999 Evidentiality and Epistemic Modality: Setting Boundaries. Southwest Journal of Linguistics 18: 83-101. 2001 The Relation between Modality and Evidentiality. Linguistische Berichte 9: 201-216. 2005 Encoding speaker perspective: Evidentials. In Linguistic diversity and language theories. Frajzyngier, Zygmunt, Adam Hodges and David S. Rood (eds.), 379-397. Amsterdam/Philadelphia: John Benjamins. Hanks, William F 1992 The Indexical Ground of Deictic Reference. In Rethinking Context, Language as an Interactive Phenomenon. Duranti, Alessandro and Charles Goodwin (eds.), 43-77. Cambridge: Cambridge University Press. Haspelmath, Martin 1999 Why is grammaticalization irreversible? Linguistics 37 (6): 1043-68. Heine, Bernd 1993 Auxiliaries: cognitive forces and grammaticalization. New York: Oxford University Press.

References 355 On the role of context in grammaticalization. In New Reflections on Grammaticalization. Wischer, Ilse and Gabriele Diewald (eds.), 83101. Amsterdam/Philadelphia: John Benjamins. 2003 Grammaticalization. In The Handbook of Historical Linguistics. Brian D. Joseph and Richard D. Janda (eds.), 575-601. Oxford: Blackwell. Heine, Bernd, Ulrike Claudi and Friederike Hünnemeyer 1991 Grammaticalization: A Conceptual Framework. Chicago: University of Chicago Press. Heine, Bernd and Tanja Kuteva 2002 World Lexicon of Grammaticalization. Cambridge: Cambridge University Press. Heine, Bernd and Hiroyuki Miyashita 2008 Accounting for a functional category: German drohen ‘to threaten’. Language Sciences 30: 53-101. Helbig, Gerhard and Joachim Buscha 2001 Deutsche Grammatik. Ein Handbuch für den Ausländerunterricht. Berlin: Langenscheidt. 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 P. Himmelmann and Björn Wiemer (eds.), 21-42. Berlin/New York: Mouton de Gruyter. 2005 Gram, construction, and class formation. In Wortarten und Grammatikalisierung. Clemens Knobloch and Burkhard Schaeder (eds.), 7993. Berlin: de Gruyter. Hopper, Paul 1990 Principles of Grammaticalization: Towards a Diachronic Typology. In Language Typology 1987. Systematic Balance in Language. Papers from the Linguistic Typology Symposium, Berkeley, 1-3 December 1987. Winfred P. Lehmann (ed.), 157–170. Amsterdam/ Philadelphia: John Benjamins. 1991 On some principles of grammaticalization. In Approaches to Grammaticalization. Elizabeth C. Traugott and Bernd Heine (eds.), 18–35. Amsterdam/Philadelphia: John Benjamins. Hopper, Paul J. and Elizabeth Closs Traugott 2003 Reprint. Grammaticalization. 2nd ed. Cambridge: Cambridge University Press. Original edition, Cambridge: Cambridge University Press, 1993. Itayama, Mayumi 1993 Werden - modaler als die Modalverben. Deutsch als Fremdsprache 30 (4): 233-237. 2002

356 References Jacobsen, William H., Jr. 1986 The Heterogeneity of Evidentials in Makah. In Evidentiality. Chafe, Wallace L. and Johanna Nichols (eds.), 3-28. Norwood/New Jersey: Ablex. Jakobson, Roman 1971 Reprint. Shifters, verbal categories, and the Russian verb. In Selected Writings, Vol. II: Word and language. Roman Jakobson (ed.), 130147. Berlin: de Gruyter. Original edition, Harvard: Harvard University, Dept. of Slavic Languages and Literatures, 1957. Johanson, Lars 2000 Turkic indirectives. In Evidentials. Lars Johanson and Bo Utas (eds.), 61-88. Berlin/New York: Mouton de Gruyter. Johanson, Lars and Bo Utas (eds.) 2000 Evidentials. Berlin/New York: Mouton de Gruyter. Joseph, Brian D. 2003 Morphologization from Syntax. In The handbook of historical linguistics. Brian D. Joseph and Richard D. Janda (eds.), 472-492. Malden, Massachusetts: Blackwell. Jung, Walter 1988 Grammatik der deutschen Sprache. Leipzig: Bibliographisches Institut. Kay, Paul and Charles J. Fillmore 1999 Grammatical constructions and linguistic generalizations: The What’s X doing Y? construction”. Language 75 (1): 1–33. Kleiner, Mathilde 1925 Zur Entwicklung der Futur-Umschreibung werden mit dem Infinitiv. University of California Publications in Modern Philology 12: 1101. Kokutani, Shigehiro 2004 Grammatikalisierung ist keine Desemantisierung. Zur Identifizierung von syntaktischen Kategorien und Hilfsverben. Neue Beiträge zur Germanistik 3 (2): 48-61. Kotin, Michail 2003 Die werden-Perspektive und die werden-Periphrasen im Deutschen. Frankfurt/Main: Lang. Kuryłowicz, Jerzy 1965 The evolution of grammatical categories. Diogenes 51: 55-71. Langacker, Ronald W. 1985 Observations and speculations on subjectivity. In Iconicity in syntax: Proceedings of a symposium on iconicity in syntax, Stanford, June 24-6, 1983. John Haiman (ed.), 109-150. Amsterdam/Philadelphia: John Benjamins.

References 357 Foundations of cognitive grammar Vol. 1: Theoretical prerequisites. Stanford: Stanford University Press. 1991 Foundations of Cognitive Grammar. Volume II: Descriptive Application. Stanford: Stanford University Press. 1999 Losing control: grammaticization, subjectification, and transparency. In Historical semantics and cognition. Andreas Blank (ed.), 147-175. Berlin: de Gruyter. 2008 Cognitive grammar: a basic introduction. New York: Oxford University Press. Lazard, Gilbert 1999 Mirativity, evidentiality, mediativity, or other? Linguistic Typology 3 (1): 91-109. 2001 On the grammaticalization of evidentiality. Journal of Pragmatics 33 (3): 359-368. Lehmann, Christian 1985 Grammaticalization: synchronic variation and diachronic change. Lingua e Stile 20 (3): 303-318. 1995 Reprint. Thoughts on grammaticalization. Revised and expanded version. First published edition. München: Lincom Europa. Original edition, Köln: Kölner Universalienprojekt, 1982. 2002 New reflections on grammaticalization and lexicalization. In New Reflections on Grammaticalization. Ilse Wischer and Gabriele Diewald (eds.), 1-18. Amsterdam/Philadelphia: John Benjamins. 2004 Theory and Method in Grammaticalization. Zeitschrift für Germanistische Linguistik 32 (2): 152-187. Leirbukt, Oddleif 1997 Untersuchungen zum bekommen-Passiv im heutigen Deutsch. Tübingen: Niemeyer. Leiss, Elisabeth 1992 Die Verbalkategorien des Deutschen. Berlin: de Gruyter. 2000 Artikel und Aspekt: die grammatischen Muster von Definitheit. Berlin: de Gruyter. 2008 The silent and aspect-driven patterns of deonticity and epistemicity. In Modality-Aspect Interfaces: Implications and typological solutions. Werner Abraham and Elisabeth Leiss (eds.), 15-42. Amsterdam/Philadelphia: John Benjamins. Lerch, Eugen 1942 Das Futurum des zu Erwartenden im Französischen und im Deutschen. Neuphilologische Mitteilungen 43: 161-191. Letnes, Ole 1997 ‚Sollen‘ als Indikator für Redewiedergabe. In Aspekte der Modalität im Deutschen – auch in kontrastiver Sicht. F. Debus and Oddleif Leirbukt (eds.), 119-134. Hildesheim: Olms. 1987

358 References Lipsky, Angela 2002 Eine semantische und pragmatische Darstellung der Konstruktion werden + Infinitiv. Deutsch als Fremdsprache 39 (2): 103-107. Lüdeling, Anke, Emil Kroymann, Sebastian Thiebes, and Ulf Leser 2004 Eine vergleichende Analyse von historischen und diachronen digitalen Korpora. Technical Report 174 des Instituts für Informatik der Humboldt-Universität zu Berlin, November 2004. Lyons, John 1977 Semantics. Cambridge: Cambridge University Press. Matlock, Teenie 1989 Metaphor and the grammaticalization of evidentials. Proceedings of the Annual Meeting of the Berkeley Linguistic Society 15: 215-25. Matzel, Klaus and Bjarne Ulvestad 1982 Futur I und futurisches Präsens. Sprachwissenschaft 7: 282-328. Meid, Wolfgang 1971 Das germanische Praeteritum: indogermanische Grundlagen und Ausbreitung im Germanischen. Innsbruck: Kowatsch 1983 Bemerkungen zum indoeuropäischen Perfekt und zum germanischen starken Praeteritum. Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 36 (3): 329-336. Melcuk, Igor A. 1998 Kurs obšcej morfologii = Cours de morphologie générale 2: Morfologiceskie znacenija. Moskva: Škola jazyki russkoj kul'tury. Michaelis, Laura 2004 Type shifting in construction grammar: An integrated approach to aspectual coercion. Cognitive Linguistics 15: 1-67. Mortelmans, Tanja 2000a Konjunktiv II and Epistemic Modals in German: A Division of Labour. In Constructions in Cognitive Linguistics. Selected Papers from the Fifth International Cognitive Linguistics Conference, Amsterdam, 1997. Ad Foolen and Frederike van der Leek (eds.), 191215, Amsterdam/Philadelphia: John Benjamins. 2000b On the evidential nature of the epistemic use of the German modals müssen and sollen. In Modal verbs in Germanic and Romance languages. Johan Van der Auwera and Patrick Dendale (eds.), 131-148. Amsterdam/Philadelphia: John Benjamins. 2004 The status of the German auxiliary werden as a ‘grounding predication’. In Modalität und Übersetzung. Heinz Vater and Ole Letnes (eds.), 33-56. Trier: Wissenschaftlicher Verlag Trier. 2009 Erscheinungsformen der indirekten Rede im Niederländischen und Deutschen: zou-, soll(te)- und der Konjunktiv I. In Modalität: Epistemik und Evidentialität bei Modalverb, Adverb, Modalpartikel

References 359 und Modus. Werner Abraham and Elisabeth Leiss (eds.), 171-187. Tübingen: Stauffenburg. Mushin, Ilana 2001 Evidentiality and Epistemological Stance: Narrative Retelling. AmsterdamI Philadelphia: Benjamins. Nordlinger, R and Elizabeth C. Traugott 1997 Scope and the development of epistemic modality: evidence from ought to. English Language and Liguistics 1 (2): 295-317. Nuyts, Jan 2001 Epistemic modality, language and conceptualization: a cognitivepragmatic perspective. Amsterdam/Philadelphia: John Benjamins. Olsen, Susan 1981 Problems of seem/scheinen. Constructions and their Implications for the Theory of Predicate Sentential Complementation. Tübingen: Niemeyer. Oswalt, Robert L. 1986 The Evidential System of Kashaya. In Evidentiality. Wallace L. Chafe and Johanna Nichols (eds.), 29-45. Norwood/New Jersey: Ablex. Pafel, Jürgen 1989 Scheinen + Infinitiv. Eine oberflächengrammatische Analyse. In Wissen, Wahrnehmen, Glauben: epistemische Ausdrücke und propositionale Einstellungen. Gabriel Falkenberg (ed.), 123-174. Tübingen: Niemeyer. Palander-Collin, Minna 1999 Grammaticalization and Social Embedding: I Think and Methinks in Middle and Early Modern English. Helsinki: Société Néophilologique. Palmer, Frank Robert 1986 Mood and Modality. Cambridge: Cambridge University Press. 2001 Mood and Modality. 2nd ed. Cambridge: Cambridge University Press. Paul, Hermann 1992 Reprint. Deutsches Wörterbuch. 9th ed. Tübingen: Niemeyer. Original edition, Halle: Niemeyer, 1897. 1989 Reprint. Mittelhochdeutsche Grammatik. 23rd ed. Tübingen: Niemeyer. Original edition, Tübingen: Niemeyer, 1881. Plungian, Vladimir A. 1998 Ƚɪɚɦɦɚɬɢɱɟɫɤɢɟ ɤɚɬɟɝɨɪɢɢ, ɢɯ ɚɧɚɥɨɝɢ ɢ ɡɚɦɟɫɬɢɬɟɥɢ. Ɇɨɫɤɜɚ, 1998. [“Grammatical categories, their analogues and alternatives”, postdoctoral lecture qualification]. 2001 The place of evidentiality within the universal grammatical space. Journal of Pragmatics 33: 349-357.

360 References Radtke, Petra 1998 Die Kategorien des deutschen Verbs: zur Semantik grammatischer Kategorien. Tübingen: Gunter Narr. Rauh, Gisa 1978 Linguistische Beschreibung deiktischer Komplexität in narrativen Texten. Tübingen: Narr. 1984 Aspekte der Deixis. Sprachwissenschaft 9: 23-84. Reis, Marga 2005 Zur Grammatik der sog. ‘Halbmodale’ drohen/ versprechen + Infinitiv. In Deutsche Syntax. Empirie und Theorie. Symposium Göteborg 13.-15- Mai 2004. Franz J. D’Avis (ed.), 125-145. Göteborg: Acta Universitatis Gothoburgensis. Saltveit, Laurits 1962 Studien zum deutschen Futur. Die Fügungen werden mit Partizip des Präsens und werden mit dem Infinitiv in ihren heutigen Funktionen und in ihrer geschichtlichen Entwicklung. Oslo: Norwegian Universities Press. Schlichter, Alice 1986 The origin and deictic nature of Wintu evidentials. In Evidentiality. Wallace L. Chafe and Johanna Nichols (eds.), 273-312. Norwood/New Jersey: Ablex. Schmid, Hans Ulrich 2000 Die Ausbildung des werden-Futurs. Überlegungen auf der Grundlage mittelalterlicher Endzeitprophezeiungen. Zeitschrift für Dialektologie und Linguistik 67: 6-27. Sennholz, Klaus 1985 Grundzüge der Deixis. Bochum: Brockmeyer. Smirnova, Elena 2006 Die Entwicklung der Konstruktion würde + Infinitiv im Deutschen. Berlin: de Gruyter. 2007 Rekonstruktion eines Grammatikalisierungsprozesses. Entwicklung der Konstruktion würde + Infinitiv zum Konjunktiv II-Grammem. Zeitschrift für Dialektologie und Linguistik 74 (1): 20-37. Speas, Margaret 2004 Evidentiality, logophoricity and the syntactic representation of pragmatic features. Lingua 114 (33): 255-276. Squartini, Mario 2001 The internal structure of evidentiality in Romance. Studies in Language 25 (2): 297–334. 2004 Disentangling evidentiality and epistemic modality in Romance. Lingua 114: 873-895. Stolz, Thomas and Benedikt Jeßing (eds.) 1994 Grammatikalisierung und Metaphorisierung. Bochum: Brockmeyer.

References 361 Sumbatova, Nina 1999 Evidentiality, transitivity and split ergativity: evidence from Svan. In Tense-Aspect, Transitivity and Causativity. Werner Abraham and Leonid Kulikov (eds.), 63-95. Amsterdam/Philadelphia: John Benjamins. Sweetser, Eve 1988 Grammaticalization and semantic bleaching. In Proceedings of the Fourteenth Annual Meeting of the Berkeley Linguistics Society. Shelley Axmaker, Annie Jaisser, and Helen Singmaster (eds.), 389-405. Berkeley: Berkeley Linguistics Society. 1990 From Etymology to Pragmatics: Metaphorical and Cultural Aspects of Semantic Structure. Cambridge: Cambridge University Press. Thieroff, Rolf 1992 Das finite Verb im Deutschen: Tempus, Modus, Distanz. Tübingen: Narr. Tournadre, Nicolas 1996 Comparaison des Systèmes médiatifs de quatre dialectes Tibétains (Tibétain Central, Ladakhi, Dzongkha et Amdo). In L' énonciation médiatisée. Zlatka Guentchéva (ed.), 47-70. Louvain/Paris: Peeters. 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/Philadelhia: John Benjamins. 1989 On the rise of epistemic meanings in English: an example of subjectification in semantic change. Language 57: 31-55. 1995 Subjectification in grammaticalization. In Subjectivity and subjectification: Linguistic perspectives. Dieter Stein and Susan Wright (eds.), 31-54. Cambridge: Cambridge University Press. 1997 Subjectification and the development of epistemic meaning: the case of promise and threaten. In Modality in Germanic Languages: Historical and Comparative Perspectives. Toril Swan and Olaf Jansen Westvik (eds.), 185-210. Berlin/New York: Mouton de Gruyter. 1999 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. 2003 Constructions in Grammaticalization. In The Handbook of Historical Linguistics. Brian D. Joseph and Richard D. Janda (eds.), 575-601. Oxford: Blackwell. 2008 Grammatikalisierung, emergente Konstruktionen und der Begriff der Neuheit. In Konstruktionsgrammatik II. Von der Konstruktion zur

362 References Grammatik. Translated by Arne Zeschel. Anatol Stefanowitsch and Kerstin Fischer (eds.), 5-32. Tübingen: Stauffenburg. 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. Vol 1: Focus on theoretical and Methodological Issues. Traugott, Elizabeth Closs and Bernd Heine (eds.), 189-218. Amsterdam/Philadelphia: John Benjamins. Vater, Heinz 1975 werden als Modalverb. In Aspekte der Modalität. Calbert, Joseph P. and Heinz Vater (eds.), 71-148. Tübingen: Narr. Verhagen, Arie 2000 The girl that promised to become something: an exploration into diachronic subjectification in Dutch. In The Berkeley Conference on Dutch Linguistics 1997. Dutch Linguistics at the Millenium. Lanham Thomas Frederic Shannon and Johan Pieter Snapper (eds.), 197-208. MD: University Press of America. Vliegen, Maurice 2006 The relation between lexical and epistemic readings: the requirements of promise and threaten in Dutch and German. In Topics in Subjectification and Modalization. Bert Cornillie and Nicole Delbecque (eds.), 73-95. Amsterdam/Philadelphia: John Benjamins. Weinrich, Harald 2003 Textgrammatik der deutschen Sprache. Hildesheim: Olms. Welke, Klaus 2005 Tempus im Deutschen: Rekonstruktion eines semantischen Systems. Berlin: de Gruyter. Westvik, Olaf Jansen 1994 Language Change and Language Structure: Older Germanic Languages in a Comparative Perspective. Berlin/New York: Mouton de Gruyter. 2000 Über Herkunft und Geschichte des werden - Futurs. Eine Auseinandersetzung mit neuerer und neuester Forschung. In Raum, Zeit, Medium. Festschrift für Hans Ramge zum 60. Geburtstag. Richter, Gerd Richter, Hans Ramge, Jorg Riecke and Britt-Marie Schuster (eds.), 235-261. Darmstadt: Hessische Historische Kommission. Whitt, Richard 2008 Evidentiality and Perception Verbs in English and German: A corpus-based analysis from the early modern period to the present. Ph.D. diss., Department of German, University of California, Berkeley.

References 363 Wiemer, Björn to appear Hearsay in European languages. Toward an integrative account of grammatical and lexical marking. In Linguistic Realization of Evidentiality in European Languages. Gabriele Diewald and Elena Smirnova (eds.). Berlin/New York: Mouton de Gruyter. Wierzbicka, Anna 1986 A semantic metalanguage for the description and comparison of illocutionary meanings. Journal of Pragmatics 10 (1): 67-107. 1991 Cross-Cultural Pragmatics. The Semantics of Human Interaction. Berlin: Mouton de Gruyter. 1994 Semantics and epistemology. The meaning of ‘evidentials’ in a cross-linguistic perspective. Language Sciences 16 (1): 81-137. 1996 Semantics, Primes and Universals. Oxford: Oxford University Press. Wiesinger, Peter 2001 Zum frühneuhochdeutschen Ausdruck der Aktionsart im Präteritum beim steirischen Dichtermönch Andreas Kurzmann um 1400. In Zur Verbmorphologie germanischer Sprachen. Watts, Sheila, Jonathan West and Hans-Joachin Solms (eds.), 175-188. Tübingen: Niemeyer. Willett, Thomas 1988 A cross-linguistic survey of the grammaticalization of evidentiality. Studies in Language 12: 51-97. Wilmanns, Wilhelm 1906 Deutsche Grammatik. Dritte Abteilung: Flexion. 1. Hälfte: Verbum. Strassburg: Trübner. Zifonun, Gisela, Ludger Hoffmann and Bruno Strecker 1997 Grammatik der deutschen Sprache. In 3 Bänden. Berlin/New York: de Gruyter.

Subject Index

abstract constructional schema 144, 155, 311, 329 agentive 108, 127, 130, 131, 135, 193f., 203, 204, 206, 207, 210, 211, 212, 270f., 280, 293, 301, 317 ambiguity 44, 49, 136f., 141, 185, 187, 198, 265 aspectuality 7 auxiliary 16, 20, 52, 108, 112, 116f., 120, 131, 137, 148, 157, 160, 188, 237, 270, 310, 330 commissive 214f., 286, 289f., 293f., 302, 307, 319 construction 26, 98, 143f., 153f., 155 construction type 34, 48, 144, 155, 158, 180, 192, 205, 230, 267, 268, 270f., 280, 291f., 321f. context 26, 33, 43, 98, 143f. critical 144, 147f., 151, 149, 260, 277, 281, 296, 312, 315f. isolating 144, 147, 152f., 279, 311f., 316, 318, 321 untypical 144, 145f., 247, 256, 274, 280, 291, 293, 295, 311f. context dependence 136 conversational implicature 17, 44, 49, 72, 80, 93f., 136, 141f., 144f., 152, 183, 256 copula 160, 177, 184, 186f., 237, 240f., 252f., 263, 303, 305, 313 compound copula 177, 179, 263 197, 210, 271 deictic dimension 11f., 54, 331 deictic function 10, 15 deictic relation 10f., 87, 88, 331

deixis 9, 12, 14, 70f., 81, 133 desired 78, 95f., 150, 224, 326, 334 evidential expression 3, 41f., 43, 51, 57, 67, 90, 93 evidential marker 3f., 10, 14, 15f., 23f., 40, 42, 44, 46f., 50, 62, 90, 157, 327, 330 evidential periphrasis 44, 103 evidential strategy 41, 43, 45f., 49 evidentiality 2f. auditory 55, 58, 332 direct 26, 53, 54f., 56f., 59, 177, 182f., 189f., 221, 225f., 266 endophoric 57-58 hearsay 5, 53, 59, 61, 64, 67, 69, 73, 227, 333 indirect 24, 51, 54f., 59f., 62-65, 69, 92, 191, 225, 227, 325, 332 inferential 18, 21, 59-65, 91f., 95 conceptual 63f., 177, 182, 189, 223, 224, 326 perceptual 63f., 112, 150, 157, 180, 191, 202, 224, 326 reportive 26, 59, 64-68, 72, 331 reflected 60 experiencer 88, 179, 279 externalization 308, 320 extraposition 109, 113, 117, 119, factuality 6, 13, 72, 77-79, 81f., 87, 89, 94f. frequency 43, 105, 142, 147, 213, 281, 287 future 104, 120, 125, 129, 161, 163f.,

Subject Index 365 167, 170, 172, 201, 232-237, 247 goal 78, 131-133, 303-307 gradience 136, 140-142, 145, 185 grammatical category 19, 40, 50f., 77 grammaticalization parameters 38, 42, 52, 105-197, 109, 116, 128, 143, 325 heterosemy 136, 138, 143, 155, 158 homogeneity of paradigm 220f. idiom 144, 150, 153 indexical-grammatical function 133, 135, 308-310, 320 indexicality 10, 331 indicative 12, 78f., 83, 102, 148, 156f., ingressive 21, 43, 78, 89, 95, 120, 150, 160, 300, 303, 313, 321, 323 innovation 157 intentional 110, 178, 188, 194, 197f., 207, 234, 237, 280, 296 mediated information 65, 69, 222, 227, 332 metaphorization 124-128, 299, 319, 328 metonymization 303, 323, 124f. modal particle 44, 129, 136, 157 modal verb 18, 20, 50, 65, 78, 82-84, 86, 88f., 104, 125, 129, 139, 145, 152f., 171, 222, 227, 232, 235, 261f., 334 modality deontic 89, 92, 139 dynamic 89, 99 epistemic 6, 17-19, 50, 68, 72, 76f., 80, 83, 86f., 91f., 176, 201, 332 volitional 89, 334 mood 20, 41, 77, 104, 156 necessity 91f. obligatoriness 99 communicative 100-102, 114

language internal 100-102 opacity 144, 147f., 151 oppositive values 19, 60, 222 origo 10-12, 326, 331f. origo displacement 71 origo-exclusive 12-14, 54, 55, 78f., 332 origo-inclusive 12-14, 54f., 78f., 332 origo-shift 69-71, 333 paradigm 24, 78, 84, 94, 99-105, 115 paradigmaticity 99, 105f., 112, 154, 221, 329 paradigmaticization 20, 106, 154, 325, 329 paradigmatic context 144, 156, 321 path 130-132, 303-307 perception verb 8, 28, 44, 48, 58 performative 61, 126, 132, 163, 167, 194, 284 peripheral evidential construction 43-48 persistence 21, 155, 183, 185, 203, 213, 265 personal evidence 60-62 personification 127, 132, 274, 291, 315 polysemy 92, 136, 138f., 141-143, 158, 185 present experience 22, 202, 223 preterito-present 88f. pre-evidential 46-49, 90 presentational construction 243 quotative 60f., 66-73, 79, 92, 156, 222, 333 referential function 133, 135, 308, 310, 320, 328 reinterpretation 22, 33, 88, 90, 124, 129, 142, 158, 198, 250, 266, 277, 282, 303, 308, 312, 327 renovation 156 subjective 23, 78, 166, 170, 182, 185,

366 Subject Index scheme retention 128, 132, 158, 308, 328 secondary event 53, 55-57, 63f., 81, 331 secondary predication 4, 41, 46, 52, 87, 111, 150, 172, 186f., 190, 196, 257, 279 semantic abstraction 302, 319 semi-auxiliary 109, 193, 207 source 11, 52, 56, 63, 81-87, 130, 320, 333 speech act 69, 110, 116, 126, 130, 132, 150, 214, 272, 286, 291, 293, 301, 307, 310, 317, 321, 323

193, 264 subjunctive 12, 72, 78f., 83, 148, 156, 162, 180, 248, 258, 263 syntagmatic pattern 230, 237, 244, 252, 262, 270, 286, 293 temporality 7, 23, 51, 75 tense 7, 219, 232f. text-connective function 135, 169, 309 visual 2, 6, 20, 44, 49, 52, 55, 57, 58, 172, 178, 180, 183, 188, 226, 254, 262, 264, 301, 304, 309, 314, 316, 321, 323, 328