Linguistic Realization of Evidentiality in European Languages 9783110223972, 9783110223965

Citation preview

Linguistic Realization of Evidentiality in European Languages

Empirical Approaches to Language Typology 49

Editors Georg Bossong Bernard Comrie Yaron Matras

De Gruyter Mouton

Linguistic Realization of Evidentiality in European Languages Edited by Gabriele Diewald Elena Smirnova

De Gruyter Mouton

ISBN 978-3-11-022396-5 e-ISBN 978-3-11-022397-2 ISSN 0933-761X Library of Congress Cataloging-in-Publication Data Linguistic realization of evidentiality in European languages / edited by Gabriele Diewald, Elena Smirnova. p. cm. ⫺ (Empirical approaches to language typology ; 49) “This volume contains a selection of contributions to the workshop “Linguistic realization of evidentiality in European languages”, held at the 30th Annual Convention of the German Society of Linguistics in Bamberg (February 27⫺29, 2008).” Includes bibliographical references and index. ISBN 978-3-11-022396-5 (alk. paper) 1. Evidentials (Linguistics) 2. Typology (Linguistics) I. Diewald, Gabriele. II. Smirnova, Elena. P325.5.E96L56 2010 415⫺dc22 2010042027

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. 쑔 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 Introduction Evidentiality in European languages: the lexical-grammatical distinction Gabriele Diewald and Elena Smirnova Types of verbal evidentiality marking: an overview Vladimir Plungian Hearsay in European languages: toward an integrative account of grammatical and lexical marking Björn Wiemer

1

15

59

Information source in Spanish and Basque: a parallel corpus study Asier Alcázar

131

Embedded evidentials in German Mathias Schenner

157

Embedding indirective (evidential) utterances in Turkish Hatice Coúkun

187

Epistemic modality and evidentiality and their determination on a deictic basis: the case of Romance languages Gerda Haßler

223

Evidentiality, polysemy, and the verbs of perception in English and German Richard Jason Whitt

249

Evidential markers in French scientific writing: the case of the French verb voir Francis Grossman and Agnès Tutin

279

An interactional approach to epistemic and evidential adverbs in Spanish conversation Bert Cornillie

309

vi

Table of contents

Revelative evidentiality in European languages: linguistic marking and its anthropological background Alexandra Kratschmer and Adriënne Heijnen

331

Subject index

369

Introduction Evidentiality in European languages: the lexicalgrammatical distinction1 Gabriele Diewald and Elena Smirnova The term “evidentiality” refers to a semantic-functional domain, which, beyond being expressed by a great variety of lexical means, is a relevant category in the grammatical systems of numerous of the world’s languages. Like “modality” or “temporality” the notion of evidentiality covers a range of meanings that may serve referential as well as non-referential purposes, or – to use a different diction – evidential meanings range from lexical to grammatical functions. In order to capture this, we speak of a “semanticfunctional” domain here. The basic characteristic of linguistic evidentiality is the explicit encoding of a source of information or knowledge (i.e. evidence) which the speaker claims to have made use of for producing the primary proposition of the utterance. The type of evidence the speaker adduces may be of various kinds (i.e., different modes of perception and cognition), which may be encoded by the evidential expressions but also may be left unspecified. This common core is reflected in most definitions of linguistic evidentiality as the following selection illustrates: 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) [Evidentials express], how the speaker obtained the information on which s/he bases an assertion. (Willet 1988: 55) 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)

In the past decades research on evidentiality and on the expression of evidential meanings in language has made enormous progress. In particular,

2

Gabriele Diewald and Elena Smirnova

there has been growing acceptance of the assumption that evidentiality is a semantic-functional domain in its own right, and not a sub-division of epistemic modality. Accordingly, it is aknowledged by many scholars that evidential markers and evidential systems in languages are in principle – i.e, notwithstanding the empirical fact of frequent overlap – an independent category (cf. de Haan 2001 and Aikhenvald 2004:7). In much of the prior work on evidentiality it has been generally assumed that European languages typically lack grammatical markers and grammatical systems of evidentiality. Despite the fact that evidential expressions of Balkan languages have been described in several studies (cf. e.g. Friedman 1986, 2000, 2003, Matras 1995, Guentchéva 1993, 1996), a detailed description of the language-specific realizations of evidentiality is yet to be carried out for most of European languages. This volume is a step towards that aim. It attempts at providing classificatory reflections as well as empirical facts about languages that have various – lexical as well as grammatical – evidential expressions. Most papers of this volume originate in presentations given at the workshop “The linguistic realization of evidentiality in European languages”, held at the 30th Annual Convention of the German Society of Linguistics in Bamberg (February 27-29, 2008). The papers by Schenner and Coúkun are additional contributions, which have been especially commissioned for this volume. The paper by Haßler was not held at the workshop but was submitted afterwards.2 The scope of the ten papers collected here ranges from broad overviews on areal and/or typological issues to in-depth studies of evidential expressions in particular languages, some taking the perspective of synchronic contrastive comparison, others focussing on diachronic investigation. Furthermore, there are contributions that take a broader view by combining linguistic investigation with cultural and anthropological aspects. Thus, beyond the typologically informed contributions, which survey a variety of languages (Plungian, Wiemer), there are studies on particular evidential phenomena in sets of Romance and Germanic languages (Haßler, Kratschmer and Heijnen), as well as detailed examinations of evidentials expressions in French (Grossmann and Tutin), German (Schenner, Whitt), English (Whitt), Spanish (Alcazár, Cornillie), Basque (Alcázar), and Turkish (Coúkun). Beyond a sample of rich and original empirical data on types of evidential expressions in European languages, their respective degrees of grammaticalization and the kinds of function evidentials fulfil, the volume provides new insights into the following more specific issues:



– – – – – – –

Introduction

3

evidential (sub)systems and paradigms in different languages: semantic distinctions and pragmatic functions; interrelations of evidential expressions with other grammatical categories (e.g. mood, tense, aspect); syntactic and pragmatics issues concerning the restrictions on evidentials in subordinate clauses; the diachronic rise of evidential markers and the degrees of grammaticalization of evidentials in different languages; lexical sources of evidential markers in the domain of verbs of perception; frequency, distribution and specific functions of evidential markers in different text types and/or registers; metalinguistic devices of marking information sources according to anthropological parameters.

Before introducing the papers individually, one central issue has to be discussed as it is addressed in virtually every paper of this volume, and is – so to speak – the red thread running through its argumentative texture. It is the dispute raised by Aikhenvald (2004) who argues for a very restrictive definition of evidentials and in all declines the justification for assuming that European languages possess evidentials. This point addresses the fundamental question of the grammatical-lexical continuum in expressions of evidentiality, and therefore is worth devoting some attention to from a theoretical viewpoint as well. Aikhenvald (2004: 3-11) takes a strong stand concerning the distinction between evidential markers proper on the one hand and what she calls evidential strategies, on the other. She claims that European languages do not display evidential markers and that most evidential phenomena in those languages are “mere” evidential strategies. In her argument Aikhenvald points out that the fact that modal verbs in European languages often have evidential overtones and usages 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. 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). This again has to be agreed upon. Nevertheless, a closer and more differentiated look at European languages is called for. In particular,

4

Gabriele Diewald and Elena Smirnova

it has to be considered that grammatical systems develop, thereby producing a full scale of phenomena with intermediate values concerning the cline between lexical and grammatical. Moreover, in the course of developing new evidential functions (which may be grammaticalized later), a particular item (or construction) is bound to run through early developmental stages where evidential functions are contextually induced via conversational implicature and non-evidentail functions still prevail to later developmental stages where evidential functions are part of the item’s inherent semantic structure (i.e. they are semanticized or conventionalized) no matter whether older, more lexical usages still co-exist beside the new meanings or not. Though Aikhenvald (2004: 11) admits that change leads to clines and intermediate values, 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 fully grammaticalized inflectional systems. The reason for this position 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: 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 conceptual domain, i.e. “evidentiality”, is analogous to equating the grammatical category “tense” with the conceptual domain “temporality”.3 This tight and impermeable association of conceptual domain and grammatical category prevents an 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. To give an example (which is discussed in detail in Diewald and Smirnova 2010): the German evidential constructions werden & infinitive and scheinen/drohen/versprechen & 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. They are not yet full-fledged grammaticalized evidential systems as compared to those systems invoked by Aikhenvald, which have inflectional or



Introduction

5

clitic evidential markers, but they are instances of evidential systems on the rise (e.g. Cornillie 2007, Dendale and Tasmowski 2001, Diewald 2004, Mortelmans 2000, Plungian 2001). Thus, we contend that, since grammaticalization is a gradual process, it would be counterproductive to restrict attention to fully developed grammaticalized systems. Linguistic research can only account for the make-up and functioning of evidential systems, if it does not exclude evolving systems and their interplay with lexical and other means in expressing evidential values from the agenda. Therefore, it does not come as a surprise that Aikhenvald’s radical position presented in her 2004 book is unequivocally contested by all contributors to this volume. Wiemer explicitly applies the definition given by Aikhenvald (2003: 1) not only to grammatical markers but also to lexical expressions in different European languages. Whitt discovers evidential functions of German and English constructions with verbs of perception which display different degrees of grammaticalization, i.e. which are situated at different points of the lexical-grammatical continuum. Alcázar shows that Basque obviously has a system of evidential markers (particles) which interacts with an array of further evidential strategies in the same language. Schenner proposes a definition of evidential expressions which is meant to equally apply to lexical as well as to grammatical linguistic elements. Schenner argues for treating German verbal periphrastic forms as an evidential system. Coúkun focuses on Turkish which without a doubt does have evidential markers in the strict sense, which however – as she shows – at the some time interact with other phenomena to produce evidential and further effects. Kratschmer and Heijnen draw a clear distinction between grammatical markers and (simple or composed) lexical expressions, and insist that both types of formal expression are valid objects of investigation for those interested in the study of evidential marking. Grossmann and Tutin concentrate on lexical means expressing evidentiality, and emphasize the importance of analyzing lexical elements as they are often potential predecessors of grammaticalized evidentials. In treating epistemic and evidential adverb and adverbials in Spanish, Cornillie assumes that evidentiality is an important linguistic phenomenon in European languages. Thus, Aikhenvald’s verdict has to be considerably relativized: European languages, or better, languages spoken in Europe may not have the most complex or formally and functionally differentiated grammaticalized systems of evidential markers found in the world’s languages, but there is a large number of them which clearly have grammaticalized systems of evi-

6

Gabriele Diewald and Elena Smirnova

dential markers, and there is an even larger number with evidential systems in their initial or intermediated stages of development and/or with a great variety of stereotypical, automatized lexical expressions for that purpose. In short, all authors of this volume consent that it is well worth studying evidential markers as well as evidential strategies in European languages, which – needless to mention it – does of course not question the fact that “true” evidential markers – i.e. maximally grammaticalized evidential markers – are (morphologically bound) grammatical formatives. The order chosen for the contributions starts from the general to the more specific. The first two papers present typological surveys (Plungian, Wiemer), followed by papers focussing on more structural, grammatical issues concerning particular languages (Spanish and Basque by Alcazár, Turkish by Coúkun, German by Schenner). The next contributions concentrate on semantic issues, in particular on evidentials having lexical sources in the domain of perception verbs (Whitt on evidentiality, polysemy, and verbs of perception in English and German; Grossmann and Tutin on French voir as a metalinguistic device in written scientific discourse), as well as semantic considerations concerning the borders between evidentiality and modality (Haßler on various issues in Romance Languages). Cornillie takes up the topic of functional extensions of evidentials (and epistemic) expressions in spoken discourse, thus shifting the focus towards central aspects of dialog analysis like the turn-taking mechanism. Finally, the field is extended to „revelative evidentiality” in an interdisciplinary approach combing linguistic and anthropological research (Kratschmer and Heijnen on linguistic marking of revelative evidentiality in Icelandic, German, and Italian). Vladimir Plungian presents a general overview on evidentiality and on the study of evidentiality. In his paper “Types of verbal evidentiality marking: an overview”, he gives a brief overview of grammatical evidential systems which is based on generalizations proposed in the typological studies. After a short description of the areal distribution of evidentialitymarking systems, he discusses the history of linguistic studies in evidentiality. Furthermore, he presents a cross-linguistic classification both of evidential values and evidential systems, whereby a “basic” system is taken as a point of departure. The basic system is considered the prototypical realization of a generalized typological idea of evidentiality in a most neutral way. Plungian concludes with a discussion of the relation between evidentiality and modality. He argues to treat them as two different semantic domains which are closely related synchronically as well as diachronically. He notes that the nature of the relation between evidentiality and



Introduction

7

modality cannot be explained in a general way since it depends on the concrete evidential system in a concrete language under investigation. In this connection Plungian proposes to distinguish between “modalized” and “non-modalized” evidential systems. The contribution by Björn Wiemer offers a first general crossclassification of linguistic devices with reportive meaning found in European languages (whereby he takes “reportive” as synonymous to “hearsay” and as a sub-domain of evidentiality). The title of this contribution “Hearsay in European languages: towards an integrative account of grammatical and lexical marking” is programmatic and takes a clear position contra Aikhenvald’s 2004 stance. The layout of this study is empirical and classificatory, and, though acknowledging the distinction between epistemic modality and evidentiality the author aims at a maximally inclusive survey, i.e. at collecting as many reportage expressions as possible. The main body of the papers shows that these – i.e. expressions of reportive functions – range from bound (inflectional and agglutinative) morphology, like e.g. in Georgian languages and Turkish, via functional extensions of various verbal grammatical categories, like tense, mood and aspect markers (e.g. the analytic perfect as in Baltic and Balkan languages), extensions from future grams, past tense grams (e.g. the Italian imperfect), moods (e.g. French and Portuguese) and modal auxiliaries (e.g. German), to lexical means and constructions (like volitional verbs, predicative constructions, sentence adverbials and particles). Due to this overall scheme, the paper is not primarily concerned with language change or grammaticalizationlexicalization clines, but instead takes a purely empirical, “surfaceoriented” perspective. Thus, it makes a valuable first move towards a comprehensive view on the areal distribution in Europe’s languages of expressions of reportive functions. In the paper “Information source in Spanish and Basque: a parallel corpus study” Asier Alcázar provides a contrastive study of grammatical evidential markers in Basque and non-grammaticalized evidential strategies in Spanish. Using a Spanish-to-Basque parallel corpus he shows that Basque – unlike Spanish – is a language with several grammaticalized evidential expressions (particles) which fit into typologically acknowledged evidential systems. Special focus is put on reported evidence expressed by the Basque particle omen appearing before the tensed verb. The investigation of the translation practice in the Spanish-to-Basque parallel corpus reveals that omen is uniformly chosen to translate a variety of Spanish evidential expressions, and, moreover, that omen is frequently inserted even if there is no evidential expression in the Spanish original. Furthermore, omen is only

8

Gabriele Diewald and Elena Smirnova

used when the relevant Spanish expression has sentential scope. On the basis of these findings the author concludes that Basque omen is clearly a grammatical evidential marker with a defined functional spectrum (‘hearsay’), obligatory sentential scope, and increased obligatoriness. Further – though less grammaticalized – members of the evidential particle paradigm in Basque are seen in the particles bide ‘apparently’, edo ‘inferential’ (‘must’), ote ‘dubitative’, ohi ‘habitual’. As Basque, in addition to the evidential particles, has further expressions to render evidential meaning, in particular in cases of constituent scope ambiguity or semantic ambiguity, Basque is an exemplar case of the interplay of grammaticalized means and lexical meanings in the expression of evidential values and functions. Based on corpus as well as questionaire data and using formal semantics as a descriptive tool, Mathias Schenner in his paper “Embedded evidentials in German” tackles two questions concerning evidentials in German. First, he takes up the discussion on whether German (representative of further „European languages”) does have evidentials in the true sense of the word. Here, again, Aikhenvald’s position is contradicted, and Schenner argues for accepting a semantically strict but formally liberal notion of evidentiality. Separating epistemic notions from evidential ones (i.e. arguing for treating evidentiality as a category of its own), he shows that the “reportive usage” of sollen in German qualifies this item as an evidential. With reference to Diewald and Smirnova [to appear] he further argues that there may be an even more elaborate paradigm of evidentials in German. Second, he takes up the question of the possibility of embedding evidentials in subordinate clauses. He demonstrates that – contrary to mainstream assumptions – embedded reportive uses of sollen can be found in German, whereby a range of different meanings of this item in embedded clauses has to be taken into account. His conclusion is that embedded reportive sollen can be used only with certain types of embedding predicates, whereby a particular type of the matrix predicate licenses a particular reading of sollen. Communication predicates correspond to the assertive reading of sollen, (semi-)factive predicates license the global reading, and certain negative predicates embed sollen in its concord reading. Hatice Coúkun, in her paper on “Embedding indirective (evidential) utterances in Turkish”, focuses on the interpretative span of the highly grammaticalized evidential marker -mIs in Turkish. This marker in usually labelled as expressing “indirectivity”, i.e. an evidential category indicating that the propositional content of the utterance is reached in an indirect way (‘it appears to the recipient that X is the case’). The author takes up the question which kinds of subordinate clauses license evidential meaning in



Introduction

9

this marker. She demonstrates that – contra prior research, e.g. by Johanson 2000, 2006, according to which evidential sentence types in Turkic languages are restricted to asserted main clauses – in analytic types of embedding of subordinate clauses evidential readings arise in specific contextual constellations, whereby the semantic type of the matrix verb and other factors of the discourse-pragmatic type are essential for the actualization of different meanings. She suggests – similarly to Schenner – that there are five types of matrix predicates selecting finite complement clauses with indirective content, namely: communication verbs (e.g. de- ‘say’ or anlat‘explain’), cognition verbs (e.g. bil- ‘know, think’), perception verbs (duy‘hear’), attitude verbs (san-/zannet- ‘think, believe, consider’), evaluative or descriptive nominal predicates (belli ‘known’, iyi ‘good’). Beyond the result that evidential meanings may be embedded, this paper underpins the fact that very often meanings which elsewhere are realized by grammatical markers may arise due to what we might call “contextual collaboration”, i.e. the interplay of syntactic, morphological and semantic-pragmatic functions. Thus, although the author does not argue herself in this direction, the data and results presented in this paper strongly point towards the usefulness of constructional approaches in the interpretation of grammatical(izing) meanings. Gerda Haßler’s contribution is entitled “Epistemic modality and evidentiality and their determination on a deictic basis: the case of Romance languages”. Beyond a succinct recapitulation of descriptions of evidential systems in typological research this paper discusses the relation between epistemic modality and evidentiality and its various expressions in Romance languages. The author applies a dialogically enriched, non-canonical notion of deixis to identify evidential meanings and demonstrates her findings using examples from French, Spanish, Italian and Portuguese corpora. The author claims that while evidentiality is a deictic category, epistemic modality is not; and that therefore “recognising the deictic character of evidentiality and describing it as deictic phenomenon helps to determine the place of evidentiality within the language and provides a basis for distinguishing it from epistemic modality.” The contribution by Richard Jason Whitt on “Evidentiality, polysemy, and perception verbs: a corpus-based analysis of English and German” summarizes the results of a corpus-based analysis of evidential meanings signified by English and German perception verbs from the Early Modern period to the present, focusing on polysemy of such expressions on one hand and on the differences between German and English on the other. Beyond semantic issues the paper also addresses the importance of con-

10

Gabriele Diewald and Elena Smirnova

structions for particular readings and specific features of perception verbs, in particular with respect to the distinction between subject-oriented and object-oriented perception (ich schmecke das Salz – die Suppe schmeckt salzig). The author shows that verbs of perception in English and German are polysemous and can express a variety of evidential meanings, i.e. their readings range from direct/perceptual to inferential and to hearsay evidential values, some of which are restricted to specific construction types. Furthermore, the similarities and differences between German and English perception verbs are highlighted, as well as the great importance of perception verbs for the expression of the speaker’s epistemology and for the development of evidential markers. The role of perception expressions within the domain of evidentiality is also the topic of the next paper. Francis Grossman and Agnés Tutin in their paper “Evidential markers in French scientific writing: the case of the French verb voir” analyze corpus data of French academic writings in linguistics and economics, thereby focusing on the different uses of the French verb voir. Different construction types of this perceptual verb are taken into account. The authors propose that voir displays five clearly distinct usages: (i) voir as a “statement marker”, where it has the clear evidential inferential meaning which can be paraphrased as “inference based on observation”; (ii) voir as a “reference marker”, where it refers to sources of knowledge and to textually present additional material; (iii) voir in the sense of ‘to examine’, which is not evidential; (iv) voir as a judgement – and non-evidential – verb; and (v) other non-evidential usages of the verb. The authors concentrate on the first two usages of the verb and their distribution and particular functions in linguistic and economic scientific texts. As voir is often used to specify the source of information in scientific texts, it is said to have an important role in validating research work. Admitting that the core semantics of the verb voir is its perceptual meaning, the authors argue that – used in academic writings – it may be used as an inferential evidential strategy, indicating the source of information the researcher has for making the claim and serving as a inter- and intratextual device to direct the reader’s attention to pieces of evidence. Shifting focus from written – monologic – language to spoken – dialogic – discourse, Bert Cornillie presents “An interactional approach to epistemic and evidentials adverbs in Spanish conversation”. Although epistemic and evidential adverbs in the majority of uses in written language have the function of qualifying the proposition according to factuality values (in the case of epistemic markers), or according to information sources (in the case of evidentials), they display a variety of additional, interactional func-



Introduction

11

tions in conversational contexts. Using corpus data on the epistemic adverb(ial)s a lo mejor, igual, quizá, tal vez (all menaing ‘maybe’, ‘perhaps’) and on the evidential expressions aparentemente, evidentemente, obviamente, supuestamente, visiblemente, the study aims at classifying various discourse functions of epistemic and evidential adverbs. The author shows that in both groups the items can be distinguished according to frequency as well as the prominences of additional interactional functions they take in dialogic discourse. Of the four the epistemics a lo mejor and igual have the additional function of marking the hypothesis as subject to a confirmation or refutation by the interlocutor, of the group of evidential adverbs, obviamente and evidentemente display interactional functions in contributing to the on-line planning concerning the turn-taking mechanism (in signalling that the speaker wants to keep the floor). Adrienne Heijnen and Alexandra Kratschmer are concerned with “Revelative evidentiality in European languages: linguistic coding and its anthropological background”. Using data from biblical texts in Romance and Germanic languages as well as interviews and other data sources, they show the interrelations between linguistic coding of revelative evidentiality, which means “seeing in a dream”, on one hand and cultural and social traditions on the other. The authors treat revelative evidentiality as a subcategory of the – semantic – domain of evidentiality and propose to define it as follows: “information linguistically marked as created inside the mind of a subject without input from the outside world”. Using this definition, Heijnen and Kratschmer investigate which linguistic means (including grammatical as well as lexical elements) can be used in different languages to mark this semantic sub-category of evidentiality. For this they use large corpus data of biblical texts and interviews with native speakers of modern Icelandic, German and Italian. The authors conclude that there are no independent grammatical markers for revelative evidentiality in European Languages. Revelative evidentiality is often associated with visual evidential expressions and with inferential evidential expressions. Most often, however, there are lexical linguistic means to mark that information was obtained in a dream. With the help of their broad and valuable data material the authors are able to connect anthropological and linguistic questions, laying open the anthropological and cognitive foundations of linguistic structure. Furthermore, the authors raise methodological questions concerning the motivational and argumentative connection between language and culture, thus touching on the issue of linguistic relativity. Moreover, the authors propose that it is important to take the interactive aspect of evidentiality into account: they encourage typologist to investigate whether

12

Gabriele Diewald and Elena Smirnova

languages have different markers for evidential meanings like ‘overhearing something’ vs. ‘(inter-)actively being told’. With this array of studies the present volume makes an attempt to show that, if one is concerned with the notion of evidentiality, European Languages should not be excluded and are worth investigating. European Languages with their different ways of the linguistic realization of evidentiality often display not only lexical means for expressing evidential meanings but also have grammaticalized markers and even tightly organized grammatical paradigms which are in line with other typologically acknowledged evidential systems. By presenting theoretical considerations and by providing empirical evidence the papers of this volume aim at establishing a coherent view on the notion of evidentiality. Namely, they all emphasize that evidentiality should be seen as a semantic-functional (conceptual) domain which is not restricted to grammatical(ized) markers but can be realized by different linguistic expressions. Furthermore, in claiming that evidentiality and epistemic modality are two categories which are largely independent from each other, though often intertwined in individual languages and in individual expressions, this volume contributes to a better understanding of this particularly complex issue.  Notes 1.

2. 3.

The workshop and this publication were funded by the Fritz Thyssen Stiftung as part of the research project “Evidential markers in German”. The work of the editors on this topic was furthermore supported by the Belgian Federal Grant P6/44 within the program of interuniversity attraction poles (IAP). The editors express their gratitude to the sponsors. Cf. the workshop discussion platform under www.gabrielediewald.de In other places in the book, Aikhenvald clearly is aware of this difference, cf.: “Saying that English parentheticals are ‘evidentials’ is akin to saying that time words like ‘yesterday’ or ‘today’ are tense markers”. (Aikhenvald 2004: 10).

References Aikhenvald, Alexandra Y. 2003 Evidentiality in typological perspective. In Studies in Evidentiality, A. Y. Aikhenvald and R. M. W. Dixon (eds.), 1-31. Amsterdam/Philadelphia: John Benjamins. 2004 Evidentiality. Oxford: Oxford U.P.



Introduction

13

Information source and evidentiality: what can we conclude? In Evidentiality between lexicon and grammar, Squartini, M. (ed.), Rivista di Linguistica 19-1: 209-227. Anderson, Lloyd B. 1986 Evidentials, Paths of Change, and Mental Maps: Typologically Regular Asymmetries. In Evidentiality: The Linguistic Coding of Epistemology, Wallace Chafe and Johanna Nichols (eds.), 273-312. Norwood, NJ: Ablex. Bybee, Joan L. 1985 Morphology: A Study of the Relation between Meaning and Form. Amsterdam/Philadelphia: John Benjamins. Cornillie, Bert 2007 Evidentiality and epistemic modality in Spanish (semi-)auxiliaries: a cognitive-functional approach. Berlin/New York: Mouton de Gruyter. Dendale, Patrick and Tasmowski, Liliane (eds.) 2001 Evidentiality (Journal of Pragmatics 33, special issue). Amsterdam: Elsevier. Diewald, Gabriele 2004 Faktizität und Evidentialität: Semantische Differenzierungen bei den Modal- und Modalitätsverben im Deutschen. In Tempus/Temporalität und Modus/Modalität im Deutschen – auch in konstrastiver Perspektive, Oddleif Leirbukt (ed.), 231-258. Tübingen: Stauffenburg. Diewald. Gabriele and Elena Smirnova 2010 Evidentiality in German: Linguistic Realization and Regularities in Grammaticalization. Berlin/New York: Mouton de Gruyter. Friedman, Victor A. 1986 Evidentiality in the Balkans: Bulgarian, Macedonian, and Albanian. In Evidentiality: The Linguistic Coding of Epistemology, Wallace Chafe and Johanna Nichols (eds.), 168-187. Norwood, NJ: Ablex. 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. 2003 Evidentiality in the Balkans with special attention to Macedonian and Albanian. In Studies in Evidentiality, Alexandra Y. Aikhenvald and R.M.W. Dixon (eds.), 189-218. Amsterdam/Philadelphia: John Benjamins. Guentchéva, Zlatka 1993 La catégorie du médiatif en bulgare dans une perspective typologique. Revue des études slaves 65(1): 57-72. 2007

14

Gabriele Diewald and Elena Smirnova

Le Médiatif en Bulgare. In L'énonciation médiatisée, Zlatka Guentchéva (ed.), 47-70. Louvain/Paris: Peeters. Haan, Ferdinand de 2001 The relation between modality and evidentiality. In Modalität und Modalverben im Deutschen, Marga Reis und Reiner Müller (eds.), 201-216. Hamburg: Buske. Johanson, Lars 2000 Turkic indirectives. In Evidentials: Turkic, Iranian and neighbouring languages, Lars Johanson and BoUtas (eds.), 61-87. Berlin/New York: Mouton de Gruyter. 2006 Indirective sentence types. Turkic Languages 10: 72-87. Lehmann, Christian 1995 Thoughts on Grammaticalization. München/Newcastle: LINCOM Europa. 2002 New reflections on grammaticalization and lexicalization. In New reflections on grammaticalization, Ilse Wischer and Gabriele Diewald (eds.), 1-18. Amsterdam/Philadelphia: John Benjamins. Matras, Yaron (ed.) 1995 Romani in Contact: The history, structure and sociology of a language. Amsterdam/Philadelphia: John Benjamins. Mortelmans, Tanja 2000 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.), 131-149. Amsterdam/Piladelphia: John Benjamins. Plungian, Vladimir A. 2001 The place of evidentiality within the universal grammatical space. Journal of Pragmatics 33: 349-357. Willett, Thomas 1988 A cross-linguistic survey of the grammaticalization of evidentiality. Studies in Language 12: 51-97. 1996

Types of verbal evidentiality marking: an overview Vladimir A. Plungian1

The paper provides a brief overview of verbal systems with grammatical marking of evidentiality, based on previous case studies and generalizations proposed in the vast literature on the subject. The following topics are discussed: definitions of evidentiality as a verbal grammatical category; areal distribution of evidentialitymarking systems; the history of linguistic studies in evidentiality; possible crosslinguistic classifications both of evidential values and evidential systems; and the relation of evidentiality to other verbal categories (mainly, person and modality).

1. Introduction 1.1. Goals and outline The primary goal of the present study is to provide an overview of the main types of existing systems where evidentiality is grammatically marked as a verbal category. This goal naturally requires a certain limitation of the material, which means that, as will be shown below, one can by no means expect an all-embracing description of the problem of evidentiality. We will primarily be concerned with the possible set of semantic parameters that are characteristic for evidentiality as a universal-linguistic grammatical category (or “cross-linguistic gram type” in terms of Bybee and Dahl 1989 or Bybee et al. 1994). Furthermore, evidentiality is construed here as a verbal category (i.e. as a category expressed by means of morphological or analytic modifications of the verbal forms). Not all generally possible types and means of expression of evidential values are accounted for with this definition. Thus, the various means of lexical expression of evidentiality in the languages of the world are not taken into consideration (for a discussion of this problem, which in the last time began to attract much more attention of typologists, see, for instance, the recently published edited volumes by Squartini 2007 and Wiemer and Plungjan 2009, as well as Squartini 2008 and Wiemer 2010). Similarly, also those markers will be

16

Vladimir A. Plungian

ignored which cannot be clearly analyzed as modifiers of the verb (for the possibility of expressing evidential values on nouns by means of different types of “sentential” markers whose categorial membership is not entirely clear, see, for example, Aikhenvald 2004; the problem needs further research). However, the remaining material, i.e. that in which evidentiality occurs as a grammatical category of the verb, is still extensive, so that in the present overview we may concentrate only on some basic phenomena that seem to be particularly important and interesting for the problem in question. There are two main questions to which we would like to provide some answers: (i) possible types of grammatical values that may be found within the universal semantic domain of evidentiality, and (ii) possible types of grammatical systems that are used for expressing grammatical values (grammemes) of evidentiality in the languages of the world. The discussion will be structured as follows. In Section 1.2 a preliminary definition of evidential values will be introduced. Furthermore, we will discuss the problem related to the distinction between phenomena that can be related to evidential functions on the one hand and those that have to be regarded as part of other functions on the other hand. (Since the distinction between evidential values and some other values is rather difficult, we will address it again at a more advanced stage, in Section 4). There will also be a brief overview of those languages in the world in which evidentiality is expressed grammatically (Section 1.3). Later, in Section 1.4, a brief outline of the history of the study of evidentiality that covers the main stages will be given. Sections two and three form the central part of the present study. Section two includes a detailed discussion of the inventory of the main types of evidential values and different suggestions concerning their classification (including our own ones, which are partly based on the works by Plungian 2000: 321-325 and Plungian 2001). Section 3 deals with a typology of evidential systems (taking the suggestions made in Aikhenvald’s 2004 book into consideration, with some modifications). Section 4 offers a detailed discussion of the interrelation between evidentiality and two other grammatical categories: person (Section 4.1) and modality (Section 4.2). The distinction between evidentiality and modality is, as it is known, difficult to be drawn and often disputed in the typological literature. In the concluding Section some further tasks of the typological description of the category of evidentiality will be formulated. Since the present study offers a generalization of the material collected by other researchers in the first way (and thus represents the result of an “indirect access” to the sources of information in terms of the category of



Types of verbal evidentiality marking

17

evidentiality) we will include a minimal quantity of linguistic examples only, restricting our references mainly to those works in which the material of the respective languages was analyzed by specialists in these languages. Our main goal is the documentation of the history of research and, related to it, the “history of ideas” on evidentiality, but also a discussion of the problem of finding the typologically most adequate universal classification of markers of evidentiality.

1.2. What is evidentiality: a preliminary characterization According to the common view in current grammatical typology, markers are related to the semantic domain of evidentiality if their basic meaning is the indication of the “source of information” of the situation spoken about. In other words, while using a marker of evidentiality speakers communicate how they came to know about what they are talking about. If the system of such marking is obligatory (for instance, if it occurs as a grammatical category of the verb) speakers have to indicate on which basis they make an assertion about a respective situation with each use of a verbal form. To be more precise, one could say that in such cases the speakers indicate in what way they had access to the information referred to in a particular speech situation. As we shall see below when discussing the classification of evidential values, the conception of “access” appears to be somewhat more elastic than the conception of “knowledge” or that of an “acquisition of knowledge”, although they are very close to each other. However, the conception of access places more emphasis on the way in which a situation is perceived than on the mere epistemic aspect of the problem, and this obviously corresponds better to the way in which the category of evidentiality is organized in the languages of the world. From this point of view, the title (or better: the subtitle) of one of the most wellknown edited volumes on the typology of evidentiality, Chafe and Nichols’ (1986) The linguistic coding of epistemology, appears to be somewhat imprecise since the problem of evidentiality is more related to the way in which an individual perceives the reality surrounding him than to the problem of epistemology as such. In order to illustrate what evidentiality is we will reproduce the wellknown example by Aikhenvald from a language which is usually considered one with a maximally differentiated expression of the values of this category, namely the languages of the Tucano family (Amazon basin).2 According to Aikhenvald (2004: 52) the meaning of the phrase ‘the dog

18

Vladimir A. Plungian

stole the fish’ can (or better: must) be translated into Tucano by means of at least one of four possible constructions, depending on the way in which the speakers had access to this information: (1)

Tucano [Aikhenvald 2004: 52]; evidential markers are highlighted a. diâyï wa’î-re yaha-ámi ‘the dog stole the fish’ (I saw it) b. diâyï wa’î-re yaha-ásƭ ‘the dog stole the fish’ (I heard the noise) c. diâyï wa’î-re yaha-ápƭ ‘the dog stole the fish’ (I inferred it) d. diâyï wa’î-re yaha-ápï’ ‘the dog stole the fish’ (I was told)

All sentences listed in example (1) exhibit the order SVO (therefore the agent is unmarked, but the patient is marked by means of the special suffixal marker -re, which Aikhenvald calls the “topic non-subject case”). As we can see, the way in which the information was accessed is clearly reflected in the morphology of the verb. The four different grammemes of evidentiality distinguished in Tucano represent rather typical instances of evidential values (they will be characterized more in detail in Section 4). They either indicate that the speakers themselves directly observed the situation as viewers (a), or that the speakers observed the situation themselves and directly earlier, but not as viewers, i.e. they did not see it, but only heard what was going on (or perceived it in any other way) (b), or that the speakers had no direct access to the situation, but observed some facts which they interpret as having caused a particular situation, i.e. based on what they observed they can assume that the situation took place (c), or, finally, that the speakers themselves had neither access to the situation, nor to its consequences, but somebody else informed them that the situation took place (d). A translation of the various sentences in example (1) into a language without grammatical markers of evidentiality would require the use of lexical equivalents in order to express the respective meaning. The most typical lexical equivalent to the meaning in (a) is the construction ‘I saw that/how P’ or ‘P in my eyes’; an equivalent to the meaning of type (b) is ‘I heard that P’; one expressing the meaning of type (c) is ‘obviously/one may notice that P’; and one to type (d) – ‘it is said that/it can be heard said that P or through the ears, P’. As lexical units such expressions are not very frequent. They are used only in such cases in which the speaker explicitly



Types of verbal evidentiality marking

19

wants to emphasize the respective meaning. As already mentioned, grammatical markers of evidentiality must occur with all forms of the verb (or at least with the majority of verbal forms marked for any other category, such as all finite forms, all forms of the past tenses, etc.), often independently of the speaker’s intention.

1.3. Evidentiality in the languages of the world: the main areas The obligatory indication of the source of information about an uttered fact in a linguistic system might seem entirely exotic to speakers of the majority of the European languages. Nevertheless, in a rather large number of languages on earth this strategy forms part of the grammatical rules of the use of verbal forms. Moreover, the number of languages in which evidentiality is grammaticalized is, as studies of the past years have shown, much larger than it is often assumed. Nowadays, evidentiality is not regarded as a rare and unusual category, as it was the case in the period following its “discovery” by European linguists. It will perhaps soon be more correct to say that the lack of grammatical marking of evidentiality in the majority of the languages of Western and Central Europe is an important linguistic particularity of this area, which distinguishes it from many others. Currently, the following linguistic areas are known as agglomerations of languages in which evidentiality is expressed grammatically with verbs:3 A. The extended Euro-Asian area, including the Balkan languages (except for Modern Greek), Asia Minor and, from there, a broad strip going to the Far East over the Caucasus, Southern Asia, the Volga district, and Southern Siberia; in the linguistic literature this area is usually called the “Great Evidential Belt” and considered to be the most significant geographical locality of evidentiality, both with respect to the size of the territory and the number of languages and their genetic diversity (the Great Evidential Belt includes Southern Slavic, Albanian, Iranian, Indo-Arian, Armenian, Kartvelian, Abchaso-Adygei, Nachsko-Daghestan, Turkish, Finno-Ugric, and some other languages). It should, however, be mentioned that the type of grammaticalized evidentiality which is mainly represented in the languages of the Great Belt belongs to the group of more simple evidential systems (the so-called binary type, cf. Section 3), that is, the area does by far not represent the full range of possible types and thus the full typological diversity. An exception are the Tibetan languages (especially those which will be discussed below), which have very unusual and com-

20

Vladimir A. Plungian

plex systems of evidentiality that are without direct analogues in other areas. The existence of evidential markers in the languages of the area sketched above has long been known in modern linguistics, and it is the profile of these languages that originally formed the point of departure for the „discovery” of the category of evidentiality in typology (next to the profile of the languages spoken in North America, see Section 1.4). Therefore, earlier descriptions of the grammatical system of the languages of the “Balkan type” usually did not treat evidentiality as an independent grammatical category, but analyzed evidential meanings as a special type of modality marking (or the marking of tense with a modal shade). Thus, the grammars of these languages referred to it by means of all-embracing categories of the type “non-obvious tenses”, “retell mood”, and so on. The motivation for such descriptions was based on the peculiar occurrence of the evidential markers in the languages of the Balkan: with respect to their expression they exhibit a close relation to markers of the perfect, regarding their content they display a close relation to the semantic domain of modality. This problem will be treated more in detail in Section 4.2. Evidentiality in the languages of the “Great Belt” has been described in many studies (beginning with the monograph by Haarmann 1970). An important role for a comprehension of the nature of these markers in the languages of the Balkan played the studies by Friedman 1986, 2000, 2003 et al. For the Balkan languages and other languages of the area see also the articles in the edited volume by Guentchéva 1996, especially those about Bulgarian (Demina 1959, Aronson 1967, Guentchéva 1993, 1996a, and Nizolova 2006). For the languages of Asia and the Caucasus cf. the specific works by Lazard 1957 (one of the first theoretical studies of evidentiality), Friedman 1979, Slobin and Aksu 1982, Tatevosov 2001, the edited volumes by Johanson and Utas 2000, Aikhenvald and Dixon 2003, Guentchéva and Landaburu 2007, Xrakovskij 2007, and others. For the languages of the Volga area, see Leinonen and Vilkuna 2000. B. Northern Siberia and Eurasia’s circumpolar zone. This includes, above all, languages of the Samodic group, which have a rather complex system of the grammatical expression of evidentiality, but also the ObUgric languages, Yukagir and some others. For a discussion of the particularities of the evidential systems in the languages of this area see, in particular, Perrot 1996, Nikolaeva 1999, Künnap 2002, Maslova 2003, Burkova 2004 and especially the edited volume by Xrakovskij 2007. As one can see, an intense typological study of these systems has started only during the past years. Unfortunately, it is accompanied by a similarly intense loss of



Types of verbal evidentiality marking

21

these languages, which results in a destruction of grammatical systems and a complete or partial loss of the category of evidentiality with the younger generation of speakers (see Il’ina 2002 and Kazakeviþ 2005). C. The Baltic Region includes, above all, Lithuanian and Latvian (and their dialects), which are Baltic languages of the Indo-European family, and Estonian, a Uralic language of the Balto-Finnic group. Evidentiality in the Baltic languages has, in all probability, a totally areal origin, although not all details of this process are entirely clear (see Koptjevskaja-Tamm and Wälchli 2001, Holvoet 2007: 81-105). Moreover, the system of the expression of evidentiality in Lithuanian exhibits a number of differences from that found in Latvian and Estonian, which are both more closely related to each other, and it generally seems to be somewhat more complex although it is, at present, also less grammaticalized than in the two other languages (for further details cf. Wiemer 2006, Wiemer 2007 and Holvoet 2007). Typologically, the systems of evidential marking found in these languages resemble the one used in the Balkan languages, both with respect to their form (use of participle forms of the verb, see Wälchli 2000) and content (the systems of evidentiality in the Baltic languages also belong to the “binary” type). However, there are also some important fine differences between the two (see Kehayov 2002, 2009, and Wiemer 2006). D. The languages of North America, Central America and South America. The American continent is one of the richest areas with respect to the types and means used to express evidentiality. Furthermore, the most complex grammatical systems of the expression of evidentiality with verbs that we know at present are located in this area (as such one may consider the systems found in the languages of California, particularly in those of the Wintu and Pomo families). Systems of evidentiality of different types (all of them including, as a rule, a large number of grammatical oppositions) are found also in the Salish and Wakashan language family and in other languages of the North-Western coast of North America (including Eskimo languages) and in a number of other areas, such as in the languages of the Caddoan family, the Muskogean family, and others (Sherzer 1976, Mithun 1999: 181-186). In South America evidential marking is particularly widespread in the languages of the Amazon Basin (Aikhenvald and Dixon 1998), which are equally complex and diverse with respect to their structure (see the example from Tucano above). Evidentiality is also well attested in the neighboring Andean languages of Colombia, Ecuador, Peru and Bolivia, in various dialects of Quechua and Aymara, in the Barbacoan languages, and others.

22

Vladimir A. Plungian

One should not forget that the knowledge of evidentiality as a special category of the verb (the peculiarity of these languages seems to reflect cultures and ways of thinking unknown to the Europeans) acquired by the European linguists began with the study of the American Indians in the works of Boas and Sapir (although the languages of the Balkan type had, of course, been known earlier in Europe). Also the term “evidentiality” originated from studies of the languages spoken in this area. More precisely, the beginning of the typological study of evidentiality in the 1980s in general turned out to be linked with the languages of the American Indians basically. Therefore, the importance of the languages of the American continent for the study of evidentiality cannot be overestimated. Except for the studies of the languages of North and South America referred to above we would also like to mention the edited volume by Aikhenvald and Dixon 1999. Specific aspects of evidential systems in these languages are also discussed in many articles included in the widely known volume by Chafe and Nichols 1986, in the study by Willett 1988 and, most importantly, in the edited volumes by Aikhenvald and Dixon 2003 and Guentchéva and Landaburu 2007. E. Australia and New Guinea. The languages spoken in this area have, without any doubt, markers of evidentiality, which exhibit different degrees of grammaticalization. However, the form of evidential marking is entirely different in the various regions. As it seems, the most complex of all the systems that are attested at present are located in New Guinea, especially in the languages of Southern Nagaur (Foe, Faso, and others; see Foley 1986: 166, Aikhenvald 2004: 62-63, 293). Next to these, there are also languages with simpler systems in New Guinea and in Australia, among them “binary” systems, and in many languages the grammatical expression of evidentiality is entirely absent. In the Australian languages in general evidentiality is weakly grammaticalized and usually expressed by means of a number of adverbial (or phrasal) units whose meaning often includes more than just a clearly identifiable element of evidentiality. The languages of New Guinea, on the other hand, are characterized by rarely occurring systems of the expression of evidentiality with untypical grammemes, a fact that requires a more detailed study. The presented survey did not include all linguistic areas in which evidentiality can be found, but all those in which the grammatical expression of evidential values is most intensely used and/or in which it is highly grammaticalized. Areas that are comparatively poor with respect to the expression of evidentiality are, for instance, Oceania (where evidentiality is attested in some single Philippinean languages only), but also tropical



Types of verbal evidentiality marking

23

Africa, although the problem of the expression of evidential values in the verb system of the African languages requires further investigation. Probably, the suggestion that there is no evidentiality in the African languages (Aikhenvald 2004: 291) will have to be corrected in the future. At present, the grammatical expression of clearly definable evidential values can be found only in some few African languages (among them Dogon, spoken in Mali, which I studied some time ago and which is one of the central dialects that have perfective forms with clearly inferential components, see Plungian 1988). However, some of the grammatical strategies that are broadly spread across Africa and which are related to the phenomenon in question deserve much more attention. Among them are, above all, the highly frequent use of logophoric pronouns and the category of temporal distance that occurs in various forms in the verbal systems of many languages. Indeed, logophoric pronouns, which indicate whether the subject of the speech (i.e. the speaker) and the subject of the governing verb of saying in a proposition coincide or not, may, in some specific contexts, be used (similar to the German Konjunktiv) for the indication of whether a given fragment of a text is the reproduction of the words uttered by a third individual and, consequently, whether given information is known to the speaker through the words of some other speaker. The close relation between logophoric markers and markers of evidentiality has been noticed in a number of typological studies (see e.g. Dimmendaal 2001 and Aikhenvald 2004: 132 ff.). Concerning the category of temporal distance, whose use (especially in discourse) has by far not been studied sufficiently, one may observe that in many African languages the verb system is characterized by a special marking of events that go back far in the past and which are not directly accessible to the perception of the speaker (for further details see Dahl 1985: 120-128, Aksenova 1997, Nurse 2008: 80-124). A relation to evidential values seems to be very likely in this case or at least deserves some more detailed investigation. For the important semantic potential of many forms that are traditionally labelled “discontinuous past tense” see also Plungian and van der Auwera 2006.

1.4. From the history of the study of evidentiality As it has already been said, in the tradition of “western” linguistics the grammatical category of evidentiality was clearly recognized only during the second half of the twentieth century. The first typological studies of this category began to be conducted towards the end of the 1960s until the

24

Vladimir A. Plungian

beginning of the 1980s. A more intense study of evidentiality in various languages of the world began even later, towards the end of the 1980s. This clearly distinguishes evidentiality from other typologically important verbal categories, which were known in the European grammatical tradition much earlier, in some cases (such as with tense or mood) as early as in the antiquity. Nevertheless, the history of the study of evidentiality is not that short that one could represent it in a nutshell. As it is often assumed, the first occurrence of evidential oppositions in forms of the past tense of Turkish verbs was noticed and analyzed already in the well-known “Collection of Turkish languages” (Arabic: DƯwƗnu LuƥƗto t-Turki) by Mahmud alKashghari in the eleventh century (see Friedman 2003: 189), with comments by Robert Dankoff, the English translator of the DƯwƗnu (Dankoff 1982). Another pioneer of the modern study of evidentiality was, undoubtedly, the French folklorist and poet Auguste Dozon, who at the end of the nineteenth century focused on special forms of the Albanian verb (see e.g. Dozon 1879). In order to describe them he introduced the term ‘admirative’ (French admiratif), which came to be widely used. Dozon took the emotional evaluation of an uttered fact as the basis of meaning (therefore the unusual internal form of the term, which indicates the meaning of “affection”), which is indeed clearly observable in the evidential forms of the Albanian verb, together with more usual inferential and reportative values that form one of the distinctive features of Albanian (for further details see Friedman 1986, 2000, 2003; Duchet and Përnaska 1996). Considering this shift of the semantic focus on modal meanings of evaluation, one may conclude that the observations by Dozon were highly important and had a strong impact on the terminology and the direction of subsequent investigations in the tradition of Balkan philology. Apparently, these two names – Mahmud al-Kashghari and Auguste Dozon – also represent the two most important landmarks of the history of the study of evidentiality. However, the establishment of this category as a research object in modern linguistics begins, as it is usually claimed, with Franz Boas. Even the term “evidential” itself goes back to Boas,4 as well as the first exhaustive description of the functioning of this category in Kwakiutl (Wakashan family) and a number of other languages of the NorthWestern coast, which was carried out at the beginning of the twentieth century. In subsequent works Boas repeatedly focused on the role of evidential meanings in what today we could label “folk semantics” (or “naïve conceptual system”) characteristic of American Indian languages. A crucial element in Boas’ argumentation was the grammatical (that is, obligatory)



Types of verbal evidentiality marking

25

character of the expression of evidential values within the verb systems of many North American languages, a phenomenon that, from his point of view, formed a marked difference between American Indian languages and languages like English concerning the grammatical strategies used to create forms of the verb. After Boas, evidential markers were discovered in and described for many languages of the American Indians, particularly in the works of Edward Sapir, Morris Swadesh and other specialists in American languages. The terminology was certainly not entirely uniform and fixed, and a conception of these markers as a special type of “mood” predominated (similarly, in earlier works in Bulgarian philology the phenomenon is often referred to as the “renarrative mood” of the Bulgarian verb). However, one may state that the specific grammatical properties of evidentiality in this period had been understood rather well, although the position of this category among other grammatical categories of the verb and its typological status in general were not subject to more detailed analyses. This was, however, a natural consequence of the fact that in the first half of the twentieth century a conception of a typology of grammatical categories in the truest sense of the word had not been established yet. The turning point in the history of the typological study of evidentiality was the year 1957 when, at the same time and independently of each other, two works appeared, both determining the direction of subsequent studies for a long time. One of them was the seminal article by Jakobson (1957), in which he introduces the term “shifter” and, based on it, the first fundamental classification of verbal categories (with an illustration of their different properties on the basis of data from Russian). The other, not less important (though, perhaps, less known) work was an article by the French specialist in Iranian (and later one of the leading French typologists) Gilbert Lazard 1957, which was highly relevant for a conception of the problem. In this article the author studied the verbal system of Tadjik (including the expression of the category of evidentiality in this language, which he particularly focused on). Both studies do not exclusively focus on evidentiality, but in each of them particular aspects of this category play an important role. In Jakobson’s article evidentiality is one of the examples used to illustrate the peculiar type of verbal categories, which include in their meaning the joint reference to what he labelled “narrated event”, “speech event” and “narrated speech event”. In his own words, “[t]he speaker reports an event on the basis of someone else’s report (quotative, i.e. hearsay evidence), of a dream (revelative evidence), of a guess (presumptive evidence), or of his

26

Vladimir A. Plungian

own previous experience (memory evidence)” (Jakobson 1971: 135)5. Furthermore, Jakobson did not only establish the term “evidentiality” as one denoting that grammatical category which indicates the source of information of the speaker’s utterance, but also integrated the category into the broader typological context, establishing its place next to a number of other verbal categories. Thus (and probably for the first time), various phenomena found in typologically different languages could be described under one label, that of “evidentiality”. Generally speaking, one could say that it was now possible to find similarities, for example, between the verbal system of Bulgarian and that of Kwakiutl, both of which are constructed in an entirely different way otherwise. Thus, we owe it to Jakobson that a general label was found for phenomena that occurred in different languages, but which until that time had not been considered as similar in nature. One may regard this moment as the beginning of the typological study of evidentiality. Furthermore, it was in the works of Jakobson that evidentiality was first clearly distinguished from modality. This distinction had not been drawn in such a clear-cut way before, neither in the tradition of the Balkan linguistics nor in American linguistics. The article by Lazard is entirely different in character. It is a detailed analysis of the verbal system of Tadjik, which has grammaticalized markers of evidentiality. However, the way evidentiality is expressed in Tadjik (which belongs to the zone of the “evidential belt” of Eurasia) differs fundamentally from the one observed by Boas and his successors in the languages of the North American Indians: it is a binary system (see below) in which evidential markers are highly polysemous. For the description of the semantic invariant of such markers Lazard suggested the term “meditative” (French médiatif), which is not entirely identical with Jakobson’s “evidentiality”: it denotes any form of “indirect” reflection of a situation, which is the case when the perception of situation is not based on direct, personal experience of the speakers or on their own conception of the world. During the following decades, Lazard’s suggestions were not immediately acknowledged in linguistic typology. However, through the accumulation of linguistic data and a deeper understanding of evidential systems in the languages of the world it became clear that Lazard’s intuitions were in many respects accurate, and the term “meditative” came to be widely used for the description of systems of the Balkan type, especially in the French linguistic tradition (see e.g. Guentchéva 1993, 1996, but also later works of Lazard himself: Lazard 1996, 1999 and 2000). The publication of Jakobson’s works gave a new impulse to the study of evidentiality in various languages, and the period from the beginning of the



Types of verbal evidentiality marking

27

1960s to the middle of the 1980s was characterized by a continuous accumulation of new findings and the publication of a remarkable number of works. In these works, the authors intended to provide a deeper and more precise description of systems of evidentiality found in individual languages and to establish a more comprehensive typological conception of evidentiality, i.e. one that was not restricted to a particular tradition, but which accounted for all languages in the world that were known at this time. Some of the most important studies of this period are Aronson 1967 (elaborating on Jakobson’s ideas of the Bulgarian system of the expression of evidentiality), Haarmann 1970 (one of the first monographs on evidentiality in the languages of the Great Evidential Belt), Friedman 1979 (analyzing data from the Caucasian languages), Hardman 1981 and Barnes 1984 (both providing a detailed analyses of data from South American languages), Slobin and Aksu 1982 (one of the first detailed descriptions of evidentiality in Turkish), and Givón 1982 (one of the first theoretical works accounting for the relation between evidentiality and modality). The studies of this period laid the foundation of the publication of the famous edited volume by Chafe and Nichols 1986 and, closely related to it, the article by Willett 1988, in which the material from the edited volume was systemized and commented by the author. Until the present day, this edited volume is one of the most frequently cited books with respect to evidentiality. It does not only include a collection of very interesting and reliable data from languages of different linguistic areas (although the languages of North America predominate), but also proposes a reasoned classification of evidential values in the languages of the world (which was supplemented in Willett’s article). Practically all subsequent studies were based on this classification in one way or the other, and many ideas suggested in the edited volume and in Willett’s article were taken up later by other authors who wrote about the classification of evidential values (for further details see Section 2.2). The following years were characterized by a rapidly growing interest in the category of evidentiality (which, strictly speaking, ceased to be “exotic” after the edited volume by Chafe and Nichols). It was a period in which more data were accumulated and in which more attention was given to the languages of the „Old World”. Such a shift of the focus was only to be expected in view of the fact that these languages were given much less room in the edited volume by Chafe and Nichols. However, the universal classification of evidential values proposed in the volume could now be applied to the new material, both to new and to known data. It does not come as a surprise that in this period a number of edited volumes dealing

28

Vladimir A. Plungian

with evidentiality in the languages of the Old World were published, such as Guentchéva 1996 and its continuation Guentchéva and Landaburu 2007, Bo and Utas 2000, but also Dendale and Tasmowski 2001 (published as a special issue of the Journal of Pragmatics). Each of these works does not only provide a collection of new data, but also attempts to make a contribution to the theory of evidentiality and the classification of evidential values in general. These attempts often end up in more or less polemic statements with reference to Willett’s classification, which had drawn largely on material from the languages of the New World. Especially important with respect to theoretical aspects are the re-introduction of Lazard’s “meditative” (or “indirective” in the terminology of Johanson and others) as a cover term for the semantic cluster formed by evidential markers in binary systems, and also the more profound discussion of the relation between modality and evidentiality (we will return to this problem in Section 4.2). All these discussions are characterized also by a renewed interest in the conception of admirativity, which some authors exclude from the semantic domain of evidentiality, whereas others do include it (see the different points of view in De Lancey 1997, 2001, Lazard 1999 and Xrakovskij 2007a; for the modal component of evidential values see especially de Haan 1999 and 2005, Xrakovskij 2007a, but also van der Auwera and Plungian 1998). The results of the study of evidentiality at this stage were analyzed in two recently published works of more general nature: the edited volume by Aikhenvald and Dixon 2003 and the monograph by Aikhenvald 2004. Currently, the latter can be regarded as the most comprehensive guide through the history of the study of evidentiality and the semantics of this category in the languages of the world. Further research on evidentiality will largely draw on this book. In the present overview we will repeatedly refer to Aikhenvald’s views and also include those ideas into the discussion which appear debatable to us.

2. Classification of evidential values 2.1. The main types of opposition As it was shown above, the generalization of existing data of the category of evidentiality suggests that the main opposition inherent in the various evidential values is based on the types of access to the information uttered by speakers, i.e. the means by which the speakers got to know about the



Types of verbal evidentiality marking

29

situation they speak about. This opposition can be of two kinds. On the one hand, one may distinguish between direct (firsthand, witnessed) and indirect access to information. Direct access refers to any means of obtaining information that presuppose a direct perception of a situation by the speakers (mainly visual, but not necessarily) and/or a direct participation of the speakers in a situation. The difference between these two types of direct access is needed because not all forms of direct participation in a situation necessarily involve its observation “from the outside”. Moreover, sometimes occurrences that are described from the “first person” perspective are generally not observable, for instance in the case of personal physiological sensations and emotions. These two types of direct access may be expressed by means of different grammatical constructions in different evidential systems, which will be discussed more in detail below. Accordingly, an indirect access to information presupposes that the knowledge about a situation was obtained through other means: the speaker did not observe the situation in a direct way and did not participate in it. On the other hand, one may distinguish between a personal and a nonpersonal access to information. This opposition is not identical with the preceding one, although it seems to be very similar to it. A personal access to information presupposes that the speakers know about a situation on the basis of facts that they personally got to know, whereas a non-personal access to information presupposes that the speakers received their knowledge, roughly, through a report in somebody else’s words. Thus, the source of a retold assertion may be either known or unknown to the speakers. The non-personal access to information forms the basis of the semantics of one of the most widespread types of evidential markers, which, in accordance with current practice, is called ‘reportative’ (one may also find the terms ‘renarrative’, ‘quotative’, ‘hearsay’, and others; for more details on the typology of these markers, see especially Wiemer to appear). For a significant number of situations the conception of direct/indirect and personal/non-personal access appears to be identical: a direct access is, for obvious reasons, always personal, and an indirect access is nonpersonal. (Note that a non-personal access is, per definition, also indirect.) However, these features may also be independent of each other. Indeed, a direct (and personal) access presupposes that the speakers perceived the situation which they speak about themselves (in person) and in a direct way. A non-personal (and indirect) access presupposes that the speakers, talking about a situation, base their utterance on the information of another person. Another possible combination is the one of an indirect and personal access to the situation. In this case the speakers obtained

30

Vladimir A. Plungian

knowledge of a situation themselves, without other persons being involved, but the knowledge of this situation has not been obtained in a direct way since the speakers did not observe the situation directly. They may, for instance, either observe any results of a situation and take these as the basis for the conclusion that it took place or argue on the basis of some other data. In any case, the speakers talk about a situation which they did not observe personally and in which they did not take part, but about which they have some evidence that allows them to assume that it took or takes place. In (1) this type of access to a situation is reflected in (c): it may be uttered by speakers if, for instance, they do not see the fish on the table, but see how the dog guiltily has its tail between its legs. Such values are usually called inferential (or inferentive), since the speakers are required to draw a logical conclusion (an inference) on the basis of observed results. Another widespread type of meaning of a personal indirect access is the so-called presumptive value, which refers to cases where speakers produce an utterance in which they refer to a situation about which they do not know through concrete observed results, but through their knowledge about particular cause-and-effect relations: compare the context of the form at this time of the year the berries should already be ripe, where the utterance about the ripeness of the berries is not based on direct observation, but, so to speak, on the natural order of things. The difference between inferential and presumptive contexts may also be illustrated in the following example. One and the same utterance, such as The neighbour is already at home, may, in a language that has the respective set of grammatical means, be produced either by using presumptive or by using inferential markers of evidentiality. The first would be used for instance in the context of eight o’clock - the neighbour should already be at home (if the speakers know that this is time when the neighbour usually comes home). The second construction would be used in the context of the light is on in the house – it seems that the neighbour is already at home (in this case the speakers conclude from an observed situation, which they interpret as evidence enough for making the utterance). In this sense, the main difference between presumptive and inferential markers is that the first are based on the speakers’ knowledge about the world and their capacity to draw logical conclusions, whereas the second are based on direct observation by the speakers (and, indirectly, on their capacity to draw logical conclusions from it). For more details on this problem, see also Tatevosov 2003. From the two main oppositions characterized above the one between the direct and the indirect access proves to be hierarchically (and typologically) more important. The indirect access stands for a very broad range of



Types of verbal evidentiality marking

31

ways to get access to information, but excludes a direct synchronous perception of a situation or participation in it. One of these ways may be a logical conclusion from an observed result, or a reasoning based on general properties of the world, or a reproduction of what other people said. In many languages, this variety of types of indirect access is reflected in the existence of finer grammatical distinctions: here, specialized markers of the inferential, presumptive, reportative and other specific categories are possible. However, in the majority of languages a different strategy can be observed, namely the use of one single marker with a diffuse meaning for the expression of the indirect access to a situation. Its basic value includes only the indication of the fact that the speakers did not have direct access to a situation, but that they got informed about it anyhow, or perhaps things cleared up (if they did at all) through the context or through pragmatic factors. Exactly this uniform type of evidential markers, i.e. the one which indicates indirect access with a wide range of values (which are, however, conceptually alike), was the type discussed by Lazard 1957 under the label “meditative” (French médiatif), which was also called “indirective” in his studies on the Turkish languages, “non-confirmative” in many studies of the Balkan languages, and in many other ways. Aikhenvald 2004 uses the term “non-firsthand” for values that are very close to the one discussed above. Since languages with this type of evidential markers form, as already said, the majority of all languages spoken on earth, the number of terms that are used to capture their meaning and that arose in various linguistic traditions independently of each other is large as well. In the Russian linguistic tradition, the term zaglaznost’ (referring literally to what happens “behind one’s back”) has been used to refer to this type of uniform markers that indicate indirect access (especially in the works of Caucasian linguistics, see, e.g., Kibrik 1977 for Archi). During the past few years also the descriptive term “indirect evidence” (kosvennaja zasvidetel’stvovannost’) came to be widely used (see Kozinceva 1994 and 2007, Xrakovskij 2007a). Markers of indirect access, which have a broad range of values in the languages of the world, have, as a rule, one property in common: they do not only express evidential values, but also different types of modal values. In other words, their inherent evidential meaning is, as a rule, not separated from the modal meaning, more precisely, from the meaning of epistemic assessment (i.e., the assessment of the degree of certainty of what is said). The pragmatic basis for this proximity is evident: speakers tend to evaluate information that they did not obtain as the result of a direct synchronous observation of a situation as less certain or less reliable. A more careful

32

Vladimir A. Plungian

formulation would probably not make use of the term “incertainty”, but rather speak of “epistemic distance”: the speakers refrain from taking over responsibility of the truth value of an utterance since the respective information did not enter into their personal sphere. It should be noted that the relation between the values of indirect witness and of epistemic incertainty (or rather epistemic “distance”) is generally not extremely close: there are evidential systems in which the fact that the speakers had no personal access to the respective information does not necessarily result in a lower degree of certainty of this information. A useful criterion for the evaluation of this relation is the type of markers used in a given language to formulate utterances of the type “general truths” (i.e. those which are unconditionally taken for granted in a given community) or utterances that refer to well-known facts (for example, facts related to the past of a given community), i.e. types of information which speakers cannot observe personally. In languages with “meditative” evidentiality, though not only in these, it often occurs that the use of markers of indirect access is omitted in such cases (although the speakers has no direct access to the described situation) in order to avoid an unnecessary epistemic load of these markers. An analysis of data according to these criteria can be found e.g. with reference to Bulgarian in Guentchéva 1996a and in Nicolova 2007; cf. also Friedman 2000 for other languages with such semantic effects. However, in languages with more complex evidential systems markers of direct access often lack such epistemic load: the use of “indirect” evidential markers does not mean that the reliability of the description of a situation that has not been observed personally is in any way affected. Rather, these markers may be used also to describe situations known as true (this fact has, for example, been considered as characteristic for Tibetan by Tournadre 1996). Confer also Mithun 1999’s observation that in Central Yup’ik, in contrast to a number of other languages, markers of indirect access may render a verbal statement even more credible than markers of a direct access since their semantics does not include reference to subjective personal experience, but to more reliable collective experience (§‘This is not what I myself thought through, this is what everyone knows.’). Various aspects of this type of marking are discussed also in Aikhenvald 2004, where the term “epistemic extensions” is used to refer to this aspect of the semantics of evidential markers. Thus, generally speaking, assertions referring to common knowledge may be expressed by means of entirely different strategies in languages of the world: in some rare cases specialized markers are used in this case (such as in the Tibetan and the Pomo lan-



Types of verbal evidentiality marking

33

guages), in all other cases such strategies may be related either to the semantic domain of indirect access (if it does not have additional shades of epistemic uncertainty) or to the semantic domain of direct access (in contrast to its literal meaning). Before we will continue with a discussion of the different approaches to a typological classification of evidential values, we should add some important remarks on that component of the semantic domain of evidentiality which is related to a direct access of the speaker to a situation. Typologically, one may observe a rather large number of independent values within this domain, which in the languages of the world may be assigned a specialized grammatical expression. In other words, the fineness of the subdivision of this semantic domain in the universal grammatical space may be rather high. The most typical division of those markers which are related to the domain of direct access is the one into markers of visual and non-visual evidence. The first means that the speaker directly observed the respective situation as a viewer, whereas in the second case the speaker got to know about a synchronous situation not through visual perception, but through any other means (through acoustic, olfactory or tactile perception, for example). In some rare cases, one may find a more precise distinction within the domain of visual markers, which depends on whether the speakers observed the situation from far or whether they were in the immediate vicinity or in contact with the participants of a situation. Sometimes visual markers may also indicate whether the speakers saw a situation in a dream, although in a number of languages markers of indirect access are used in this case (cf. n. 4). However, reliable evidence for the idea that there must be a grammatical opposition of different ways of sensory non-visual perception (say: auditory and tactile) in the languages of the world is not available in the existing typological studies on evidentiality (cf. the data in Aikhenvald 2004: 63-64). In this sense, the term “auditive”, which is used in descriptions of evidential systems in some languages (e.g. the Samoyedic), is somewhat misleading: actually, in Samoyedic the different auditive forms refer both to auditory and to other types of non-visual perception of a situation through the speaker (for example, tactile or olfactory perception). For further details see, for instance, Il’ina 2002, Burkova 2004, Gusev 2007, Ljublinskaja and Mal’þukov 2007. Another well-known and important opposition within the group of nonvisual markers is that between sensory markers (referring to all emotions that are not related to visual perception) and the so-called endophoric markers (this term, which was suggested by Claude Hagège, was used for

34

Vladimir A. Plungian

the description of evidential systems in the works of Tournadre 1996, among others). Endophoric markers are used when speakers have to describe their own mental, emotional or physiological state. Such situations cannot be perceived through any of the five “external” senses, but may be accessed only on the basis of “internal” sensations, which shows that it seems to be a special type of direct access. Correspondingly, in languages with a system of endophoric markers expressions like I am hungry and you are hungry have different grammatical forms. The first one is expressed by means of a special endophoric marker (the speakers “know” about their own physiological sensations, because they feel them, although they do not perceive them visually or acoustically etc.), being in contrast to, for instance, markers of a direct-visual or direct-sensory access. The second expression is, however, realized with one of the markers of indirect access, depending on the concrete evidential system: the speakers cannot have equally reliable information about the physiological state of another person as they may have about their own state. They may only draw conclusion on this state on the basis of some indirect signs. The specialized expression of endophoric evidentiality is typical, above all, for the Tibetan languages, but occurs also in other languages of South-East Asia and the Far East, in which evidentiality is grammaticalized. However, next to visual and non-visual markers there is also another, but much more rarely attested type among the group of markers of direct access: the so-called “participatory” markers (also known as “performative”, “personal agency”, “constative”, and others with different authors). They are used in those cases in which the speakers themselves are participants of the situation which they speak about: they do not know about a situation because they observed it or perceived it in any way, but because they were directly involved in it. Participatory evidential markers, which do not coincide with visual or sensory markers, can be found above all in the Pomo languages (Oswalt 1986, Mithun 1999: 181), but also in Oksapmin and a number of other New Guinean languages (Loughnane 2007). The participatory value is close to the endophoric one, but, as far as one can conclude from existing descriptions, it is much more likely to occur with predicates denoting activities and less so with predicates referring to internal states of the subject. Moreover, these two types of markers of direct access can be regarded as the realization of one and the same strategy of obtaining information about a situation, namely one that is not based on an observation of this situation, but on the direct participation in it.



Types of verbal evidentiality marking

35

2.2. The main types of classification Many different classifications of all the evidential values that are possible in the languages of the world have been suggested in the literature, but two of them are most commonly known (and very similar to each other in various ways). The first one goes back to Willett 1988 and is, in principle, a generalization of empirical data presented in the volume edited by Chafe and Nichols 1986. Willett distinguishes between three basic types of evidential values: direct ones (subdivided into visual, auditory and other sensory ones), indirect inferential ones (based on conclusions drawn by the speaker), and indirect reportative ones (based on a transmission of someone else’s words; one may distinguish between markers indicating different degrees of quotation and markers whose use is based on an unspecified source of information, e.g. a folklore text). In a somewhat simplified way, Willett’s system looks as illustrated in Figure 1. Direct/Attested: – visual – auditory – other sensory Indirect/Inferring: – results – reasoning Indirect/Reported: – second-hand/third-hand/folklore Figure 1. Summary of Willett 1988’s classification

The basic types of grammatically relevant oppositions within the semantic domain of evidentiality are very well captured in his system, but their internal subdivision seems to be insufficient. The most controversial aspect of this classification is the separate place accorded to the specialized auditive within sensory non-visual values, which is a typological claim that may easily be disproved (for further details see Section 2.1). One may also notice a somewhat too detailed subdivision of the reportative domain: whether the various values distinguished by Willett are really universal ones requires a more profound argumentation. At the same time, a number of important evidential values discussed in the preceding Section are missing in this classification. Another widely known typological classification of evidential values is the one proposed by Aikhenvald 2004 (see Figure 2).

36

Vladimir A. Plungian – – – – – –

Visual Sensory Inference Assumption Hearsay Quotative

(direct/firsthand) (inferred/non-firsthand) (reported/non-firsthand)

Figure 2. Summary of Aikhenvald 2004’s classification

Aikhenvald’s classification is based on a distinction into three basic groups of evidential markers as well, similar to the groups proposed by Willett (i.e. values of direct access and values of indirect access, which in turn are split into inferential and reportative ones). The basic types distinguished by Aikhenvald reflect the typological reality as it is known today in a better way: a number of typologically doubtful values (or values for which there is not sufficient evidence) occurring in earlier classifications are removed in her classification. Equally reasonable seems also the internal subdivision of evidential values proposed by Aikhenvald. Important for a typology of evidentiality is, of course, the opposition of visual and sensory types of direct access, and the one between inferential and presumptive types of indirect access, but also the subdivision of reportative values into those which refer to utterances of a concrete person (“quotatives”) and those referring to utterances of a person who is unknown or whose identity is unimportant (“hearsay”: rumors, stories, generally known opinions etc.).6 This terminology reflects the typological reality better than the distinction into “second-hand” and “third-hand” (as well as folklore) information proposed by Willett: although, for instance, in some South American languages one may find strategies by which different degrees of distance with reference to the source of information are distinguished (cf. Aikhenvald 2004: 178-179), their place within the system of grammatical means used for the expression of evidentiality is rather marginal. Aikhenvald’s classification could be supplemented by some subtypes that are not listed among the basic types of evidential values (although direct or indirect reference to their existence is made in Aikhenvald’s monograph). This refers, above all, to the values of direct access, which should also include participatory and endophoric values.7 A special discussion also deserves the value of “generally known truths”, which in the majority of cases is expressed by one of the two types of evidential markers: either by means of markers of direct access, or by markers of the reportative group. This way, a slightly modified and supplemented classification of those evidential values which are the most important cross-linguistically



Types of verbal evidentiality marking

37

would look as shown in Figure 3 (the values printed in italics are those which are either missing or not sufficiently integrated into the general schema by Willett and Aikhenvald; the terms occurring in square brackets are the most common terminological variants). Note the proposed division of evidential values into an “upper” (with a distinction between direct/indirect and personal/non-personal access) and a “lower” level. This way, a much more balanced classification of evidential values in the languages of the world compared to the existing ones can be derived since all evidential values are divided into three large classes that result from the interSection of the two partly independent binary features ‘direct/indirect’ and ‘personal/non-personal’ access to information: values of direct access (direct witness), indirect personal access and indirect nonpersonal access (the two latter groups are subsumed under the category “indirect evidence”, which could also be labelled “meditative”, “nonconfirmative”, or differently). Direct/Personal [= Attested, Witnessed, Firsthand, Confirmative] – Participatory/Endophoric; – Visual (with subtypes) – Non-visual [= Sensory] Indirect/Personal – Inferential (based on observed results) – Presumptive (based on plausible reasoning); Indirect/Non-personal [= Secondhand] – Reported (with subtypes) Figure 3. Updated classification of evidential values

Among the values of the group of direct access one may distinguish, on the one hand, participatory and endophoric values, which characterize a fact as being based on the personal participation of the speaker in a given situation (in the case of endophoric values: as experiencer), and, on the other hand, values which characterize a fact as being based on the speaker’s personal perception of a situation. If the type of perception of a situation is grammatically relevant, one may further distinguish between visual and non-visual (sensory) markers of direct access. Among the values of the group of indirect personal evidence one may distinguish, above all, between inferential values, which indicate that an utterance about a situation is based on what the speaker interprets as its observable result, and presumptive values, indicating that the utterance about a situation is based on cause-and-effect relations known to the speaker. As already stated above,

38

Vladimir A. Plungian

in languages with grammaticalized evidentiality utterances referring to the category of general truths can be formulated in different ways: if they do not require the use of specific markers, they can be formulated either by using markers of direct access, or through presumptive markers (if these do not carry the additional epistemic load of incomplete certainty of what is said). Finally, reportative values of the group of non-personal (and thus indirect) evidence describe different variants of accessing a situation on the base of the words uttered by another person. Within this group one may distinguish between markers indicating a specific source of information known to the speaker, an unknown source of information, and others.

3. Types of evidential systems in the languages of the world The semantic domain of evidentiality in general was briefly characterized above, i.e. the semantic domain that is related to the type of access to a situation and from which the languages of the world draw grammaticalized markers of evidentiality of the verb. We will now deal with another question, which is one that is not less important for grammatical typology: How can the different grammatical markers of evidentiality occurring in the languages of the world be organized within verbal systems? This question can be divided into two parts. Firstly, it has to be shown how systems of specialized evidential grammatical markers of the verb can be constructed (“typology of evidential systems”). Secondly, it has to be shown in what way verbal grammatical markers of evidentiality may be built into the grammatical system of verbal markers in general (“the place of evidential markers within the grammatical system”). Until the present day, these and other questions have attracted little interest in grammatical typology. This is related to the well-known fact that modern grammatical typology began to develop mainly as a typology of universal grammatical values (“cross-linguistic gram types”) and practically did not pay any attention to the question of how universal grammatical values correspond to each other in the verbal systems of particular languages. Most of the studies on grammatical typology share a higher or lower degree of “antistructuralist” ideology (the most typical representative of this direction is certainly Joan Bybee), and the question of languageinternal interrelations of grammatical values must have occurred too “structuralist” and thus of secondary importance to these scholars. However, the position within the language system (as well as the typology of such systems in general) is a very important topic both for purely descriptive



Types of verbal evidentiality marking

39

purposes and for many other aspects related to the function of grammatical markers. Thus, it is not a coincidence that research on the typology of grammatical systems has begun to accept these aspects as one of the most important tasks of grammatical typology in general in the past years, together with a description of the composition and structure of a universal grammatical inventory, i.e. the totality of values that in principle might be grammaticalized in natural languages. With respect to evidential markers, Aikhenvald’s 2004 monograph is the only one in which the task of setting up a typology of evidential systems is addressed. The problem is treated in a number of independent Sections in this monograph. In this sense, one of the strengths of Aikhenvald’s pioneering study is to have brought this topic into the center of interest, although the typology proposed in this study is, as one may assume, not the only possible one (some of its shortcomings will be discussed more in detail below). The main (and practically only) principle that determines the classification of evidential systems in Aikhenvald 2004 is very simple: it is based on the number of grammemes of the category of evidentiality that are formally distinguished in a given language. Accordingly, Aikhenvald distinguishes between “systems with two choices” (the simplest case), “systems with three choices”, “systems with four choices” and more complex systems (the latter occurring very rarely, and the data of such systems often need additional analyses). Each of these systems may be divided further into subtypes, depending on the types of grammemes of evidentiality that are formally distinguished in them. One objection that might be put forward against this method used for the classification of evidential systems is more general in character: the quantitative principle, which appears to be clear and simple at first sight, is somewhat too mechanical and leads to an analysis by which entirely different systems may fall together into one and the same class, if they happen to exhibit the same “arithmetical” number of grammemes of evidentiality. Therefore it seems to be useful to propose an alternative classification of evidential systems in which, next to the number of evidential grammemes, also the type of evidential semantics expressed in each system will be given an equal consideration. First of all, we will proceed from the distinction between “basic” evidential systems and their variants, modified in one way or another. It seems that the systems which have the best reason to be considered as basic are ternary evidential systems. Here, all three types of cross-linguistically relevant grammemes of evidentiality discussed above are included, i.e. markers

40

Vladimir A. Plungian

of direct access, indirect personal access (inferential) and non-personal access (reportative) are formally opposed to each other. In this sense, the basic system is in a certain way the prototypical realization of a generalized typological idea of evidentiality in a most balanced and “symmetric” way. Furthermore, the basic ternary systems are not necessarily the most widespread ones, but systems that are sufficiently well represented in the languages of the world. An example is the system of Quechua, but also that of a number of other languages (for a more detailed overview see Aikhenvald 2004: 43-46). Interestingly, according to the descriptions in Wiemer 2006 and Wiemer 2007, present-day Lithuanian also belongs to the languages in which such a system can be found, considering that in standard written Lithuanian two different strategies of the marking of evidentiality are distinguished. The reportative is realized by means of constructions with genuine particles (of the type ji dar ne-perskaiþiusi laiško ‘it is said that she has not read the letter yet’, literally: ‘she yet not-having-read ofletter’), whereas the inferential is realized by means of constructions with “impersonal” negative particle that do not exhibit agreement (of the type þia vaiko miegota ‘it seems as if a child slept here’, literally: ‘here of-child it-was-asleep’). Still, the systems represented in the majority of Lithuanian dialects can most probably be treated as binary. All the other systems of evidentiality can be regarded either as being the result of a reduction of the basic system or, in contrast, of its extension. Binary evidential systems, for instance, belong to the systems of the reduced type, and are apparently the ones which are most widespread on the linguistic map of the world. A more usual type of binary systems is one with a single generalized evidential marker of the “meditative” type (i.e. one which represents all types of indirect access to a situation), which is opposed to a marker of direct access or to an unmarked form (although it is difficult to draw a clear border line between these two cases; see especially Aikhenvald 2004: 39-41 for a discussion of this issue). One may also find binary systems in which only reportative value occurs as a marked element (Latvian and Estonian seem to belong to this type, since both lack the grammatical expression of inferential values). In the latter case the “reduction” of the basic system was not brought about by the generation of a polysemous marker with a broad range of values, but through a procedure by which one of the basic evidential values is simply not expressed by means of grammatical elements. On the other hand, all evidential systems with additional contrasts, i.e. those that go beyond the inferential and the reportative values, belong to systems of the extended type. As we have already seen, a rather large in-



Types of verbal evidentiality marking

41

ventory of such additional evidential values can be found in the languages of the world: they may be found within the domain of direct access (participant or visual vs. sensory), and within the domain of indirect access (inferential vs. presumptive or various types of reportative markers). More finegrained additional distinctions of evidential markers are also possible, some of which have already discussed above. The observed diversity of evidential systems basically corresponds to the proposed taxonomy, i.e. they all correspond to either the basic ternary type, the reduced binary type, or to the extended type with more than three grammemes of evidentiality. However, another issue should be considered in this respect. An evidential system that appears to be reduced is, generally speaking, not always and necessarily a binary one. In some (however rare) cases we may find systems with more than two grammemes of evidentiality, but where the basic contrast between direct, inferential and reportative values is nevertheless absent. Such effect may arise when, for instance, a system includes two different reportative markers, but none for the marking of inferential values. Formally, such a system should be classified as ternary (as it is indeed done in Aikhenvald’s monograph), but there are still important semantic differences between such systems and the basic ternary ones. Therefore, it is more appropriate to regard such systems as subtypes of binary ones, to which they are typologically closer, and to speak of modified binary systems in this case. In this sense, the classification we propose is not based on the mere number of evidential grammemes, but more on formal and semantic oppositions (see Figure 4). Ternary systems (“basic”): – Direct vs. Inferred vs. Reported [Quechua-type] Binary systems (“generalized”): – broad Indirective/Mediative or Secondhand [Balkan-type] Complex systems (“elaborated”): – ternary systems with further distinctions and/or additional parameters such as ‘endophoricity’, ‘participation’, ‘assumption’, ‘common knowledge’, etc. NB: Modified binary systems (e.g. with two types of secondhand evidence) vs. “true” ternary or complex systems Figure 4. Types of evidential systems

The problem of the position of evidential markers within verbal systems is, at present, a minor topic, and in this survey we may only add some remarks about it (useful additional facts are provided in Aikhenvald 2004). On the

42

Vladimir A. Plungian

one hand, the degree to which evidential markers occupy a central position in verbal systems may vary. There are verbal systems (e.g. those of a number of languages spoken in North and South America) in which evidentiality is a central (but not the only) grammatical category of the verb, and their expression is, this way or the other, subordinated to all other verbal categories. In other cases evidentiality has a status that is equal to that of other verbal categories (and is expressed independently), such as aspect, mood, and others. On the other hand, evidentiality often occurs, so to speak, on the periphery of some other category within the verbal system and may be analyzed as a special value of this category or a special way of using it. Such cases of “dependent“ expression of evidentiality are called “evidential strategies” in Aikhenvald 2004, although, as one may conclude from existing descriptions, it is not always easy to draw the line between evidential strategies and the expression of evidentiality by means of full-fledged grammemes. One of the most common strategies of expressing evidentiality is the use of forms of the perfect with evidential functions (either derived etymologically from the perfect, or still preserving a parallel use for marking the perfect). The use of such “perfectoid” forms is typical for binary systems of the “Balkan type”, with a broad mediative marker. In systems with perfect-based strategies of expressing evidentiality this contrast is usually realized only in the forms of the past tense of the verb (perfectoid evidential systems may over time undergo an evolution into a fullfledged evidential system, but such an evolution presupposes a considerable reorganization of the entire verbal paradigm). There is a large body of literature on perfectoid evidential systems; see, above all, Guentchéva 1996, Johanson and Utas 2000, Tatevosov 2001, 2003. One may also observe the tendency to use constructions with infinite forms of the verb for the expression of evidentiality (participles, converbs, and infinitives), which in initial stages are usually weakly grammaticalized and which get engaged into the verbal system only gradually (for more details see, for instance, Wälchli 2000, who offers an overview of data from the Baltic languages).

4. Evidentiality and other verbal categories In this section we will briefly discuss the interrelation between evidentiality and two verbal categories, namely person and modality. As it has often been mentioned in the literature on typology, both types of interrelations



Types of verbal evidentiality marking

43

are highly important (although the interpretation of the data differs tremendously with different authors). Below we will provide a brief overview of some well known facts.

4.1. Evidentiality and person A verb form of the first person singular marked for an indirect evidentiality displays a rather paradoxical combination of grammemes, since this form relates to a situation in which the subject is the speaker. Note that markers of both indirect and direct access are equally paradoxical here (of course, except for participant and endophoric markers, which are specifically used for such situations). Actually, if the speaker is the subject of a situation, the access to the given situation is normally direct. However, the use of many markers of direct access – especially those which presuppose a direct observation of the event – is not necessarily obligatory in this situation. In this sense, one could expect that in many languages forms of the first person cannot be combined with markers of evidentiality. Nevertheless, in the majority of languages such forms exist and are actively used, and the above-mentioned paradox is overcome in such a way that the given forms assume a number of additional semantic features. Such additional features are called ‘first person effect’ in Aikhenvald (2004: 237). In combination with different markers of indirect access the forms of the first person adopt, as a rule, the value of an uncontrolled and unintentional action (the speakers do not remember what they did, they acted in a dream, in a drunken state, etc.). This way, the subjects of a situation split up into an unconsciously and a consciously acting figure and look at the preceding unconscious occurrences as if they did not take part in them, as mere observers.8 An important semantic property of these forms is, that, for the speakers (and in some languages also for the addressees, see Aikhenvald 2004: 239), the information about their actions turns out to be new and often unexpected. It is this semantic peculiarity in which the admirative value of evidential forms, which will be discussed in the following section, manifests itself. The first-person semantic effect is also connected with the so-called “conjunct/disjunct marking” which, strictly speaking, is not so much related to the grammatical expression of evidentiality, but to the expression of person (i.e. that of the speaker). The expression denotes a strategy (particularly characteristic for Barbacoan, Tibeto-Burman, and some other languages), by which verb forms are opposed to each other on the basis of

44

Vladimir A. Plungian

two markers of person: one marker is used for utterances with a first person subject (or the second person in questions), the other one for utterances with a non-first person subject. The latter may also be used in utterances from a first-person perspective, but in this case the speaker causes the effect already mentioned, namely the characterization of an action as uncontrolled or unintentional (for further details, see Curnow 2002, 2003, but also the overview in Aikhenvald 2004: 123-130), by which the speakers adopt the role of outside observers of themselves. The semantic “first person effects” occurring in the context of evidential markers were discussed in a number of special studies; next to some generalized observations in Aikhenvald (2004: 219-239), the effects are also referred to in the works by Guentchéva et al. 1994, Majsak and Tatevosov 2000, Curnow 2003, Nicolova 2006.

4.2. Evidentiality and Modality The interrelation between evidentiality and modality is, in all likelihood, one of the most complex problems of all the theoretical difficulties related to the description of the category of evidentiality. Therefore, it is no coincidence that this question has been dealt with from all possible points of view so that one may find strict positions (evidentiality is a type of modality; evidentiality and modality have nothing in common) and more relaxed ones (evidentiality and modality are different categories, but there are semantic overlaps). Apparently, the problem became especially difficult to deal with because different authors had different things in mind when talking about modality and evidentiality (although they often simply suggest different interpretations of one and the same phenomenon). Evidentiality in general is a very diffuse category with respect to its semantic nature (especially if we do not analyze markers of evidentiality in isolation, but in relation to the concrete verbal systems in which they function), and it is not easy for a researcher who is familiar with only one type of evidential systems to formulate general statements that would be valid for all observed types and not only for one of them. In what follows, we will present our ideas related to this confusing matter, knowing that we cannot, of course, offer a final solution of the problem. One may find a rather large number of concurring views on the nature of modality and the question of where the line between the semantic domain of modality and that of other categories has to be drawn (we will not



Types of verbal evidentiality marking

45

present a more detailed summary of this problem here, but restrict ourselves to a reference on the monographs by Bybee et al. 1994, Bybee and Fleischmann 1995, Palmer 2001; cf. also van der Auwera and Plungian 1998, Plungian 2000). In the strictest sense, the conception of modality includes, above all, the expression of possibility and necessity (including also epistemic values, i.e. an assessment of the certainty of an event); a broader conception would also include the expression of volition and all other types of assessment, as well as the expression of the speakers’ subjective attitude in general. Volitive modality may, in a certain way, be regarded as a central element of the modal semantic domain since it includes an evaluative component (people usually evaluate those things as positive which coincide with their desires) and the concepts of necessity/possibility (people usually want things that are not available in a given moment, but which will probably be so in the future). Volitive modality is the basis for grammatical categories of the verb like different kinds of the optative or the imperative. In this sense, necessity, possibility, epistemic assessment (and other types of evaluation) as well as volition all form a legitimate part of the semantic domain of modality. As for the inclusion of broader semantic components into this domain, it is by far more debatable and requires a special justification. Our discussion of the semantics of evidentiality with all its typologically diverse forms has shown that one can by no means claim that evidentiality is a subclass of modal values, i.e. that it is entirely subordinated to the semantic domain of modality. Such a conception of evidentiality was, however, characteristic for the initial stages of the study of this category and reflects an insufficient knowledge of all its different forms in the languages of the world.9 Evidentiality as a whole, i.e. the indication of the type of access to a situation (or that of the source of information underlying an utterance about a situation), can by no means be equated with modality, not even in the widest sense or by referring to some of the values of evidentiality that have a prototypical modal character. It would, however, be too early to stop the discussion at this point, since the problem is not that easy and one-sided. In spite of the fact that evidentiality and modality are different – even entirely different – categories, it is also difficult to accept the opposite, radical view according to which these categories have generally nothing in common. On the contrary, it would be much more useful to adopt an approach that accounts for the large number of diverse and complex relations that may be observed between evidentiality and modality.

46

Vladimir A. Plungian

One example for these relations is the fact that particular values within the semantic domain of evidentiality also occur in the domain of modal values, as it is, above all, the case with presumptive values (of the group of values of personal indirect access; see also van der Auwera and Plungian 1998). Indeed, an utterance which refers to the fact that a situation takes or took place, due to the existence of convincing reasons for it, is actually not different from one referring to the epistemic necessity of this situation: in both cases the speakers do not intend to become personally convinced of the fact a situation takes or took place, but consider it as highly credible, due to certain cause-and-effect relations known to them. The difference between presumptive evidentiality and epistemic necessity (if it exists at all) is often blurred through the merger of one of these meaning components, namely that which was emphasized, with the other. If we regard such values as modal, we stress one of the basic characteristics of modality, namely the assessment of a situation (as highly probale); regarding it as evidential, we stress one of the basic characteristics of evidentiality, namely the reference to logical conclusions as a source of information about a situation. This way, markers of presumptive evidentiality are the only evidential markers with inbuilt modal components and the only modal markers with inbuilt evidential components. In a language that lacks other markers of evidentiality the markers of epistemic necessity are not regarded as evidential ones, of course, since the inference of an evidential fragment turns out here as the natural pragmatic consequence of this type of epistemic modality. The existence of a marker of epistemic necessity is therefore, if taken for itself, not an indicator for the presence of the grammatical expression of evidentiality within the system of a language. However, markers of this kind always exhibit an intersection of modal and evidential values. There are, however, also other factors to be discussed, which are probably even more interesting for the study of grammatical typology. One of them is the occurrence of secondary modal values with evidential markers in the languages of the world, another one the occurrence of secondary evidential values with markers of modality. These values occur, at least in the initial stages of their existence in a language, as derived and secondary, and in this sense one may, following Aikhenvald 2004, speak of different “evidential strategies” of using modal markers. However, one may also speak of “modal strategies” of using evidential markers. Both strategies are not universal, but each of them occurs frequently enough to allow for the claim that the relation between evidentiality and modality is very close and not coincidental.



Types of verbal evidentiality marking

47

We already discussed one of the most widespread modal strategies of using evidential markers (i.e. the occurrence of secondary modal values with evidential markers) in the beginning of the present article (see Section 2.1). The basic idea was that markers of indirect access convey the value of epistemic uncertainty which, in the weak form, occurs as “epistemic distance”, i.e. the speakers are released from the responsibility for the truth of the utterance. Such modal strategies are characteristic for the large majority of binary evidential systems with generalized markers of the mediative type. The fact that such systems were well known to researchers already in the earliest periods of the study of evidentiality explains why in many of the earlier studies of evidentiality this category is treated mainly as a modal category. However, we have seen that such a relation does not always exist. Furthermore, modal strategies of the type described above may be entirely absent in evidential systems of a different kind (especially with complex evidential systems where the value of epistemic certainty does not necessarily correlate with markers of direct access). Another widespread type of modal strategies is the occurrence of socalled admirative (or mirative) values with evidential markers. The latter term, suggested by Scott De Lancy, is more recent and obviously preferred in typological studies. Its meaning is, taken for itself, undisputedly modal, since it reflects one of the various kinds of epistemic evaluation, namely a contradiction to the expectation of the speakers or, in other words, the fact that the speakers were not prepared to cognitively process the situation they observed.10 However, mirativity is not used to indicate the source information of a situation. Furthermore, a mirative situation is, as a rule, accessed by means of a direct observation by the speakers. However, in spite of this fact, mirativity may in the majority of languages be expressed by means of markers of indirect access, which constitutes an important descriptive and theoretical problem. Linguists who defend a rigid separation of mirative and evidential values, due to the legitimate semantic differences between both, cannot explain why these values are regularly expressed by means of one and the same marker. On the other hand, linguists who claim that mirativity is a part of the various values of evidentiality cannot offer a typologically correct and useful definition of the category of evidentiality itself, since from a semantic perspective mirativity is probably too far away from the prototypical evidential values. From our point of view (cf. also Plungian 2001) the relation between mirativity and evidentiality is not a direct, but an indirect one: the mirative effect occurs, as a rule, in binary evidential systems, where the markers of indirect access already have a strong modal character. The val-

48

Vladimir A. Plungian

ue of unexpectedness then derives from the further development of the value of epistemic uncertainty, but not from evidential values as such. However, what is important is that all of these values form part of one single mediative cluster in languages with evidentiality of the “Balkan type”. A more detailed discussion of the problem of mirativity can be found in the works of De Lancey 1997, 2001, Lazard 1999, 2000, Friedman 2005, and others. In this sense, it is important to note that modal strategies of different kinds are most characteristic for binary systems in which the unmarked form of the verb or the form that indicates direct access is opposed to a form that represents a wide range of both truly evidential (relating to indirect access) and modal values (epistemic distance/epistemic uncertainty and mirativity). A semantic cluster of this type is usually called mediative (if we use the “Iranian-based” terminology by Lazard), non-confirmative (if we use the “Balkan-based” terminology by Friedman) or indirective (if we use the “Turkic-based” terminology by Johanson). One should also pay attention to a phenomenon that is, in a certain sense, opposed to the one discussed above, namely the use of modal markers for the expression of secondary evidential values. In other words, we should also consider cases in which evidential strategies are realized through modal expressions. Such occurrences are well known since they are widespread, above all in the Romance and the Germanic languages. We will refer to only some of these. (2) illustrates the use of the English modal marker of necessity in an inferential context; (3) shows the German modal marker of necessity in a reportative context. A close equivalent to (3) in French would be the use of the form of the conditional (aurait tué). One of the basic values of the conditional is indeed modal in character, but the category is generally more related to epistemic possibility (hypothetic consequence) and optativity. Similar strategies may be found also in other Romance languages (for further details see, e.g., Mortelmans 2000 and Pietrandrea 2005). (2) (3)

inferential use of necessity marker Engl. A dog must have run here reportative use of necessity marker Germ. Eine Schülerin aus Thüringen soll den Lebensgefährten ihrer Mutter erschlagen haben. (from a newspaper) ‘A schoolgirl from Th. is reported to have killed the boyfriend of her mother’



Types of verbal evidentiality marking

49

Summing up what has been said so far, one may conclude the following. Evidentiality and modality are two different semantic domains which are closely related from a synchronic and a diachronic perspective. There is a small semantic room (presumptive evidentiality – epistemic necessity) in which these two zones intersect. It is important to note that the nature of the relation between evidentiality and modality cannot be explained in a general way since it depends on the concrete evidential system. In this sense it might be useful to distinguish between “modalized” and “nonmodalized” evidential systems, as introduced in Plungian 2001. In modalized systems markers of indirect access obtain secondary values of epistemic distance and, deriving from them, mirative values. In non-modalized systems such a strict relation does not exist. Modalized systems include, above all, binary evidential systems of the “Balkan type” (which are also found in Central and South Asia, in the Caucasus etc.), with mediative markers that represent a wide range of modal-evidential values. To conclude our – necessarily imcomplete and fragmentary – overview of the problem related to the grammatical expression of evidentiality on verbs we will mention some guidelines for the study of evidentiality which in our view are essential for the further development of research in this domain. The first one is the continuation of descriptive studies of such evidential systems whose working mechanism is not sufficiently or badly understood (this refers to systems as those found in the languages of South America, New Guinea, Tropical Africa and the Utmost North). Such investigations may, above all, contribute to a significant enrichment of the universal set of parameters used for the classification of evidential systems. An important aspect that has been neglected in the past is the function of evidentiality as a core element of discourse and pragmatic mechanisms (which is also shown in Aikhenvald 2004). Finally, further research on the interrelation between markers of evidentiality and other grammatical markers of the verb within a universal typology of verbal systems is needed. In this respect, only the first step has been taken.

Notes 1. 2.

The article is translated from Russian by Alexander Haselow. For a detailed discussion of evidential systems in Tucano cf. the frequently quoted articles by Barnes 1984 and Malone 1988; cf. Aikhenvald (2003a: 294ff.) in relation to a description of the system of evidentiality in Tariana, an Arawak language that is subjected to intense influence from Tucano.

50

Vladimir A. Plungian

3.

Obviously, the category of evidentiality in general belongs to those types of linguistic phenomena which have a clearly areal character: it is easily diffused by means of language contact, even if the languages do not share a close genetic relation. Therefore, evidentiality is often considered as an important “areal-forming” feature in studies of areal typology. For a more detailed discussion of the areal specifics of evidentiality cf. especially Aikhenvald and Dixon 1998, also Aikhenvald 2003 and 2004: 288-299. However, Boas did not use the term evidential to refer to the grammatical category itself (for which he used terms of the type ‘source of information’, which reflects its characteristic feature much more precisely). Rather, he used the term to refer to one of the values of this category for which in current works the term “inferential” is used. Not all of the special values of the category of evidentiality distinguished by Jakobson have an undisputedly typological status and can be equally well corroborated by current investigations. Especially questionable is the occurrence of grammemes specialized for the indication of the ‘revelative’ in the language of the world, which Boas postulates for Kwakiutl. Events seen in a dream may, of course, be expressed by specific grammatical means, including evidential markers of the verb, but these markers usually also fulfil other functions than that of indicating the ‘revelative’ only (for further details see Aikhenvald 2004: 344-347). The term ‘reportative’ (or ‘reportive’) seems to be the most convenient as a cover term for the various markers of non-personal access in general, and in this sense Aikhenvald’s (2004: 25) terminology is much more adequate. The terms like ‘quotative’ or ‘hearsay’, also widely attested, should better be used as referring to various sub types of reportative values (in this sense, I would suggest to change the terminological practice, which I myself have followed earlier, as e.g. in Plungian 2001). In Wiemer’s article, this volume, the terms ‘reportive’ and ‘hearsay’ are used synonymously, which is, in my view, somewhat less felicitous. The existence of endophoric markers means that Aikhenvald’s claim, according to which “no language has two visual or two non-visual evidentials” (Aikhenvald 2004: 177) should be tested more accurately. Furthermore, languages may also have a number of visual markers which distinguish, for instance, additional deictic parameters. An analogous effect may, according to several descriptions, be observed in the Tibetan languages, which manifests itself in the forms of the first person and their combination with markers of direct access. Here, the construction is used to describe a situation in which an event is observed visually (see Aikhenvald 2004: 237). Note, however, that these languages also have endophoric markers, which are used for descriptions of events or feelings of the speaker, and visual markers used for those contexts in which the speakers “observe them-

4.

5.

6.

7.

8.



Types of verbal evidentiality marking

51

selves from the outside”. In languages with other evidential systems markers of indirect access would be used in this case. 9. One should add that there were also approaches which were closer to the current stage of the development of grammatical typology, such as those proposed by Givón 1982 and Chafe 1986. According to these authors, evidentiality is part of a more general mechanism of the expression of epistemic values (“attitude towards knowledge”, according to Givón). In modern studies of evidentiality such radical points of view are, however, not supported. 10. In descriptions of the semantics of the mirative one may often find the cliché of an “unprepared mind”, as in the study by Slobin and Aksu 1982, which since then was continuously reproduced. One should consider that in a number of languages (as e.g. in Albanian, in which this value was observed first) the semantic evolution of mirative markers went even farther: they may not only indicate the unexpectedness of a situation, but a general positive emotional evaluation of a situation by the speaker; see also Section 1.4.

References Aikhenvald, Alexandra Y. 2003 Evidentiality in typological perspective. In Aikhenvald, Alexandra Y. and R.M.W. Dixon (eds.), 1-32. 2003a A Grammar of Tariana, from Northwestern Amazonia. Cambridge: Cambridge University Press. 2004 Evidentiality. Oxford: Oxford University Press. Aikhenvald, Alexandra Y., and R.M.W. Dixon. 1998 Evidentials and areal typology: A case study from Amazonia. Language Sciences 20 (3): 241-257. Aikhenvald, Alexandra Y., and R.M.W. Dixon (eds.) 1999 The Amazonian Languages. Cambridge: Cambridge Univ. Press. 2003 Studies in Evidentiality. Amsterdam/Philadelphia: John Benjamins. Aksënova, Irina S. 1997 Kategorii vida, vremeni i naklonenija v jazykax bantu (= The categories of aspect, tense and mood in the Bantu languages). Moskva: Nauka. Aronson, H.I. 1967 The grammatical category of the indicative of the contemporary Bulgarian literary language. In To honor Roman Jakobson, v. 1. The Hague: Mouton, 82-98. Barnes, Janet 1984 Evidentials in the Tuyuca Verb. International Journal of American Linguistics 50: 255-271.

52

Vladimir A. Plungian

Burkova, Svetlana I. 2004 Evidencial’nost’ i èpistemiþeskaja modalnost’ v neneckom jazyke (= Evidentiality and epistemic modality in Nenets). In Issledovanija po teorii grammatiki 3: Irrealis i irreal’nost’ (= Studies in the theory of grammar 3: Irrealis and irreality). J. A. Lander et al. (eds.), 353-374. Moskva: Gnozis. Bybee, Joan L. and Östen Dahl 1989 The creation of tense and aspect systems in the languages of the world. Studies in Language 13 (1): 51-103. Bybee, Joan L. and Suzanne Fleischman (eds.) 1995 Modality in grammar and discourse. Amsterdam/Philadelphia: John Benjamins. Bybee, Joan L., Revere Perkins and William Pagliuca 1994 The Evolution of Grammar: Tense, Aspect and Modality in the Languages of the World. Chicago: The University of Chicago Press. Chafe, Wallace 1986 Evidentiality in English conversation and academic writing. In Chafe, Wallace and Johanna Nichols (eds.), 261-272. Chafe, Wallace and Johanna Nichols (eds.) 1986 Evidentiality: The Linguistic Coding of Epistemology. Norwood, NJ: Ablex. Curnow, Timothy J. 2002 Conjunct/disjunct marking in Awa Pit. Linguistics 40 (3): 610-627. 2003 Non-volitionality expressed through evidentials. Studies in language 27: 39-60. Dahl, Östen 1985 Tense and aspect systems. Oxford: Blackwell. Dankoff, Robert (with J. Kelly) 1982 Mahmud al-KƗšƥarƯ. Compendium of the Turkic dialects. (DƯwƗn luƥƗt at-turk). Part I. Cambridge: Tekin. De Haan, Ferdinand 1999 Evidentiality and epistemic modality: Setting boundaries. Southwest Journal of Linguistics 18 (1): 83-101. 2005 Encoding speaker perspective: evidentials. In Linguistic diversity and language theories. Zygmunt Frajzyngier, Adam Hodges and David S. Rood (eds.), 379-397. Amsterdam/Philadelphia: John Benjamins. DeLancey, Scott 1997 Mirativity: the grammatical marking of unexpected information. Linguistic Typology 1 (1): 33-52. 2001 The mirative and evidentiality. Journal of Pragmatics 33 (3): 369-382. Dendale, Patrick and Tasmowski, Liliane 2001 Introduction: evidentiality and related notions. Journal of Pragmatics 33 (3): 339-348.



Types of verbal evidentiality marking

53

Dëmina, Elena I. 1959 Pereskazyvatel'nye formy v sovremennom bolgarskom literaturnom jazyke (= Reported forms in contemporary Bulgarian). In Voprosy grammatiki bolgarskogo literaturnogo jazyka (= Issues in the grammar of the Bulgarian literary language). S. B. Bernstein (ed.), 313378. Moskva: Nauka. Dimmendaal, Gerrit J. 2001 Logophoric marking and represented speech in African languages as evidential hedging strategies. Australian Journal of Linguistics 21 (1): 131-157. Dozon, Auguste 1879 Manuel de la langue chkipe ou albanaise. Paris: Ernest Leroux. Duchet, Jean-Louis and Remzi Përnaska 1996 L’admiratif albanais: recherche d’un invariant sémantique. In Guentchéva, Zlatka (ed.), 31-46. Foley, William A. 1986 The Papuan Languages of New Guinea. Cambridge: Cambridge University Press. Friedman, Victor A. 1979 Towards a typology of status: Georgian and other non-Slavic languages of the Soviet Union. Papers from the conference on nonSlavic languages of the USSR. Chicago, 338-349. 1986 Evidentiality in the Balkans: Bulgarian, Macedonian, and Albanian. In Chafe, Wallace and Johanna Nichols (eds.), 168-187. 2000 Confirmative/nonconfirmative in Balkan Slavic, Balkan Romance, and Albanian with additional observations on Turkish, Romani, Georgian and Lak. In Johanson, Lars and Bo Utas (eds.), 329-366. 2003 Evidentiality in the Balkans with special attention to Macedonian and Albanian. In Aikhenvald, Alexandra Y. and R.M.W. Dixon (eds.), 189-218. 2005 Admirativity: between modality and evidentiality. STUF 58. Givón, Talmy 1982 Evidentiality and epistemic space. Studies in Language 6 (1): 23-49. Guentchéva, Zlatka 1993 La catégorie du médiatif en bulgare dans une perspective typologique. Revue des études slaves 65 (1): 57-72. 1996a Le médiatif en bulgare. In Guentchéva, Zlatka (ed.), 47-70. Guentchéva, Zlatka; A. Donabédian; M. Meydan, and R. Camus 1994 Interactions entre le médiatif et la personne. Faits de langues 3: 139148. Guentchéva, Zlatka (ed.). 1996 L’énonciation médiatisée. Paris: Peeters.

54

Vladimir A. Plungian

Guentchéva, Zlatka and J. Landaburu (eds.) 2007 L’énonciation médiatisée II. Le traitement épistémologique de l’information: illustrations amérindiennes et causasiennes. Paris: Peeters. Gusev, Valentin J. 2007 Evidentialnost’ v nganasanskom jazyke (= Evidentiality in Nganasan). In Xrakovskij, Viktor S. (ed.), 415-444. Haarmann, Harald 1970 Die indirekte Erlebnisform als grammatische Kategorie: eine eurasische Isoglosse. Wiesbaden: Harrassowitz. Hardman, Martha James (ed.) 1981 The Aymara language in its social and cultural context. Gainesville: University Press of Florida. Holvoet, Axel 2007 Mood and Modality in Baltic. Kraków: Wydawnictwo Uniwersytetu JagielloĔskiego. Künnap, Ago 2002 On the Enets evidential suffixes. Linguistica Uralica 2: 145-153. Il’ina, Ljudmila A. 2002 Auditivnaja forma glagola v sel'kupskom jazyke: tradicionnye funkcii i èvoljucija (= Auditive forms of the verb in Selkup: traditional functions and evolution). Gumanitarnye nauki v Sibiri. Ser. Filologija (= Human Sciences in Siberia. Series Philology), 2002 (4): 31-34. Jakobson, Roman 1957 Shifters, Verbal Categories and the Russian Verb. Cambridge (MA): Harvard U.; also in: Selected Writings. The Hague: Mouton, vol. II, 130-147. Johanson, Lars and Bo Utas (eds.) 2000 Evidentials: Turkic, Iranian and Neighbouring Languages. Berlin: Mouton de Gruyter. Kazakeviþ, Olga A. 2005 Izmenenie struktury jazyka s ograniþennoj sferoj upotreblenija (= The structural change of a language with a restricted domain of usage). In Malye jazyki i tradicii: sušþestvovanie na grani. Vyp. 1: Lingvistiþeskie problemy soxranenija i dokumentacii malyx jazykov (= Minority languages and tradition: existence at the borderline. Vol. 1: Linguistic problems of the preservation and documentation of minority languages). A. E. Kibrik (ed.), 122-134. Moskva: Novoe Izdatel’stvo. Kehayov, Petar 2002 Typology of grammaticalized evidentiality in Bulgarian and Estonian. Linguistica Uralica 2, 126-144.



Types of verbal evidentiality marking

55

to appear Interactions between grammatical evidentials and lexical markers of epistemicity and evidentiality: a case study of Bulgarian and Estonian. In Wiemer, Björn and Vladimir A. Plungjan (eds.), to appear. Kibrik, Aleksandr E. 1977 Opyt strukturnogo opisanija arþinskogo jazyka. t. II: Taksonomiþeskaja grammatika (= An essay of the structural description of Archi, Vol. II: Taxonomic grammar). Moskva: MGU. Koptjevskaja-Tamm, Maria and Berhard Wälchli 2001 The Circum-Baltic languages: an areal-typological approach. In Circum-Baltic languages: Typology and contact. Vol. 2: Grammar and typology. Ö. Dahl and M. Koptjevskaja-Tamm (eds.), 615-750. Amsterdam/Philadelphia: John Benjamins Kozinceva, Natalija A. 1994 Kategorija èvidencial'nosti (Problemy tipologiþeskogo analiza) (= The category of evidentiality. Problems of typological analysis). Voprosy jazykoznanija (= Questions in linguistics) 1994 (3), 92-104. 2007 Tipologija kategorii zasvidetel'stvovannosti (= A typology of the category of evidentiality). In Xrakovskij, Viktor S. (ed.), 13-36. Lazard, Gilbert 1957 Caractères distinctifs de la langue tadjik. Bulletin de la Société de Linguistique de Paris 52 : 117-186. 1996 Le médiatif en persan. In: Guentchéva, Zlatka (ed.), 21-30. 1999 Mirativity, evidentiality, mediativity, or other? Linguistic Typology 3(1): 91-109. 2000 Le médiatif: considerations théoriques et application à l’Iranien. In Johanson, Lars and Bo Utas (eds.), 209-228. Leinonen, Marja and Maria Vilkuna 2000 Past tenses in Permic languages. In Tense and Aspect in the Languages of Europe. Östen Dahl (ed.), 495-514. Berlin: de Gruyter. Ljublinskaja, Marija D. and Mal’þukov, Andrej L. 2007 Èvidencial’nost' v neneckom jazyke (Evidentiality in Nenets). In Xrakovskij, Viktor S. (ed.), 445-468. Loughnane, Robyn 2007 Expanding the typology of evidentiality: the participatory/factual in Oksapmin. Paper presented at the ALT 7 conference (Paris). Majsak, Timur A. and Tatevosov, Sergej G. 2000 Prostranstvo govorjašþego v kategorijax grammatiki, ili þego nel'zja skazat' o sebe samom (= The speaker’s area in the categories of grammar, or what one cannot say about oneself). Voprosy jazykoznanija (= Questions in linguistics) 2000 (5): 68-80. Malone, Terrell 1988 The origin and development of Tuyuca evidentials. International Journal of American Linguistics 54: 119-140.

56

Vladimir A. Plungian

Maslova, Elena S. 2003 Evidentiality in Yukaghir. In Aikhenvald, Alexandra Y. and R.M.W. Dixon (eds.), 219-235. Mithun, Marianne 1999 The Languages of Native North America. Cambridge: Cambridge University Press. Mortelmans, Tanja 2000 On the ‘evidential’ nature of ‘epistemic’ use of the German modals müssen and sollen. In Modal verbs in Germanic and Romance languages. J. van der Auwera and P. Dendale (eds.), 131-148. (= Belgian Journal of Linguistics 14). Nicolova, Ruselina 2006 Vzaimodejstvie èvidencial'nosti i admirativnosti s kategorijami vremeni i lica glagola v bolgarskom jazyke (= The interrelation between evidentiality and admirativity with the categories of tense and person of the verb in Bulgarian) Voprosy jazykoznanija (= Questions in linguistics) 2006 (4), 27-45. 2007 Modalizovannaja èvidencial’naja sistema bolgarskogo jazyka (= Modalized evidential system of Bulgarian). In Xrakovskij, Viktor S. (ed.), 107-196. Nikolaeva, Irina 1999 The semantics of Northern Khanty evidentials. Journal de la Société Finno-Ougrienne 88: 131-159. Nurse, Derek 2008 Tense and aspect in Bantu. Oxford: Oxford University Press. Oswalt, R.L. 1986 The evidential system of Kashaya. In Chafe, Wallace and Johanna Nichols (eds.), 29-45. Palmer, F. R. 2001 Mood and modality. Cambridge: Cambridge University Press (2nd ed). Perrot, Jean 1996 Un médiatif Ouralien: L’auditif en Samoyède Nenets. In Guentchéva, Zlatka (ed.), 157-68. Pietrandrea, Paola 2005 Epistemic modality: Functional properties and the Italian system. Amsterdam/Philadelphia: John Benjamins. Plungian, Vladimir A. 1988 Resultative and apparent evidential in Dogon. In Typology of Resultative Constructions. V. P. Nedjalkov (ed.), 481-493. AmsterdamIPhiladelphia: John Benjamins. 2000 Obšþaja morfologija: vvedenie v problematiku (= General morphology: an introduction). Moskva: Editorial URSS (2nd edition, 2003).

 2001

Types of verbal evidentiality marking

57

The place of evidentiality within the universal grammatical space. Journal of Pragmatics 33 (3): 349-357. 2006 Towards a typology of discontinuous past marking. Sprachtypologie und Universalienforschung 59 (4): 317-349. Slobin, Dan I. and Ayhan Aksu 1982 Tense, aspect, and modality in the use of the Turkish evidential. In Tense-aspect: Between Semantics and Pragmatics. P. Hopper (ed.), 185-200. Amsterdam/Philadelphia: John Benjamins. Squartini, Mario 2008 Lexical vs. grammatical evidentiality in French and Italian. Linguistics 46 (5): 917-947. Squartini, Mario (ed.) 2007 Evidentiality between lexicon and grammar. Italian Journal of Linguistics 19.1. Tatevosov, Sergey 2001 From resultatives to evidentials: Multiple uses of the Perfect in Nakh-Dagestanian languages. Journal of Pragmatics 33(3): 443-464. 2003 Inferred evidence: Language-specific properties and universal constraints. In Meaning Through Language Contrast. Jaszczolt, K.M. and K. Turner (eds.), 177–192. Amsterdam/Philadelphia: Benjamins. Tournadre, Nicolas 1996 Comparaison des systèmes médiatifs de quatre dialectes tibétains (tibétain central, ladakhi, dzongkha et amdo). In Guentchéva, Zlatka (ed.), 195-213. Van der Auwera, Johan and Vladimir A. Plungian 1998 Modality’s semantic map. Linguistic Typology 2 (1): 79-124. Wälchli, Bernhard 2000 Infinite predication as a marker of evidentiality and modality in the languages of the Baltic region. Sprachtypologie und Universalienforschung 53: 186-210. Wiemer, Björn 2006 Grammatical evidentiality in Lithuanian (a typological assessment). Baltistica 41 (1): 33-49. 2007 Kosvennaja zasvidetel’stvovannost’ v litovskom jazyke (= Indirect evidentiality in Lithuanian) In Xrakovskij, Viktor S. (ed.), 197-240. 2010 Hearsay in European languages: towards an integrative account of grammatical and lexical marking. This volume. Wiemer, Björn and Vladimir A. Plungjan (eds.) 2009 Lexikalische Evidenzialitäts-Marker in slavischen Sprachen. München: Sagner. (Wiener Slawistischer Almanach, Sonderband 72). Willett, Thomas L. 1988 A cross-linguistic survey of the grammaticization of evidentiality. Studies in Language 12 (1): 51-97.

58

Vladimir A. Plungian

Xrakovskij, Viktor S. 2007 Èvidencial'nost’, èpistemiþeskaja modal’nost’, (ad)mirativnost’ (= Evidentiality, epistemic modality, (ad)mirativity). In Xrakovskij, Viktor S. (ed.), 600-632. Xrakovskij, Viktor S. (ed.) 2007 Èvidencial’nost’ v jazykax Evropy i Azii. Sb. statej pamjati N. A. Kozincevoj (= Evidentiality in the languages of Europe and Asia. In the memory of N. A. Kozinceva). Sankt-Peterburg: Nauka.

Hearsay in European languages: toward an integrative account of grammatical and lexical marking Björn Wiemer This article aims to survey and classify expressions with reportive (= hearsay) function used in European languages. Hearsay is a subdomain of evidentiality, which in turn is understood to be a functional domain pertaining to the cognitive and/or communicative ground on the basis of which the speaker makes an assertion. It therefore differs from epistemic modality. The purpose of this investigation is to look at hearsay expressions from an integrative perspective allowing on the one hand for them to be unified from a functional point of view and, on the other hand, for a typology to be developed that accounts for differences of formal properties like syntactic distribution and morphological format. The survey includes any kind of distinct units regardless of whether they are to be considered functional lexemes (e.g., particles, complementizers), auxiliaries, grammatical affixes or complex TAM markers. These expressions are located on a lexicon-grammar cline. The decisive criterion for inclusion is that they be sufficiently conventionalized in the given language and not just used as “ordinary” speech act verbs or other lexical items with a full argument structure. On this basis, each class of units is subclassified according to their morphosyntactic properties, diverse kinds of semantic and pragmatic properties are taken into account and illustrated, and gross areal clusterings are discussed.

1. Introduction The purpose of the present article is to provide a survey and classification of expressions with reportive function used in European languages. Reportivity is a sub-domain of evidentiality. Hearsay is understood to be a synonym, a cover term for different kinds of functions subsumable under reportive evidentiality. These functions, however specific they may be, all indicate that the speaker of the actual utterance bases his/her assertion on previous utterances, usually made by another person (other persons) but not necessarily so (see 3.5). I am not going to discuss some of the notorious principal issues that have lately been raised such as the relationship of evi-

60

Björn Wiemer

dentiality to epistemic modality, or the question of what counts as a conventionalized, holistic construction in this or that language. For the purposes of this paper, suffice it to say that I advocate a strict separation of evidentiality and epistemic modality on the conceptual and analytic level1 and am aware of the fact that many units marking evidential functions carry epistemic components (“overtones”), either additionally or even as a feature more salient than the evidential function. On the one hand, I have been eager to take into account as many types of reportive marking devices as we can find (with one regular exception, see Section 2) by including diverse kinds of distinct units for which a hearsay meaning can be considered as a stable meaning component, regardless of genre or register variation and frequency, and regardless of whether other components, e.g. epistemic ones, co-occur with the hearsay meaning. However, the range of meanings other than reportive will be accounted for, too. On the other hand, I have been anxious to include only those devices for which a reportive component can be accepted firmly enough, i.e., considered as a conventionalized part of its meaning. There are some borderline cases which will be looked at critically during the discussion of data in Section 3. I conceive of evidentiality as a conceptual domain. My understanding actually fits the definition given by Aikhenvald (2003: 1): “Evidentiality proper is understood as stating the existence of a source of evidence for some information; that includes stating that there is some evidence, and also specifying what type of evidence there is.” However, contrary to Aikhenvald, I apply this definition not only to markers that are considered to be grammatical in the strict sense, but also to lexical means, more specifically: to various classes of function words (see Section 2) whose semantics contains a stable, non-detachable reference to hearsay. This broader functional notion corresponds to Aikhenvald’s recently coined term ‘information source’ (Aikhenvald 2007). From this extension of the range of linguistic devices for marking functions from a given conceptual domain, we should distinguish the question of the degree to which a linguistic device can be regarded as sufficiently conventionalized (and not only as an evidential strategy), i.e., with a stable, non-detachable indication of an evidential or, more precisely, hearsay function. Therefore, the approach taken here starts from a functionalonomasiological perspective, i.e. we are searching for all kinds of distinct means serving to express evidential functions, especially if they convey some specific meanings from a taxonomy of evidential functions. Such (partial) taxonomies have been established, inter alia, by Aikhenvald 2004, Plungian 2001 and, at least indirectly, by Palmer (22001: 35-52), and more



Hearsay in European languages

61

specifically for hearsay by Anderson 1986. However, as soon as we have found such means, the onomasiological perspective changes into a semasiological one since we want to describe and classify these means on the basis of their semantics and structural properties such as scope, morphosyntactic format, and distribution (see below). Thus the classification grid to be introduced in Section 2 scrutinizes the formal properties of devices which serve as indicators of hearsay. These formal properties will, as far as possible at the present stage of our knowledge, be correlated with (i) subfunctions of reportive evidentiality and (ii) possibly interfering inferential functions and functions from other domains (see Section 3). I will group marking devices according to their morphological format and syntactic behavior; the principles will be argued for in Section 2, the classification along these principles will be carried out in Section 3, which constitutes the main body of this paper. The part of the overview in Section 3 which deals with grammatical marking sensu stricto (bound morphology, extensions of TAM grams), can be read as a critical assessment of the present state of investigations insofar as it concerns grammatical evidentiality marking, that has already been studied for an honorable span of time. The chosen procedure should allow these devices to be described and compared to one another independent of decisions as to whether the given marker has undergone a process of grammaticalization or lexicalization. I thereby want to avoid an involvement in ongoing disputes concerning the delimitation of grammaticalization from lexicalization (and other phenomena of language change and their results). From the perspective taken and the purpose of a morphosyntactically oriented classification of marking devices, such an involvement is unnecessary; we need not wait until such disputes might eventually have been abandoned or solved. Thus, although in Section 2 I will briefly “out” my stand regarding the difference between grammaticalization and lexicalization (viz. the relation between grammar and lexicon), first and foremost the prime purpose is a descriptive and exploratory one, namely: to work out a representative picture of form: function-correlations of contemporary languages in a delimited geographic area without any bias toward theories concerning specific processes of linguistic change. Here Europe is conceived of as the whole geographically defined subcontinent including its eastern edges since I want to integrate one better known language each from the Turkic family (Turkish in Turkey), the Indo-European family in the Caucasus (Armenian), from Kartvelian (Georgian), and from Finnic (Estonian). The survey is short of data on Scandinavian (except of Swedish) and also lacks information on Celtic, Basque,

62

Björn Wiemer

Hungarian, Greek and practically all languages with minor numbers of speakers, such as numerous Turkic, Finnic and Indo-European varieties in the Balkans and the European part of Russia. To my knowledge, the present study is the first attempt at a comprehensive account of a sub-domain of evidentiality (but see the preliminary study based on questionnaire data reported in Ramat 1996). I do not intend to draw any more far-reaching conclusions on areal patterns of formal classes of hearsay marking. An areal assessment would be premature given the fact that we are only beginning to understand along which parameters, both functional and structural, evidential marking ought to be estimated and classified, let alone the question of what can count as a sufficiently conventionalized structural or lexical indicator of, or functional extension into, evidentiality and, more specifically, into hearsay. Only a few salient features from among an enormously intricate network of factors can be illustrated, others can only be “scratched” upon, many others will remain unmentioned.

2. Hearsay markers on a lexicon-grammar cline What we are looking for presently is a viable classification of marking devices for reportive functions comprising not only the traditionally investigated grammatical means, but also distinct classes of words carrying hearsay meaning. By “viable” I mean that there should be a unified and theoretically justifiable basis applicable to differently structured (ideally to all) languages. From the start, we can exclude “trivial” means of marking hearsay like illocutive verbs, i.e., verbs with a ‘say’-component in their semantic description, with an “ordinary” argument structure implying a sentential complement and various subject NPs (e.g., People say/Somebody said that P) as well as steady collocations, or phraseological syntagms, like Legend has it/There are rumors (that P) etc. Among the first to formulate criteria for defining evidential units in both semantic and structural terms was Anderson 1986. His semantic criterion corresponds neatly to the approach advocated here: “Evidentials show the kind of justification for a factual claim which is available to the person making that claim”, as well as the following caveat: “Evidentials have the indication of evidence as their primary meaning, not only as a pragmatic inference” (1986: 274). In fact, Anderson restricted the notion of ‘evidentials’ to grammatical units (1986: 275): “Morphologically, evidentials are inflections, clitics, or other free syntactic elements (not compounds or de-



Hearsay in European languages

63

rivational forms).” He did not consider whether there might be a lexicalgrammatical distinction for other “free syntactic elements”. The same holds true for de Haan who, in refining Anderson’s criteria, added that evidentials do not show agreement with the speaker and cannot be in the scope of negation2 (de Haan 1997: 147-150; 2000: 75f.). Semantic (or functional) criteria like those provided by Anderson have hitherto been used to find evidential markers from among grammatical morphemes in the narrow sense; such criteria have been generally accepted. Since they are semantic ones, nothing hinders us from applying them to units other than grammatical morphology too, first of all to morphologically independent lexical units. Given this, we can “browse” through all sorts of morphemes in languages and then arrive at a cline of the shape illustrated in Figure 1. distinct lexical unit

1. particles

grammatical morphology 5. predicatives3)

2. complementizers2) 3. adpositions4) 4. sentential adverbs

require ‘hosts’ to take scope over, but non-relational (with the exception of complementizers)

8. functional extensions of TAM-paradigms1)

6. copular participles 7. auxiliaries1) 9. bound morphology (inflexion, agglutination) operate on heads (in a strictly morphosyntactic sense)

Remarks: 1) Often hearsay functions of modal auxiliaries show up saliently only in marked mood forms (see Romance modals discussed in 3.3.2). 2) This includes conjunctions. 3) Function as heads of PPs or of sentential dependents; in the latter case they combine with complementizers. 4) Heads of NPs (= PPs), PPs headed by them have propositional scope. Figure 1. Types of evidential markers on a morphosyntactic cline

64

Björn Wiemer

Presenting the types of units on a cline does not imply that units on one pole are the input of diachronic processes yielding an output at the other pole. The cline is simply meant as a very rough (and preliminary) representation of the fact that some morphemes (those approaching the left side) have independent word status and thus belong to the lexical stock of a language, whereas other units (closer to the right end) are bound morphemes and function as operators on heads (in a strictly morphosyntactic sense). In-between are units with an independent morphological status (in this respect, separate lexical units) which, however, function only as components of complex predicates (auxiliaries, copular participles) or are themselves heads of sentential arguments (predicatives). Finally, among functional units with distinct word status (on the left end), complementizers are the only class that takes part in narrowly understood dependency relations: they relate superordinate predicates to their propositional arguments which, as a rule, have a predicate-argument structure of their own. The other functional classes are used either as adjuncts (PPs headed by adpositions, sentential adverbs) or are outside clausal syntax (particles). For this reason, complementizers are put slightly more to the right than the other distinct function words. Parentheticals are not included in this cline because they are highly heterogeneous in terms of their morphosyntactic format; functionally they are closest to particles since, as a rule, they do not partake in clausal syntax. The status and role of evidential parentheticals will be discussed more thoroughly in Sections 3.5-3.6. The distinctions just mentioned should become clear in the course of the data discussion in Section 3. Between different morphosyntactic classes of possible evidential markers, manifold relations appear, arising mostly as the result of parallel or subsequent diachronic development from an earlier unit. Here is no place to dwell on this issue systematically, but see the remarks on ‘heterosemy’ in Section 3.9. Notice that on the cline there is no obvious place for monoclausal constructions like NcI-constructions with certain epistemic or illocutive verbs (suppose, claim etc.) or with perception verbs (like seem, appear) which, for different languages, have been claimed to develop reportive meanings (as for SEEM/APPEAR-verbs see Section 3.6). I have deliberately excluded monoclausal (or, following generative terminology, ‘raising’) constructions; not because they were not worth deliberating, but because even more space would be required to clarify the relation of these constructions to particles and “ordinary” sentential complements of the respective verbs (as well as the significance of constructions for evidentiality marking).



Hearsay in European languages

65

Returning to the cline, one might object that those items appearing as ‘lexical units’ on the left pole can, at least partly, be re-termed as ‘function words’. In fact, grammars and textbooks usually refer to adpositions, conjunctions (complementizers) and particles under this rubric. However, what would be gained by saying that bound morphology, TAM paradigms and auxiliary constructions belong to a language’s grammar as well as function words (adpositions, conjunctions etc.)? Such a qualification amounts to saying that all functional units of a language belong to its grammar. This sounds like a trivial, if not analytic assertion. It would not supply us with a useful basis for discriminating between different expression classes of functional elements (morphemes, word classes). Instead, a useful discriminatory basis should allow for operative and verifiable decisions about whether a given item is (rather) a distinct unit of its language’s lexicon (= inventory of lexemes) or (rather) a formative of more complex units on which it depends3. Certainly, there may be cases in which a definite and clear decision is difficult to take; but this is just a corollary of clines with focal points as the one introduced above. I have now come to the point where I have to take a concise position as to what shall be considered as grammatical(ization) vs. lexical(ization). The one does not exclude the other; the relationship between both has to be seen as the relation of a holistic unit (morphologically either simple or complex) to constituents of a larger syntactic format. I adhere to Lehmann (2002: 1) who formulated this relationship in the following condensed way: “A sign is lexicalized if it is withdrawn from analytic access and inventorized. On the other hand, for a sign to be grammaticalized it means for it to acquire functions in the analytic formation of more comprehensive signs.” Exactly such a withdrawal from analytic access is what happens when, e.g., a particle or a preposition evolves from petrified verb forms, regardless of whether it underlies erosion or not; compare particles like Cz. prý < praví.3.PL.PRES.IND ‘they say’, Russ. mol < †molvilɴ.3.SG.PAST.IND ‘he said’, various forms of the descendents of Latin dicere in Romance and also particles deriving from SEEM/APPEAR-verbs (see 3.5-3.6) or the Lithuanian hearsay preposition pasak ‘according to’ < pasakyti ‘to say’ (see 3.8). Since, for instance, hearsay particles have scope over whole propositions, in order to claim that they are “grammaticalized” we must first show what the more comprehensive signs look like, which format they have and in which sense the particles can be considered as inextricable parts of these comprehensive signs. Do we want to say that since particles are operators on whole sentences (or conjunctions on clauses) they are grammatical formatives (or components) of the latter? To my mind, to do so would mean to

66

Björn Wiemer

overuse the notion of ‘grammatical marker’. Independent of this, we still would have to describe every particle (conjunction, preposition, etc.) as a distinct unit of the lexical inventory of the language. This is what usually is neglected or overlooked in recent research. The following remark from Lehmann (2002: 16) applies to all functional word classes distinguished in this paper: “Those numerous complex prepositions and conjunctions which constantly come and go do not indicate incomplete grammaticalization processes but are simply products of lexical change.” (the emphasis is mine, BW). This makes them no less worthy of thorough investigation, and it is time to start doing this research. Consequently, the following survey and remarks concerning the classification of hearsay marking devices are also meant as a step toward the creation of a cross-linguistic data base establishing and describing the inventory of evidentiality markers. From this angle, this is a problem paper which certainly digs out more questions than supplies answers; hopefully, however, it helps to find the way toward solutions for at least some nagging problems.

3. Distribution of markers across European languages With respect to the areal distribution of morphological vs. lexical marking of evidentiality in the languages of Europe, we can generalize some observations that will be discussed for more detail in the main body of the following text in advance. In the eastern half of Europe there are three hotbeds of evidential extensions of perfect paradigms: the Balkans, the Caucasus, and the Baltics. Perfect paradigms of languages in the western “hemisphere” hardly ever show signs of evidential extensions. Instead, Romance and some Germanic languages (German, Dutch) show evidential extensions of marked moods (subjunctives) or their analytical substitutes formed with an auxiliary. This area overlaps with evidential extensions of modal auxiliaries, which are quite commonplace in the western part of Europe and pertain even into Western Slavic. Copular particles seem to be more or less restricted to the Baltic languages, whereas predicatives are typical for Slavic and Baltic and appear to be virtually absent in other language groups of Europe (for possible exceptions in Sardinian and Romanian, see 3.9). Evidential complementizers likewise seem to be rather restricted areally to the Baltic-Slavic area (see 3.7), whereas evidential particles can be found almost everywhere throughout Europe; in fact, this situation is less clear because of problems related to delimiting particles



Hearsay in European languages

67

from sentential adverbs, which form a frequent class of evidential modifiers, too (see 3.5). In a sense, bound morphemes indicating hearsay, either as a central function or as a functional extension, are common in at least some parts and/or language families of Europe. If we disregard tense-aspect and mood grams, which had acquired the status of inflectional endings long before they underwent evidential extensions, the number of languages with bound morphology specializing in evidential functions diminishes drastically. Within Europe we find them only on its eastern, especially its southeastern periphery (Eastern shores of the Baltic Sea region, Balkan, Southern and Eastern shores of the Black Sea region). But again, even here, in practically no case do these markers specialize in hearsay, instead the reportive function is only a contextually conditioned reading of a broad evidential meaning potential. All bound morphemes of this sort represent agglutinative morphology rather than inflectional affixes and almost all derive from the (present) perfect. A remarkable exception, both in terms of etymology and the range of evidential functions, is the Estonian suffix -vat which, like the predicative active participles used in Estonian and the two Baltic languages, is restricted to hearsay (see 3.1 and 3.2.1). The following overview starts with hearsay marked on, or by, the predicate (verb phrase). Bound morphology which has acquired evidential functions will be looked at first (3.1) and separately from cases in which TAM paradigms have experienced functional extensions into evidentiality but not necessarily turned into unequivocal evidential markers (3.2). Clitics will be dealt with in 3.5 and 3.9. I am aware that in many cases it is arguable whether a clitic has turned into an agglutinated affix. As long as the unit in question does not attach in a fixed position to a specific kind of word form (without any “material” intervening its connection with the stem), I will regard it as a clitic and not as a word-form internal morpheme.

3.1. Bound morphology: inflection and agglutination The only example of an inflectional gram used for evidential purposes in European languages is the Georgian perfect. It forms part of a threefold series of stems (Boeder 2000: 278). In the standard language evidential readings (hearsay or inferential) are not marked by any additional morphology (Boeder 2000: 277). It is only in dialects of Western Georgia and adjacent Kartvelian languages where we find “a richer system which provides special evidential verb forms as counterparts not only for the non-

68

Björn Wiemer

evidential aorist but also for the present, imperfect, and future” (Boeder 2000: 309).4 If used with values of indirect evidentiality, the standard Georgian perfect allows for definite time adverbials and can thus principally be used as the “propulsive tense” of a narrative (Boeder 2000: 296 and 317, f. 39). An analogous release of restrictions in evidential narratives can be observed with the Turkish mIú-suffix (Johanson 2000) as well as in Baltic, where perfect participles are encountered as the basic tense of narrative texts in folklore (cf. Wälchli 2000). This suffix is probably the most frequently mentioned case of an agglutinated evidential marker in Europe. In fact, cognates of -mIú are attested in practically all Turkic languages. But even if we restrict ourselves to standard Turkish (being spoken on the edge of Europe), it is essential to distinguish the agglutinated suffix from the nearly homonymous (and cognate) copula particle ImIú. These morphemes differ both in their meaning potential and their distributional properties, including their impact on word prosody (cf. Csató 2000: 36-39; Johanson 2000: 80f.). A short comparison of both units will be given in 3.9. At the moment it is important to realize that the suffix -mIú can be added only to primary verb stems, i.e., stems that are not modified by other grammatical affixes. It is itself an anteriority marker able to focus on resultative states and therefore tightly associated to core meanings of perfect grams; this obviously gave rise to its contemporary meaning of ‘indirectivity’ (according to Johanson 2000), among which hearsay and inferential interpretations figure prominently. The connection with the perfect is never dissociated. Since -mIú can also be used for different kinds of inferential judgments, we can qualify it as a general marker of indirect evidentiality. In this respect the meaning potential of -mIú closely resembles the Georgian perfect (see above) as well as Armenian, Balkan Slavic (see 3.2.1) and Albanian (see below).5 Another agglutinated suffix exists in standard Estonian, namely -vat. In contrast to Turk. -mIú (and its cognates in other Turkic languages), Est. vat is totally unrelated to the perfect system; it originates from a merger of the marker -v(a) of the present active participle and the partitive ending -t (Kehayov 2002: 129); compare, e.g., Ta tule-vat ‘S/he has come, they say’ with the non-reportive past Ta tule-b ‘S/he has come/came’). Another difference in comparison to the Turkic (and Albanian kam ‘have‘, see below) construction is the restriction of -vat to hearsay, i.e., it is not used in inferential function. This parallels the outstanding functional specialization of active participle constructions in Estonian and the Baltic languages (see 3.2.1).



Hearsay in European languages

69

The last language to be mentioned in this subSection is Albanian. Obviously, it is the only European language in which the functional extension from perfect to evidentiality has to some degree been accompanied by coalescence (in the sense of Lehmann 1995): the inflected HAVE-auxiliary agglutinates to the abridged participle, and it does so in reverse order compared to the perfect proper6, which is still productively used as an analytical tense: (1)

Albanian perfect kam hapur > have.PRES.1.SG open.PART ‘I have opened’

admirative hap-kam open.have.PRES.1.SG ‘I have opened’ (ADM/EVID).7

Albanian is striking as it is the only language of Europe (including its “edges” in the Black Sea region) in which the original perfect auxiliary incorporated so intimately with the verb stem (the participle) that this morpheme combination has been reinterpreted into a marked form of the p r e s e n t tense and given rise to a whole set of admirative (and evidential; see f. 8) paradigms combinable with any tense and mood except the aorist (cf. Breu 2009: 3.3 and Friedman 2000: 342f. for details). None of the following languages show verb morphology marked for evidentiality in the present tense: Balkan Slavic, Turkish, standard Georgian, and Armenian. This contrasts with Georgian dialects (see above), Estonian and Baltic (see 3.2.1).

3.2. Functional extensions of TAM-paradigms Since the TAM domain comprises an agglomerate of all kinds of grams with different combinations of tense, aspect and mood distinctions, I will divide these devices of hearsay marking into subtypes: first come grams dominated by tense-aspect components, then grams for which mood distinctions are central. The tense-aspect part further splits into perfect grams, future grams and past grams; the last mentioned are the most problematic ones (see 3.2.3). A decision as to whether these extensions have become new meanings of the paradigm or whether they are to be treated just as evidential strategies in Aikhenvald’s sense cannot be made in a general manner but must be considered for every single case.

70

Björn Wiemer

3.2.1. Extensions from analytic perfects Beside agglutinative perfect grams (Albanian and Turkish) and the inflectional perfect of Georgian, treated in 3.1, evidential extensions are attested for languages with perfect grams of an analytical structure based on participles. These are: (a) the extant Baltic languages (Lithuanian, Latvian) and some Finnic languages, first of all Estonian (but not Finnish!), (b) Balkan Slavic (Bulgarian, Macedonian), and (c) West Armenian. The arrangement of these languages (a-c) is meant to indicate that they belong to three areal clusters, noticeably all of them on the Eastern periphery of Europe. At least Balkan Slavic and Armenian form part of a much larger convergence area (see f. 6). Armenian has to be treated with some reservation, insofar as Modern East Armenian has lost its core perfect character, since it can be used also “in narration as a nonevidential past tense” (Kozinceva 2007: 81; cf. also Kozintseva 2000: 407f.). This does not apply to Modern West Armenian, in which the perfect is not used as a narrative tense and closely resembles both the paradigmatic and functional distinctions of Balkan Slavic: their perfects have not lost their character of typical perfects (Donabédian 1996: 88f.).8 As concerns Balkan Slavic, Bulgarian and Macedonian conform to each other in almost every respect; their “dialects show considerable morphological variation in their treatments of evidentials [as do varieties of other Balkan languages; BW] (…), but the underlying semantics are fairly consistent” (Friedman 2003: 192). The Balkan evidential system has sometimes been characterized as a highly “epistemicized” one, i.e., that an epistemic judgment usually accompanies the evidential function (Plungian 2001: 354); others, like Friedman (2000, 2003), maintain that it is the “speaker’s attitude” toward the information conveyed, not reference to the “source”, that is the decisive motif behind these forms. The same has been claimed for Vlach Romani. This Balkan Romani variety shows a split of inflected past tense forms vs. bare active past participles. This split is restricted to the 3rd person singular (cf. Matras 1995), but it nonetheless very much resembles the Balkan Slavic one to be discussed further. The problem with the alleged epistemic component of evidential perfect forms on the Balkans is that time and again investigators have come up with examples and context types in which no necessary epistemic load can be imputed, e.g., into the Bulgarian (or Macedonian) forms (see, for instance, ex. 2, where epistemic connotations are absent). This notorious bone of contention has remained extremely confusing. Whatever stance



Hearsay in European languages

71

one might take towards this issue, hearsay is only one possible interpretation of the paradigmatic forms referred in Bulgarian grammar as ‘preizkaznoto naklonenie’ (‘renarrative mood’), in Macedonian grammar as ‘prekažuvanje’ (‘renarration’). Thus in an utterance like the following one, the speaker indicates only that s/he has not witnessed the described event him/herself; out of context it remains indeterminate as to the more specific evidential value, so that the hearer is free to interpret the form peel ‘sung’ as either inferential or reportive: (2)

Djado mi (e) pe-e-l COP.PRES.3.SG sing.IMP.PPA:SG.M Grandpa me.DAT v cărkva-ta. in church.DEF i. ‘(My) Grandpa is said to have sung in the church (choir).’ ĺ hearsay ii. ‘Obviously, (my) grandpa sang in the church (choir).’ ĺ inferential, or perfect (‘indefinite past’)

The form of the predicate peel goes back to the Common Slavic lparticiple, originally the active participle of the indefinite past used as a perfect. In most Slavic languages, this form has ousted the synthetic preterits (aorist, imperfect) and changed into a general past tense, but in Bulgarian and Macedonian the paradigmatic opposition between perfect, aorist and imperfect has survived without restrictions. The meaning of the (indicative) perfect often interferes with evidential readings (plus a possible mirative one); absence vs. presence of the copula (e in ex. 2) is no reliable criterion (see below). There is, however, a point concerning the formal expression of the functionally extended perfect which makes it clearly distinguishable and which can likewise be illustrated from ex. (2). In contrast to Georgian, Armenian and the Baltic languages, Bulgarian and Macedonian have introduced a new combination of the perfect marker (the l-suffix) with a derived stem of the imperfect. The “evidential” participle peel is made up of the imperfect stem pe-e- plus the l-participle. Both suffixes (i.e., an allomorphic ‘a/’ealternation for the imperfect expanded from the infinitive stem, and the lsuffix) belong to common Slavic heritage, whereas the combination of both does not. This innovation occurred relatively late and was conditioned by the establishment of a series of evidential forms (for a simplified, yet instructive table showing the paradigm structure cf. Friedman 2000: 330).

72

Björn Wiemer

The two main differences between the Bulgarian and Macedonian perfect series (analytic tenses) showing evidential extensions (Friedman’s ‘Nonconfirmative’) are the following: (i) Macedonian never uses a copula in the third person, even if used without evidential function (Friedman’s ‘Unmarked Past’); (ii) Macedonian has developed a resultative perfect with the HAVE-verb (imam.PRES.1.SG + indeclinable participle) whose l-participle (i.e., actually the perfect) forms the basis of an additional evidential paradigm (e.g., Si imal napraveno ‘[Apparently] you did /have done (it)’). The rise of the HAVE-perfect began in the Southwestern area (Ohrid, Struga) and was obviously triggered by the model of the Albanian HAVE-perfect (see 3.1; cf. also Breu 1987: 293f.). In sum, a functional characteristic concerning most languages mentioned so far is that the reportive function of the perfect can be qualified only as a subfunction within indirect evidentiality (including admirativity). Exceptions are the northern-most languages; Latvian and Estonian as well as Lithuanian, as far as active participles are concerned (see below). Furthermore, in all these languages, most (in standard Georgian even all) forms of the indicative tenses are “homonymous” with the ‘evidential series’. The “copula criterion” stated by normative grammars does not stands the test of empirical distribution in discourse in Baltic, Balkan Slavic, and Georgian. This fact corresponds to the weak paradigmatic distinction of the evidential series in these languages.9 Nevertheless, the distinction from the respective perfect paradigms is greater than in the case of mood or aspecttense paradigms with evidential extensions in other European language groups (see the following subSections). Finally, in all languages mentioned in this subSection, paradigmatic weakness corresponds to functional weakness: in no case are the forms obligatory (cf. also Friedman 2003: 193, 209f. on Balkan Slavic and Albanian). Now let us have a closer look at the Baltic languages and Estonian. All three languages possess series of active participles from the present and past – in the Baltic languages also from the future – stem, with which the copula is normally absent or itself coded in a non-finite (participle) form (see, for instance, f. 13, also Wiemer 2007a: 207-210). In this sense we may say that these languages display greater “paradigmatic consistency” over tenses than Balkan Slavic, Georgian, Armenian and Turkish, in which participle-based evidentiality marking is restricted to forms deriving from past tense and present perfect. Diachronically, in Baltic the present-tense based verb forms with reportive functions do not derive from reinterpretations of the perfect; the Baltic and Estonian present tense evidentials should rather be explained by a converging effect of two factors: (i) analo-



Hearsay in European languages

73

gy with an evidential reinterpretation of the perfect (which is formally identical with the Balkan Slavic one) plus (ii) logophoric constructions which in time acquired independent syntactic status (for the details cf. Wälchli 2000: 191-197; Wiemer 1998: 233-239; 2007a: 228-234). As a consequence, the forms used in Baltic and Estonian vary quite considerably in the non-past domain: Latvian and Estonian are actually more similar to each other than the genetically closely related Latvian and Lithuanian (cf. Holvoet 2007: 81-91 and Wälchli 2000: 188-190 for details). This, by the way, clearly indicates that language contact has played a considerable role in the spread of evidential marking by grammatical means in the eastern part of the Circum Baltic Area. Compare the following examples which illustrate the past evidential, whose form is closest to the perfect indicative: (3)

(4)

(5)

Lithuanian Tơvas sugrƳžĊs iš miško. return.PPA:NOM.SG.M father.NOM.SG.M ‘Father apparently (they say) has returned from the forest.’ Latvian ViƼš dzƯvojis. live.PPA:NOM.SG.M he.NOM ‘He has lived, they say.’ Estonian Ta elanud. live.PPA s/he.NOM ‘S/he has lived, they say.’

There is only one basic formal difference in Estonian as compared to Latvian and Lithuanian: the active participle functioning as evidential marker is uninflected for those nominal categories which are usually marked on Estonian participles. Compare ex. (5): the form elanud does not show any agreement, contrary to sugrƳž-Ċs (3) in Lithuanian and dzƯvoj-is (4) in Latvian. Different kinds of reinforcement by a doubling of participles on the copula occur (as they do in Balkan languages, too); cf. Holvoet (2007: 9799), Wiemer (2007a: 215). Forms based on a pluperfect are also created, but their time reference quite often does not differ clearly from the time reference of reportive forms based on the present perfect (cf. Kehayov 2002: 129f., 136f. on Estonian). Lithuanian (more precisely, standard Lithuanian) is specific insofar as it has two different participial constructions which tend to be distributed

74

Björn Wiemer

nearly complementarily in terms of evidential subdomains. It thus differs from the two other languages both functionally and with respect to concurring constructions. All participle constructions illustrated in (3-5) are basically restricted to hearsay. But Lithuanian has an alternative construction with an invariant participle for the present and the past (or perfect) tense, ending in -ma (simultaneous to reference time) and -ta (anterior to reference time), respectively: (6)

Krnjmuose vištǐ perơta/perima. brood.TA/brood.MA bush.LOC.PL hen.GEN.PL ‘In the bushes (some) hens (must) have brooded/are [must be] brooding their eggs.’

Sentence (6) could be uttered after the speaker has found a couple of eggs in the bushes but does not see the hens; s/he thus concludes that hens must have sat there. This construction is used almost exclusively for inferential evidentiality; it is, therefore, in (almost) complementary distribution with the agreeing participle construction illustrated in (3). From the structural point of view, this construction does not have any agreeing NP. Instead, the highest-ranking (= most agent-like) argument must be coded with the genitive (vištǐ).10 Latvian and Estonian do not have a specialized construction for inferential evidentiality, nor do the other European languages with evidential extensions of perfect paradigms. There is, however, another detail worth mentioning. Latvian and Estonian can make reportive marking more explicit by using an uninflected form of the copula BE.11 Actually, this is just an expansion of the constructions in (4-5): Latv. ViƼš esot dzƯvojis, Est. Ta olevat elanud ‘He lived, they say‘. This construction is likewise restricted to hearsay function. In Lithuanian esą, the lexical and morphological cognate of Latv. esot, is used for this purpose, too: (7)

Jis esą gyvenĊs COP.INDECL live.PPA:NOM.SG.M he.NOM Vilniuje dešimt metǐ. ‘In Vilnius he (has) lived, they say, 10 years.’12

In this case, Lith. esą sort of reinforces (or stresses) the reportive value (whatever that may mean), as do its Latvian and Estonian “cousins”. But, contrary to them, esą can occur alone and behaves like a particle, sometimes even like a conjunction or the predicate (cf. Holvoet 2007: 85f.;



Hearsay in European languages

75

Wiemer 2007a: 217-220; see 3.6, 3.7 and 3.9). This makes it differ from Latv. esot and Est. olevat in grammatical and lexicographic terms. We may say that the particle and complementizer uses of Lith. esą are the result of the loss of paradigmatic relations with the inflected forms of the copula proper (including its inflected participles!), whereas in Latvian and Estonian this relation obviously has not been cut off.

3.2.2. Extension from future gram Future markers, in particular grams that can be classified as future perfect (futurum exactum) as indicators of inferential evidentiality, are widespread in Europe. They are commonplace in Romance, Germanic, Baltic and South Slavic. But practically nowhere do they expand into the domain of hearsay, the only exception being Portuguese. As Squartini (2001: 319) writes, this is “a case which (...) is unique in Romance and has never been pointed out before“. Compare one of his examples quoted below: (8)

Segundo fontes que lhe são próximas Soares terá dito a Gomes Motas ea Carlos Monjardino que as criticas à liderança de Guterres foram “pura locura”. ‘According to sources which are quite close to him, S. said [lit. will have said] to G.M. and C.M. that the criticism of G.’s leadership was pure nonsense.’

Obviously, this case is unique not only when compared to other Romance languages, but also to all other European languages. 3.2.3. Extensions of past tense grams Squartini (2001: 308-314) gives a conscientious account of usage types of the Italian imperfect (Imparfetto) which have repeatedly been considered to carry reportive meaning: “the evidential value expressed by the Imperfect is a form of report of what the speaker knew was due to happen and about which confirmation is lacking”. An important condition necessary for the Imparfetto to acquire this function is that “the reported information refers to a scheduled situation” (Squartini 2001: 311). Examples which corroborate this conclusion are utterances in which the speaker presupposes knowledge shared with the interlocutor. It is important that aspectual values usually ascribed to the Imparfetto can be cancelled, as in (9) where the

76

Björn Wiemer

adverbial fino alle sei del pomeriggio ‘until six o’clock in the evening’ delimits the time span and thereby normally requires the Passato Remoto (= aorist) or Passato Prossimo (= perfect or generalized past): (9)

Paolo quel giorno lavorava fino alle sei del pomeriggio. ‘On that day Paolo was to work [lit. worked.IMP] until six o’clock in the evening.’

Here the Imparfetto can only convey reportive meaning, which arises by implicature from the fact that the speaker possesses only indirect knowledge about the described state of affairs and that s/he could obtain this knowledge only via hearsay. What further corroborates Squartini’s conclusion is the fact that the assumed reportive use of the Imparfetto patterns exactly like the conditional, which in modern Italian conveys only a reportive but no inferential function (see 3.2.4). However, despite this clear parallel and the fact that the aspectual core meaning of the imperfect can be overridden, there remains the question of whether we can really ascribe a new conventionalized meaning to the imperfect (and not just a contextual implicature).13 This question can be answered in the affirmative for the Estonian modal auxiliary pidama14 whose past tense form can acquire reportive meaning (see 10a). According to Erelt (2001: 16), this functional extension is to be observed in the fact that past tense15 forms of pidama (+ infinitive) “can be replaced by the form of the morphological quotative” (= reportive), as in (10b): (10) a. Praegugi pid-i neid mitme-s many-INE now_too must-PAST.3SG they.PART koha-s maa-s vedele-ma. ground-INE lie-INF. place-INE b. Praegugi vedele-vat neid mitme-s many-INE now_too lie-PRES.REP they.PART koha-s maa-s. ground-INE. place-INE ‘Even now some of them [little slips of papers] are reported to be lying on the ground in various places.’ As Erelt (2001) stresses, only the past tense of pidama can be used with hearsay meaning, at least in contemporary standard Estonian. Usage patterns typical of the older written language and some Northern dialects also



Hearsay in European languages

77

show indicative present forms in this function which have been related to German influence (compare the modal sollen; see 3.3.2). Pidama can occur together with other grammatical forms that are associated with reportive meaning, first of all with the vat-suffix (see 3.1); cf. Erelt (2001: 16f.), Kehayov (2002: 136). Notably, the past tense form pidi evokes an imminence effect (‘was about to /on the verge of’; cf. Erelt 2001: 12-15). This effect can probably be linked with the reportive function on the basis of intentional stage and presupposed knowledge, which have been shown to be crucial components for a reportive interpretation of the Italian Imparfetto, too.

3.2.4. Extensions from moods Let us now look at hearsay extensions of mood grams. A clear case in point is German, which uses the subjunctive (Konjunktiv I and II) for this purpose. Typically it occurs in the news, as in the following example with the Konjunktiv I from a TV report (January, 13th, 2007): (11)

Angela Merkel hielt sich bedeckt //die arbeite gut und gerne mit Stoiber zusammen ‘Angela Merkel remained guarded //she is said to cooperate well with Stoiber.’

Here the subjunctive form arbeite (vs. indicative arbeitet) serves as an indicator that the author of this utterance did not produce the assertion ‘AM likes to cooperate with St.’ herself. Konjunktiv II works the same way; actually the only difference between both subjunctives lies in (relative) time reference: Konjunktiv II is based on the pluperfect and thus can always be recognized from the auxiliary (hätte vs. hatte and wäre vs. war), whereas the forms of Konjunktiv I are to a large part homonymous with the indicative (present and perfect);16 cf. Roels et al. 2007. The hearsay function of the German subjunctives most clearly arises in unembedded clauses (see ex. 11) where it developed from its original, merely syntactic function, namely: as an indicator of indirect speech. This diachronic background certainly explains why the evidential extension of the subjunctive is restricted to hearsay and does not include inferentiality. The same holds for some dialects of Rhaeto-Romance, in particular for Surselvan, where the subjunctive is exploited as a means of marking hearsay in main and subordinate clauses (Grünert 2003: 536-540). This beha-

78

Björn Wiemer

vior clearly departs from the usual pattern in Romance (see below) and can most probably be explained by contact with German. In Dutch there is no f o r m a l equivalent of the German situation, since Dutch does not have a subjunctive, or conditional, as an inflectional category of the verb. Instead, an analytical device is exploited, for which see 3.3.1. In Romance the situation differs from German in some significant respects. In these languages it is the conditional which is the basis for evidential extensions; and in most Romance languages, these extensions are not restricted to hearsay but result in an undifferentiated indirect evidential function (see Section 4). See the following examples (quoted from Squartini 2001: 318, 306): (12)

(13)

Portuguese Segundo certas vozes, teu pai teria reunido a esta altura um bom pecúlio. ‘According to some rumors, your father had saved [lit. would have saved] a lot of money at that time.‘ French Aux dernières informations, les concurrents auraient franchi le Cap Horn. ‘According to the latest information, the competitors have rounded [lit. would have rounded] Cape Horn.’

Among the contemporary Romance languages, Italian stands out insofar as its conditional can no longer be used with inferential function, being restricted to reportive evidence. However, this situation must have occurred only recently, because during the 19th century the Italian conditional could still mark inferential evidentiality. Modern Italian appears to be almost unique among contemporary Romance, since it no longer shows an overlap between the evidential functions of the future tense (ĺ inferential) and the conditional (ĺ reportive); cf. Squartini (2001: 307f., 314f., 324-327). It is only in Italian where we observe a clear-cut complementary distribution between mood and tense grams: the French and Portuguese conditionals allow for both interpretations whereas in Spanish (and Catalan), the conditional as a means of marking evidentiality seems to be stigmatized in general, and the Romance future is generally restricted to inferential evidentiality (with the remarkable exception of Portuguese; see 3.2.2); cf. the systematic and comprehensive study Squartini 2001.



Hearsay in European languages

79

Only Romanian comes close to the Italian distributional pattern (Squartini 2005: 252), provided we account for the additional complication that in Romanian both conditional and future can combine with the presumptive. This is basically a modal category, which in Romanian can be exploited for evidential purposes but obviously only as a strategy that allows for either reportive or inferential interpretations. Its formal expression is based on the combination of a gerund of the lexical verb with an indeclinable (petrified) form of an auxiliary verb usually called a (modal) particle; the gerund usually derives from present tense (-înd), although many researchers admit that past gerunds are possible, too (cf. Friedman 2000: 350f.; Squartini 2005: 252-263). This formal structure bears similarity to the structure of the Baltic and Estonian active participle constructions, which can be based on the present tense, too (see 3.2.1).

3.3. Auxiliaries Similarly to TAM-paradigms, this subSection divides into tense-aspect auxiliaries, mood auxiliaries, modal auxiliaries (briefly: modals) and volitional verbs. However, within the domain of tense and aspect the only auxiliaries relevant would be copulae as components of analytic perfects. These were already discussed in 3.3.1 where we saw that the distinction between perfect (indicative) and evidential (reportive) paradigms tends (or, has been claimed) to hinge on the presence (ĺ perfect) vs. absence (ĺ evidential) of the copula. We may, therefore, immediately turn to mood auxiliaries.

3.3.1. Extensions from mood auxiliaries Cases in point exist in Germanic. Germ. würde (+ infnitive) is used as an analytical substitute of the subjunctive (see 3.2.4). Historically, this is the morphological past subjunctive of werden ‘to become’ (Roels et al. 2007: 189; Smirnova 2006). We can regard würde as an auxiliary because it has a defective paradigm (no infinitive or other infinite forms, only present tense forms) and combines only with the infinitive of the lexical verb. This combination can be considered as an analytical mood, which has increasingly been replacing the simple subjunctive (discussed in 3.2.2) both in sentential arguments of illocutive verbs (indirect speech) and in syntactically independent sentences in which it clearly marks hearsay.17 Compare the

80

Björn Wiemer

following example in which the würde-construction co-occurs with the Konjunktiv I (underlined): (14)

Vor allem der Kommunikationsgedanke stehe im Vordergrund, betonte auch Jeanette Klee. Schließlich gebe es einige ältere Menschen, die zum Teil auch nicht mehr in der Lage seien, für sich selbst zu kochen, oder Gesellschaft während des Essens bevorzugen würden. ‘Above all the idea of communication is in focus, also Jeanette Klee emphasized. It is said that after all there are some elderly people who are no longer capable of doing the cooking for themselves, or who would rather prefer to take their meals in company.’ (Mannheimer Morgen, 17.12.2004)

In Dutch we find a structural equivalent which functions in a similar, though not identical manner. The auxiliary zou(den)18 is the usual device to form an analytical conditional; morphologically these are the past tense forms of the auxiliary zullen, which is etymologically related to Germ. sollen and Eng. shall. As a conditional auxiliary, zou(den) is used both in main and subordinate clauses. Thus, syntactically, it behaves like Germ. sollen. Zou(den) is also used as a marker of reported speech like the Germ. sollen which, however, is not a mood marker, but a modal auxiliary (see 3.3.2). As a mood marker, zou(den) parallels Germ. würde, and also serves as a reportive marker; but, in contrast to würde (and sollen), zou(den) indicates reported speech only in main, not in embedded clauses.19 (15)

Hij zou de moordenaar van het meisje zijn. ‘He is supposed to be the murderer of the girl.’

In sum, as reportive marker German würde + infinitive is less restricted as for its admissible grammatical contexts and, to this extent, can be considered as more grammaticalized than zou(den), which, thus, shares only the semantic restrictions in the domain of evidentiality (i.e., only hearsay) with Germ. würde and sollen. Its categorial function as mood auxiliary differs from its German cognate sollen (which is a typical modal) but overlaps considerably with würde (cf. the systematic Dutch-German comparison in Roels et al. 2007).



Hearsay in European languages

81

3.3.2. Extensions from modal auxiliaries This pattern is fairly well attested in Germanic and Romance. The modals concerned in Germanic all belong to the field of deontic necessity (MUST, SHOULD), whereas in Romance modals of possibility (CAN) are also encountered. Furthermore, evidential extensions of German (but not Dutch) modals are clearly restricted to hearsay, whereas in Romance they consistently lead to undifferentiated indirect evidentiality. Another difference between German and Romance lies in the fact that in Romance, evidential meanings of modal auxiliaries are very intimately bound to grammatical mood since it is first and foremost their conditional forms that carry such meanings (B. Staib, p.c.), in particular a reportive one. In German, by contrast, reportive extensions of mood and modal auxiliaries can be kept apart neatly. This will be illustrated in the following. Germ. sollen is used not only as a deontic modal (‘should, ought to‘), but also as a marker of hearsay. This use, however, is restricted to the present tense forms20 (see 16-20). The forms sollte(n) – morphologically the past tense of soll(en) – can be used only to mark a conjecture (or assumption), apart from a deontic use.21 Potential ambiguities with a deontic reading are dissolved22 in accordance with the (un)controllable character of the denoted state of affairs: if the intended illocution refers to a controllable situation, the deontic meaning prevails, if the situation is conceived of as uncontrolled (i.e., as a simple description), the reportive reading comes to the fore (for a detailed analysis of context factors cf. Diewald 1999: 278282). Compare (17a-b): without context, (17a) is ambiguous, but the slight change in (17b) forces a controlled action so that the hearsay reading is ousted (or at least heavily backgrounded): (16)

In China sollen sie ganz andere Feiertage haben. ‘It is said that in China they have totally different feasts.’

(17) a. Er soll im Bett liegen.

(i) ‘He is said to be lying in bed.’ ĺ hearsay (ii) ‘He has to lie [= stay] in bed.’ ĺ demand

b. Er soll im Bett liegen bleiben. ‘He has to stay [lit. stay lying] in bed.’ ĺ demand (deontic) /hearsay

82

Björn Wiemer

Conversely, if sollen combines with an anteriority (“perfect”) infinitive, which marks an event prior to, or a resultant state simultaneous to, reference time, a reportive reading is triggered, a deontic blocked (see 18). Notice, therefore, that this distribution is not restricted to telic verbs since it does not hold solely for resultative states (as in 18) but applies to any situation anterior to the time of reference, as shown by (19-20): (18)

(19) (20)

Er soll liegen geblieben sein/hingefallen sein/ein schönes Bild gemalt haben. ‘He is said to have remained lying/to have fallen/to have painted a beautiful picture.’ Er soll (schon einmal) Schach gespielt haben. ‘He is said to have played chess (once before).’ In früheren Jahrhunderten sollen die Menschen mehr auf dem Land gearbeitet haben. ‘In earlier centuries, people are said to have worked more in the country.’

Sollen never implies an inferential meaning, i.e., it patterns complementarily in comparison to its superficial past tense form (see f. 22). Sollen can be used with reference to anonymous, generalized or contextually unspecified original speakers as well as to specified speakers mentioned in the closest context (Mortelmans 2000: 134). In this respect, it is slightly more restrictive than the subjunctive, which can be used even if the actual speaker refers to an earlier assertion of him/herself (Diewald 1999: 229f.). A deontic-reportive polysemy characteristic of Germ. sollen is observed also for the Polish, Czech and Slovak verb ‘to have‘ (Pol. mieü, Cz. mít, Slov. maĢ) used as an auxiliary. We encounter the same restriction to reportive (and exclusion of inferential) meaning; most probably the functions of HAVE were calqued from Germ. sollen. Curiously, in all three West Slavic languages the HAVE-verb can be used as hearsay device also in its past tense forms (21), in contrast to German. Cf. an example from Polish (cf. also Hansen 2001: 137f. as well as Rytel 1982: 49f. on Polish and Czech): (21)

Miał zapomnieü swój parasol na dachu samochodu. ‘He is said to have forgotten his umbrella on the roof of the car.’

However, the Upper Sorbian HAVE-verb (mČü), which also “copies” the deontic-reportive polysemy of Germ. sollen, shows the same restriction to



Hearsay in European languages

83

present tense if used as a hearsay device (Faßke 1981: 77f.). Here the parallel with the German modal is even stronger and corroborates the assumption that the reportive function of West Slavic HAVE-modal has been calqued from German sollen. Let us now come back to Germanic. Based on the data presented here, it is worth remarking that in modern English no single modal belonging to the former series of praeteritopraesentia (see f. 21) shows hearsay extensions. Notably, Old English did know sculan ‘should‘ as a reportive marker, although this usage was predominantly bound to subordinate clauses embedded in contexts of overt speech acts (Traugott 1989: 41f.). Since in this paper we are not concerned with diachronic issues, I am not going to dwell further on the relation of OE. sculan to Eng. should. There is, however, a secondary modal with a different etymological background, which in modern English has been gaining hearsay function, namely: be supposed to; cf. a short remark in Chafe (1986: 268) and the corpus-driven analysis by Moore 2007. Her analysis makes one understand that a possible reportive reading of utterances like (22)

The Brownings were supposed to be in Italy at that time.

arises from the speaker‘s (writer‘s) wish to evoke “an indeterminate but external source for the belief or expectation” (2007: 119). In 3.3.1, I mentioned the hearsay use of the Dutch mood auxiliary zullen, a cognate of Germ. sollen. Notice that in terms of Dutch, only the past tense forms zou(den) are used reportively, whereas with Germ. sollen, reportive use is restricted to the present tense forms. These cognates are therefore diametrically opposed in terms of their categorial distribution. From this perspective, the Swedish cognate modal ska occupies an intermediate position, behaving in a way slightly closer to its German than to its Dutch cousin. Ska is the present tense form and used as a reportive marker. According to Kronning (2007: 300-303), its past tense form skulle can likewise be used for the purpose of hearsay. This reading proves, however, to be difficult to get – contrary to ska – if there is neither a lexical specification of hearsay in the context nor any such hint in the speech situation (2007: 302f.). Evidently the past tense form skulle does not mark hearsay by itself, but it is compatible with such a reading, similar to German sollte(n) (see f. 22). Following Kronning’s exposition, both tense forms, ska and skulle, are likely to have developed as an extension either from deontic or from alethic (viz. dynamic) modality.

84

Björn Wiemer

Apart from ska, Swedish knows another modal used as hearsay marker, namely: lär. Etymologically, it derives from lärä ‘to seize, apprehend, learn’ and has become void of all its paradigmatic forms except the present indicative. Like ska, its evidential function must have evolved from modal meanings, either deontic or epistemic. The reportive interpretation of both ska and lär, if used alone, is easily available for contexts in which the reported event is anterior or simultaneous to the actual speech event; only quite specific contexts trigger a reportive interpretation of ska and lär if the reported situation is posterior to the speech event. In this respect, these Swedish auxiliaries are similar to Fr. devoir (cf. Squartini 2004: 874-879), albeit not wholly identical to it (Kronning 2007: 293-296; cf. also Dooley/de Haan 2006). However, lär and ska can be combined to one complex unit whose unequivocally reportive meaning does not depend on the linguistic context, also with regard to situations posterior to the speech event. Other modals of NECESSITY seem to be less well represented. Dutch moeten ‘must’ is a potential case in point; its hearsay reading seems to be triggered only by favorable context conditions, whereas the evidential default interpretation is the inferential one (F. de Haan, p.c.). Cf. the following example from de Haan (1999b: 76): (23)

Het moet een goede film zijn. i. ‘This must be a good film.’ ii. ‘This is said to be a good film.’

Neither Germ. müssen nor Engl. must show any remarkable signs of an extension from inferential to reportive function. Mortelmans (2000: 137f.), who claims that this happens, does not give any English example; nor is the single German corpus example she adduces convincing, although Mortelman’s implicit intuition concerning discourse conditions favoring a shift from inferential to reportive meaning is probably correct. The example constructed by Ehrich (2001: 150) deserves an analogous remark: Das Testament muß unterschrieben sein (habe ich gehört) ‘The testament must have been signed (as I heard)’. Here müssen acquires a reportive interpretation only by virtue of the added comment habe ich gehört ‘(as) I have heard’. What the speaker is primarily saying in such a case is that s/he infers that the testament has been signed; as a kind of “evidential corroboration” s/he can support this inference by reference to hearsay although this is not inherent to müssen as such. This indifferent evidential meaning corresponds to what we encounter in some geographically close Romance languages. Fr. devoir ‘must’ has



Hearsay in European languages

85

been evolving a hearsay function via favorable contexts of report; its Italian cognate dovere has done so to a more restricted extent; even more restrictive are the conditions under which Sp. deber can acquire an air of hearsay. With both the Italian and especially with the Spanish modal, these conditions are characterized by the non-factual status of the state of affairs which is being conveyed as (having been) bound to occur. Their status as reportive markers is thus much more arguable than that of Fr. devoir (cf. Squartini 2004). In contrast to German and English but in accordance with Dutch, we find the auxiliary for MUST with hearsay function in Romance. Again, this function seems to have arisen, under favorable context conditions, out of a broader meaning of undifferentiated indirect evidentiality (cf. Squartini 2004). Compare an Italian example, which corresponds exactly to the Dutch one in (23) (same for Fr. devoir): (24)

Dev’essere un buon film. i. ‘It is said to be a good film.’ ii. ‘It appears to be a good film.’

Excepting Germanic, Romance and West Slavic, modal auxiliaries hardly ever seem to have acquired a reportive function. The only case known to me is the past tense of Est. pidama ‘must’, already discussed in 3.2.3. Remarkably, like the Germanic and Romance modals mentioned above, pidama belongs to the field of (originally deontic) necessity. Apart from the reportive past tense (pidi), the present tense form pidama if affixed by the reportive vat-suffix can also convey a reportive function, but it will then evoke epistemic overtones (Kehayov 2002: 136).

3.3.3. Extensions from volitional verbs The only indisputable case seems to be Germ. wollen ‘to want’. The conditions of its use are identical to those of sollen (cf. Diewald 1999: 283f.; Remberger, to appear), as is the restriction to the present tense forms (see 3.3.2). A reportive reading is hardly possible if agentive control is obvious. Thus an utterance like Sie will auf der Couch liegen will be interpreted volitively (‘She wants to lie on the sofa.’), and a reportive reading (#‘She claims to be lying on the sofa.’) appears to be rather far-fetched. In comparison to sollen, there is only one crucial additional restriction: with wollen the speaker of the original utterance must be identical with the referent

86

Björn Wiemer

of the subject NP of the actual utterance, whereas it is exactly this referential identity, which is excluded for sollen. We can easily see this from a comparison of the following examples. If in the made-up examples (19-20) wollen replaces sollen, only (19’) can be given a plausible interpretation with (20’) being extremely odd. The reason for this is that the identity condition just formulated for wollen requires there to be a real referent of the subject NP existing at the time of utterance:23 (19’) (20’)

Er will (schon einmal) Schach gespielt haben. ‘He is claiming to have played chess (once before).‘ ?? In früheren Jahrhunderten wollen die Menschen mehr auf dem Land gearbeitet haben. ?? ‘In earlier centuries, people claim to have worked more in the country.’

Furthermore, wollen is hardly imaginable as a reportive marker in the first person. First-person use would yield a volitional reading. In minimal pair conditions, the substitution of wollen for sollen would render the sentence unintelligible or force a very complex interpretation due to which the actual speaker verbalizes what somebody else has claimed him/herself (i.e., the actual speaker) to have told about him/herself; simultaneously, the actual speaker denies having uttered the alleged assertion; see (25b). Occasionally, first-person wollen could also lead to the pragmatically undesirable effect of unmasking oneself; see (26b). This contrasts with first-person usage of sollen, which is invariably interpreted as a simple reference to other people‘s assertions (utterances) about the actual speaker (25a, 26a) (# means pragmatic unintelligibility or inappropriateness): (25) a. Ich soll im Bett liegen geblieben sein. ‘I am said to (*am obliged to, *should) have stayed in bed.’ b. ? Ich will im Bett liegen geblieben sein. ‘I am said to have said that I have stayed in bed.’ #?‘I want to have stayed in bed.’ (26) a. Ich soll mit der Mafia unter einer Decke stecken. ‘I am said to be cooperating with the mafia.’ b. ? Ich will mit der Mafia unter einer Decke stecken. i. #‘I am said to be cooperating with the mafia.’ ii. ‘I want to cooperate with the mafia.’



Hearsay in European languages

87

In sum, wollen and sollen distribute complementarily with respect to a property that might be called ‘referential selectivity’. Remberger (to appear), in a recent, thorough study on evidential and modal shifts of WANT-verbs, has compared Germ. wollen with It. volere and Fr. vouloir and their alleged hearsay meanings. She concluded that quasi-evidential WANT “is restricted to a specific class of non-animate nouns (...) or folklore” (p. 17f.) in the two Romance languages. Such uses can be found not only in German, but also in English (The tradition wants..., Die Tradition will (es), daß ...). For this reason we cannot count these verbs as sufficiently conventionalized markers of reportive evidentiality. 3.4. Predicatives The class of lexemes to be treated in this subSection is peculiar, insofar as they are neither modifiers of the verb phrase, nor of clauses (or sentences), but constitute sentences themselves. In this view, predicatives are intermediate between modifiers of the nucleus and lexemes that are not integrated into clausal syntax but take scope over whole sentences (as do particles, see 3.5). Predicatives are a specific word class of heterogeneous morphological origin whose members share three properties: (i) they are uninflected (including originally finite verbal or nominal forms deprived of their paradigms), (ii) their only syntactic function is a predicative one (thence their name), i.e., they constitute the nuclei of (simple or complex) sentences, (iii) they imply an animate argument (sentient being, perceiver) whose syntactic realization is either blocked or relegated to an oblique case (mostly dative) or a PP. Instead, many predicatives require a second argument, very often of a propositional nature, which is syntactically realized by complementizer + clause; exceptionally this argument may be encoded with a PP which results in a specific evidential function (see ex. 33 and the comment on it). In any case, sentences with predicatives as nucleus do not have a nominatival subject, i.e., no NP to trigger agreement on the predicate (verb). Predicatives are numerous in Slavic and Baltic languages where we also observe certain tendencies toward morphological unification. As concerns lexical semantics, the predominant number of predicatives denote transient emotional or physical states (e.g., Pol. Przykro nam.DAT, Īe nie zdałeĞ egzaminu ‘We are sorry that you didn‘t pass the exam‘, Zimno mi.DAT ‘I‘m cold‘), attitudes (e.g., Lith. Visiems.DAT buvo gơda, kad teko sumelu-

88

Björn Wiemer

oti ‘Everybody was ashamed that they couldn‘t but lie‘) or modal notions like obligation, dynamic and deontic possibility or their negative counterparts (e.g., Russ. možno, nel‘zja, nado, Pol. trzeba, Lith. galima). Beside the lexical groups just mentioned, there is a small subgroup of predicatives related to perception which have evolved into evidential markers; an even smaller number specializes in hearsay. These are: (27)

Russian

reportive: inferential:

(28)

Polish

reportive: inferential:

(29)

Lithuanian

reportive: inferential:

slyšno ‘to be heard (to be told)’ vidno ‘to be seen, inferred’ (in the same meaning also dialectal vidat‘; cf. Kuz‘mina 1993: 129f.) słychaü ‘to be heard, told’ widaü ‘to be seen, inferred’, czuü ‘to be sensed‘, znaü ‘to be known; as is known’ girdơti ‘to be heard, told’ matyti ‘to be seen’

The tiny lists given here are most probably exhaustive; apparently, Latvian does not have a single such unit (though, like the three other languages, it has predicatives describing emotional and physiological states). From their morphological form, the Polish and Lithuanian units in (2829) are infinitives. In Polish, these forms have been petrified because they have lost all other forms of their former paradigms. On the contrary, Lith. girdơti and matyti still function as the main, stylistically neutral verbs of seeing and hearing, respectively. It is the infinitive (which does not allow for a nominatival subject) and the way how it codes a sentential argument which set them apart as hearsay units from the “homonymous” verbs with intact inflectional paradigms and a diversified array of syntactic patterns (see below; on matyti cf. Usonienơ 2001). Contrary, the Russian predicatives slyšno and vidno go back to adjectives (or adverbs) derived from the same common Slavic roots vid- ‘see’ and slyš-/slych- ‘hear’ (as do the Polish predicatives in 28). In modern standard Russian, slyšno still occasionally shows up in adverbial usage (see 30a), whereas vidno can no longer be used as an adverbial at all; thus both units have split off semantically from the adjectives vidnyj ‘1. visible, 2. prominent, 3. stately (about goodlooking men)’ and slyšnyj ‘audible’, which are tightly associated to direct evidentiality. Depending on the type and coding of the argument, slyšno (as well as vidno) mark different evidential functions. With a nominal object NP,



Hearsay in European languages

89

slyšno refers to direct perception (30b); if the argument is propositional and adjoined by a complementizer, slyšno can still refer to direct perception (or inferences based thereof, see 30c), but it then can also acquire reportive function (30d): (30) a. Na berezax ele slyšno oživajut razmjakšie vetki. ĺ adverb ‘On the birch trees the withered twigs come back to life hardly audibly.’ b. Marfen‘ku vsegda slyšno i vidno v dome. Ona to smeetsja, to govorit gromko. ‘Marfen’ka can always be heard and seen in the house. Either she is laughing or she is shouting around.’ (I. Gonþarov, “Obryv”) ĺ predicative, direct perception c. Gluxota moja byla nepolnoj: skvoz’ nee bylo slyšno, þto golos Beloku rogo vysok i nazojliv. ‘My hearing loss was not profound; it was still possible to hear that Belokuryj had a high and piercing voice.’ (I. Grekova, “Bez ulybok”,1975) ĺ predicative, direct perception (inference ?) d. Poslednee vremja tol’ko i slyšno, þto oligarxam vredno zanimat’sja politikoj. ‘Recently it can only be heard that it is detrimental for the oligarchs to engage in politics.’ (“Sovetskaja Rossija”, Feb., 15th, 2003) ĺ predicative, hearsay Predicatives functioning as markers of indirect (reportive or inferential) evidentiality never allow for a realization of the implied animate argument. This salient syntactic feature distinguishes them from predicatives that belong to other semantic domains (in particular from those denoting emotional or physical states, but see also oligarxam.DAT.PL vredno ‘for the oligarchs it is detrimental’ in the second part of ex. 30d). Cf., for instance Lith. girdơti (31) and Pol. słychaü (32): (31)

Gaila, kad Žemaitơs mes negalim skaityti žemaitiškai... – Gal galơsim? Girdơti (*visiems), kad susipratĊ leidơjai ketina parnjpinti autentiškai parengtus jos "Raštus". ‘Too bad that we cannot read Žemaitơ in Samogitian … – Maybe it will be possible one day? (*To all) there are rumors that publish-

90

(32)

Björn Wiemer

ers who are skilled in it intend to get the authentically revised “Writings”.’ (“Šiaurơs Atơnai”, 1995/5-7) Slychaü (*nam), Īe w Iraku znowu dokonano zamachu. ‘It is said (*to us) that yet another assault has been plotted in Iraq.’

Like Russ. slyšno, these predicatives convey a reportive meaning only if they code a propositional argument; see (31-32). Token frequency of such items is, however, very low (cf. Wiemer 2006b: 29 for Polish) and outranked by a pattern in which these predicatives take a nominal argument and, consequently, mark direct auditory evidentiality; compare Pol. Na ulicy było.PAST.N słychaü krzyki.ACC.PL ‘On the street one could hear shouts’, Lith. Už kalno girdơti griausmas.NOM ‘Behind the hill thundering can be heard’. Somewhat more frequently, Pol. słychaü is used with the valence frame o+LOC ‘about’. In this case it does not refer to the content of reported speech but only its topic; e.g. (33)

CzĊsto było słychaü o nim. REP about he.LOC often COP.PAST.3.SG.N lit. ≈ ‘Often (it) was heard about him’, i.e. ‘Often he was spoken about.’

The same applies to Russ. slyšno, which is, however, more colloquial (noticeably with the preposition pro+ACC)24 than Pol. słychaü o, whereas Lith. girdơti practically does not realize this syntactic pattern at all. The units listed in (27-29) have “homonymous” particles. The distinctive criterion between particles and predicatives is that particles lack argument structure and, consequently, cannot be combined with a complementizer.

3.5. Particles and sentential adverbs I will now proceed with those considerably more numerous expressions that belong to classes which lack an argument structure of their own: particles and sentential adverbs. Since, on the basis of the existing literature, I have found myself unable to figure out any fail-safe difference (either functional or structural) between sentential adverbs and (modal) particles, I will treat them here in one rubric, leaving the question of whether there is any tenable functional difference that can be generalized over languages for future investigations.25 I will refer to the relevant units as ‘particles’,



Hearsay in European languages

91

although in various descriptions some of them might have been named ‘adverbs’ (for a survey cf., for instance, Ramat 1996); terminology varies considerably also due to national linguistic traditions. The medley bag of words called ‘particles’, which we are concerned with here, consists of sentential modifiers (as opposed to focus particles – see f. 25 – and particles that function as signals of turn-taking or attentionregulating devices in dialogue). Whether in European languages hearsay particles can take scope not only on the propositional, but also on the illocutionary level (for the difference cf. Faller 2006), cannot be investigated here. As far as I can see, no cross-linguistically applicable criteria have been formulated for what should count as particle; often particles are “defined” in negative terms: they are uninflected; they are neither part of the clause’s constituent structure, nor do they have to occupy fixed positions; they are highly heterogeneous in terms both of syllable structure, cliticizability and morphological provenance, etc. Ultimately, the only valid criteria appear to be located on the level of pragmatic functions (for overviews and/or attempts toward defining particles as well as parentheticals cf., e.g., Burkhardt 1999; Grochowski 1986; 1997: 22-24; 2007; Mosegaard Hansen 1998: ch. 3). More or less identical criteria are applied to determine parentheticals; these are functionally closest to sentential particles, and this is reflected in their shared scopal properties and syntactic independence from the rest of the utterance (see 3.6). The reason why parentheticals should, nonetheless, be treated apart is that parentheticals constitute an expression class that embraces not only words (word forms), but also units consisting of more than one word (form). Often they are even composed of whole syntagms (e.g., small clauses). For this reason, one often encounters difficulties in drawing a divide between (rather) occasional and (rather) conventionalized parentheticals. Thus there is no appropriate place for parentheticals on the lexicon-grammar cline in Fig. 1; they have to be considered on a different level because units of virtually any morphological format and syntactic status can be used parenthetically. Therefore, before “uniting” particles and parentheticals on functional grounds, it seems advisable to account for particles as one-word units first. Admittedly, this dividing line appears somewhat arbitrary and in some cases, it will even be violated, at least insofar as we would need a universal definition of ‘word’ (in morphological and prosodic terms). We should, however, notice that one-word units, being short, often underlie rules of cliticization specific for the respective language. Beside that, particles can take parenthetical clauses into their scope (see ex. 47); they must therefore

92

Björn Wiemer

be distinguished from each other. Thus, in order to supply a transparent description, I prefer, for the time being, to provide arguments and empirical data for parameters according to which reportive particles can be subdivided. Whether and which of these parameters can be applied to subclassify parentheticals, too – and how we may unite both classes – has to remain the task of a (hopefully) near future. Since structurally particles are not integrated into the syntax of the clause which they modify, their scopal behavior can vary, and since among all reportive markers they display probably the widest array of semantic-pragmatic diversification, they are best suited for the task of catalogizing structural and functional parameters along which reportive markers can be classified. Particles, as it were, can be considered as the best accessible playground of multifarious factors that influence the usage of reportive markers and, therefore, supply the empirical basis on which I will try to work out an important part of the complex network of these factors. For the present purposes, it should suffice to conceive of ‘particles’ as units which fulfill the usual criteria mentioned above. I will first give some representative examples of reportive particles from different European languages and then elaborate on parameters of their subdivision. (34)

(35)

(36)

(37)

English apparently (cf. Wierzbicka 2006: 278ff., from where this example is quoted): Apparently there’s eight times more sheep than there are people in New Zealand. German angeblich Die Bundesregierung arbeitet angeblich an einem zweiten Konjunkturprogramm im Umfang von mindestens 30 Milliarden Euro. ‘The Federal Government is reported to be working [is allegedly working] on a second economic stimulus package totaling at least 30 billion Euros.’ (ZEIT online, Dec., 13th, 2008) Romanian (by courtesy of A. Merlan, Konstanz) Cică se căsător-eúte cu o franĠuzvaică REP RM marry.3.SG.PRS with INDF.ART French.F ‘He is said to be marrying a French woman.’ Spanish dizque Esto dizque va a ser pantano. REP go.PRES.3.SG to be swamp this ‘This is going to be swamp, they say.’



Hearsay in European languages

93

For this and other examples from different varieties of Romance cf. the systematic account of dizque-units in Cruschina and Remberger (to appear). (38)

(39)

(40)

(41)

Polish jakoby (cited from Wiemer 2006b: 41) List stryja wiele wyjaĞniał, ale jeszcze wiĊcej gmatwał; czego tam nie było! Potopy, które co tysiąc piĊüdziesiąt lat nawiedzają PolskĊ; przodkowie rodem z Egiptu; historia Sieciecha, palatyna Władysława Hermana, który jakoby pragnął przekazaü przyjacielowi brata koronĊ; przyszłe córki Alicji w roli zbawczyĔ ojczyzny... ‘The uncle’s letter explained much, but then it caused even more confusion: there was all sorts of stuff in there! Floods that hit Poland every one thousand and fifty years; ancestors of Egyptian origin; the story of Sieciech, Wladyslaw Herman’s Palatine, who allegedly wished to pass on the crown to his brother’s friend; the future daughters of Alicja as saviours of the nation…’ (T. Mirkowicz: “Pielgrzymka do Ziemi ĝwiĊtej Egiptu”, 1999) Croatian/Serbian navodno Sve bi nas valjalo ponovo vratiti iza Karpata, odakle smo, navodno, došli. ‘We all should be sent back beyond the Carpathian Mountains, from where we are said to come.’ (Internet) Lithuanian esą Daugeliui teko ilgai laukti savo moksliniǐ laipsniǐ ir pedagoginiǐ mokslo vardǐ patvirtinimo, nes, esą, jie „neturƳ” sovietinio diplomo. ‘Many had to wait long for an approval of their academic titles or teaching diplomas, because, so it was said, they did “not have” a Soviet degree.’ (Zinkeviþius “Prie lituanistikos židinio”; 1999, 76) Latvian it kƗ ‘as if’ Šobrid it kƗ viss ir, lai izglƯtƯba atbilstu darba tirgus prasƯbƗm. ‘Allegedly, there is everything available now to render education fit for the job market.’ (www.politika.lv, December 2006; by courtesy of I. KƺƝvere-Wälchli)

Further cases: It. dice (Giacalone Ramat and Topadze 2007: 27; Pietrandrea 2007: 55-57), Gr. léi (K. Stathi, p.c.), Russ. mol, deskat’, -de (Arutjunova 2000; Plungjan 2008), Russ. kažetsja, budto by, vrode (Letuþij 2008; Wiemer 2008b), Pol. podobno, ponoü, rzekomo (Rytel 1982: 47f.; Wiemer

94

Björn Wiemer

2006b), Cz. prý, údajnČ (Hoffmannová 2008; Karlík et al. 1995: 626f., 664; Rytel 1982: 46, 48f.), Slovak vraj, Croat./Serb. kaže, Macedonian vele (Mushin 1997: 294-296), Lith. tarytum, girdì, neva (cf. Wiemer 2007b), Georg. turme, titkos, vitom (Topadze 2008: 51f., 56-60). We should add Arm. eϻer, which is an etymological equivalent of Lith. esą (see ex. 40), as it is an isolated participle form of the copula. In semantic terms, however, eϻer differs from Lith. esą insofar as it is not restricted to hearsay (cf. Donabédian 1996: 95-97). Parameters pertinent to a sub-classification of reportive particles can be divided and ordered into (at least) five global groups: 1) diachronic background, with a further subdivision into: a) etymological and semantic evolution, b) changes of morphosyntactic properties (along the “cline” in Fig. 1), 2) criteria of formal (morphosyntactic) properties and behavior, 3) criteria of semantic-pragmatic properties, 4) distributional criteria, i.e., how do such particles combine a) with other evidential markers (regardless of their “morphosyntactic format”), b) with other units that function as sentential modifiers (e.g., epistemic ones)? In this paper I will focus only on criteria of groups (ii-iii). A fifth group of criteria concerns (v) the variability of grammatical status (syntactic distribution), i.e., the question of whether a unit with a particular phonological shape can occur in different types of constituency (according to the classification substantiated in Section 2) without noticeable changes of lexical meaning (sc. evidential function). This question as to the proportion of units concerned is not to be neglected, nor should we dismiss it in our theorizing. It is a kind of synchronic correlate to the diachronically oriented issue in (i-b). The manner we handle in issue (v) will be crucial for any solid lexicographic account of evidential markers in a language as well as for approaches towards the lexicon—syntax-interface. I will pick up this issue in 3.9. Probably the most important criterion of group (ii) concerns the scopal behavior of a unit, first of all whether its scope is only sentential or variable as in the latter case; it can narrow its scope down to even attributes of a noun (see 42b, 43, 44). Although one needs to have checked a sufficient amount of relevant data in order to be able to state firmly that the predominant number of reportive particles displays variable scope, I have by now



Hearsay in European languages

95

got the general impression that reportive particles are capable of variable scope by default. Let us assume this for the time being. Another default rule to hold for most hearsay particles appears to be that they need not specify the author of the original utterance. All units illustrated so far can have wide (propositional) scope; most, but not all of them can occur with narrow scope. If they do, they normally take the narrowest possible scope, i.e., scope only over a part of a NP (an attribute). At least this is the most frequent case with text tokens known to me. In this respect, Germ. angeblich ‘allegedly’ is representative; cf.: (42) a.

wide scope Gestern [kam er angeblich erst um fünf mit seiner bereits sechsjährigen Tochter].26 ‘Allegedly, yesterday [he arrived only at five with his already sixyear-old daughter].’ narrow scope b. Er kam gestern erst um fünf mit seiner angeblich [bereits sechsjährigen] Tochter. ‘Yesterday he arrived only at five with his allegedly [already sixyear-old] daughter.’

Here and in the following examples, narrow scope is indicated by square brackets. See some further typical illustrations: (43)

(44)

(45)

Polish jakoby Wiele osób (...) w dyskusjach nad problemem przywrócenia do Īycia owego czołowego pomnika pamiĊci narodowej podnosi [zbyt długi] jakoby okres odbudowy oraz ogromne jej koszty. ‘In debates on the issue of the revitalization of this outstanding site of the national memory many have addressed the allegedly [too long] period of reconstruction and the enormous costs.’ (“Młody Technik”, 1971-3) Russian jàkoby (cited from Rakhilina 1996: 299) Vy, razumeetsja, slyšali o moem jakoby [romane] s vami? ‘For sure, you have heard about my alleged [affair] with you.’ Russian budto by ‘as if, apparently’ Muxin do six por budto by [niþego ne znaet ob ơtom]. ‘Apparently, until now Muxin knows nothing about it.’

96

Björn Wiemer

Utterances like (45) allow for inferential readings, too.27 Pol. jakoby and Russ. budto by are also encountered as complementizers (see 3.7, 3.9). Also attested, though less often, are cases in which a particle takes an adverbial PP into its scope (46) or a parenthetical expression (small clause, see 47): (46)

(47)

Lithuanian esą (cited from Wiemer 2007b: 177) Pirmą kartą nepažƳstamas vyriškis Juozui paskambino šiǐ metǐ balandžio pabaigoje. Užuominomis jam pasakơ, kad turơsiąs mokơti pinigus esą [už blogą tarpininkavimą]. Buvo leista suprasti, kad reikia duoklơs. ‘The first time an unknown man phoned up Juozas at the end of April this year. Using allusions he told him that he would have to pay money supposedly [for bad mediation]. They let him know that a tribute was required.’ (“Lietuvos Aidas” 1995-8) Croatian/Serbian navodno (by courtesy of A. Drobnjakoviü) Povodom teksta u estradnom þasopisu “Mega hit”, u kojem je osvanu la informacija da se razvodi, navodno [zbog jedne plave žene], upitali smo Ljubu Aliþiüa da li je to istina. „Naravno da je u pitanju traþ”, kaže Ljuba. ‘On the occasion of the article in the magazine ‘Mega hit’, where the information appeared that he is going to divorce his wife, allegedly [because of a blond woman], we asked Ljuba Aliþiü if that is true. ‘Of course it’s a gossip,’ says Ljuba.’

There is at least one peculiar case of a reportive particle which can also function as a sort of word-formation device. This is Pol. niby ‘as if’: like jakoby it can be used both as particle and conjunction (see 3.7, 3.9); it then not only refers to a foreign speech act, but also implies that the actual speaker does not subscribe to the implicit consequences which the original speaker might have had in mind (cf. Wiemer 2006b: 45-48). If used like a prefix, niby- adds an ironical touch: (48)

Widocznie te twoje niby-badania w tej szopie to wcale nie taki cymes. ‘Apparently, your so-called [= pseudo-]research in that shed there is not such a big deal.’ (R. Antoszewski: “Kariera na trzy karpie morskie”, 2000)



Hearsay in European languages

97

This is reminiscent of what Brendel et al. (2007: 6) have called ‘modalized citation’ (“modalisierendes Zitat”) and similar to Travis’ (2006: 1278) ‘labelling’. Thus niby- raises the general problem of whether it is justified to include adnominal modifiers of the SO CALLED-type into an inventory of reportive markers. There are reasons in favor of such a decision. First, labeling is not restricted to word-formation devices like Pol. niby- or attributive adjectives like Germ. so genannter, It. cosidetto, Russ. tak nazyvaemyj, but is commonplace among particles with variable scope, i.e., to the assumed absolute majority of them. Second, adpositions – which are adnominal modifiers par excellence – can also be considered reportive units (see 3.8), although they do not participate in word-formation. Third, notice that SO CALLED-modifiers have a specific quotative function sensu stricto. Meibauer 2003 argued that quotative meaning components28 can be integral parts of words. If we accept these arguments, we should, however, be aware that we are likely to be running into problems with delimitation both in semantic terms (how deeply can the element referring to a previous utterance be anchored in the semantic explication of a linguistic item?) and in morphosyntactic terms. As for the latter, remember Anderson (1986: 275), who excluded compounds and derivational forms from the inventory of evidential markers (see his fourth defining criterion). A further problem arises: if niby becomes a bound morpheme and thereby loses its syntactic freedom even more than clitics, how should its categorial status be handled and, consequently, how many lexical entries of niby are there? (On this general issue see Section 3.9.) Apart from this, notice that niby used as a free morpheme cannot be regarded as a clitic (just like most markers in Polish, Russian and Lithuanian do not behave like clitics); we cannot, therefore, assume a cline: free > cliticized > agglutinated (> fused) morpheme, as might be natural in other languages with a more salient propensity towards clitics (e.g., Romance or South Slavic); instead, we would have to assume that the “clitic stage” is circumvented. Although most particles with hearsay meanings supposedly have variable scope, there are some of them which demonstrate only wide, sentential scope. On preliminary terms, two rather systematic types figure saliently. Remarkably, both can be characterized as quotatives in the narrow sense (cf. Aikhenvald 2004: 64, 177f.). First, reportive markers which have only recently split off from complementizers; in such cases it may even be arguable whether they have to be treated as units distinct from complementizers at all. A nice case to illustrate this is supplied by colloquial Czech že. See the following examples from spontaneous discourse: in (49), že occurs after a subject pronoun (voni ‘they’) and we can assume an ellipsis of the

98

Björn Wiemer

predicate. In (50), however, že occurs so frequently without a speech act verb, both after intonational breaks and without any breaks (in the following indicated by a dash ‘/’), that we get the impression that it is becoming a syntactically independent unit marking quoted speech: (49)

(50)

mČla sem pČkný seno (...) a tak sem šla Kostelce/a voni že už maji všechno plný a že už to seno nebudou brát jo /(...) ‘I had some nice hay (…) and so I went to Kostelec/but they že already had enough of it, and že they won’t take any more hay, will they’. vona mČ prosila (...) že zkrátka bych mohla že þeká rodinu že bych mohla u nich bejt/že se bude mít dobĜe //ale že byli sami chudáci/ že sem se nic moc nemČla/že sem se tam hroznČ zhubla (...) ‘she asked me (…) že I could, at short notice že she is expecting a child že I could stay with them/že s/he will be well off //but že they were all poor/že I wasn’t well off/že I terribly lost weight there’.

Hoffmannová (2002: 374) treats the repeated use of že as hypertrophy which leads to a mixture of quoted, indirect and free-indirect speech. Now, on account of the status of že as oscillating between a complementizer “floating” without its predicate and a particle, it is unlikely that že gains scope over constituents on levels lower than the clause. Admittedly, some reportive particles which arguably began (and can still be used) as complementizers with an ‘as if’-meaning are now used with variable scope; compare Russ. jakoby and kak budto, budto by, Pol. jakoby (see ex. 43), probably also Latv. it kƗ. These units, however, are likely to have gained the status of particles a more considerable time ago than Cz. že. Most plausibly, all of them started as clause connecting devices in constructions of comparison with an ‘as if’-meaning.29 To this group we can add Lith. esą, although its history is different: it derives from a participle used in complementation (see 3.2.1). It is reasonable to assume that all these items acquired variable scope only after they had separated from “their” subordinating verbs, although this assumption has not yet conclusively been tested on diachronic data. Second, another group of quotative units lacking variable scope originates from syntactic positions other than clause connectives. Colloquial quotatives like Engl. like, Russ. tipa (= tip.GEN ‘type, sort’) should be named here in the first place. Again, in this group, expressions used for comparison are particularly commonplace, but (combinations of) demonstratives, pronouns and quantifiers (in a broad sense) are also mentioned quite often such as, for instance, Germ. und sie so, lit. ‘and she so’ (+



Hearsay in European languages

99

quote). As a rule, such syntactically indeterminate units precede the reproduction not only of a literal quote, but also of paralinguistic features, mimics and gestures of the original speech event; cf. the title of Golato’s 2000 article dealing with the reference to “embodied actions”. This distinguishes them from units like Cz. že, which started their lives as connectives. Since quotatives of the LIKE-type are so tightly associated to the complex reproduction of speech, it is hard to imagine that they will ever occur with narrow scope. Moreover, since they refer to the entire “behavioral accompaniment” of utterances (the manner of pronunciation, characteristics of voice, etc.), the notion of sentential (or propositional) scope sounds somewhat misplaced. Utterances need not take on the shape of sentences, and in spoken language (i.e., where quotatives of this type occur) they notoriously decline from models of standard syntax. Thus if we want to say that these quotatives have scope over propositions, they additionally, and on equal terms, mark illocutionary and emotional aspects of speech. Some reportive particles can behave like clitics, e.g. Cz. že, prý, Russ. mol, probably also It. dice. Notice that the degree of coalescence (in terms of Lehmann 1995) does not automatically correlate with a reduction of scope. For instance, the Georgian hearsay morpheme -o is a clitic, but it nonetheless operates on phrase level. According to Boeder (2000: 279), “in colloquial speech [it] can occur on each constituent boundary”. The following example is quoted from Giacalone Ramat and Topadze (2007: 18): (51)

gaero-m, v -i -ziareb -t 1 VERS share:PRS PL UN.ERG mc’uxareba-s -o REP grief.DAT ‘The United Nations said, we share your grief!’

tkven-s your.DAT

Furthermore, certain reportive clitics show tendencies towards agglutination. As cases in point we can count Pol. niby mentioned above and Georg. metki (see ex. 55). A further item which might belong into this subclass is Russ. -de (cf. Plungjan 2008). Arutjunova (2000: 438) cites examples like the following from the Dal’-dictionary (composed during the 19th c.), in which de occurs repeatedly on successive syntagms (and which can barely be translated into English other than with the aid of a colloquial quotative marker): (52)

On govorit, ja-de ne pojdu-de, chot’-de, þto choš’ delaj. ‘He goes “I won’t go, you can take it or leave it”.’

100

Björn Wiemer

Such occurrences have, however, clearly diminished; (52) sounds rather hypertrophic in contemporary Russian. Let us now turn to the criteria of group (iii). Among them, three are salient: (a) the relation to speaker’s assessment of the trustworthiness or veracity of the utterance(s) referred to (absence vs. presence of “epistemic overtones”); (b) known – anonymous speaker – folklore; (c) the question of which level, or aspect, of the reported utterance(s) is highlighted by the respective marker: propositional content vs. topic of speech vs. manner of speaking; propositional content vs. literal quote; can the speaker refer to an utterance of him/herself?; must the speaker have heard the reported utterance(s) him/herself? etc. As concerns the relation of hearsay to epistemic assessment, I refrain here from any discussion. Factor (b) has casually been commented on above and in Sections 3.2.4, 3.3.2 and 3.3.3. In the following, I will concentrate on criteria of group (c). We can distinguish a bunch of aspects within the reference to speech acts which prove to be relevant at least for some units. The criteria named immediately below do not pretend to be exhaustive, but they are exceptional provided we regard the kind of reportive markers as the most trivial (widespread) that (i) focus on the propositional content of speech (ii) uttered previously (iii) by another person (other people) (iv) in the absence of the actual speaker. Based on of these four “default conditions”, the following more specific aspects can be singled out: 1. Literal reproduction of speech. This is what quotatives proper do. We have already mentioned a few. Notice that, as a rule, quotatives usually imply that the actual speaker him/herself has witnessed the utterance s/he is referring to. We may further ask whether particles that normally fulfill the four default conditions just formulated can also be used with direct quotes. Theoretically, nothing precludes such a usage; however, in practice this seems to happen rarely, and labeling appears to be an exclusion (see ex. 48). This shows that reference to utterances with “embodied action” (Golato’s 2000 term) differs in principle from reference to denominations (etiquettes). Again, these impressions open up further issues remaining for empirical investigation. In this connection it seems sensible to also test whether the following unidirectional implication holds: if a reportive particle (or, generally, a reportive marker) is normally used under the four



Hearsay in European languages

101

default conditions, it can also be used with quotes, the reverse does not apply (i.e., quotative markers cannot replace less specific reportive markers if these refer only to the propositional content). 2. There are markers that allow referring not only to previous speech acts, but also to their semiotic substitutes. In fact, quotatives often do refer to semiotic substitutes or thoughts (assumed by the speaker), too. But this is not an exclusive characteristic; other, non-quotative subgroups of hearsay markers are able to highlight semiotic substitutes of speech as well. I have found the following two. First, the Russian “xenomarkers” mol and deskat’, which do not primarily refer to the propositional content of the reported message but to the manner in which it was formulated (Plungjan 2008: 306f.). In the following example (from Arutjunova 2000: 444) mol does not refer to any speech act but to another person’s way of behavior (delaja vid, þto ‘pretending /showing that’) interpreted by the actual speaker: (53)

Na lice raboþego prenebreženie, stal vozle tolpy bokom, delaja vid, þto on priostanovilsja na minutu, dlja zabavy: mol, zaranee znaju, þto vse govorjat þepuxu. ‘The worker’s face shows contempt. He positions himself at the edge of the crowd, giving the impression as if he had stopped only for a minute, just for fun: {mol} I know from the start that everybody talks rubbish.’ (Bunin)

Second, reference to semiotic substitutes of speech occasionally occurs with general hearsay markers that focus on the content of speech. See an example of Pol. jakoby ‘as if’ (quoted from Wiemer 2006b: 40): (54)

Innym znów razem, gdy upuszczony ołówek niefortunnie potoczył siĊ za daleko, podniosła go niezwłocznie, nie przerywając wykładu i, jakby nigdy nic, włoĪyła do szuflady. Na rozpaczliwe gesty poszkodowanego (Īe nie ma jakoby czym pisaü, Īeby mu oddaü ołówek) nie zwracała najmniejszej uwagi, kontynuując lekcjĊ. ‘Some other time, when a dropped pencil rolled too far away, she picked it up immediately, without interrupting her lecture, and, as if nothing had happened at all, put it in her desk drawer. She paid no attention whatsoever to the injured party’s desperate gestures (meaning that he allegedly had nothing to write with, so that he would need the pencil back) and she carried on with the lesson.’ (A. Libera “Madame”, 1998)

102

Björn Wiemer

In both cases given, an air of irony arises. It remains an open question as to whether this is an accidental effect or the consequence of a more stable implication of the “semiotic transfer” of the respective hearsay markers. 3. As shown for the Polish predicative słychaü (see 3.4), reportive markers happen to focus only on the topic of speech, provided their argument is coded with a PP. I am unaware of similar cases from among particles or any other group of markers (in terms of their “format”). 4. Usually reportive markers are used by speakers who have not participated in the speech events on which they report, they rather exclude the actual speaker’s personal witness. An exception to this rule is found in the Russian “xenomarkers” already mentioned. Not only must the speaker have witnessed him/herself the reported utterance, but s/he can even also refer to his/her own speech uttered previously. In addition, mol, deskat’ and -de require that the original speaker not be anonymous (or be at least easily retrievable from the linguistic context), that reference to a reported speech act has been prepared in the preceding context and the conditions under which it occurred are described; cf. Arutjunova (2000: 442-447), Letuþij (2008: 223f.), Plungjan (2008: 306f.). 5. Reportive markers happen to underlie restrictions with respect to grammatical person and/or to participants of the reported or the reporting speech act. A certain kind of ‘referential selectivity’ was already demonstrated for the reportive use of the German auxiliaries sollen and wollen (see 3.3.3). There are, however, other dimensions of referential peculiarities, one of them is characteristic of the Georgian particle -metki (< me vtkvi ‘I said’; cf. Topadze 2008: 44f.). It marks “exclusively quotations in the first person singular when the speaker reports an utterance s/he had already made or reflected on in the past” (Giacalone Ramat and Topadze 2007: 16). The following example given there can be characterized as an indirect directive speech act: (55)

ramdendžer g -i -txar-i k’ar-i door.NOM how many times 2.SG -VERS.tell.AOR.1SG ar dak’et‘-o – metki! NEG close.OPT.2.SG – QUOT ‘How many times have I told you not to close the door (I said)!’

Georgian has another marker, -tko (< -tkva ‘s/he said’), which “is used when the addressee is a mediator between the speaker and a third person. The utterance is addressed to the third person, the information source is the speaker” (Giacalone Ramat and Topadze 2007: 16):



(56)

Hearsay in European languages

103

utxar-I male mod-i-tko. soon come-IMP.2.SG-QUOT tell.2.SG-IMP ‘Tell him/her, to come soon.’ (literally: ‘tell him/her, come soon, s/he said’)

In this case, the reporting utterance has not yet been realized and is, as it were, made in advance; by the same turn, the addressee of the speech event uttered in (56) is “circumvented”. The suffix -tko functions as a quotative proper, i.e., with literal quotes, whose reference to discourse participants does not agree with the indication of grammatical person: -tko is a quotative marker reserved for the “prospective”, intended addressee who is absent in the speech event of (56). This usage resembles Free Indirect Speech insofar as there is a deictic switch of person reference which, in the second part of (56), re-aligns the deictic system from the point of view of the prospective addressee. In a sense, this reminds us of some (not all!) uses of the Russian xenomarkers, namely those which seem to be typical of an older, 18th-19th century pattern; cf. an example from Arutjunova (2000: 441): (57)

Poslušaj, podi skaži Petru, þtoby on skoree sjuda prišel: barynja, deskat‘, sprašivaet. ‘Listen, go and tell Peter that he should come here as quickly as possible: the landlady {deskat’} asks (for him).’ (Krylov)

Here the speaker is the landlady (the Barin’s wife, barynja) herself, but in the last part of the utterance she refers to herself by using the third person (sprašivaet ‘asks’) and marks the phrase she wants the servant to convey with deskat’.30 Another item showing restrictions regarding grammatical person appears to be It. dice. Having been isolated from its paradigm, this originally verbal form can be used in clauses with predicates (finite verbs) in the 3rd person, both singular and plural (see ex. 58a-b), but neither with 1st nor 2nd person predicates (see 58c-d). This restriction holds at least if dice is used after or within direct speech (C. Guardamagna, p.c.) so that it itself might turn out to be restricted to the quotative use of dice: (58) a. Mi ha chiesto/detto – dice – [continuation of quote]. ‘S/he has asked/said to me – dice …’ b. Mi hanno chiesto/detto – dice – [continuation of quote]. ‘They have asked/said to me – dice …’ c. Ho/Abbiamo chiesto – *dice – [continuation of quote].

104

Björn Wiemer

‘I/we have asked – *dice …’ d. Hai/Avete chiesto – *dice – [continuation of quote]. ‘You (SG/PL) have asked – *dice …’

3.6. Parentheticals As already mentioned, from a functional point of view, particles and parentheticals31 are hardly distinguishable: they both (a) take scope over whole utterances, (b) their position is not fixed (neither is their scope), and (c) their relation to the sentences (or constituents) they modify cannot be captured in terms of dependency.32 Here are some examples of markers where the decision as to whether the given unit counts as parenthetical or particle proves to really be arbitrary: (59)

(60)

(61)

French paraît-il33 (…) le charmant roi mage (…) avec lequel on lui avait trouvé au trefois – paraît-il – une grande ressemblance. ‘The charming king of magicians (…), whom people once had found him to resemble closely – so it seemed [as they said].’ (M. Proust: À la recherche du temps perdu; cited from “Le grand Robert…” 1985/VII: 74) Alemannic (Swiss etc.) German schints (schynts etc.) Vermutlich, weil das Senioren-Konzert in Dübendorf zu Ende war, Herr Woody. Dort waren ja schins nicht alle gleich zufrieden mit dem Gebotenen. ‘Presumably because the concert for the elderly people in Dübendorf was finished, Mister Woody. Apparently not everybody was equally content with what they had presented.’ (Campionatischer, 2006.08.07) Russian kažetsja (cf. Bulygina and Šmelev 1993: 80) Ivan, kažetsja, uexal v Kiev. ‘Ivan apparently left for Kiev.’

It is no accident that all examples given contain units deriving from verbs with the meaning ‘seem, appear’. Other examples are Sp. al parecer (Cornillie 2007: 34-36), It. a quanto pare (Squartini 2008: 932f.), Pol. zdaje siĊ (Wiemer 2006b: 53-59; 2008b: 356-358), Engl. it seems (Chafe 1986: 268). Characteristically, all these units seem to have acquired hearsay function



Hearsay in European languages

105

via inferential meanings, and most of them must now be qualified as undifferentiated markers of indirect evidentiality34 (see Section 4). Interestingly, the 3rd person singular form of the present indicative of SEEM-verbs35 highlights a reportive function only when it loses its dependency relations with a host sentence, i.e., when it becomes petrified as a particle viz. parenthetical. As long as it has not lost syntactic links with their immediate surrounding – in particular, if it is still used with a complementizer (basically as the head of a bi-clausal sentence) or in constructions known as raising – it functions only as a marker of inferences or direct perception (often with epistemic “overtones”). Diewald 2001 and de Haan 2007, in their presentation of data from Germanic languages, do not concentrate on reportive meaning, nor do they point to this tendency, but it can be read between the lines. It becomes evident also in Cornillie’s 2007 discussion of constructions and items based on Sp. parecer, and I received the same impression from my own preliminary investigation of Russian and Polish forms based on SEEM-verbs. An analogous tendency can be observed with ‘as if’connectives (see 3.7). Of course, at present, these observations allow only for the careful formulation of a hypothesis to be verified by diachronically oriented in-depth analyses of every single item. Other source expressions frequently encountered in reportive parentheticals (particles?) are the basic illocutive verbs of a language. Compare again It. dice which, in colloquial speech, is still often used as a unit petrified jointly with the complementizer che (Giacalone Ramat and Topadze 2007: 27; Pietrandrea 2007: 55-57). In Southern Italy, however, it occurs also without a complementizer (C. Guardamagna, p.c.); see the above comment on (58a-d). In fact, at least some parentheticals differ from particles in at least one important respect. Parenthetical clauses can contain anaphoric expressions which link the whole clause to their ‘host’ without exerting an influence on the syntactic structure of the latter; they can thus exhibit one-sided syntactic dependencies on their hosts (cf. Dehé and Kovalova 2007: 6, basing on Ackema and Neeleman 2004: 96-99). This property brings them close to non-restrictive relative clauses as it is clearly associated to their function as metatextual comments (see f. 32). However, I am unaware of any such cases from among parentheticals with a reportive function. Of course, we know of numerous cases in which a grammatical form of a SAY/TELL-verb combines with connectives like Engl. as, Germ. wie, Russ. kak, Pol. jak, It. come ‘as (people say)’. The question, however, is whether (and to which degree) these syntagms can count as conventionalized holistic units and not

106

Björn Wiemer

just as productively used combinations of words to be analyzed compositionally.

3.7. Conjunctions (complementizers) In some of the languages of the Eastern part of Europe we encounter hearsay units that have to be qualified as complementizers or conjunctions. These units center around the meaning ‘as if’ and introduce sentential arguments of speech act verbs or link sentential modifications to NPs. Clear cases in point are Pol. jakoby (62) and Lith. esą (63): (62) (63)

Wspomniano mi, jakoby miał taki zamiar. ‘I was told that apparently [lit. as if] he had such an intention.’ Seniai girdơti ir noriai cituojami argumentai, esą Baltijos valstybơs yra Maskvos išlaikytinơs. ‘For a long time we have been hearing arguments, as if the Baltic states were nurtured by Moscow.’ (“Lietuvos Aidas” 1995-8)

Lith. esą is the petrified obsolete neuter form of the present active participle of the copula bnjti, its etymology thus differs from units meaning ‘as if’ (Wiemer 2007b: 177-179). All these units, however, have developed further into “homonymous” particles (see 3.9); as for Lith. esą see (40), as for Pol. jakoby see the following example, in which jakoby combines with the complementizer Īe (cf. StĊpieĔ 2008: 327-330; Wiemer 2006b: 40-45; 2008b: 365-368): (64)

Paweł twierdzi, Īe Michał jakoby znał Marka. ‘Paweł claims that Michał allegedly/apparently knew Marek.’

As for the genetically closest languages (Russian, Czech, Slovak, Ukrainian, Belarusian and Latvian, respectively), I have by now been unable to find text instances with cognate or functionally similar units that are used both as particles and conjunctions (complementizers) and in either case would mark hearsay. In other words: hearsay conjunctions seem to belong to a rare species of lexical units, whereas etymologically related particles are more numerous since they prove to be more apt to develop a reportive function (see 3.9). For instance, Letuþij 2008 does not give a single example of Russ. kak budto as a conjunction in speech act contexts; Letuþij (p.c.) considers such a use to be ungrammatical, contrary to budto by (e.g.,



Hearsay in European languages

107

Govorili, *kak budto /budto by on uexal ‘They said that [lit. as if] he had left’; cf. also Wiemer 2008b: 347-353). In the same vein, Latv. it kƗ ‘as if’ functions as a complementizer, but, again, according to examples known to me (thanks to I. KƺƝvere-Wälchli and N. Nau) and confirmed by J. Chojnicka (p.c.), it then introduces sentential complements in exclusively inferential contexts (perception verbs etc.), whereas as a particle it kƗ can mark hearsay (see ex. 41). It seems thus that, contrary to Polish and Lithuanian, the syntactic distribution of the Russian and Latvian ‘as if’-connectives (particle vs. complementizer) tightly correlates with the evidential function (perceptive-inferential vs. reportive).

3.8. Adpositions The last word class to be considered in connection with hearsay is adpositions. In European languages, relevant units are encountered throughout with prepositions being in the majority. Semantically, adpositions modify NPs, whereas syntactically they are their heads, rendering PPs. Consequently, we have to distinguish their constituent-internal scope, which is purely adnominal, from the scope taken by the PP as a whole. PPs with reportive meaning always function as adverbials, their scope is therefore propositional. Adpositions themselves show various restrictions as for the admissible lexical noun class. Here I must restrict myself to a general remark on a coarse level of classification, which will be illustrated from Polish. Among adpositions with a reportive function, we should distinguish at least two types. In the first type the hearsay meaning is the only possible one, i.e., the adposition collocates only with names of persons (or groups of people), see (65a); this collocation restriction is loosened only in favor of names of institutions or bodies of people with some official function (65b). Adpositions of this sort cannot be used with names of products of speech, see (65c). Compare Pol. zdaniem: (65) a. Zdaniem rzecznika rządu (przewodniczącego/naszych specjalistów/profesora...) kryzys gospodarczy nie powinien wprowadzaü nas w panikĊ. ‘Due to the opinion of the government spokesman (the chairman/ our specialists/the professor…) the economical crisis should not cause panic.’

108

Björn Wiemer

b. Zdaniem zarządu (redakcji/komisji/rady...) kryzys gospodarczy nie powinien ... ‘In the opinion of the board (the editorial office/the commission/ the council…) the economical crisis …’ c. *Zdaniem ogłoszenia rzecznika rządu (artykułu w gazecie/ raportu...) kryzys gospodarczy nie powinien ... *‘In the opinion of the governmental spokesman’s release (the newspaper article/the report…) the economical crisis …’ Units belonging to the second type mark hearsay only by virtue of a specialized instantiation of a more global meaning consisting in a reference to the basis of judgment or measurement. For this reason, markers of this type can also collocate with names of products of speech, e.g., written texts, announcements, and dictionary entries (see 66c). They can occur even with names of other products of intellectual activity such as research results, public surveys, etc. (see 66d). This is where, for instance, Engl. according to belongs to (see the translations). Compare examples with its Polish equivalent według:36 (66) a. Według wychowawczyni (przyjacół/babci...) dziewczynka ma wielki talent do rysowania. ‘According to the house-mistress (the friends/grandma...) the girl is very gifted in drawing.’ b. Według komisji (rady nauczycielskiej…) dziewczynka ma … ‘According to the commission (the teachers’ council) the girl …’ c. Według przepisu (podrĊcznika kucharskiego/wskazaĔ sąsiadki...) nie moĪna dawaü tak duĪo droĪdĪy do ciasta. ‘According to the recipe (the cooking guide/the neighbour’s advices…) one must not put that much yeast into the dough.’ d. Według wyników badaĔ (ostatnich sondaĪy/wniosków...) posuniĊcia rządu spotykają siĊ z coraz mniejszym poparciem obywateli. ‘According to the research results (the last public surveys/the conclu-sions ) the measures taken by the government are receiving less and less support from the people.’ As far as I can see, beside Pol. Zdaniem, only Lithuanian happens to have units fitting to the first type, namely: pasak, anot (cf. Wiemer 2007b: 183f.).



Hearsay in European languages

109

Contrastingly, examples of units of the second type are commonplace. To these belong Germ. zufolge, gemäß, laut, Engl. according to, It. secondo, Sp. según, Fr. selon, Russ. soglasno, Bulg. spored, Croat. prema as well as Georg. mixedvit, tanaxmad, cnobit (for the last three ones cf. Topadze 2008: 49).

3.9. Intermediate cases In the preceding subsections I have rather casually pointed at cases in which the morphosyntactic classification of the units varies or causes problems. To a large extent, these cases can be characterized as “heterosemy”. One of the first to use this term was Lichtenberk (1991: 476); for him heterosemy occurs “where two or more meanings or functions that are historically related, in the sense of deriving from the same ultimate source, are borne by reflexes of the common source element that belong in different morphosyntactic categories”. Heine 1990 used the term ‘heterosemy’ to capture the result of a split of a unit’s status on a lexicon-grammar cline if this split was not accompanied by erosion. Correspondingly, I apply ‘heterosemy’ to label a unit which, in an unaltered phonological shape, occurs in variable syntactic environments and thereby belongs to different syntactic morpheme or word classes without however changing its semantic value. As a rule, this syntactic variation reflects different layers of diachronic development (a phenomenon which in work on grammaticalization has been called ‘divergence’ or ‘leveling’). For a similar approach cf. Autenrieth 2002. We will first look at units that are on the verge of focal points of a morphological cline. In 3.5 we mentioned the Russian xenomarker -de. Normally it is called a particle, but we could equally well treat it as an agglutinative suffix. In fact, very few units in Russian behave this way oscillating, as it were, between bound morpheme and clitic status, and this is consonant with the fact that Russian generally has very few true clitics. We face an analogous phenomenon with the Turkish ‘copula particle’ ImIú (a term used among others by Csató 2000). ImIú is cognate to the mIúsuffix treated in 3.1. In contrast to this suffix, the evidential meaning of the copula particle is temporally indifferent, i.e., not restricted to anterior events and their results (Csató 2000: 37f.; Johanson 2000: 80). This functionally stable feature correlates with its distributional properties. Look at the following pair of examples given by Csató (2000: 37; her translations):

110 (67)

(68)

Björn Wiemer

Ali hasta-ymıú. (< hasta + ımıú) Ali ill:IMIù ‘Ali is/was/has (apparently) been ill.’ Ali hasta-la-n-mıú. Ali become_ill:DNV.RM.MIù(= ANT) ‘Ali has (apparently) become/became ill.’

The ImIú particle can be attached to nominal stems (including extended non-finite verb forms) and thereby marks nominal predicates (67); this property brings it into complementary distribution with mIú, which is attached to primary and finite verb stems (the latter may be extended and derived from nominal stems); see (68). Furthermore, ImIú cannot be accentuated, whereas -mIú is capable of carrying pitch accent. From this we see that ImIú is less integrated into the word form than -mIú. Actually, this property seems to be the only argument in favor of counting ImIú as a clitic, not as an agglutinated affix. Now let us continue with items difficult to classify in terms of syntactic distribution, inter alia because changes of word or morpheme class do not hinge solely on morphologization (coalescence). Such a case was noticed for Cz. že. Originally being (and still used as) the main complementizer void of any additional semantic elements, in colloquial speech it can be disconnected from speech act verbs after which it introduces sentential arguments (see 3.5). This can be considered as a change into particle status which, however, is still bound to clause initial position. We can say that the categorial change of že has resulted from the loss of the syntactic link with a subordinating verb. The mirror image to this process, i.e., the loss of an otherwise necessary complementizer following a speech act verb, has been repeatedly attested in Romance. Here we come across reportive modifiers on clause level, which originate from the fusion of a petrified form of a general SAY-verb (as a rule, the form of 3.SG.PRES.IND) with a complementizer. These are typical cases of lexicalization as understood by Lehmann 2002 and Himmelmann 2004 and referred to in Section 2. If these modifiers lose all their verbal features, they should be classified as particles (or sentential adverbs; see 3.5). Interestingly, some of the relevant items have not stripped off all verbal features: although they have ceased to be analyzable into their original morphemes (root/stem+COMP); in some Romance varieties such uninflected function words can still take a sentential complement. This becomes obvious if they occur with another complementizer. According to



Hearsay in European languages

111

Cruschina/Remberger (to appear, p. 11f.), such cases are attested in Sardinian and Romanian. Compare an example they give for Sardinian: (69)

issa E nachi chi REP COMP she and andaiada nudda. go.IMPF.3.SG nothing ‘And it is said that she didn’t go there at all.’

no NEG

b’ there.CL

Analogous behavior can be observed in Rom. cică and colloquial Italian dice (che); for the latter see 3.5. On the basis of the criteria formulated in 3.4, such units should best be classified as predicatives; complementselecting clausal modifiers of this kind do not allow naming the implicit human referent. This is also characteristic of particles (see 3.5), but normally particles do not combine with complementizers. There is another class of predicates coming close to the definition of predicatives, which is enormously widespread in European languages: the 3.SG.PRES.IND form of SEEM/APPEAR-verbs, e.g. Eng. it seems that P, Germ. es scheint, daß P, Fr. il paraît que P,37 It. pare/sembra che P, Pol. zdaje siĊ/wydaje siĊ, Īe P, Russ. kažetsja, þto P, Lith. atrodo, kad P, as well as Gr. fénome (K. Stathi, p.c.), etc. Their etymology denotes nonagentive, non-specific perception; their “impersonal” forms regularly evolve into inferential matrix predicates with sentential complements, in many cases further into particles (cf. de Haan 2007 and Diewald 2001 on Germanic). Not every item among these crosslinguistic equivalents develops into a marker of hearsay, and there are borderline cases for which it is difficult to decide whether a reportive interpretation is to be considered still as a context-conditioned implicature or already as a conventionalized meaning (as for Russian and Polish cf. Wiemer 2006b: 55f.; 2008b). But remarkably, the transition from inferential to hearsay meaning seems to correlate with a change from predicative to particle status. A similar observation was made in 3.7 with respect to ‘as if‘complementizers from which hearsay particles evolve. From the syntactic point of view they “mirror” units lexicalized from SEEM/APPEAR-verbs since ‘as if’ often links complements to such verbs. We have here another prominent type of heterosemy, the most complex case probably being Lith. esą; although its etymology does not derive from ‘as if’, it functions like an ‘as if’-unit and deserves special attention.

112

Björn Wiemer

Compare (40), where esą was used as a particle, with example (63) and the following one, where it links a clause to the verb priekaištauti ‘to reproach’ (cf. Wiemer 2007b: 177): (70)

R. Ozolas sakơ, kad jam bandoma priekaištauti, esą pasitraukus A. Šleževiþiui, Ƴ jo vietą ateis kitas pareignjnas, kuris darytǐ lygiai tą patƳ. ‘R. Ozolas said that there had been attempts to reproach him, (for) as people say when A. Šleževiþius will have retired, another official will occupy his place,who will do exactly the same.’(Lietuvos Aidas1995-9)

Moreover, esą can be encountered in contexts where it can be regarded as existential or copular predicate within a hearsay-context, although this happens much more rarely; see, for instance the following (cf. also Wiemer 2007b: 178, ex. 4): (71)

Per patƳ šaudymosi ir sprogimǐ ƳkarštƳ (…) pro mus praơjo du vokieþiǐ kariai ir pasakơ, kad už jǐ jau slenka rusai, kuriǐ esą labai daug. ‘In the heat of the gunfire and between explosions two German soldiers passed by saying that behind them the Russians are advancing, and allegedly there were many of them.’ (Z. Zinkeviþius “Prie lituanistikos židinio”, Vilnius 1999; p. 55)

This usage reflects the original use from which the conjunctional and particle use must have developed. Furthermore, a challenge for an account in terms of word and morpheme classes is offered by some quotative markers. Engl. like is a good case in point. See the following example from colloquial English (Google, by courtesy of M. Lampert): (72)

How did you come up with the concept for “I Don’t Need a Hook”? The concept came from one of my friends. My friend rec orded it in his studio in Patterson, he's an upcoming producer, Dustin Hill, and he was like, “Oh, yeah, I have this beat for you!” And I’m like, “OK, let me hear it.” And I’m like, “Wow, that's really a hot beat. I have to really get on that beat.” He like, “Yeah, I think this song should be called ‘I Don’t Need a Hook.’” And I’m like, “OK.” So, from there I just wrote the song and it’s all history from there.



Hearsay in European languages

113

Should we say that like is an adjective in predicative use and that the direct quote follows on it asyndetically? Or should we, instead, assume there to be an elliptical (or zero) predicate, just as this could be argued for in the case of the Czech complementizer že when it started losing its connection with the speech act verb (see ex. 49-50)? Based on the latter analysis, like could be considered as a particle. In contrast to Cz. že, it does not originate from a complementizer, but from an expression used in comparisons. We may say that its semantic evolution into a quotative marker has been accompanied by a syntactic change into an adverbial modifier of a zero predicate with very general meaning (‘to say’ or ‘to do’).38 A similar point can be made for colloquial Russ. tipa ‘like, as if’; morphologicaly, this is the genitive of tip ‘type, sort’. See the following internet example (Google, by courtesy of M. Makarcev; idiomatic translation approximative): (73)

Blin, devki, mne segodnja mama takoe zarjadila. Posmotrela v telefon, a tam Tomik i govorit takaja tipa na tebja poxož!!! Ja þut‘ ne vypala!!! ‘Hi guys, fuck you, today my mum just knocked me flat out. She looked at the phone, and there was Tomik and she was like “he’s like you”!!! It almost threw me for a loop!!!’ (http://forum.tokiohotel.ru/showthread.php?t=1316&page=28)

From the syntactic point of view, we can hardly qualify tipa as the predicate, even if we do not assume a syntactic break after the preceding speech act verb govorit ‘(she) says’. We can regard tipa as being coded along with a zero predicate. Zero predicates are a typical phenomenon of colloquial Russian, and speech acts are one of three semantic domains in which they occur particularly often.39 If the zero interpretation proves an adequate option, Russ. tipa should be treated as a particle, too.

4. In place of a conclusion: recollecting the threads and data As mentioned at the beginning, only some factors from the enormously complex picture which condition the meaning and the behavior of hearsay marking devices could be discussed and shown to be relevant for a comprehensive typology of such markers, which would include not only grammatical means (bound morphology, auxiliaries, periphrastic TAM grams), but also lexical units. Until very recently, the latter ones have largely been

114

Björn Wiemer

neglected, the remarks made above could, to a lesser extent, only be based on firmly achieved linguistic knowledge. Many times for these, I could only “scratch” the surface of observations whose empirical validity must be ascertained not only by thorough comparative in-depth studies, but also on the basis of a transparent methodology which seeks to distinguish properly evidential from epistemic meaning components, pragmatic implicatures from non-detachable semantic elements as well as diverse subfunctions within evidentiality, and which is anxious to unite markers of different morphosyntactic format and status under a common functional roof. The concluding resumption to follow now is meant only to highlight some important, recurrent phenomena and problems mentioned in 3. First, we have to stress that many markers with hearsay interpretations are not reportive units per se, but should rather be characterized as general markers of indirect evidentiality since they do not differentiate between reportive and inferential functions. These indiscriminate semantics are a recurrent pattern of the following classes of markers: (i) evidential extensions of the (present) perfect; the Baltic languages and Estonian are an exception to this “rule”, insofar as their respective active past participles seem to be restricted to hearsay. (ii) evidential extensions of the conditional: modern Italian is an exception. Since, however, the restriction of the Italian conditional to hearsay is a recent development, it rather confirms the indiscriminate nature of this mood within evidentiality. In this respect, the Romance conditional strikingly differs from the German subjunctive, which is restricted to hearsay as is its “analytic substitute” würde + infinitive. (iii) evidential extensions of modal auxiliaries of deontic necessity; compare Romance descendents of Latin debƝre as well as Dutch moeten. For the latter, a hearsay function is debatable, even more so for Germ. müssen and Sp. deber. It remains an open question as to whether evidential extensions of MUST-auxiliaries, if they occur at all, are acquired via epistemic necessity or directly from deontic necessity. Notice further that Italian and French modals of strong obligation rather acquire evidential functions in their conditional forms. Again, their pattern remarkably differs from Germ. sollen and West Slavic HAVEverbs, which are representatives of weakened obligation and have extended only into hearsay, not into inferentiality. (iv) Throughout Europe, the form of the 3rd person singular present indicative of SEEM/APPEAR-verbs extends into evidentiality, but a hearsay function becomes more salient only when this form, by losing its ar-



Hearsay in European languages

115

gument structure, starts being used as a particle (or parenthetical), i.e. without any complementizer. This allows for a careful hypothesis saying that reportive meanings correlate with the least possible integration of paradigmatically isolated verb forms into clausal syntax, whereas inferential functions of the same units are available if the respective unit still functions as a predicate with a sentential argument in a syntactically definable dependency relation. (v) The same applies to sentential adverbs like Engl. apparently, units in which a form of a SEEM/APPEAR-verb has fused with a complementizer, e.g. Sic. parica ‘apparently‘ (Cruschina/Remberger, to appear: p. 21), or stable collocations like It. a quanto pare Squartini (2008: 927939), which should be regarded as parentheticals. (vi) An analogous remark holds for ‘as if’-connectives. They have to be seen as the syntactic mirror image of petrified forms of SEEM/APPEARverbs, insofar as they can serve as their complementizers. Again, hearsay function is typical for particles with an ‘as if’-etymology rather than complementizers. If ‘as if’-units function as evidential complementizers, they are either indifferent, since an inferential vs. reportive reading has to be figured out from the context (Pol. jakoby), or they are in almost complementary distribution with cognate units; e.g. Russ. kak budto (inferential, never reportive as complementizer and seldom so as particle) vs. budto by (rather reportive if used as a particle, but also inferential if used as a complementizer). Georg. turme, titkos, vitom as well as Latv. it kƗ mark hearsay only as particles; Russ. jàkoby (a cognate of Pol. jakoby) has become obsolete as a complementizer with inferential function and developed into a particle marking only hearsay (accompanied by epistemic overtones). Notice that Lith. esą shows the same heterosemy as ‘as if’-units but is restricted to hearsay, probably due to its different etymology (petrified copula going back to logophoric constructions). Curiously, I have not come across instances of evidential ‘as if’-units outside East Slavic, Polish, Baltic and the Caucasus (Georgian, Armenian); apparently they are typical for (or even restricted to?) the eastern half of Europe. On the basis of these observations we can hypothesize that a lack of discrimination between inferential and reportive functions is the result of a slow transition from inferential to reportive meaning, in other words: that inferential meaning is acquired before shifts into reportive interpretations take place; in some cases, this shift eventually ended with a stable, nondetachable hearsay component of the respective unit’s meaning. A plausi-

116

Björn Wiemer

ble pragmatic mechanism favoring such a transition is based on the need to ground one’s inference on other people’s assertions. For diachronic and discourse pragmatic arguments in favor of this explanation cf. Wiemer 2005; 2008b based on Russian and Polish. As concerns grammatical marking in the strict sense, the overview in 3.1-3.2 showed that both the distribution of evidential extensions of TAMgrams (however marked) and the distribution of inflectional vs. agglutinative morphology with evidential functions do not scatter randomly. Instead, within Europe, clear areal biases can be ascertained: the eastern half of Europe “prefers” an extension of the perfect, whereas in the western “hemisphere” (continental Germanic and Romance), marked moods (the subjunctive and its analytical substitutes in German and Dutch, and the conditional in Romance) are chosen as the basis of evidential extensions. The fact that Romanian in this regard patterns basically like the other Romance languages, although it otherwise has acquired many Balkanisms and Slavicisms due to language contact, demonstrates that contact-conditioned features (among others, functional extensions of TAM-paradigms) can be overridden, or resisted by remarkable genetic stability (for a survey on Romance cf. Squartini 2001; 2005). On the other hand, Rhaeto-Romance uses the subjunctive, not the conditional, contrary to the rest of Romance; a fact which could be viewed as a result of German influence. Agglutinated morphemes indicating hearsay, either as a central function or as a functional extension, are rare in Europe; they appear only in three languages of its South-Eastern and Eastern periphery: in Estonian, Albanian and Turkic.40 Only Est. -vat specializes in hearsay, whereas the reportive function of Turk. -mIú and Alb. -kam is only a contextually conditioned reading of a broad evidential (including mirative) meaning potential. Georgian is the only language using inflectional morphology, as its perfect forms part of a system based on alternations of stems. Finally, notice that there is no tight correlation between type of morphology and gram type: evidential extensions from the perfect can remain analytic constructions (Balkan Slavic, Baltic languages, Armenian), can betray signs of coalescence (Albanian), or be as agglutinative (Turkish) or inflectional (Georgian) as the respective perfect proper. Est. -vat, in turn, does not derive from the perfect but is nonetheless agglutinated. Finally, we must remain aware that the extensions of mood paradigms into evidentiality attested in Germanic and Romance can count as inflectional markers of hearsay (or evidentiality in general) only insofar as these inflections developed long before evidential extensions started, and the latter ones



Hearsay in European languages

117

have by no means led to a change in the morphonological features of these moods.

List of abbreviations Languages Alb. = Albanian, Croat./Serb. = Croatian/Serbian, Cz.= Czech, Eng. = English, Est. = Estonian, Georg. = Georgian, Germ. = German, Gr. = Greek, Fr. = French, It. = Italian, Latv. = Latvian, Lith. = Lithuanian, OE. = Old English, Pol. = Polish, Russ. = Russian, Sic. = Sicilian, Slov. = Slovak, Sp. = Spanish

Grammatical comments 1 = first person, ACC = accusative, ANT = anterior, AOR = aorist, ART = article, = copula, DAT = dative, DEF = definite, DNV = derived nominal verb, ERG = ergative, F = feminine, GEN = genitive, IMP = imperfect, IMPF = imperfective, INDF = indefnite, IND = indicative, INDECL = indeclinable, INE = inessive, INF = infinitive, LOC = locative, M = masculine, MA = indeclinable present tense participle (Lithuanian), N = neuter, NEG = negation, NOM = nominative, OPT = optative, PART = partitive, PAST = past tense, PL = plural, PPA = participle past active, PRES = present tense, QUOT = quotative, REP = reportive marker, RM = reflexive marker, SG = singular, TA = indeclinable past tense participle (Lithuanian), VERS = verbal version (Georgian). COP

Acknowledgments I have to thank very many colleagues and informants who helped gathering material, testing assumptions and clarifying misunderstandings. I am especially indebted to Walter Breu, Joanna Chojnicka, Matthias Grünert, Caterina Guardamagna, Maksim Makarcev, Tanja Mortelmans, Nicole Nau, Elena Petroska, Eva Maria Remberger, Julian Rentzsch, Birutơ Ryvitytơ, Mario Squartini, and Manana Topadze – to name only those whom I have consulted most frequently and/or with regard to most intricate details. My thanks are also due to an anonymous reviewer of the first version. I am nonetheless the only person responsible for any kind of incorrect data, misinterpretation or misguided claim that might have remained unnoticed so far in this article. Finally, I owe my gratitude also to Beth Martin for having “streamlined” the overall paper into an idiomatic English as well to

118

Björn Wiemer

Tina and Günther Lampert who, beforehand, had essentially contributed to rendering examples in a more readable English, facing the numerous cases in which the transfer of evidential marking into another language can hardly be at once adequate and idiomatic.

Notes 1.

2.

3.

4. 5.

6.

7.

For arguments in favor of such a separation cf., inter alia, Aikhenvald 2004, Anderson 1986, de Haan 1999a, and Ehrich 2001, to name but a few. Cf. also Wiemer 2006b; 2008a: 6-10. This restriction can be violated only under very specific discourse conditions (de Haan 1997: 154f.), which, however, do not bear on the argument developed in this article. Apart from this, if we do not accept a lexicon-grammar cline based on an opposition between an inventory of lexical units (functional or autosemous) and an inventory of morphemes that are highly dependent on their immediate morphosyntactic context with which they form analytically accessible, productive complex units, other methodological problems arise: (a) How are we to distinguish whether a one-word or a multiple-word form is conventionalized holistically? (b) How are we to distinguish units of different “formats” with identical functions (in particular, in discourse)? Both problems arise with respect to particles and parentheticals. Further analyses on such a system in Kartvelian and neighboring Caucasian languages can be found in Friedman (2000: 353-355) and Topadze 2008. Actually, the functional similarity of grammatical evidentiality marking and its tight connection to the perfect in the languages of this large area are obviously a result of areal convergence. It is very likely that the Turkic system rendered the model on which evidential extensions of the perfect spread via contact into adjacent language varieties. Megleno-Romanian, which is spoken in the vicinity of Albanian (in the GreekMacedonian border region), also uses an inverted perfect with an inflected (agglutinated or cliticized?) HAVE-auxiliary as an admirative (Friedman 2003: 190). As for Aromanian, however, only the Frasheriote dialect spoken in southwestern Macedonia appears to have developed an admirative marked by the invariant “particle” -ka, borrowed from Albanian kam ‘(I) have’ and attached to the inherited Romance HAVE-auxiliary or the inflected verb in the present tense (Friedman 2000: 347f.). For the classification pursued, it is of no importance whether the Albanian evidential forms should better be qualified as ‘admiratives’ (or ‘nonconfirmatives’, in Friedman’s terms) or whether the (ad)mirative function has to be regarded as an evidential one. For different views on the relation be-



8.

9.

10. 11. 12.

13.

14. 15. 16.

17.

Hearsay in European languages

119

tween evidentiality and (ad)mirativity in Albanian cf. Breu (2009: 4.3) vs. Friedman (2000: 342f.). The main difference with regard to Lithuanian (and Latvian) lies in the fact that, contrary to Armenian, Baltic has no aorist-imperfect system, only a general past. The paradigmatic and functional link between perfect and evidential usage is however the same, as is the “ban” on the perfect as a narrative tense (unless in re-narration, i.e. when hearsay occurs). As for Bulgarian cf. Friedman (2000: 334-338), and Kehayov (2002: 127f.), for Lithuanian Wiemer (1998: 232f.; 2006a; 2007a: 201-210), and for Georgian Boeder (2000: 285-294). For a systematic account cf. Wiemer (2006a; 2007a: 213-216). In Latvian, the future form of this uninflected participle (ending in -šot) is used, too (cf. Holvoet 2007: 82f.). The temporal value of such a sentence varies as it corresponds either to the past or to the present perfect. Informants confirm that esą combines with a finite verb or participle in any tense. Perfect evidentials (reportives) are characterized by the declinable (agreeing) form esąs.M, esanti.F etc. (e.g., Jis esąs gyvenĊs Vilniuje ‘He has lived in Vilnius, people say’), but these are very uncommon in colloqial speech (informants reject them) and seem to be restricted to some dialects. Such an implicature can be detected in other languages as well, regardless of their tense-aspect system. See, for instance, utterances characteristic of the German past tense (Präteritum, aspectually indifferent) formulated as questions like those adduced by Squartini: Wer bekam das Schnitzel? ‘Who is to get the escalope?’ (lit. ‘Who got the escalope?’), which covertly refers to a speech act (ordering of a meal). A reportive implicature is also possible with Russian (or Polish) imperfective aspect (both in present and past tense), e.g. Russ. (Ja znaju, þto) ix poezd otpravljalsja v vosem‘ þasov ‘(I know that) their train was to leave at eigth o’clock‘ (lit. left.IPF.PAST). Like the Imparfetto, this function of imperfective verbs (often called ‘prospective’) is typical of scheduled situations, but, contrary to the Imparfetto, it is also compatible with knowledge deriving from speaker’s eyewitnessing (...i ja videl, kak poezd uxodil ‘...and I saw how the train was leaving‘). The stem is pida-. The form used here represents the so-called ma-infinitive. Traditionally, and confusingly, this tense has been called “imperfect”; notice that it does not bear any specific aspectual value. This formal (and diachronically motivated) difference is usually neutralized, since in real discourse the time reference of these forms is indistinguishable (cf. Thieroff 1992: 223-227). In this respect, the form:function-relations of the subjunctive paradigms in contemporary German very much resemble the situation in Estonian and Baltic (see 3.2.1). Since würde + infinitive is connected not only with the subjunctive, but also with the future (werden + infinitive), its paradigmatic status in the German

120

18. 19.

20.

21.

22.

23.

24.

25.

Björn Wiemer tense-mood system has been an issue of constant debate, as has the question as to which degree this auxiliary construction can be considered a reportive marker (as opposed to the mere syntactic function of a subjunctive). The most recent overview of research is given in Smirnova (2006: 19-40). In the conclusion of her primarily diachronic investigation Smirnova subsumes that the evidential (i.e., reportive) reading of this construction is not as grammaticalized as its subjunctive function but should nonetheless be figured out as a distinct function (2006: 334). The form zou is for the singular, zouden for the plural. The only case when zou(den) can be imagined in indirect speech is found in sentences in which the content of speech refers to an interval posterior to the original speech act. Thus a German sentence like Er sagte, er würde nie mehr wiederkommen lit. ‘He said he would never return again’ can be translated with zou(den): Hij zei dat hij nooit nog zou terugkomen. This, however, is only a manifestation of the regular grammatical correspondence between future tense (in direct speech: Ik zal nooit terugkomen ‘I shall not [= don’t want to] come back’) and past conditional/subjunctive in indirect speech (T. Mortelmans, p.c.). From a diachronic viewpoint, the contemporary present tense of modals derives from the paradigm of former praeteritopraesentia. However, for the present classification this is irrelevant since these forms were reinterpreted as a special present tense conjugation a long time ago. Corpus examples with sollte like those adduced in Faller (2006: 17) are no proof against this restriction to the present tense form since sollte goes together here with the sentential adverb angeblich ‘allegedly’; without the latter a reportive reading is hardly available. Actually, the difference in functional distribution yields an argument for considering sollte(n) not as the past tense of soll(en), but as a separate lexeme. Disambiguation is possible with negation because the deontic reading requires external negation (i.e., negation includes the modal) whereas with evidential (reportive) reading, internal negation holds (i.e., negation is in the scope of the “evidential modal”). Cf. the thorough analysis on this matter by Ehrich 2001. (20’) could be interpreted at best in the sense that the people spoken about left some written documents behind about, say, their conditions for and habits of labor. But how could those people have compared their way of life with conditions at the beginning of the 21st century? Compare an instance from Šoloxov (“Tixij Don“): – Pro Aksin’ju ne slyšno? – sprosil Grigorij s delannym ravnodušiem v golose. ‘“Is there nothing to be heard about Aksin’ja?”, Grigorij asked in a voice that pretended indifference.’ Notice that focus particles belong to a slightly different species and are not considered here because – contrary to modal particles – they usually focus on a specific part of a sentence (uttterance). Although there have been attempts at analysing focus particles as sentential modifiers (cf. König 1991: 11, 19;



26.

27. 28.

29.

30.

31.

32.

33. 34.

Hearsay in European languages

121

Mosegaard Hansen 1998: 41-49), they obviously differ in their scopal behavior from sentential adverbs and particles, which always take propositional scope but are often able to take narrow scope, too (see below on variable scope). Actually, scope may be ambiguous and embrace only the adjacent constituents to the right (the adverbial erst um fünf ‘only at five’ together or without the comitative PP). Prosody (sentence stress) plays an important role in disambiguating scope. We must, however, leave this general problem, as it does not contribute essentially to the present points. For an analysis of budto by and other etymologically closely related units cf. Letuþij (2008: 222-232) and Wiemer (2008b: 347-353). Such a component need not be marked morphologically, it can also manifest itself as a prosodic (or graphic) feature in combination with the conversion of whole phrases. But with SO CALLED-units, they are part of their lexical meaning. A preliminary investigation of Russ. jakoby and its Polish cognate jakoby corroborates this assumption (for details cf. Wiemer 2005: 115, 120-123; 2008b: 362-369). As for the development of evidential marking from comparison constructions, in general, and the Russian series of expressions centering around budto ‘as if’, in particular, cf. Letuþij 2008. During the last centuries Russian xenomarkers must have experienced a functional change. In former times (obviously until the 19th c.) they were quite common as markers of direct quotes, whereas in recent times they have been moving out of this domain and been specializing as markers of speech that, on the contrary, has to be reformulated by the actual speaker (cf. Plungjan 2008). The German equivalent is ‘Schaltwörter, Schaltsyntagmen’ (literally ‘switching words/syntagms’), the Russian one, ‘vvodnye slova’ (lit. ‘introductory words’). Cf. Kaltenböck (2007: 31) on parenthetical clauses: “parentheticals cannot be defined by themselves. (…) They derive their existence, as it were, from their interaction with a host clause. This interaction, however, takes place purely on the linear plane, not on a relational (i.e., dependency) level.” (Similarly in Dehé and Kovalova 2007: 9 and Grochowski 2007: 71.) The same two characteristics – a functional one: their interpretation depends on a host to whose truth-conditional semantics they bear some relevance, and a negative syntactic one: no dependency relations – are typical of modal, or epistemic, particles (under which evidential particles have usually been subsumed). The same intermediate status can be ascribed to dit-on. This is probably also true of some units which are usually classified as sentential adverbs but have identical etymology, like Engl. apparently, Dutch schijnbaar and their German counterpart anscheinend, which lately seems to be increasingly used for the purposes of hearsay.

122

Björn Wiemer

35. To this extent, Sp. al parecer (= infinitive) is an exception. 36. For further details cf. Wiemer (2006b: 32-38). In the last resort the semantic difference between these two types of hearsay adpositions can be captured by differences in metonymic shifts and conceptual distance: units of the first type allow only for shifts from ‘person’ to ‘institution embodied by that person (or several persons)’, i.e., to an immediately contiguous subdomain. Units of the second type allow for more drastic shifts (or chains of metonymic transfers). 37. The appearance of an expletive subject pronoun in English-type languages is due to their non-PRODROP-character and irrelevant to the argument because expletive subjects do not fill argument positions. 38. The same applies, e.g., to It. dice (‘say’), which presents us with the syntactic “mirror image” of what happens to complementizers or adverbials. 39. The other two domains are movements (as a rule, one-directional) and intense physical actions like beating and hitting. 40. From a broader areal perspective the Turkic and Estonian (at the least) should be considered as the western-most representatives of a vast Eurasian territory (cf. Johanson 2000: 83f.; Wälchli 2000 and the contributions to Xrakovskij (ed.) 2007).

References Achema, P. and Neeleman, A. 2004 Beyond Morphology. Interface Conditions on Word Order. Oxford: Oxford University Press. Aikhenvald, AlexandraY. 2003 Evidentiality in typological perspective. In Aikhenvald, Alexandra Y. and R.M.W. Dixon (eds.), 1-31. 2004 Evidentiality. Oxford: Oxford University Press. 2007 Information source and evidentiality: what can we conclude? In Evidentiality between lexicon and grammar. Squartini, M. (ed.), 209227. Aikhenvald, Alexandra Y. and R.M.W. Dixon (eds.) 2003 Studies in Evidentiality. Amsterdam/Philadelphia: John Benjamins. Anderson, Lloyd B. 1986 Evidentials, Paths of Change, and Mental Maps: Typologically Regular Asymmetries. In: Chafe, Wallace and Johanna Nichols (eds.), 273-312. Arutjunova, N.D. 2000 Pokazateli þužoj reþi de, deskat‘, mol. K probleme interpretacii reþepovedenþeskix aktov. In Jazyk o jazyke. Arutjunova, N.D. (ed.), 437449. Moskva: Jazyki russkoj kul’tury.



Hearsay in European languages

123

Autenrieth, Tanja 2002 Heterosemie und Grammatikalisierung bei Modalpartikeln (Eine synchrone und diachrone Studie anhand von »eben«, »halt«, »e(cher)t«, »einfach«, »schlicht« und »glatt«). Tübingen: Niemeyer. Boeder, W. 2000 Evidentiality in Georgian. In Johanson, Lars and Bo Utas (eds.), 275328. Brendel, Eva, Jörg Meibauer and M. Steinbach 2007 Aspekte einer Theorie des Zitierens. In Zitat und Bedeutung. Brendel, E., Meibauer, J., and Steinbach, M. (eds.), 5-25. Hamburg: Buske. Breu, W. 1987 Albanisches und makedonisches Verbalsystem in kontrastiver Gegenüberstellung. Balkan-Archiv, Neue Folge 12: 279-305. forth. Mood in Albanian. In Mood in the Languages of Europe. Rothstein, B. and Rolf Thieroff (eds.). Amsterdam/Philadelphia: John Benjamins. Bulygina, T.V. and A.D. Šmelev 1993 Gipoteza kak myslitel’nyj i reþevoj akt. In Logiþeskij analiz jazyka: Mental’nye dejstvija. Arutjunova, N.D. (ed.), 78-82. Moskva: Nauka. Burkhardt, D. 1999 Govorjašþee lico und þasti(cy) reþi: Überlegungen zu den vvodnye slova. In Beiträge der Europäischen Slavistischen Linguistik (POLYSLAV). Böttger, K., M. Giger and Björn Wiemer (eds.), 8392. München: Sagner. Chafe, Wallace 1986 Evidentiality in English Conversation and Academic Writing. In Chafe, Wallace and Johanna Nichols (eds.), 261-272. Chafe, Wallace and Johanna Nichols (eds.) 1986 Evidentiality: The Linguistic Coding of Epistemology. Norwood, NJ: Ablex Publ. Cornillie, Bert 2007 Evidentiality and Epistemic Modality in Spanish (Semi-)Auxiliaries. A Cognitive-Functional Approach, Berlin/New York: Mouton de Gruyter. Cruschina, Sanrdo and Eva-Maria Remberger to appear In Selected Proceedings of the 34th Incontro di Grammatica Generativa. Benincà, Paola, Federico Damonte and Nicoletta Penello (eds.). Padova: Unipress. Csató, É.Á. 2000 Turkish MIù- and IMIù-items. Dimensions of a functional analysis. In Johanson, Lars and Bo Utas (eds.), 29-43.

124

Björn Wiemer

de Haan, Ferdinand 1997 The Interaction of Modality and Negation (A Typological Study). New York, London: Garland Publ. 1999a Evidentiality and Epistemic Modality: Setting Boundaries. Southwest Journal of Linguistics 18: 83-101. 1999b Evidentiality in Dutch. Berkeley Linguistics Society 25: 74-85. 2007 Raising as grammaticalization: the case of Germanic SEEM-verbs. In Squartini, Mario (ed.), 129-150. Dehé, Nicole and Y. Kovalova 2007 Parentheticals (An introduction). In Dehé, Nicole and Y. Kovalova (eds.), 1-22. Dehé, Nicole and Y. Kovalova (eds.) 2007 Parentheticals. Amsterdam/Philadelphia: John Benjamins. Diewald, Gabriele 1999 Die Modalverben im Deutschen: Grammatikalisierung und Polyfunktionalität. Tübingen: Niemeyer. 2001 Scheinen-Probleme: Analogie, Konstruktionsmischung und die Sogwirkung aktiver Grammatikalisierungskanäle. In Müller, Reimar and Marga Reis (eds.), 87-110. Donabédian, A. 1996 Pour une interprétation des différentes valeurs du médiatif en arménien occidental. In Guentchéva, Zlatka (ed.), 87-108. Dooley, Sh. and Ferdinand de Haan 2006 Epistemic Modality and Evidentiality: Swedish lär. University of Arizona. http://u.arizona.edu/~fdehaan/papers/laer.html Ehrich, Veronika 2001 Was nicht müssen und nicht können (nicht) bedeuten können: Zum Skopus der Negation bei den Modalverben des Deutschen. In Müller, Reimar and Marga Reis (eds.), 149-176. Erelt, M. 2001 Some notes on the grammaticalization of the verb pidama in Estonian. Estonian Typological Studies 5: 7-25. Faller, Martina 2006 Evidentiality below and above speech acts. http://personalpages.manchester.ac.uk/staff/Martina.T.Faller/docume nts/Evidentiality.Above.Below.pdf Faßke, H. 1981 Grammatik der obersorbischen Schriftsprache der Gegenwart (Morphologie). Bautzen: VEB Domowina. Friedman, Viktor A. 2000 Confirmative/nonconfirmative in Balkan Slavic, Balkan Romance, and Albanian with additional observations on Turkish, Romani, Georgian, and Lak. In Johanson, Lars and Bo Utas (eds.), 329-366.



Hearsay in European languages 2003

125

Evidentiality in the Balkans with special attention to Macedonian and Albanian. In Aikhenvald, Alexandra Y. and R.M.W. Dixon (eds.), 189-218. Giacalone Ramat, Anna and Manana Topadze 2007 The coding of evidentiality: a comparative look at Georgian and Italian. In Squartini (ed.), 7-38. Golato, Andrea 2002 An innovative German quotative for reporting on embodied actions: Und ich so/und er so ‘and I’m like/and he’s like’. Journal of Pragmatics 32: 29-54. Grochowski, M. 1986 Polskie partykuły (Składnia, semantyka, leksykografia). Wrocław etc.: Ossolineum. 1997 WyraĪenia funkcyjne (Studium leksykograficzne). Kraków: IJP PAN. 2007 Partykuły, parenteza a wyraĪenia metatekstowe. Z polskich studiów slawistycznych, seria XI: JĊzykoznawstwo. Warszawa, 69-74. Grünert, M. 2003 Modussyntax im Surselvischen (Ein Beitrag zur Erforschung der Morphosyntax des Verbs im Bündnerromanischen). Tübingen/Basel: Francke. Guentchéva, Zlatka (ed.) 1996 L’énonciation médiatisée. Louvain/Paris: Peeters. Hansen, Björn 2001 Das slavische Modalauxiliar (Semantik und Grammatikalisierung im Russischen, Polnischen, Serbischen/Kroatischen und Altkirchenslavischen). München: Sagner. Heine, Bernd 1990 Grammaticalization Chains as Linguistic Categories. Duisburg. (L.A.U.D., Series A, Paper No. 291.) Himmelmann, Nikolaus P. 2004 Lexicalization and grammaticization: Opposite or orthogonal? In What makes Grammaticalization? A Look from its Fringes and its Components. Bisang, Walter, N.P. Himmelmann and Björn Wiemer (eds.), 21-42. Berlin/New York: Mouton de Gruyter. Hoffmannová, J. 2002 Intertextualita v mluvených projevech (Úvahy o metodologii výzkumu a konceptuální soustavČ). Stylistyka XI, 371-381. 2008 Role þeského prý/prej/pré pĜi vyjadĜování reproduktivních a modálních významĤ. In Wiemer, Björn and Vladimir A. Plungjan (eds.), 149-165. Holvoet, A. 2007 Mood and modality in Baltic. Kraków: Wyd-wo UJ.

126

Björn Wiemer

Johanson, Lars 2000 Turkic indirectives. In Johanson, Lars and Bo Utas (eds.), 61-87. Johanson, Lars and B. Utas (eds.) 2000 Evidentials. Turkic, Iranian and Neighbouring Languages. Berlin/New York: Mouton de Gruyter. Kaltenböck, G. 2007 Spoken parenthetical clauses in English (A taxonomy). In: Dehé, Nicole and Y. Kovalova (eds.), 25-52. Karlík, P., M. Nekula and Zd. Rusínová (eds.) 1995 PĜíruþní mluvnice þeštiny, Praha: Nak-ví Lidové noviny. Kehayov, Petar 2002 Typology of grammaticalized evidentiality in Bulgarian and Estonian. Linguistica Uralica 38(2): 126-144. König, Ekkehard 1991 The Meaning of Focus Particles: A Comparative Perspective. London/New York: Routledge. Kozintseva [= Kozinceva], N. 2000 Perfect forms as a means of expressing evidentiality in Modern Eastern Armenian. In Johanson and Utas (eds.), 401-417. 2007 Perfect forms as a means of expressing evidentiality in Modern Eastern Armenian. In Xrakovskij, Viktor S. (ed.), 67-84. Kronning, H. 2007 Les auxiliaires «mediatifs» en suédois (à la lumière du conditionnel épistémique et du futur aléthique en français). In Problèmes de sémantique et de syntaxe. Hommage à André Rousseau. Begioni, L. and Cl. Muller (eds.), 287-309. Lille : Éd. du Conseil Scientifique de l’Université Charles-de-Gaulle. Kuz’mina, I.B. 1993 Sintaksis russkix govorov v lingvogeografiþeskom aspekte. Moskva: Nauka. Lehmann, Christian 1995 Thoughts on Grammaticalization. München: LINCOM Europa. 2002 New reflections on grammaticalization and lexicalization. In New reflections on grammaticalization. Wischer, Ilse and Gabriele Diewald (eds.), 1-18. Amsterdam/Philadelphia: John Benjamins. Letuþij, Alexandr 2008 Sravnitel’nye konstrukcii, irrealis i ơvidencial’nost‘. In Wiemer, Björn and Vladimir A. Plungjan (eds.), 217-238. Lichtenberk, F. 1991 Semantic Change and Heterosemy in Grammaticalization. Language 67: 475-509. Matras, Yaron 1995 Verb evidentials and their discourse function in Vlach Romani narratives. In Romani in contact: The history, structure and sociology of a



Hearsay in European languages

127

language. Matras, Yaron (ed.), 95-123. Amsterdam/Philadelphia: John Benjamins. Meibauer, Jörg 2003 Phrasenkomposita zwischen Wortsyntax und Lexikon. Zeitschrift für Sprachwissenschaft 22: 153-188. Moore, C. 2007 The spread of grammaticalized forms: The case of be+supposed to. Journal of English Linguistics 35(2): 117-131. Mortelmans, Tanja 2000 On the ‘Evidential’ Nature of the ‘Epistemic’ Use of the German Modals müssen and sollen. In Modal verbs in Germanic and Romance languages. van der Auwera, Johan and Patrick Dendale (eds.), 131-148. (= Belgian Journal of Linguistics 14). Mosegaard Hansen, Maj-Britt 1998 The function of discourse particles: A study with special reference to spoken standard French. Amsterdam/Philadelphia: Benjamins. Müller, Reiner and Marga Reis (eds.) 2001 Modalität und Modalverben im Deutschen. Hamburg: Buske. Mushin, Ilana 1997 Maintaining Epistemological Stance: Direct Speech and Evidentiality in Macedonian. In Chicago Linguistic Society 33: The Main Session. Singer, K. et al. (eds.), 287-300. Chicago. Palmer, Frank R. 2001 Mood and Modality, Cambridge: Cambridge University Press. Pietrandrea, Paola 2007 The grammatical nature of some epistemic-evidential adverbs in spoken Italian. In Squartini, Mario (ed.), 39-63. Plungian, Vladimir A. 2001 The place of evidentiality within the universal grammatical space. Journal of Pragmatics 33 (3): 349-357. 2008 Ob ơvidencial’nyx pokazateljax þužoj reþi i nedostovernosti v russkom jazyke: mol, jakoby i drugie. In Wiemer, Björn and Vladimir A. Plungjan (eds.), 285-311. Rakhilina, E.V. 1996 Jakoby comme procédé de médiatisation en russe. In Guentchéva, Zlatka (ed.), 299-304. Ramat, Paolo 1996 “Allegedly, John is ill again”: strategies pour le médiatif. In Guentchéva, Zlatka (ed.), 287-298. Remberger, Eva-Maria to appear The Evidential Shift of WANT. In Proceedings of the GLOW Workshop on evidentiality. UBC Working Papers in Linguistics. Dechain, R.-M., T. Peterson, M. Schenner, U. Sauerland (eds.). Utrecht.

128

Björn Wiemer

Roels, Linde, Tanja Mortelmans, and Johan van der Auwera 2007 Dutch equivalents of the German past conjunctive: zou + infinitive and the modal preterit. Cahiers Chronos 17: 177-196. Rytel, D. 1982 Leksykalne Ğrodki wyraĪania modalnoĞci w jĊzyku czeskim i polskim. Wrocław etc.: Ossolineum. Smirnova, Elena 2006 Die Entwicklung der Konstruktion würde + Infinitiv im Deutschen. Berlin/New York: de Gruyter. 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. 2005 L’evidenzialità in rumeno e nelle altre lingue romanze. Zeitschrift für romanische Philologie 121 (2): 246-268. 2007 Evidentiality between lexicon and grammar (= Rivista di Linguistica 19-1). 2008 Lexical vs. grammatical evidentiality in French and Italian. Linguistics 46 (5): 917-947. StĊpieĔ, M. 2008 O wzajemnym przenikaniu siĊ ewidencjalnoĞci i modalnoĞci (na przykładzie wybranych polskich czasowników i wyraĪeĔ funkcyjnych). In Wiemer, Björn and Vladimir A. Plungjan (eds.), 313-333. Thieroff, Rolf 1992 Das finite Verbi m Deutschen: Tempus-Modus-Distanz. Tübingen: Narr. Topadze, Manana 2008 Mezzi di espressione dell’evidenzialità in Georgiano. Pavia: Università degli Studi di Pavia, Facoltà di Lettere e Filosofia. (unpubl. ms. of doctoral dissertation) Traugott, Elizabeth C. 1989 On the Rise of Epistemic Meaning in English: An Example of Subjectification in Semantic Change. Language 65 (1): 31-55. Travis, C. 2006 Dizque: a Colombian evidentiality strategy. Linguistics 44(6): 1269-97. Usonienơ, Aurelia 2001 Veiksmažodžio matyti komplementǐ tipai: formos ir reikšmơs sąveika. Baltistica 36(1): 115-124. Wälchli, B. 2000 Infinite predication as a marker of evidentiality and modality in the languages of the Baltic region. Sprachtypologie und Universalienforschung 53(2): 186-210.



Hearsay in European languages

129

Wiemer, Björn 1998 Pragmatical inferences at the threshold to grammaticalization − The case of Lithuanian predicative participles and their functions. Linguistica Baltica 7: 229-243. 2005 Conceptual affinities and diachronic relationships between epistemic, inferential and quotative functions (preliminary observations on lexical markers in Russian, Polish and Lithuanian). In Modality in Slavonic languages: New perspectives. Hansen, Björn and P. Karlík (eds.), 107-131. München: Sagner. 2006a Grammatical evidentiality in Lithuanian: a typological assessment. Baltistica 41(1): 33-49. 2006b Particles, parentheticals, conjunctions and prepositions as evidentiality markers in contemporary Polish: A first exploratory study. Studies in Polish Linguistics 3: 5-67. 2007a Kosvennaja zasvidetel’stvovannost’ v litovskom jazyke. In Ơvidencial’nost’ v jazykax Evropy i Azii. (Sbornik statej pamjati Natalii Andreevny Kozincevoj). Xrakovskij, Viktor S. (ed.), 197-240. SanktPeterburg: Nauka. 2007b Lexical markers of evidentiality in Lithuanian. In Squartini, Mario (ed.), 173-208. 2008a Lexikalische Markierungen evidenzieller Funktionen: zur Theoriebildung und empirischen Erforschung im Slavischen. In Wiemer, Björn and Vladimir A. Plungjan (eds.), 5-49. 2008b Pokazateli s citativnoj i inferentivnoj funkcijami v russkom i pol’skom jazykax – kommunikativnye mexanizmy semantiþeskogo sdviga. In Björn Wiemer and Vladimir A. Plungjan (eds.), 337-378. Wiemer, Björn and Vladimir A. Plungjan (eds.) 2008 Lexikalische Evidenzialitätsmarker im Slavischen. (= Wiener Slawistischer Almanach, Sonderband 72.) Wierzbicka, Anna 2006 English. Meaning and Culture. Oxford: Oxford University Press. Xrakovskij, Viktor S. (ed.) 2007 Ơvidencial’nost’ v jazykax Evropy i Azii (Sbornik statej pamjati Natalii Andreevny Kozincevoj). Sankt-Peterburg: Nauka.

Information source in Spanish and Basque: a parallel corpus study1 Asier Alcázar

This paper is a preliminary study of Basque evidentiality that builds on empirical data from a Spanish-to-Basque parallel corpus. The focus of the study is omen, an alleged hearsay evidential marker. The corpus data confirms that omen is an expression of grammatical evidentiality. Basque translators also resort to lexical evidentiality in particular cases: to express hearsay at the constituent level, to avoid scope ambiguity (parentheticals) and to preserve ambiguity in the original (vs. inference, doubt).

1. Introduction Does Basque have an evidential system? This paper is the first in a series of investigations dedicated to answering the question (see Alcázar 2009a, to appear, 2009b, 2009c). To this end I have studied a collection of sentence pairs extracted from a Spanish-to-Basque parallel corpus, where the same sentence is expressed in both languages; Spanish being the original and Basque the translation. In light of the empirical data, I take Spanish and Basque to represent two different evidential systems, broadly conceived as lexical and grammatical, respectively. What these two labels mean, and the predictions that may follow from them, is the subject of a lively and productive discussion in the field (see, among others, the contributions to this volume on evidentiality in Western European languages, and the individual contributions of Squartini, Pietrandrea and Aikhenvald to the special issue on evidentiality in Rivista di Linguistica). I will illustrate the object of the study, the sentence pairs, with an example momentarily, which I hope will at least vindicate that Spanish and Basque express evidentiality differently, if many questions remain about the lexicalgrammatical divide2. The following example illustrates the sentence pairs under investigation. The first part of the sentence pair is the Spanish original (1), which is followed first by a line indicating the grammatical gloss and secondly by a line providing a non-literal English translation. Bold

132

Asier Alcázar

font identifies words that we assume to express information source. A list of abbreviations used in the glosses can be found in this endnote3. (1)

aún está lejos la que parece the that seem.3SG yet be.3SG far ‘what seems to be [the best solution] is yet to come: la solución definitiva: el que sea the solution ultimate the that be.3SG.SUBJ ‘the best solution:’ escribir al dictado write to.the dictation ‘the day one could write by dictating.’

posible possible

In (1) the use of a clause with the verb parecer ‘seem’ points to the reader that the writer is either presenting this information third hand, or not fully vouching for it, among other possible interpretations. From the choices made by the Basque translators, I infer that adverbs such as todavía and aún ‘yet’ may on occasion serve a similar role. The second part of the example, indicated by (1’), comes from the Basque section of the corpus, which was aligned to its Spanish original by means of computational tools. The Basque example is followed by a grammatical gloss, which is not too detailed to enable the examples to fit in one or two lines, and also to facilitate their inspection for readers unfamiliar with the language. It seems fitting to provide a second, semi-literal translation for the Basque example to elucidate its syntax. For instance, in contrast to English and Spanish, Basque relative clauses precede the noun they modify (cf. first line 1’), the unmarked word order is SOV (vs. SVO), postpositions are used instead of prepositions, etc. (1’)

Azken-eko aukera omen den option HEARSAY be.3SG.REL last-GEN ‘which seems to be the best option’ esan-era idatzi ahal izate-a, write able be-ABS say-ADL ‘being able to write by dictating’ ordea, urruti omen dugu orain-o however far HEARSAY have.1PL.3SG now-ADL ‘however, is yet far from us’ (lit.: we have it far up to now)

Beyond salient differences in sentence structure between the two languages, one other thing stands out. Basque may resort to a different mechanism

Information source in Spanish and Basque

133

to indicate information source. In (1’) omen, a verb particle, is used in both instances to relay a similar, though more specific meaning. As we will see, omen expresses hearsay evidence. What is more, the translator feels at liberty to dispense with the elements in the Spanish original that signal a potential hearsay interpretation: the verb and the adverb. (1’’) presents a comparison of the two sentences. (1’’)

Proposition 1: Spanish: Basque: Proposition 2: Spanish: Basque:

[it is far] neutral (or lexically cued by aún) /omen/ [(be) the best solution] /parece/ /omen/

The corpus data reveals that Spanish resorts to a diverse set of lexical means, or evidentiality strategies in the sense of Aikhenvald 2004, to express meanings that are reduced to the use of omen in Basque (see Section 2.2). In this paper I adopt the term evidentiality strategy (henceforth ES for short). I describe the use of omen as grammatical evidentiality for several cogent reasons: omen is a particle (a member of a closed set), it is dedicated to the expression of hearsay, and it is a proclitic. In effect, omen may be considered part of verbal inflection. […] particles ‘count’ as inflection: they are the only elements which may stand between the participle and its auxiliary in positive analytic verbal forms, and they are also the only elements that may ‘intervene’ between the focalized element and a synthetic verb. Phonetically, they also form an accentual group with the tensed verb. (Hualde et al. 1994: 317)

This is not to say that Basque will not make use of lexical means to express information source (pace Jendraschek 2003). However, lexical strategies complement Basque hearsay evidence, and they respond to syntactic and semantic considerations (see Section 3). This paper is dedicated to the hearsay particle omen, for several practical reasons. First and foremost, omen has been accepted as a hearsay evidential, even if in passing, by numerous scholars (de Rijk 1972, Trask 1981, Jacobsen 1986, Ortiz de Urbina 1989, Cinque 1999, Rebuschi 1997, Jendraschek 2003, Aikhenvald 2004). From this view, Basque would only mark hearsay, and everything else would be unmarked. The corpus data provides confirmatory evidence that omen is a hearsay evidential, although it only expresses sentential scope (see Section 2.3). Incidentally, it seems fair to say that the idea that the object of evidentiality is a proposition is a

134

Asier Alcázar

common de facto assumption that transcends linguistic framework (Cinque 1999’s Evidential Mood Projection pp. 85-6; Speas 2006; Aikhenvald 2004; Higginbotham 2009; inter alia). This is again consistent with grammatical evidentiality, as we will see that various Spanish ES, which can vary in scope (from sentences to constituents) are only rendered as omen if the scope is at the sentence level. If, in the Spanish original, the scope is not expressed at the sentence level, but rather at the constituent level, then Basque cannot resort to omen and uses various ES instead. We will witness an interesting interplay between lexical and grammatical evidentiality in Basque articulated on the basis of the scope of the evidential expression.4 A second important reason to begin the study of Basque evidentiality with hearsay evidence has to do with visibility: hearsay evidence is expressed in both languages. This is an important aspect to consider since direct correspondence between ES and Basque particles is hard to find in other areas. In the conclusion I briefly refer to further research that extends this preliminary assessment to other possible evidential particles in Basque. The outline of the paper is as follows. Section 2 introduces the parallel corpus. It then presents the corpus data for omen and discusses the conclusions that can be drawn from it. Section 3 considers an extended data set, where we first search for particular Spanish ES to understand under what circumstances these ES will be translated as omen or as a Basque ES. The conclusion is that scope is a crucial factor. Nonetheless, translators also resort to parenthetical expressions and semi-grammaticalized forms such as the verb seem to render hearsay evidence with sentential scope. We reflect on semantic considerations that may be involved in these choices. Section 4 ends the paper with the conclusions and questions for further research on Basque evidentiality.

2. Hearsay evidence in Spanish & Basque: what does omen translate? In this section we report our findings on a set of sentence pairs that contain the particle omen in the translation. That is, the data set was retrieved by searching for the particle in the Basque side of the corpus, then retrieving the original sentences in Spanish. Section 2 now opens with background information about the corpus that I have used for the study.

Information source in Spanish and Basque

135

2.1. Consumer Eroski Parallel Corpus The Consumer Eroski magazine is a monthly paper publication in Spain that is dedicated to consumer advocacy. Every month it features an investigative report. The December 2008 issue, for example, features an investigation on 010 phone numbers, which provide citizens with practical information such as bus schedules. The headline warns that 1 in 3 cities with this service offers insufficient information. Because of its thorough investigative reporting, the magazine enjoys a certain level of prestige in Spain. Of interest to us is the fact that the magazine also exists as an electronic publication since 1998 (revista.consumer.es). Soon after the magazine began to exist online, it started to be translated to three other languages spoken in Spain: Basque, Catalan and Galician. In the course of approximately two years, every monthly issue was simultaneously published in the four languages and it continues to do so. The Consumer Eroski Parallel Corpus, from which I extracted the sentence pairs, is the magazine from 1998 to 2006 turned into a collection of Spanish sentences aligned with their Basque, Catalan and Galician translations (a personal project I undertook as a graduate student). The corpus is accessible online via the University of Vigo (http://sli.uvigo.es/CLUVI/index_en.html). For example, here is one of the results returned by the search interface if one searches for aeropuerto ‘airport’ in the Spanish section and wishes to recover the translations to the other languages (2). Since the example is intended for illustration, I spare the glosses (2a: Spanish; 2b: Basque; 2c: Catalan; 2d: Galician). (2) a.

b.

c.

Me disponía a viajar a Caracas desde Madrid y, pese a estar en el aeropuerto con tres horas de antelación, me vi afectado por el overbooking. ‘I meant to travel from Caracas to Madrid and, even though I was in the airport three hours in advance, I was a victim of over booking.’ Madrildik Caracaserako hegazkina hartzeko aireportura hiru ordu lehenagotik joan behar izan nuen baina, halaz ere, zorioneko overbooking horrek harrapatu ninduen. Em disposava a viatjar a Caracas des de Madrid i, tot i arribar a l’aeroport amb tres hores d’antelació, em vaig veure afectat per la sobrereserva.

136 d.

Asier Alcázar

Dispúñame a viaxar a Caracas desde Madrid e, aínda que estaba no aeroporto con tres horas de antelación, vinme afectado polo overbooking.

The Consumer Eroski Parallel Corpus fills an important gap in the resources available to computational linguistic research on Basque (e.g., see Diaz de Ilarraza et al. 2008). With the advent of statistical methods for natural language processing, Basque is at a disadvantage in that few monolingual text collections exist, much less bilingual. Consider, for example, the accessibility of monolingual Spanish corpora, with hundreds of millions of words (CREA & CORDE, Royal Academy of the Spanish Language: www.rae.es, or Corpus del español: www.corpusdelespanol.org, Mark Davies), which have served as resources for evidentialists (Olbertz 2007, Cornillie 2007). The Basque section of the corpus nears 1 million words. Still, it is the largest of its kind for the Spanish-Basque pairs, where it is used for developing and testing applications or algorithms that tackle general purpose language (as opposed to domain specific ones, such as legal proceedings www.deli.deusto.es/AboutUs/Projects/Legebiduna, for which parallel texts existed). It is also quite attractive in size for research on Galician and Catalan (Xavier Gómez Guinovart, Felipe Sánchez Martínez p.c.), particularly considering that it is freely available.

2.2. Corpus data I have carefully inspected 173 sentence pairs with omen: Spanish original and Basque translation. There was a handful more that the alignment software did not group into pairs. Sentences were aligned with Moore’s bilingual sentence alignment system from Microsoft labs. This alignment system achieved to align 83% of Spanish-Basque sentences. The remaining 13% did not attain the necessary confidence level. I verified all the corpus data for omen manually and no alignment errors were found. Alignment at the sentence level is almost error free. The new research frontiers are phrase and word level alignment. It is worth noting that the density of omen in the corpus is approximately 173 out of 70000 sentences, or 1 in every 400 sentences. Intuitively, it feels low for such a common word. There are several possible reasons for this. One would be that the nature of the magazine being consumer reports may have to do with the lower use of this particle. Investigative reporting is assertive. On the other hand, investigative reporting does make use of re-

Information source in Spanish and Basque

137

ported information. More importantly, as anticipated, we will see that not all ES expressions meaning hearsay are translated as omen. The nature of the data being translated text, coupled with the ambiguity of the expressions that in the Spanish original may signal hearsay (see Section 3), may be ultimately responsible for this lower frequency. Table 1. ES replaced by omen Spanish ES NB: In 23% of all cases there is no ES in Spanish parece (18), parecen (4),… it seems, they seem, ... se estima it is estimated al parecer (5), aparentemente (2) apparently, apparently se calcula it is calculated se considera (4), considerarse (1), ... it is considered, to consider.refl se prevé (4), se prevén (1) it is/they are anticipated dicen (3), dice (1) they say, he/she says se atribuyen (3), se atribuye (1) they are/it is attributed to pretende (2), se pretende (1), ... plans, it is planned todavía (2), aún (1) yet supuestamente supposedly NB: other forms (mostly single occurrences of verbs in various tenses) Total

N

%

40

23.1

26

15.0

11

6.3

7

4.0

6

3.4

6

3.4

5

2.8

4

2.3

4

2.3

4

2.3

3

1.7

2

1.1

55

31.7

173

100

It is important to clarify that omen is not a rarely used particle. An exact search in the internet for omen da ‘supposedly is’ (be also functions as an intransitive auxiliary) returns over 200,000 hits in Google; omen zen ‘supposedly was’ returns over 70,000. The search omen ditu ‘supposedly has(plural object)’ (have also functions as a transitive auxiliary) returns over 100,000. Basque verbs are heavily inflected with agreement markers

138

Asier Alcázar

for person and number for the subject, direct object, indirect object argument and the addressee (for the latter, see Ohiarçabal 1993, Alcázar and Saltarelli 2009, to appear). The searches could go on and on. The point is that these simple searches suggest that the frequency of use of the particle omen in non-translated text and other registers is definitely higher than 1 in 400 sentences. Why study the Spanish and Basque evidential systems in parallel? A first argument would be that we can benefit from the comparison, in that lexical means can provide shades of meaning to generalized, grammaticalized forms. These, in turn, can help us classify or bring together seemingly disparate forms. Extra-grammatical ways of marking information source may allow more detailed specification of various degrees of assumption, inference, opinion than do grammatical evidential systems, and often reliability and speaker’s evaluation of information. (Aikhenvald 2007: 1) There are many more options in the details one may want to express through lexical means than through grammar. (p.10)

In the case of a sentence-aligned parallel corpus, we can correlate grammatical markers with sets of ES. This is what I do in Table 1. The table presents Spanish ES that are replaced by omen, sorted by frequency and with a percentage value for reference. I procured the data for this section searching for the particle in the Basque section, and subsequently retrieving the Spanish originals. We are thus looking at hearsay evidence only when it is translated as omen. Note that the nature of the data is translated text. The discussion in this paper thus concerns not only structural or morphosyntactic differences in the means to express evidentiality, but also a meta-level of evidentiality marking. The choices made by the translators reflect their understanding of the original text, and their attempt to a most appropriate translation. By cross-referencing these choices, a parallel corpus reveals the instrumentarium in each language that writers can draw from to express evidentiality.5 Next I illustrate several sentence pairs from the corpus to exemplify the most frequent types (4-12) and the least frequent ones (13-14) reflected in Table 1. Then, I discuss the implications of these preliminary findings. The examples in (3) and (4) are representative of the use of omen when, in effect, there is no ES in Spanish (n = 40). The Spanish sentence would appear to be neutral. The translator is qualifying information source at their own discretion. These examples with no ES in the original account for about a quarter of the uses of omen. In all other cases omen replaces an ES

Information source in Spanish and Basque

139

in Spanish, as we will see momentarily. We have introduced an ES in the English semi-literal translation for the Basque examples (apparently in (3’) and seems in (4’)), since the translation to Basque does make information source explicit. (3)

(3’)

(3’’)

(4)

(4’)

(4’’)

En la actualidad, toda cita que se precie in the modern.times, every event that 3SG.REFL value.3SG.SUBJ ‘Nowadays, every [contemporary] art show of a certain caliber’ con el arte contemporáneo está obligada a incluir fotografía. with the art contemporary be.3SG obligated to include photography ‘must include photography.’ Gaur egun, bere buru-a halakotzat dauka-n as.such has.3SG.3SG-REL today day its head-ABS ‘Nowadays, [any contemporary art show] of a certain caliber’ arte garaikide-ko zeinahi erakusketa-k display-ERG art contemporary-GEN any ‘any contemporary art show’ argazki-a agertu behar omen du. show need HEARSAY photography-ABS have.3SG.3SG ‘must apparently include photography.’ Proposition: [every event must include photography] Spanish: /neutral/ Basque: /omen/ Lo que sí es común es be.3SG the what yes be.3SG common ‘What is indeed common is‘ un cierto grado de rigidez en la estructura de la personalidad, a certain level of inflexibility in the structure of the personality ‘a certain level of inflexibility in the nature of their personality,’ Ba omen dago ezaugarri orokor bat: HEARSAY be.3SG characteristic general one yes ‘There seems to be a general characteristic:’ nortasuna-ren egitura zurrun samarra, structure inflexible rather personality-GEN ‘a personality of rather inflexible nature,’ Proposition: [it is common] Spanish: /neutral/ Basque: /omen/

140

Asier Alcázar

Meaningful translation from Spanish to Basque is a sort of balancing act that requires the translator to do some restructuring. Careful inspection of the examples will show that the translation is both idiomatically correct (idioms are translated) and faithful to the original, where it is syntactically feasible. Incidentally, the quality of the translation is highly regarded in the Basque Country, and the corpus constitutes a reference tool for translators and Basque journalists alike (cf. Iker Merchán, General Coordinator of Consumer Eroski magazine, p.c.). Examples like (5) and (6) represent the most frequent Spanish ES that omen replaces. These are primarily finite forms of the verb parecer ‘seem’ (n = 26). (5)

(5‘)

(5’’)

(6)

(6’)

El ahorro en fármacos parece llegar al 40% reach to.the 40% the saving in drugs seem.3SG ‘The savings in drugs seem to reach 40% ’ en los antirreumáticos, in the rheumatoid.arthritis.drugs ‘in the case of rheumatoid arthritis drugs’ Farmazi gaie-tako aurrezki-a %40-ra 40%-ADL Pharmacy good-from saving-ABS ‘The savings in pharmacy goods up to 40%’ iris-ten omen da produktu HEARSAY be.3SG product reach-IMP antirreumatiko-etan, rheumatoid.arthritis.drug-LOC ‘seem to reach in the case of rheumatoid arthritis drugs’ Proposition: [the savings reach 40%] Spanish: /parecer/ Basque: /omen/ Lo que más parece inquietar a la sociedad trouble to the society the what most seem.3SG ‘What seems to trouble society most’ respecto de este nuevo modelo de familia regarding of this new model of family ‘with respect to this new family model’ Famili eredu berri horre-tatik gizarte-a family model new that-from society-ABS ‘From that family model’ gehien asalda-tzen omen dena[sic] HEARSAY have.3SG.3SG(duena) most trouble-IMP

Information source in Spanish and Basque

(6’’)

141

‘what seems to trouble society most’ Proposition: [what troubles society most] Spanish: /parecer/ Basque: /omen/

The following two examples illustrate the equivalence between omen, on the one hand, and it is estimated (7; n = 11) and it is calculated (8; n = 5), on the other. Because of the structural differences in the translation, we have used adverbs expressing hearsay in the semi-literal translation to Basque (7’, 8’). (7)

(7’)

(7’’)

(8)

Se estima que un sistema de 4 m2 de placa 3SG.IMPER estimate.3SG that a system of 4 m2 of panel ‘It is estimated that a system (consisting of) a 4 m2 sun panel’ solar y un acumulador de la energía generada es suficiente solar and an accumulator of the energy generated be.3SG sufficient ‘and an accumulator to store the energy generated is enough’ para cubrir el consumo de agua caliente de for cover the consumption of water hot of una familia de cuatro personas. a family of four people ‘to cover the hot water needs of a family of four.’ Lau metro karratu-ko eguzki-plaka-k eta and four meter square-GEN sun-panel-ABS.PL ‘[a system composed of] A four square meter sun panel and’ sortutako energía-ren meta-tzaile-ak osatutako sistema generated energy-GEN accumulat-or-ERG.SG composed system ‘an accumulator of the energy generated...a system composed of’ nahikoa omen da lau kide-ko familia bat-en sufficient HEARSAY be.3SG four member-GEN family one-GEN ‘is apparently sufficient...of a family of four’ ur bero kontsumo-a horni-tzeko. provide-to water hot consumption-ABS ‘to cover for the hot water consumption [of a family of four]’ Proposition: [a system is enough] Spanish: /estimar/ Basque: /omen/ Se calcula que el primer año de un bebé calculate.3SG that the first year of a baby 3SG.IMPER ‘It is calculated that a baby’s first year of life

142

(8’)

(8’’)

Asier Alcázar

cuesta a los padres, sin hacer gran alarde de gastos, great show of expenses cost.3SG to the parents without do ‘costs the parents, without incurring into exaggerated expenses,’ más de 6.000 euros more of 6,000 euros ‘in excess of 6,000 euros.’ Aparteko gastu-rik egin gabe ere, without even extraordinary expense-PART do ‘Even without incurring into any sort of extraordinary expenses,’ haurra-ren lehen urtea guraso-ei 6.000 euro-tik first year parent-DAT.PL 6,000 euro-ADL child-GEN ‘the baby’s first year of life to the parents over 6,000 euros’ gora kosta-tzen omen zaie. above cost-IMP HEARSAY AUX.3PL.3SG ‘supposedly costs’ Proposition: [a baby’s first year of life costs over 6,000 euros] Spanish: /calcular/ Basque: /omen/

Next, we have less frequent examples that involve se prevé(n) ‘it is/they are anticipated’ (9; n = 5) and dice(n) que ‘(they) say that’ (10; n = 4). It is noteworthy that, despite its lower frequency, dice(n) que is the source of a (pseudo-)grammatical form in the spoken language (see Travis 2006 on Colombian Sp.; Olbertz 2007 on Mexican Sp.). The written register does make use of the verb parecer ‘seem’ as we saw earlier, but the verbal expression of hearsay evidence is quite diversified (n = 57 for the mostly singleton set). (9)

En el trópico, por ejemplo, aunque este calentamiento in the Tropics for example although this temperature.increase ‘In the Tropics, for ex., even though this temperature increase’ será leve, se prevé que moderate 3SG.IMPER anticipate.3SG that be.FUT.3SG ‘will be moderate, it is anticipated that’ habrá alteraciones importante-s en la cantidad important-PL in the quantity have.FUT.3SG changes ‘there will be important changes in the quantity’ y frecuencia de las lluvias. and frequency of the rains ‘and frequency of precipitations’

Information source in Spanish and Basque

143

Tropikoan, adibidez, berotzea arina example-INSTR temperature.increase light tropics-LOC ‘In the Tropics, for ex., [even though] this temperature increase’ izan-go den arren, aldaketa handi-ak be.3SG.REL even change big-ABS.PL be-IRR ‘even though it will be, great changes’ izan-go omen dira eurite-en kopuru eta maiztasunean. be-IRR HEARSAY be.3PL-REL rain-GEN quantity and frequency.in ‘will presumably be in the quantity & frequency of precipitation’ (9’’) Proposition: [there will be important changes] Spanish: /prever/ Basque: /omen/ (10) Por ejemplo, siempre hay una universidad desconocida for example, always there.is a university unknown ‘For example, there is always an unknown university’ que dicen que avala sus tesis. back.3SG their theses that say.3PL that ‘that is said to support their claims.’ (10’) Adibide-z, beti dago tesi horiek example-INSTR always be.3SG thesis those ‘For example, there is always those theses’ babes-ten omen ditu-en unibertsitate ezagunen bat. unknown one back-IMP HEARSAY aux-REL university ‘that supposedly backs’ ‘some unknown university’ (10’’) Proposition: [an unknown university supports their claims] Spanish: /decir/ Basque: /omen/ (9’)

As representatives of the group with mostly single occurrences, we have chosen the following two examples. In (11) omen replaces señalan ‘(they) note’ and in (12) se cree ‘it is believed’. Señalan que no existe tal capa, exist.3SG such layer note.3PL that NEG ‘They note that such layer does not exist,’ (11’) Ez omen dago geruza-rik, NEG hearsay be.3SG layer-PART ‘They note that such layer does not exist,’ (11’’) Proposition: [there is no such layer] Spanish: /señalar/

(11)

144

Asier Alcázar

Basque: /omen/ (12) Se cree que es un elemento relevante be.3SG an element relevant 3SG.IMPER believe.3SG that ‘It is believed that it is a relevant element’ en la mitad de los casos de dispepsia. in the half of the cases of dyspepsia ‘in half of dyspepsia cases.’ (12’) Dispepsia kasuen erdi-an half-LOC dyspepsia case-GEN ‘in half of dyspepsia cases’ funtsezko elementua omen da hau. HEARSAY be.3SG this essential element ‘it is believed that this is an essential element’ (12’’) Proposition: [it is a relevant element] Spanish: /creer/ Basque: /omen/

2.3. Discussion The corpus data presents two distinct strategies in the expression of hearsay evidence. On the one hand, Spanish resorts to 60+ ES, where Basque may employ just one particle. The status of omen as a grammatical means to express hearsay information seems confirmed in the empirical data in this many-to-one mapping. On the other hand, this does not mean to say that Basque can only resort to a grammatical strategy to express hearsay evidence, as we will see in Section 3. It is noteworthy that in 23% of the inspected examples (n=40) Basque translators introduce omen without an ES being present in Spanish. The result is consistent with grammatical evidentiality. One of the parameters articulating the lexical vs. grammatical distinction is that the disclosure of information source is optional in the former and obligatory in the latter (Chafe and Nichols 1986, Aikhenvald and Dixon 2003). This could explain a more frequent disclosure of information source in the translation. Following up on this idea, if we assume an absolute distinction between lexical and grammatical evidentiality, does it imply that Basque speakers must consistently and systematically disclose information source in every clause? The answer to this question is going to lead us to a delicate point in the literature on evidentiality: whether speaker attitude on the reliability of information is to be separated from information source (e.g., Chafe 1986

Information source in Spanish and Basque

145

vs. Aikhenvald 2004). Basque cares to express certainty (perhaps the core idea behind the system, source becoming prominent as an indicator of degrees of certainty; see Mithun 1986 on Northern Iroquoian). And certainty in the Basque evidential system is the unmarked value (see Faller 2002 on her analysis of unmarked sentences in Quzco Quechua as Best Possible Grounds). Here we have to take sides with Basque and insist on the relevance of certainty in its system (whether this trait should or should not be universal is beyond the scope of our contribution). By way of example, we will adduce a textbook for teaching Basque to adults (HABE, Barrene 13, 1B maila, http://www.ikasbil.net/idatziak/barrene/barrene_13ik.pdf). In contrast to descriptive grammars of Basque, which deal with evidential particles under different grammatical points that echo the European tradition (e.g., modality: Azkue 1925, Zubiri and Zubiri 2000, Jendraschek 2003), other types of text are free to explain the language less formally but more meaningfully. The textbook introduces the topic of expressing ziurtasuna or certainty, which is split in three levels: absolutely certain, certain, and somewhat certain. The level in the middle, certain, is unmarked, and is reported to be the most frequent (“even if we lack reliable data or information, we do not question the truth of what we say in any way”, p. 34; my translation). The level somewhat certain is the domain of the particles we are concerned with, such as omen when the speaker is not certain but knows that someone else is, or bide when direct evidence is not available but inference is possible. The first level absolutely certain would require explicit expressions such as “I am sure that”, “I have no doubt that”, and the like (also relative marker -en (very certain) as opposed to -ela (certain)). We concur with the textbook description on the last two levels. Sentences expressing that the speaker is certain of the information are not marked, regardless of how this certainty was attained (through the senses, by a third person, etc.). Accordingly, to say “Bogotá is the capital of Colombia” or “John is at work” does not require the particle omen, even if we learnt these facts from a teacher long ago or we learnt about John’s schedule recently and directly from him. That said, omen signals that the information is second-hand, hearsay evidence. A relevant question here is whether, and to what extent, a speaker may assimilate someone’s belief or evidence as his own (see Givón 1982). In Basque this assimilation, for lack of a better term, is possible and may spear the use of omen and other particles. But, then again, we will not be able to solve here the question of

146

Asier Alcázar

whether speaker attitude must be an integral part of the characterization of evidentiality across languages. We just note its relevance for Basque. In this regard, Jendraschek (2003: 51) warns that omen may also express doubt (la fonction de omen n’est pas purement citationnelle. Son usage implique la non-vérification et, en conséquence, un doute de la part du locuteur; referring to Zubiri and Zubiri 2000: 580). We agree with Jendraschek on this. In our corpus we have two examples with a first person subject, which we reproduce here (13, 14). We take it that the use of first person cancels hearsay. Although we feel like the translator employed omen as a mitigator in these examples in view of the use of the emphatic pronoun neu(k) in both. Creo que yo seguiré utilizando el soporte químico, think.1SG that I continue.1SG.FUT using the support chemical ‘I think I will continue using chemical support,’ aunque mi postproducción es digital. although my postproduction be.3SG digital ‘although my postproduction is digital.’ (13’) Ni-k neu-k euskailu kimikoz baliatuz support chemical use-INSTR I-ERG I.EMP-ERG ‘I myself’ ‘using chemical support’ segitu-ko omen dut, have.1SG.3SG continue-IRR HEARSAY ‘will continue’ ene postprodukzioa digitala izanagatik. my postproduction digital in.spite.of.being ‘in spite of my postproduction being digital.’ (13’’) Proposition: [I will continue] Spanish: /creer/ Basque: /omen/ (14) Me parece que sólo consigo enseñar desde ese me seems that only manage.1SGteach from that ‘It seems to me that I only suceed in teaching from that‘ punto de vista cuando yo misma trato de aprender y de point of view when I myself try.1SG of learn and of ‘point of view when I myself try to learn and’ adaptarme a mis alumnos. to my students adapt.REFL.1SG ‘adapt myself to my students.’ (14‘) Ikuspegi horretatik, neu ikas-ten saia-tzen (13)

Information source in Spanish and Basque

147

try-IMP perspective that-from I.EMP learn-NOM ‘From that perspective, [it is only] when I myself try to learn’ naizenean, ikasle-ei egokitzen ahalegin-tzen make.an.effort-IMP aux.1SG.WHEN student-DAT.PL adapt-NOM ‘only when I make an effort to adapt to my students’ naizenean bakarrik irakas-ten omen dut. AUX.1SG.WHEN only teach-IMP HEARSAY have.1SG.3SG ‘[that] I teach’ (14’’) Proposition: [I suceed in teaching] Spanish: /parecer/ Basque: /omen/ It has been noted that first person can signal a scenario describing a dreamlike experience where a first person does not recall some event they participated in (Aikhenvald 2004). The corpus data does not provide a similar example. But it is possible to use omen in that capacity. It can also be employed as a disclaimer (an epistemic extension, following Aikhenvald 2004), although this use is not attested in the corpus either. Incidentally, according to various traditional sources, ei is an equivalent form of omen in Vizcayan Basque (e.g., Zubiri and Zubiri 2000). Thanks to Ikuska Ansola and Karlos Arregi (p.c.), we have been able to verify that ei (in Ondarroa Basque) is equivalent to omen with one exception: the disclaimer interpretation is not possible for ei. This is compatible with the idea that the disclaimer interpretation is an epistemic extension that may arise or not. It has for omen. Finally, going back to table 1, the reader may have noticed that the majority of ES are verbs, and wonder whether this is indeed the case more generally. Table 2 can answer this question, which displays the information relating to part of speech for the 133 instances of ES. In effect, verbs account for about 9 in 10 cases. Table 2. ES by part of speech and their frequency ES: Part of Speech Verb Adverb Prepositional phrase Total

N 118 9 6 133

% 88.7 6.8 4.5 100

Does this data mean that Spanish primarily uses verbs to express hearsay evidence? Not necessarily, since the data set was generated by searching in

148

Asier Alcázar

the Basque section of the corpus. To begin with, it would be necessary to perform additional searches in the Spanish section to establish whether the adverbs and prepositional phrases are used in contexts where omen cannot. As anticipated, we will see in Section 3 that ES based on adverbs and prepositional phrases are not translated as omen if their scope is not sentential.

3. Understanding the role of lexical evidentiality in Basque In this section we will show that Basque translators also employ ES to express hearsay evidence. It is our interest to understand under what circumstances Basque resorts to lexical evidentiality. There are three contexts where I can anticipate Basque translators to deploy an ES. One has to do with syntactic scope, another with semantic considerations, and a third seems to be a combination of the two. I examine each of these factors in turn.

3.1. Sentential scope vs. constituent scope When the scope of the evidential expression is not sentential, translators never resort to omen. This can be shown by inspecting evidential expressions that may vary in scope, such as the adverbs aparentemente ‘apparently’ and supuestamente ‘supposedly’. We begin with the latter. There are 6 occurrences of supuestamente in the corpus. In two of them the adverb has sentential scope, since it modifies the verb, and it is translated as omen (15; n=2). What the other four cases have in common is that the adverb modifies some other constituent. Then translators resort to ES (16; e.g., ustez (n=2), an adverb, or balizko (n=1), an adjective meaning supposed; and once it is not translated (n=1)). (15)

En mayo de 1998 una entidad bancaria incluyó in May of 1998 an entity bank include.3SG.PAST ‘In May 1998 a bank listed one of its clients‘ a un cliente en un fichero de morosos; to a client in a file of delinquent.client ‘in a file of delinquent accounts’ supuestamente debía al banco 821 euros. to.the bank 821 euros supposedly owe.3SG.PAST.IMP ‘(s)he supposedly owed the bank 821 euros.’

Information source in Spanish and Basque

149

Banku erakunde bat-ek 1998-ko maiatzean 1998-GEN May-LOC bank entity one-ERG ‘in May 1998 a bank’ bezeroetako bat berankor-ren fitxategi-an sartu put client.of one delinquent-GEN file-LOC ‘listed a client in a file of delinquent accounts’ zuen, banku-ari 821 euro AUX.3SG.3SG bank-DAT 821 euro ‘to the bank 821 euros’ zor omen zizkion eta. HEARSAY AUX.3SG.3PL because owe ‘because (s)he supposedly owed’ (15’’) Proposition: [she owed the bank 821 euros] Spanish: /supuestamente/ Basque: /omen/ (16) Por este motivo, es muy importante que si encontramos for this reason, be.3SG very important that if find.1PL ‘For this reason, it is very important that if we find’ un cadáver o cebo supuestamente envenenado, se a corpse or bait supposedly poisoned 3SG.REFL ‘or a supposedly poisoned bait,’ apliquen unas líneas de actuación que eviten defectos de forma apply.3PL some lines of action that avoid.3PL.SUBJ defects of form ‘for a protocol to be implemented to avoid these defects.’ (16’) Horregatik, oso garrantzitsua da animalia because.of.that very important be.3SG animal ‘Because of that, it is very important bat-en gorpua edo amu ustez pozoitu bat supposedly poisoned one one-GEN body or bait ‘an animal corpse or supposedly poisoned bait’ aurki-tzen ba-dugu, formazko ez izateko if-AUX.1PL.3SG form NEG be find-IMP ‘if [we] find, ...it does not conform to specifications.’ moduan joka-tzea komeni da. be.3SG manner respond-NOM convenient ‘it is appropriate to respond like [it does not conform to specs.]’ (16’’) Constituent: [poisoned] Spanish: /supuestamente/ Basque: /ustez/ (15‘)

150

Asier Alcázar

On the other hand, aparentemente has 27 occurrences. Table 3 presents a frequency breakdown of the translations of this adverb to Basque. Table 3. Aparentemente ‘apparently’ as an adjective or verb modifier ES aparentemente apparently

aparentemente apparently

Total

Modifies Verb

Adjective

Translation omen itxura batean in appearance itxuraz in appearance seem: badirudi (2), iruditu (1), diruditen (1) itxuraz behintzat at least in appearance without translation berez in itself ustez in appearance itxura batean behintzat at least in appearance

N

%

2

7.4

8

29.7

5

18.5

4

14.8

3

11.1

2

7.4

1

3.7

1

3.7

1

3.7

27

100

We know from Table 1 that in two instances aparentemente is translated as omen (n = 2). In these two cases the adverb modifies the verb directly, thus expresses sentential scope. In the reminder of cases the adverb modifies an adjective. Then the translators resort to various ES, including syntactic restructuring of the constituent as a finite clause with the verb iruditu ‘seem’ (n = 4). For reasons of space, we will spare further examples in this subsection. It seems that these two adverbs in the corpus are more frequently used to express constituent scope, rather than sentential scope. The reason for this may be that verbs expressing hearsay evidence are fixed in scope, so to speak, while adverbs may modify different linguistic elements. For this reason it could be hypothesized that adverbs, together with prepositional phrases and adjectives, are the workhorse of constituent scope in Spanish.

Information source in Spanish and Basque

151

3.2. Ambiguity in scope: parentheticals A relatively frequent ES in Spanish is al parecer (Table 1, n = 5), a prepositional alternative to aparentemente ‘apparently’ (Table 1, n = 2). In this subsection I report the results of searching for this ES in the Spanish section, then retrieving the Basque originals, regardless of whether omen was used for the translation. This prepositional phrase occurs 27 times in the corpus. Overall, it is translated as various parenthetical expressions of equivalent meaning (n = 20; antza den-ez [appearance be.3SG-INSTR], dirudien-ez [seem.3SG-INSTR], etc.), as omen (n = 5), or it is dropped in the translation (n = 2). The scope of the prepositional phrase is difficult to determine. It is either sentential or it expands over several verbs, and it is perhaps feasible to interpret it as a constituent modifier in some cases. I have observed that the only cases where translators resort to omen, the prepositional phrase directly precedes or follows the verb (half of the cases where the modifier is adjacent to the verb: n = 10). In all other cases the prepositional phrase is placed somewhere else in the sentence (n = 17) and omen is not used. My interpretation is that, because omen is restricted in its scope to the sentence where it occurs, it is less desirable as a translation equivalent to parentheticals, which are potentially ambiguous in scope. By choosing a parenthetical, the translators preserve the potential ambiguity of the Spanish original. This reflection leads me to a third possible factor: semantic ambiguity in the original ES.

3.3. Ambiguity in the original ES: is it hearsay or inference, doubt…? The verb parecer ‘seem’ has close to 200 occurrences in the third person. Out of these, the verb is translated as omen in about 13% of cases. Why translators would not resort to omen is clear in many cases, namely when the semantics of the verb does not express hearsay evidence but rather ‘looks like’ or ‘personal opinion’ (we saw an example of personal opinion before in (14), which was translated as omen, but this is a special mitigator case in my opinion). For the semantics of seem as a main verb, the translators understandably resort to an equivalent verb in Basque: iruditu ‘seem’. This verb is also called upon in the translation when perhaps omen could be potentially used. But we need to bear in mind that, while omen may be restricted to hearsay, parecer ‘seem’ is not (Bert Cornillie, p.c.), as it may be employed to express inference or doubt (see Cornillie 2007). In Basque inference and doubt, which have various shades of certainty, may be ren-

152

Asier Alcázar

dered by particles such as bide, edo and ote. For this reason, I assume that the translators often exercise caution in dealing with such evidential expressions in Spanish, and decide not to restrict their interpretation in the translation. Consequently, if my interpretation of the data is right, the nature of our data (parallel text) imposes considerable restrictions on the use of grammatical expressions of evidentiality that, if used, narrow down the semantics of Spanish ES to non-ambiguous evidential expressions.

4. Conclusion I have shown that omen is a grammatical form to express evidentiality in Basque, thus providing empirical confirmation to previous claims that Basque contains at least one evidential marker. The arguments to identify omen as a grammatical form are numerous: (i) it has no other meaning; (ii) it belongs in a closed set (the particles); (iii) like other clitic particles, it may be better seen as part of verbal inflection; (iv) it is restricted to sentential scope (but see n.5); (v) it may be generally used as a disclaimer, and with first person subjects it may project doubt (both potential epistemic extensions); (vi) it is introduced in contexts where the Spanish original has no ES. Perhaps translation practice is the clinching argument to identify omen as a grammatical strategy. Even though translators may resort to the same type of ES in Basque, (vii) they decide to use omen in place of over 60 different Spanish ES. Instead, the Basque ES are called upon in contexts where scope is at the constituent level, or where omen may narrow down scope ambiguity and/or semantic ambiguity in the original. Because of this, I view lexical evidentiality as playing an important role in Basque hearsay evidence (pace Jendraschek 2003), as it enables constituent scope and affords ambiguity. On the Spanish side, we have gathered a diverse collection of ES mostly consisting of verbs in the third person. This we take to show that the written register is far from a grammaticalization process that seems already underway in several varieties of Latin American Spanish with verbs of saying (dizque, quesque). Although our data is limited, we have also noted a preference to use adverbs for constituent scope, on the logical assumption that not all hearsay expressions are able to have various scope possibilities. The promising results on the Basque side of the corpus invite further investigation into other particles that may be related to evidential meaning. The suspicion has been there for quite some time. For example, Jacobsen points to all of the particles mentioned in this paper in a footnote (1986: 7, fn. 3).

Information source in Spanish and Basque

153

I assume that the reason why the particles remain in the shadow has to do with the ability to render their meaning explicitly in the languages geographically adjacent to Basque: Spanish and French (although I myself do not speak French; it is a bona fide assumption). By way of example, consider edo. Said particle is an inferential evidential equivalent to epistemic must (e.g., “my keys must be here”; with Azkue 1925: 470, §689) or probably (a weaker form of inferential) if it modifies a constituent (with Azkue 1925: 252, §409). I find that edo is capable of constituent scope, where its meaning shades into dubitative, similarly to Latin American dizque and quesque. Edo is also equivalent to approximately or about/around in expressions related to time and distance. I pursue an analysis of this particle in relation to the standard assumption that grammatical evidentiality must be sentential (Alcázar 2009b), and sketch a diachronic analysis of the evolution of edo from a disjunctive conjunction to an inferential in a two step subjectification process in Alcázar (2009a). Other expressions are more cryptic. Ote seems a mirative evidential in root questions and indirect questions, and a weak inferential in statements (see DeLancey 1997). As a mirative ote may show a strong connection to nervousness or anxiety (also noted by Zubiri and Zubiri 2000), but not so as an inferential, a use attested in the internet, which I have not seen described in grammars. Unfortunately, ote is simply regarded as a rhetorical question marker (e.g., Hualde and Ortiz de Urbina 2003). I am not aware of such clear-cut emotion-mirativity connection in other languages. In the corpus we have found no Spanish ES equivalent to ote. Another candidate is ohi, a likely syncretism between habitual aspect and general knowledge (see Alcázar 2009c). While ohi may express habituality, it is used with verb forms that cannot express this aspect (the trinko class, Zubiri and Zubiri 2000; Alcázar 2003). ohi co-occurs with be-support 25% of the time in our corpus. Taken together, the sentence would be regarded as highly reliable. Separately, either ohi or be-support seems to convey a strong form of certainty. If indeed ote and ohi express other evidential notions, they are notions that have been less richly documented than hearsay or sensory evidence in the evidential literature (Willet 1988, Aikhenvald 2004) and, as a result, harder to pin down. For now, these meanings are lost in translation. Yet over time they may be better understood and incorporated into the Basque evidential system.

154

Asier Alcázar

Notes 1.

2. 3.

4.

5.

I would like to thank the organizers, Gabriele Diewald and Elena Smirnova, an anonymous reviewer, and the participants at the conference, especially Bert Cornillie and Björn Wiemer for comments and suggestions. Prior to the conference, colleagues and informants helped me shape the presentation at DGfS 2008 through discussion of ideas and data: Rosaura Alcázar, Ikuska Ansola, Karlos Arregi, Vicki Carstens, Asunción Estela, Roberto Mayoral Hernández. I also appreciate the early feedback received from Alexandra Aikhenvald and Gerd Jendrascheck and their encouragement to pursue this line of research. None of the persons mentioned necessarily agrees with the ideas defended here. I am responsible for all possible inaccuracies or errors. For a different view on the Basque evidential system, see Jendraschek (2003). Key to the glosses: ADL adlative, ABS absolutive, EMP emphasis, ERG ergative, FUT future, GEN genitive, HARSAY hearsay evidential, IMP imperective, IMPER impersonal, IRR irrealis, LOC locative, PAST past, PL plural, PRES present, SUBJ subjunctive, REFL reflexive, REL relative marker, SG singular. Omen can also be used as a constituent modifier in the absence of a finite verb (e.g., see Hualde and Ortiz de Urbina 2003: 318). The inferential particle edo can also be used as a constituent modifier with or without a finite verb (Alcázar 2009a). These facts, coupled with constituent modification of other likely evidential particles in Basque, beg the question of whether the object of evidentiality must indeed be a proposition (see Blain and Déchaine 2007 on Cree dialects; Alcázar 2009b). I would like to thank an anonymous reviewer for clarification on this point.

References Aikhenvald, Alexandra Y. 2007 Information source and evidentiality: what can we conclude? In Squartini, Mario (ed.), 209-227. 2004 Evidentiality. Oxford: Oxford University Press. Aikhenvald, Alexandra Y. and R. M. W. Dixon (eds.) 2003 Studies in Evidentiality. Amsterdam/Philadelphia: John Benjamins. Alcázar, Asier 2009a Subjectification in Basque evidential particles. Cambridge Ocassional Papers in Linguistics. 2009b On the scope of evidential markers. Ms. University of Missouri, Columbia.

Information source in Spanish and Basque 2009c

155

The truncation/expansion of evidential meaning in Quechua and Basque: The case of general knowledge. Ms. University of Missouri, Columbia. 2005 Consumer Eroski Parallel Corpus. http://sli.uvigo.es/CLUVI/ 2003 Two Paradoxes in the Interpretation of Imperfective Aspect and the Progressive. Journal of Cognitive Science 4 (1): 79-105. Alcázar, Asier and Mario Saltarelli 2009 Why imperative sentences cannot be embedded. Selected Proceedings of Chronos VIII. Azkue, Ressurreción María de 1925 Morfología Vasca. Bilbao: La Gran Enciclopedia Vasca. Blain, Eleanor and Rose-Marie Déchaine 2007 Evidential types: Evidence from Cree dialects. International Journal of American Linguistics 73: 257-291. Chafe, Wallace L. and Johanna Nichols (eds.) 1986 Evidentiality: The Linguistic Coding of Epistemology. Norwood, NJ: Ablex. Cinque, Guglielmo 1999 Adverbs and functional heads: a cross-linguistic perspective. Oxford: Oxford University Press. Cornillie, Bert 2007 The continuum between lexical and grammatical evidentiality: a functional analysis of Spanish parecer. In Squartini, Mario (ed.), 109-128. DeLancey, Scott 1997 Mirativity: The grammatical marking of unexpected information. Linguistic Typology 1 (1): 33-52. Díaz de Ilarraza A., G. Labaka, and K. Sarasola. 2008 Statistical Post-Editing: A Valuable Method in Domain Adaptation of RBMT Systems. MATMT2008 workshop: Mixing Approaches to Machine Translation. Givón, Talmy 1982 Evidentiality and Epistemic Space. Studies in Language 6: 23-49. Higginbotham, James T. 2009 Evidentials: Some Preliminary Distinctions. In Compositionality, Context, and Semantic Values. Rob Stainton and Chris Viger (eds.), 221-235. Berlin: Springer Verlag. Hualde, José Ignacio and Jon Ortiz de Urbina 2003 A Grammar of Basque. Berlin/New York: Mouton de Gruyter. Jacobsen, William H. 1986 The Heterogeneity of Evidentials in Makah. In Chafe and Nichols (eds.), 3-28. Jendraschek, Gerd 2003 La modalité épistémique en basque. München: LINCOM Europa

156

Asier Alcázar

Mithun, Marianne 1986 Evidential Diachrony in Northern Iroquoian. In Chafe and Nichols (eds.), 89-112. Oyharçabal, Bernard 1993 Verb Agreement with Nonarguments: On Allocutive Agreement. In Generative Stuides in Basque Linguistics, J. I. Hualde and J. Ortiz de Urbina (eds.), 89-114. Amsterdam/Philadelphia: John Benjamins. Olbertz, Hella 2007 Dizque in Mexican Spanish: the subjectification of reportative meaning. In Squartini, Mario (ed.), 151-172. Ortiz de Urbina, Jon. 1989 Parameters in the Grammar of Basque. Dordrecht: Foris. Pietrandrea, Paola 2007 The grammatical nature of some epistemic-evidential adverbs in spoken Italian. In Squartini, Mario (ed.), 39-64. Rebuschi, Georges (ed.) 1997 Essais de linguistique basque. Bilbao: Université du Pays Basque and Saint-Sébastien: Diputación Foral de Gipuzkoa (Supplements of Anuario del Seminario de Filología Vasca “Julio de Urquijo”, XXXV). 1997a Énoncés et formes hypothétiques en basque contemporain. In Rebuschi, Georges (ed.), 139-154. 1997b Temps, mode et aspect dans les indications scéniques. In Rebuschi, Georges (ed.), 155-160. De Rijk, Rudolf P.G. 1972 Partitive assignment in Basque. Anuario del Seminario de Filología Vasca ‘Julio de Urquijo’ 6: 130-73. Squartini, Mario (ed.) 2007 Evidentiality between lexicon and grammar. Rivista di Linguistica 19 (2). 2007a Investigating a grammatical category and its lexical correlates. In Squartini, Mario (ed.), 1-6 Trask Robert L. 1981 Basque verbal morphology. Euskalarien nazioarteko jardunaldiak, 285-304. Travis, Catherine 2006 Dizque: a Colombian evidentiality strategy. Linguistics 44: 12691297. Willet, Thomas 1988 A cross-linguistic survey of the grammaticization of evidentiality. Studies in Language 12: 51-97. Zubiri, Ilari and Entzi Zubiri 2000 Euskal Gramatika Osoa. Bilbao: Didaktiker.

Embedded evidentials in German Mathias Schenner This paper addresses two widespread preconceptions: First, there are no (“real”) evidentials in German. Second, evidentials cannot be embedded. These issues are taken up in turn. The first part introduces a semantically narrow but morphologically broad notion of linguistic evidentiality that renders the set of evidentials in German non-empty, in accordance with recent studies like Diewald and Smirnova (forth.). The second part focuses on embedded reportative modals in German and demonstrates some of the semantic complexities that embedded evidentials can give rise to.

1. Introduction The past few decades have witnessed a remarkable increase of research on evidentiality. However, most studies focus on evidentials in one particular environment, namely declarative root clauses. Of course, it’s a perfectly reasonable strategy to start investigations of evidentials by studying their behavior in maximally innocent and unmarked environments like declarative root clauses. But one should be aware of the fact that it’s impossible to draw final conclusions about the semantics of evidentials unless one considers their behavior in other clause types and in complex sentences. By limiting one’s attention to declarative root clauses, one might be tempted to characterize evidentials as linguistic markers that indicate the type of evidence the speaker has for the proposition expressed by her utterance. Despite the fact that this characterization is actually rather widespread in the literature, it drastically fails for many uses of evidentials. For instance, evidentials in interrogative root clauses obviously do not indicate the speaker’s type of evidence for “the proposition expressed” (however construed), nor the addressee’s type of evidence for some proposition (as is informally suggested sometimes), but rather (in most cases) the type of evidence the speaker expects the addressee to have for the true answer. Another counterexample are evidentials in embedded clauses that may not indicate the speaker’s type of evidence for some proposition, but rather that of the matrix subject (e.g. in Tibetan, cf. Garrett 2001, ch. 5).

158

Mathias Schenner

This paper will focus on the semantics of evidentials in complement clauses in German. Since many linguists seem to take it for granted that there are no (“real”) evidentials in German, we start by laying out our conceptual and terminological background assumptions in Section 2, including a well-motivated definition of evidentality that renders the set of evidential expression in German clearly non-empty (in accordance with recent studies like Diewald and Smirnova forth.). Section 3 discusses evidentials in German, focusing on embedded occurrences of the reportative modal sollen.

2. Towards a definition 2.1. Evidentiality vs. epistemic modality It is still a matter of debate whether evidentiality is a linguistic category in its own right. Some researchers argue that evidentiality is a special kind of epistemic modality. For example, Willet (1988: 52) states that “evidential distinctions are part of the marking of epistemic modality”. There are two widespread definitions of expressions of epistemic modality. According to the first, they indicate the degree of commitment of the speaker to the embedded proposition. According to the second, they mark the necessity or possibility of the embedded proposition (the prejacent) relative to some body of evidence or knowledge (von Fintel and Gillies 2007). Using the first conception of epistemic modality, Palmer (1986: 54) argues that evidentials are epistemic modals, because their whole purpose is to provide an indication of the degree of commitment of the speaker: he offers a piece of information, but qualifies its validity for him in terms of the type of evidence he has.

On the other hand, de Haan 2001 and Aikhenvald 2004 argue at length that evidentiality and epistemic modality are (in principle) independent categories, because in a number of languages, evidentials don’t carry any epistemic overtones. Their slogan is: Evidentiality is a category in its own right, and not a subcategory of any modality. (Aikhenvald 2004: 7)

These two points of view correspond to two competing construals of the notion of evidentiality. The narrow one, given in (1a), is defended by de Haan 2001, Lazard 2001, Faller 2002, Aikhenvald 2004 and others. The



Embedded evidentials in German

159

wide one, given in (1b), is adopted by Palmer 1986, Ifantidou 2001 and Rooryck 2001 among others. (1) Two popular conceptions of evidentials: a. The narrow conception: Evidentials are linguistic markers that indicate the speaker’s type of evidence (or source of information) for her claim. b. The wide conception: Evidentials are linguistic markers that indicate the speaker’s type of evidence for her claim and/or the degree of its reliability, probability or certainty. Quite some ink has been spilled discussing which of these construals is to be preferred. But a closer look at the kinds of arguments offered raises the suspicion that there’s no consensus about what exactly is at issue. Part of the confusion is that the question “What is the relation between evidentiality and (epistemic) modality?”, as it’s usually stated, is highly ambiguous, because it leaves implicit what exactly is meant by the terms ‘evidentiality’ and ‘modality’: Concepts? Subsystems of Universal Grammar? Subsystems of the grammar of a single language? This lack of precision invites misunderstandings and often leads to unnecessary debate at an abstract level. Once we replace the original question by at least the following three more precise ones, we get a surprisingly clear picture, and most disagreements automatically disappear. (2) a. Conceptual relation: What is the relation between the concept of evidentiality and the concept of epistemic modality? b. Encoding relation: What is the relation between the encoding of evidential concepts and the encoding of modal concepts in language L? c. Formal semantic relation: Can evidentials in language L be analyzed in parallel to epistemic modals in formal semantics (involving quantification over possible worlds) or do evidentials require a different kind of analysis? A satisfactory answer to the first question (2a) (understood as a psychological, not as a philosophical question) requires a model of the human mind that integrates and explains the interactions between a person’s (type of) source of a piece of information and various attitudes towards that piece of information. This question, though important, is seldom explicitly addressed in research on evidentiality, noteworthy exceptions being Chafe 1986, Willet (1988: 85-89) and Nuyts 2001. Among the main components

160

Mathias Schenner

of such a model are the individual’s specific source (source token), e.g. a certain event in the past in which the individual has been told the relevant piece of information. Source tokens can be classified in various ways at various levels of abstraction. Some common high-level classifications are based on the distinctions between direct and indirect types of source, and between reported and inferred information. Two notions that are sometimes not properly distinguished in this context are (subjective) probability and reliability. Whereas subjective probability relates an individual and some piece of information, reliability relates an individual and a source type or token, as illustrated in Figure 1. They are related, since (ceteris paribus) a piece of information will have a high degree of subjective probability, if it stems from a highly reliable source token (both as judged by the individual). Various types of background knowledge are at work that complicate the picture, but we can safely claim that the reliability of the source token and the source type as judged by the individual, are among the main components that influence the individual’s judgment regarding the necessity or possibility of a proposition expressing the relevant piece of information and, if uttered, the degree of the person’s commitment to the proposition expressed.

Figure 1. Accessing information: Conceptual relations

Thus, without going into details, it is clear that at the conceptual level, evidentiality (understood as a person’s type of source of information) and epistemic modality (understood as the person’s judgment regarding the necessity or possibility of a proposition expressing the relevant piece of information) are different but related: The type of source influences the



Embedded evidentials in German

161

strength of belief in the truth of the proposition, but the two concepts are not interchangeable because other contextual factors intervene. If a person considers a proposition necessary, this by itself does not allow any conclusion as to the type of evidence the person has. Similarly, the fact that a person has reportative evidence for the truth of a proposition does not determine a certain strength of belief in this proposition. The second question (2b) demands linguistic investigation. The relativization to a specific language is essential, because cross-linguistic research has shown that languages differ in this respect. On the one hand, there are languages that possess clearly distinct systems for evidentiality and epistemic modality. For example, Pawnee, Wintu and Makah have distinct sets of morphemes for coding modal and evidential meanings (cf. Mithun 1999, Stenzel 2004: 338). On the other hand, there are languages that use a single set of markers to denote evidential source and epistemic strength, e.g. Haanis (Mithun 1999). In general, epistemic modals often acquire evidential meaning extensions (e.g. English must, cf. von Fintel and Gillies 2007) and evidentials may have epistemic extensions (e.g. Cuzco Quechua -chá, cf. Faller 2002: 171-189). Sometimes it’s almost impossible to determine whether an expression is primarily evidential or primarily modal (see Aikhenvald 2004: 147-151 for some borderline cases). This suggests that evidentials and epistemic modals form the end points of a continuum allowing expressions at intermediate positions to simultaneously encode evidential and epistemic meaning components. In the formal semantics literature the question concerning the relation between evidentiality and epistemic modality is almost always understood as in (2c): Are the same formal tools adequate for analyzing both epistemic modals and evidentials? The influential formal analysis of epistemic modals involving quantification over possible worlds by Kratzer 1991 has first been used by Izvorski 1997 to analyze evidentials in Bulgarian, inspiring similar analyses of evidentials in other languages, e.g. Garrett 2001 on Tibetan, McCready and Ogata 2007 on Japanese, Chung 2005 on Korean, Matthewson et al. 2007 on St’át’imcets and Sauerland and Schenner 2007 on Bulgarian. However, Faller 2006 argued convincingly that evidentials are not a homogeneous class cross-linguistically and showed that evidentials in Cuzco Quechua cannot be analyzed as epistemic modals. To sum up, evidentiality and epistemic modality are conceptually distinct (though related) and encoded independently in the grammars of some (but not all) languages. The linguistic question whether the encoding of evidentiality is part of the encoding of epistemic modality cannot be ans-

162

Mathias Schenner

wered without reference to the grammar of a specific language since there is substantial cross-linguistic variation. 2.2. Defining ‘evidential’ The previous section established that evidential and epistemic categories may be encoded independently (though they need not). Thus it makes sense to adopt some version of the narrow concept of evidentials in (1a). Let’s now try to make this more precise. The basic plan is as follows. We first define the conceptual notion of an ‘evidential condition’. Then we define the linguistic notion of an evidential via conditions on the mapping of natural language expressions to this concept. (3) An evidential condition (EC) is a proposition of the form ‘EVID(s, x, p, i)’ that is true iff x has in s evidence of type i that p is true, where s is a situation, x is an agent, p is a proposition, and i is a type of evidence. This definition incorporates several non-trivial decisions (see Schenner 2009 for a more extensive discussion). First, note that an evidential condition is simply a proposition that can be true or false; it’s not a speech act. Second, the definition is based on the assumption that the kinds of things we can have evidence for are propositions, or rather the truth of propositions, in accordance with most of the recent formal semantic literature on evidentials. This might turn out to be too restrictive, though (cf. e.g. Hengeveld 2006, who argues that there are clear differences between reportatives and (other) evidentials that are due to the fact that the former operate at a higher, interpersonal level, well above the level of propositional content). Third, the definition in (3) makes use of a variable i that ranges over types of evidence. Let’s assume that i is taken from a set I that contains all available types of information sources. Now what exactly is in this set? There are at least two ways to go. One way would be to relativize the set I to particular languages or language types. However, this would seriously undermine our attempt to define the notion of an evidential condition in a language-independent way. Thus we will assume at this point that I is a stable set of universally available types of information sources. Since it has to reflect some rather fine-grained distinctions that are not encoded grammatically in every language, we expect that linguistic evidentiality markers



Embedded evidentials in German

163

are typically associated with bundles of elements of I rather than individual members.1 We can now proceed to defining the linguistic property of having an evidential component and the linguistic notion of an evidential in (4): (4) a. A lexical item L has an evidential component iff one of the functions of L is to introduce an evidential condition with background status. b. A lexical item L is an evidential iff one of the main functions of L is to introduce an evidential condition with background status. Although this is quite a liberal definition of ‘evidential’, it still sides with the narrow concept of evidentiality mentioned in (1a) that excludes purely epistemic expressions. An important feature of (4) is that evidentials are not required to be grammatical elements (or grammaticalized). It thus makes sense to talk about lexical vs. grammatical evidentials. Another thing to note about (4) is that there are no restrictions as to the level of meaning to which evidentials contribute. Evidentials are not required to be speech act modifiers, presuppositions or conventional implicatures. The only thing required is that the evidential condition be backgrounded. This is intuitively clear, since evidentials indicate the source of information for something that’s at issue. A consequence is that there cannot be inherently negative evidentials, something like ‘nobody told me that p’ or ‘there is absolutely no evidence that p’. Evidentials serve to supply the basis of some claim (or presented content), they cannot express the lack of every kind of evidence. In other words, evidentials indicate a source for the truth of the embedded proposition, never for its falsity. Content can be backgrounded in various ways and at various levels of meaning. Among the most promising candidates are presuppositions (Izvorski 1997), conventional implicatures (in the sense of Potts 2005) and felicity conditions (Faller 2002). Lexical items may have both backgrounded and non-backgrounded uses, a prominent example being clausal complement-taking predicates. Many clausal complement-taking predicates have both non-parenthetical and parenthetical uses. In their latter use embedding verbs are backgrounded and in fact often explicitly analyzed as evidentials (cf. Rooryck 2001, Simons 2007).2 Table 1 relates our notion of linguistic evidentiality to other popular definitions in the literature (cf. Schenner 2009 for a more extensive comparison). We opted for a notion that is semantically narrow (excluding purely epistemic elements, marked in the table as [+epistemic]), but morphologi-

164

Mathias Schenner

cally broad (including elements that do not have the status of highly grammaticalized markers, marked in the table as [+lexical]). Table 1. Definitions of ‚evidential‘

source grammatical

Aikhenvald (2004)

Ļ +lexical

ĺ

+epistemic Palmer (1986) Rooryck (2001)

our notion in (4-b)

Ifantidou (2001)

3. Evidentials in German German is one of the least likely languages one thinks of when talking about evidentiality. However, even if we restrict our notion of an ‘evidential’ from Section 2 to grammaticalized means of expressing the type of source of information, we can find several expressions in German that seem to fit this description. The availability of reportative evidence can be conveyed by the modal verbs sollen ‘should’ and wollen ‘want’, inferential evidence may be indicated by means of the evidential constructions scheinen ‘seem’, drohen ‘threaten’, versprechen ‘promise’ plus zu-infinitive and werden ‘become’ plus infinitive. Diewald and Smirnova forth. argue that the latter four constructions even build a paradigm in present-day German. This section will focus on the reportative modal verbs, especially on some of the complexities involved when they occur in embedded contexts.3

3.1. Reportative modals German modal verbs are polyfunctional: They systematically allow for both a circumstantial and an epistemic interpretation. The modals sollen ‘should’ and wollen ‘want’ are special in that they give rise to evidential instead of epistemic readings, in addition to their circumstantial readings. Both indicate that there is reportative evidence for (the truth of) the prejacent proposition. In the case of sollen the source of the report is subjectexternal (as is the source of the obligation in the circumstantial reading),



Embedded evidentials in German

165

cf. (5). In the case of wollen the source is the sentential subject itself (as is the source of the volition in the circumstantial reading), cf. (6). (5)

(6)

Anna soll in Oslo sein Anna should in Oslo be a. ‘Anna should be in Oslo (in view of her obligations)’ b. ‘Anna is said to be in Oslo’ Anna will in Oslo sein Anna want in Oslo be a. ‘Anna wants to be in Oslo’ b. ‘Anna claims to be in Oslo’

The following sections will concentrate on the reportative use of sollen ‘should’ (henceforth, sollenREP) as illustrated in (5).

3.2. A standard modal analysis Intuitively, by uttering sollenREP(p) a speaker conveys that there is reportative evidence for p. But what does this exactly mean? In order to make this intuition more precise, the following questions are addressed in this section: (a) What exactly is the content of the reportative component? (b) What is the semantic status of this component: truth-conditional or illocutionary? (c) Are additional meaning components conveyed, e.g. reduction or suspension of speaker commitment? The first question is taken up in Section 3.2.1, the second in Section 3.2.2, the third in Section 3.2.3. The discussion results in a preliminary lexical entry for sollenREP in the final Section 3.2.4.

3.2.1. Characterizing the reportative component There are various ways to think about and to formally analyze utterances and reports. In a simple case (sufficient for our purposes), an event of reporting involves a speaker, an addressee and a proposition that is conveyed. Like any event, a report is located at some spatiotemporal location (in some possible world). In our simplified setting, a report can be construed as a four-place relation, as in (7a), abbreviated in (7b) (‘¨’ for ‘dicendi’).

166 (7)

Mathias Schenner

General form of a report: a. x tells y in e that p b. ¨(e, x, y, p)

Reports about reports differ in whether both the speaker and the addressee of the reported report are specified as in ‘Anna told me that p’, or only the addressee as in ‘I’ve heard that p’, or only the speaker as in ‘Anna said that p’. There are also reports about reports where neither the speaker nor the addressee of the original report is explicitly expressed. A special case are reports about rumors, as in ‘It is said that p’ or ‘There are rumors that p’. These are not reports about specific reports, but involve quantification over report events – very roughly, ‘There are report events (in some contextually salient spatiotemporal region) that involve members of some (contextually salient) speech community and convey that the proposition p is true’. This rumor reading seems to be the default interpretation of sollenREP, e.g. in (5). But the reportative component conveyed by sollenREP is compatible with many other kinds of reports. For example, it can be used to report a specific utterance whose producer (and/or recipient) is explicitly mentioned, e.g. by an adverbial laut X ‘according to X’, as in (8a), or anaphorically inferred, as in (8b). (8)

a. Bea solli laut Annai in Oslo sein. Bea should according-to Anna in Oslo be ‘Anna said that Bea is in Oslo.’ b. Annai hat uns von Bea erzählt. Bea solli in Oslo sein. Anna has us of Bea told. Bea should in Oslo be. ‘Annai told us about Bea. Shei said that Bea is in Oslo.’

This being said, we will not be concerned with distinguishing the various types of reported reports in the following. The simplified abstract utterance predicate ¨(x, p) (roughly, ‘x said that p’) is sufficient for the purposes of this paper and will be uniformly used to represent the reportative component of sollenREP.

3.2.2. Truth-conditionality There is a long-lasting and still unresolved debate on whether epistemic modals are truth-conditional, i.e. contribute to the proposition expressed (cf. e.g. Papafragou 2006). For evidentials like sollenREP the same issue



Embedded evidentials in German

167

arises. One standard test for truth-conditionality is the so-called “embedding test”, according to which an item is truth-conditional iff it can be semantically embedded in the antecedent of a conditional. When we try to apply the test to sollenREP, we find both cases where sollenREP falls within the scope of wenn ‘if’ and hence is truth-conditional, e.g. in (9), and cases where it does not fall within the scope of wenn and hence qualifies as “non-truth-conditional”, e.g. in (10) (cf. Faller 2006 for similar examples). In the consulted corpora (cf. Section 3.3.1), the latter cases are much more frequent, but there are also many cases that allow for both a truthconditional and a non-truth-conditional reading. (9) a. Wenn es morgen regnen soll, müssen wir die Fahrräder abdecken. ‘If it is said that it is going to rain tomorrow, we have to cover the bicycles’ b. Ich habe es nicht gerne, wenn es hinterher nur einer gewesen sein soll. ‘I don’t like it, if afterwards it is said that it had been only one’ [Berliner Zeitung, 02.07.2003, p.23] (10) a. Wenn Herr Schröder das gesagt haben soll, dann müßte er die Konsequenz daraus ziehen und sagen ... ‘If Mr. Schröder said this (as it is alleged), he should draw the consequence and say ...’ [Die ZEIT 32/1985] b. Die Dame müßte mindestens um zehn Jahre älter sein, als sie [tatsächlich] ist, wenn sie zu dem Bilde Modell gestanden haben soll. ‘The woman would have to be at least ten years older than she actually is, if she had acted as a model for this painting (as it is alleged)’ [Vossische Zeitung (Morgen-Ausgabe), 03.03.1903, 5-6] According to the embedding test, sollenREP has both truth-conditional and non-truth-conditional uses.4 However, it can be argued that the seemingly non-truth-conditional uses in (10) are rather parenthetical uses, as their English translation by means of as-parentheticals also suggests. Parentheticals fail the embedding test, but they can nevertheless be handled in truthconditional semantics (cf. e.g. Asher 2000, Potts 2005). The conclusion is that sollenREP is truth-conditional, but has assertive (non-parenthetical) and parenthetical uses (more on these in Section 3.3.2).

168

Mathias Schenner

3.2.3. Speaker commitment Evidentials are often taken to not only indicate the type of source of evidence, but also a certain (increased or decreased) degree of speaker commitment. Scalar hierarchies have been built that order evidentials according to their strength, i.e. the degree of speaker commitment they convey. A typical example would be ‘direct > inferred > reported’. Given such a scale, by using a direct evidential marker a speaker indicates a high degree of commitment, whereas using a reported evidential marker would indicate a low degree of commitment. However, these hierarchies are best conceived of as partial orders and as context-dependent, as reflected in the formal model of speaker commitment (changes) recently proposed by Davis et al. 2007. Turning to sollenREP, does it indicate (as part of its lexical meaning) a reduced degree of speaker commitment, or even doubt or skepticism, as is sometimes suggested? Here we can rely on Mortelmans (2000: 136), who showed in a corpus study that, while sollenREP is compatible with speaker skepticism, this usage is in practice very rare (in 5 out of 137 considered cases, only one of which was a declarative clause). In addition, the speaker’s skepticism is usually explicitly marked. We conclude that sollenREP does not lexically encode speaker doubt. Skeptic overtones are pragmatic effects. The shift of responsibility conveyed by sollenREP arises as part of the truth-conditional reportative meaning: The speaker is not committed to the reported proposition, but to the existence of a report of the embedded proposition.

3.2.4. A standard modal account There are surprisingly few formal accounts of the evidential readings of German modals, a notable exception being Ehrich 2001. She proposes roughly the following lexical entry for sollenREP (cf. Ehrich 2001: 168): (11) ||soll||w = Ȝp . [ for every world w'Rw in which the claims of xc in w are true, it holds that w'úp] (where xc is understood as the contextually supplied source of the relevant claims) The basic idea behind (11) seems to be that ‘sollenREP(p)’ is equivalent to ‘xc said that p’ or, using the abstract utterance predicate introduced in Section 3.2.1, ‘¨(xc, p)’. Framing the analysis in Discourse Representation



Embedded evidentials in German

169

Theory (DRT) (Kamp and Reyle 1993), we get the Discourse Representation Structure (DRS) in (12b) for (12a), where ¨ is understood as a relation between an individual and a DRS. Reportative wollen receives a parallel treatment, with the minimal difference that the source of the reported speech act is not a contextually supplied individual or group but rather the sentential subject itself. (12) a. Anna soll in Oslo sein. ‘Anna is said to be in Oslo’ b. [a o x: Anna(a), Oslo(o), ¨(x, [: in(a,o)])] (13) a. Anna will in Oslo sein. ‘Anna claims to be in Oslo’ b. [a o: Anna(a), Oslo(o), ¨(a, [: in(a,o)])] This analysis of German reportative modals correctly captures that sollenREP and wollenREP are truth-conditional (as noted in 3.2.2) and do not indicate a reduced degree of speaker commitment (as noted in Section 3.2.3). Without further assumptions, it predicts that embedded occurrences of sollenREP are grammatical and receive the same modal interpretation as unembedded occurrences. These predictions will be tested in the following section.

3.3. Embedded sollen: data and generalizations Evidentials are typically considered to operate at the speech act level and hence to be unembeddable (cf. e.g. Aikhenvald 2004: ch. 8.1.3) for a list of languages that do not allow their evidentials to occur in embedded contexts). However, there are exceptions to this cross-linguistic tendency. Evidentials are embeddable in complement clauses in Tibetan (Garrett 2001), in Bulgarian (Sauerland and Schenner 2007) and in German, as shown below. In all of these languages, the embeddability of evidentials is subject to certain restrictions. Reportative evidentials occur most naturally under verba dicendi, but there are additional types of embedding predicates that license evidentials in their complements. Two questions will guide our investigation of the distribution of embedded sollenREP in German. First (in Section 3.3.1), which embedding predicates license sollenREP in their complement clauses? Second (in Section 3.3.2), how is embedded sollenREP interpreted?

170

Mathias Schenner

3.3.1. The distribution of embedded sollenREP In order to determine whether sollenREP can occur in complement clauses, two strategies have been deployed: (a) a corpus study and (b) a questionnaire study. In the corpus study, occurrences of embedded sollenREP in the IDS and DWDS corpora5 of written German were identified and collected. In total, about 300 corpus examples of sollenREP in complement clauses of 160 different complement-taking predicates were considered. Some typical matrix predicates are listed in (14) in order of decreasing frequency: (14)

bekannt sein ‘to be known’ (9%), kaum/schwer (zu) glauben ‘hard to believe’ and nicht glauben können ‘cannot believe’ (7%), berichten ‘to report’ (6,5%), es heißt ‘they say’ (3%), schwer vorzustellen ‘hard to imagine’ (3%), behaupten ‘to claim’ (2,5%), erfahren ‘to find out’ (2,5%), hören ‘to hear’ (2,5%), abstreiten ‘to deny’ and leugnen ‘to deny’ (2,5%), dementieren ‘to deny’ (2%), wissen ‘to know’ (2%), kolportieren ‘to hawk’ (1,5%), erzählen ‘to tell’ (1,5%), lesen ‘to read’ (1%), sagen ‘to say’ (1%), bezweifeln ‘to doubt’ (1%), unwahrscheinlich sein ‘to be unlikely’ (1%)

In addition, a questionnaire study was conducted. 18 native speakers of German were asked to rank the acceptability of a total of 25 test sentences on a scale ranging from 1 (totally unacceptable) to 5 (perfect). The main goals were to confirm the results of the corpus study and to identify matrix predicates that do not allow for embedded sollenREP. The main results are summarized in (15), where the matrix predicates are grouped according to the mean acceptability value of sentences with sollenREP in their complement clauses. (15) a. 5-4: hören ‘to hear’, seltsam sein ‘to be odd’, sagen ‘to say’, lesen ‘to read’, erzählen ‘to tell’, erinnern ‘to remember’, entdecken ‘to discover’ b. 4-3: interessant sein ‘to be interesting’, wissen ‘to know’, bedauern ‘to regret’ c. 3-2: glauben ‘to believe’, träumen ‘to dream’, fühlen ‘to feel’, Hinweise geben ‘there be indications’, bezweifeln ‘to doubt’, lügen ‘to lie’



Embedded evidentials in German

171

d. 2-1: möglich sein ‘to be possible’, überzeugt sein ‘to be convinced’, wünschen ‘to wish’, vermuten ‘to suppose’, hoffen ‘to hope’, befürchten ‘to fear’, beobachten ‘to observe’ The results of the corpus study and the questionnaire study match in the following sense: The predicates that frequently occurred with embedded sollenREP in the corpora received a high acceptability rank in the questionnaire study (e.g. hören ‘to hear’), while low ranked predicates did not occur in the corpora at all (e.g. hoffen ‘to hope’). The lists in (16) and (17) summarize and tentatively systematize these findings by grouping the relevant predicates. (16) a.

b. c.

d. e. f.

(17) a. b. c. d. e.

Predicates that allow sollenREP in their complement clause speech/text production (utterance) predicates: e.g. behaupten ‘to claim’, erzählen ‘to tell’, berichten ‘to report’, kolportieren ‘to hawk’ speech/text perception predicates: e.g. hören ‘to hear’, lesen ‘to read’ epistemic (semi-)factives: e.g. wissen ‘to know’, bekannt sein/werden ‘to be/become known’, erfahren ‘to find out’, erinnern ‘to remember’ emotive (semi-)factives: interessant sein ‘to be interesting’, seltsam sein ‘to be odd’, bedauern ‘to regret’ negative utterance (denial) predicates: e.g. abstreiten ‘to deny’, leugnen ‘to deny’ negative epistemic predicates: e.g. kaum/schwer zu glauben ‘hard to believe’, nicht glauben können ‘to cannot believe’, bezweifeln ‘to doubt’ Predicates that do not (or only marginally) allow sollenREP in their complement clause direct perception predicates: e.g. beobachten ‘to observe’, fühlen ‘to feel’ desire predicates: e.g. wünschen ‘to wish’, hoffen ‘to hope’ (non-factive, positive) epistemic predicates: e.g. glauben ‘to believe’, vermuten ‘to suppose’, überzeugt sein ‘to be convinced’ (non-factive) emotive predicates: e.g. befürchten ‘to fear’ predicates of (low positive) likelihood: e.g. möglich sein ‘to be possible’

172

Mathias Schenner

It is a non-trivial task to identify necessary and sufficient conditions for the embeddability of sollenREP, given the heterogeneity of the licensing predicates in (16). However, we can identify three main groups that might allow embedded sollenREP for different reasons (see below): (18) a. communication predicates b. (semi-)factive predicates c. negative (denial/doubt) predicates It is clear that the set of predicates that license embedded sollenREP is distinct from the set of predicates that license embedded root phenomena, e.g. verb-second (V2) complement clauses in German. There are both predicates that allow embedded V2 but not sollenREP (e.g. befürchten ‘to fear’) and predicates that allow embedded sollenREP but not V2 (e.g. interessant sein ‘to be interesting’). However, there is some kind of interaction. It has been argued that an embedded clause can have V2 order if and only if the containing sentence can be used in such a way that the embedded clause constitutes the main point of utterance (cf. Bentzen et al. 2007). In such cases, where the embedding predicate is used parenthetically, sollenREP can even occur in complement clauses of predicates in (17), especially nonfactive epistemic and emotive predicates like glauben ‘to believe’ or befürchten ‘to fear’, as illustrated in (19a) and (19b). (19)

(Anna does not want to meet Charly at the party today, and Bea knows this. Anna asks Bea, whether Charly will come. Bea answers:) a. Ich glaube/befürchte, Charly soll kommen. ‘I think/fear Charly should come.’ b. Charly soll kommen, glaube/befürchte ich. Charly should come think/fear I ‘I think /I’m afraid it is said that Charly will come’

This does not show that we should add these predicates to the list of sollenREP licensers. It rather shows that sollenREP occurs in (19a) and (19b) essentially unembedded. To conclude, the parenthetical use of matrix clauses can render sollenREP acceptable under certain predicates in (17) that allow for such a use.



Embedded evidentials in German

173

3.3.2. The meanings of embedded sollenREP In the previous subsection it was shown that sollenREP can occur in complement clauses of a number of embedding predicates. But how is embedded sollenREP interpreted? In order to answer this question, all of the corpus examples were semantically evaluated and categorized, which often required a closer inspection of the broader linguistic context. As a result, the following three kinds of reading have been identified: (20) a. A type reading: assertive (non-parenthetical, truth-conditional) b. G type reading: global (parenthetical, non-truth-conditional) c. C type reading: concord The A type reading (assertive) is the one that the standard semantics for sollenREP in Section 3.2.4 predicts: sollenREP(p) simply means ‘it is said that p’. However, in embedded contexts this reading is surprisingly infrequent. While, by introspection, many corpus examples are in principle compatible with an assertive reading, this interpretation is in most cases contextually clearly dispreferred. There are three factors that seem to favor an assertive reading: (a) if the embedding predicate is used parenthetically (cf. (19) above), (b) if the embedding predicate is factive and/or the embedded clause discourse-old or even echoic (cf. (21a)), and (c) if the embedded clause is an indirect question (cf. (21b)). A real life example is given in (22). (21) a. A: Anna soll in Oslo sein. ‘It is said that Anna is in Oslo.’ B: Ich weiß, dass Anna in Oslo sein soll. ‘I know that it is said that Anna is in Oslo.’ b. Anna fragte, ob Charly zur Party kommen soll. ‘Anna asked whether it is said that Charly is coming to the party’ (22) 90 mal 190 Zentimeter: Das waren die Abmessungen von Goethes bescheidenem Bett. Auf den Betrachter wirkt es heute ziemlich kurz, vor allem wenn er weiß, dass Goethe groß von Statur gewesen sein soll. ‘90 x 190 cm: That was the size of Goethe’s humble bed. To the beholder it seems quite short today, especially if they know that it is said that Goethe had been tall’ [Die ZEIT 11/2004]

174

Mathias Schenner

The C type reading (concord) of sollenREP(p) is simply p, provided that it is embedded under a communication predicate. The existence of this very frequent reading, illustrated in (23), has been noted before by Letnes 1997. While an A type reading is in principle available for these sentences, it is contextually strongly dispreferred. For example, the author of (23a) clearly didn’t intend to express that the newspaper had wrongly claimed that it was said that the princess gained her peerage dishonestly. (23) a. Die Zeitschrift hatte fälschlicherweise behauptet, daß sich die Prin zessin ihren Adelstitel unredlich erworben haben soll. ‘The newspaper had wrongly claimed that the princess gained her peerage dishonestly’ [Die Presse, 19.12.1992] b. Es ist irgendwie kindisch, daß gleich behauptet wird, daß MS dahinterstecken soll. ‘It is somehow childish that it is immediately claimed that MS is behind it’ [http://www.pro-linux.de/news/2002/4353.html, acc. 04.04.2007] The G type reading (global) of embedded sollenREP(p) can best be paraphrased by a parenthetical construction: ‘p, as it is alleged’. Albeit its availability is somewhat unexpected, this type of reading is quite pervasive in all of the corpora that have been looked at. Some examples are given in (24). The term ‘non-truth-conditional’ for this reading is somewhat misleading and will be avoided in the following, but has been mentioned, because sollenREP in the G type reading fails the well-known embedding test for truth-conditionality, as mentioned in Section 3.2.2. (24) a. Daß er dem Schüler auch auf den Kopf geschlagen haben soll, streitet der Lehrer entschieden ab. ‘The teacher resolutely denies that he hit the pupil also on the head (as it is alleged)’ [Salzburger Nachrichten, 18.01.1997] b. Daß es in ganz China im Vorjahr “nur” etwas mehr als 60.000 Verkehrstote gegeben haben soll, erscheint angesichts dieser rauhen Sitten wie ein Wunder. ‘In view of these tough customs it seems like a miracle that there were “only” slightly more than 60.000 traffic deaths in China last year (as it is alleged)’ [Salzburger Nachrichten, 26.11.1994] c. Daß Legrenzi sein Lehrer gewesen sein soll, ist unwahrscheinlich. ‘That Legrenzi had been his teacher (as it is alleged), is unlikely’ [Salzburger Nachrichten, 27.07.1991]



Embedded evidentials in German

175

d. Es ist schwer zu glauben, dass ich der Vater Deines Kindes sein soll. ‘It is hard to believe that I am the father of your child (as it is alleged)’ [Berliner Zeitung, 07.06.2005, p.17] To summarize, embedded sollenREP can be used in the following three ways (see Table 2, where ‘CTP’ stands for the complement taking predicate that embeds sollenREP).6 The contextually preferred type of reading depends on a variety of factors, the probably most important being the type of the embedding predicate. Even the few examples given above suggest that there are correlations between the type of the matrix predicate and the available readings of embedded sollenREP. The three main types of sollenREP licensing predicates listed in (18) seem to be associated with the three types of reading distinguished in (20) as indicated in Table 2. The unembedded use of sollenREP patterns with the embedding under (semi-)factives. Table 2. Readings of embedded reportative sollen

Readings of CTP(sollenREP(p)) a. A (assertive) CTP(¨(p)) b. G (global)

¨(p) ú CTP(p)

c. C (concord) CTP(p)

typical environment unembedded, under (semi-)factives under negative (doubt/denial) predicates under communication predicates

3.4. Analysis revisited The semantics of sollenREP introduced in Section 3.2.4 wrongly assigns the A type (assertive) reading to all occurrences of sollenREP. There are two main options for revising the analysis: (a) an ambiguity analysis that treats sollenREP as lexically ambiguous between A/G/C readings, and (b) a nonambiguity analysis where the various readings of sollenREP are derived from a single lexical entry. These two options are explored in the following sections.

176

Mathias Schenner

3.4.1. Ambiguity analysis One way to account for the additional readings of embedded sollenREP is to argue that it is lexically ambiguous between the standard semantics stated in Section 3.2.4, a concord and a parenthetical reading. In the latter reading, the reportative component is not added to the local DRS, but to the global DRS. Informally stated and ignoring concord readings for the moment, we get the following two entries for sollenREP: (25) a. sollenREP:1(p): add the condition ‘¨(xc, p)’ to the local DRS b. sollenREP:2(p): add the condition p to the local DRS and the condition ¨(xc, p) to the global DRS For example, using sollenREP:1 we can derive the A reading of (26a), shown in (26b), and using sollenREP:2 we can derive the G reading, shown in (26c). (26) a. Bea sagt/weiß, dass Anna in Oslo sein soll. ‘Bea says/knows that Anna in Oslo be should.’ b. [a b o: Anna(a), Bea(b), Oslo(o), say/know(b,[x: ¨(x,[: in(a,o)])])] c. [a b o x: Anna(a), Bea(b), Oslo(o), say/know(b,[: in(a,o)]), ¨(x,[: in(a,o)])] There are other ways of implementing the basic idea that sollenREP has a non-parenthetical and a parenthetical reading, depending on one’s favorite theory of supplements. For example, using the multidimensional framework of Potts 2005, we could replace (25) by (27). (27) a. sollenREP:1 Ȝp Ȝxc Ȝw . [¨(p)(xc)(w)] : b. sollenREP:2 Ȝp Ȝxc Ȝw . [¨(p)(xc)(w)] : The difference between (27a) and (27b) is that the assertive (nonparenthetical) entry (27a) contributes the reportative component to the atissue content, while the parenthetical entry (27b) contributes it as a conventional implicature in the sense of Potts 2005. No matter what version, the ambiguity approach suffers from several problems. Without further assumptions, it radically overgenerates in two cases. First, it does not predict that (and hence cannot explain why) sollenREP:1 cannot be embedded in many (especially non-factive) contexts. Second, it does not predict that sollenREP:2 cannot be used in matrix clauses.



Embedded evidentials in German

177

Of course, we could come up with some principles that restrict possible disambiguations of sollenREP, e.g. along the lines in (28). (28) a. Do not commit the speaker to p, if she uttered ‘... sollenREP(p) ...’ b. Prefer the strongest meaning, i.e. prefer sollenREP:2 to sollenREP:1 However, this line of thought will not be pursued in this paper, since there is an additional reason to disfavor the ambiguity approach. By economy considerations, a non-ambiguity approach that does not require a duplication of lexical entries is to be preferred over the ambiguity approach. Hence we shift our endeavors to developing a non-ambiguity account of sollenREP in the next section.

3.4.2. Non-ambiguity analysis: a presuppositional account If we want a single entry for sollenREP, its meaning has to be, in a sense, positionally flexible, since the reportative component conveyed by sollenREP sometimes seems to be contributed to the local DRS, sometimes to the global DRS. This kind of flexibility is reminiscent of the projection behavior of presuppositions, “agile creatures eager to leave their homes” (Geurts 1999: 114). In presuppositional DRT, DRSes are constructed in two steps. First, a preliminary DRS for a sentence is built based on the lexical meanings of its parts. Presuppositions are explicitly represented where they are triggered. Second, the sentence is put in context, its presuppositions are resolved, ultimately leading to the final DRS of the sentence. There are two basic options for the resolution of presuppositions: Binding, as in (29a), and accommodation, where we can further (minimally) distinguish between global (non-local) accommodation as in (29b) and local (non-global) accommodation as in (29c) (cf. e.g. Geurts 1999). (29) a. If Anna owns a cat, Anna’s cat is black b. If Anna’s cat is black, she must be happy c. Either Anna doesn’t have a cat or Anna’s cat is in hiding The basic idea of our non-ambiguity analysis of sollenREP is that it triggers a reportative presupposition ‘¨(xc, p)’. It turns out that the three readings of (embedded) sollenREP correspond to the three basic projection possibilities of this presupposition, as shown in Table 3.

178

Mathias Schenner

Table 3. Readings of embedded reportative sollen type of reading

resolution

configuration

typical environments

A (assertive)

local accomm.

[CTP](¨(xc, p))

unembedded, under know

G (global)

global accomm. ¨(xc, p) ú CTP(p)

under doubt

C (concord)

binding

under say

CTP(p)

There is one complication: In the G reading of sollenREP(p), the proposition p plays a double role, i.e. it is used twice in the semantic representation. This double usage is typical for supplemental expressions; cf. Potts 2005 for discussion. The proposed semantics of sollenREP (somewhat simplified: extensional and ignoring tense) is stated in (30). It consists of two parts: (a) a reportative presupposition, (b) an assertive part that is only activated if the resolution of the reportative presupposition violates local informativity. The second part is required for deriving the G reading, as shown below. (30)

sollenREP(p): (a) ˜[xc | ¨(xc, p)] (b) p, if the resolution of (a) violates local informativity

The idea that evidential expressions contribute a presupposition is not new (cf. e.g. Izvorski 1997). However, as will become clear in a moment, the presupposition of sollenREP in (a) does not behave exactly like a run-of-themill presupposition (if there is such a thing). More specifically, the projection profile of the sollenREP presupposition features a low accommodation threshold (thus the possibility of binding does not strictly exclude the possibility of accommodation). The second component in the semantics of sollenREP in (30) is an instantiation of the idea that an expression has to have some effect on its local DRS (local informativity). This condition is violated, for example, if the reportative presupposition of sollenREP(p) is non-locally accommodated. In such a case, local informativity is rescued by adding p to the local DRS (stripping off sollenREP). Let’s look at some applications. In the following examples, presupposed material is underlined, conditionally activated material is in italics. The simplest cases are occurrences of unembedded sollenREP, as in (31a). Since binding is not an option here, the reportative component has to be accommodated in the local (= global) DRS, satisfying local informativity.



Embedded evidentials in German

179

(31) a. Bea soll in Oslo sein. Bea should in Oslo be ‘It is said that Bea is in Oslo.’ b. [b o x: Bea(b), Oslo(o), ¨(x,[: in(b,o)]), in(b,o)] c. [b o x: Bea(b), Oslo(o), ¨(x,[: in(b,o)])] If sollenREP is embedded under an utterance predicate, as in (32a), its reportative presupposition can be bound to it. The presence of the conditionally activated complement of sollenREP might facilitate this process which results in the concord interpretation in (32c). (32) a. Anna sagt dass Bea in Oslo sein soll. Anna says that Bea in Oslo be should ‘Anna says that Bea is in Oslo.’ b. [a b o: Anna(a), Bea(b), Oslo(o), say(a,[x: ¨(x,[: in(b,o)]), in(b,o)])] c. [a b o: Anna(a), Bea(b), Oslo(o), say(a,[: in(b,o)])] If the reportative presupposition cannot be bound, global accommodation is the preferred option, as illustrated in (33a). Since global accommodation is non-local here (in contrast to (31a)), local informativity is violated in (33c), which triggers the (b) component in (30). The resulting DRS in (33d) correctly captures the interpretation of (33a). (33) a. Es ist schwer zu glauben dass Bea in Oslo sein soll. It is hard to believe that Bea in Oslo be should ‘It is hard to believe that Bea is in Oslo (as it is alleged).’ b. [b o: Bea(b), Oslo(o), hard-to-believe([x: ¨(x,[: in(b,o)]), in(b,o)])] c. [b o x: Bea(b), Oslo(o), hard-to-believe([: in(b,o)]), ¨(x,[: in(b,o)])] d. [b o x: Bea(b), Oslo(o), hard-to-believe([: in(b,o)]), ¨(x,[: in(b,o)])] If sollenREP occurs in embedded contexts, local accommodation is also an option, albeit usually a dispreferred one (cf. Section 3.2). For example, (32a), repeated as (34a), can get the interpretation in (34c), if local accommodation is enforced. (34) a. Anna sagt dass Bea in Oslo sein soll. Anna says that Bea in Oslo be should ‘Anna says that it is said that Bea is in Oslo.’

180

Mathias Schenner

b. [a b o: Anna(a), Bea(b), Oslo(o), say(a,[x: ¨(x,[: in(b,o)]), in(b,o)])] c. [a b o: Anna(a), Bea(b), Oslo(o), say(a,[x: ¨(x,[: in(b,o)])])] In Section 3.3.2 it was noted that (semi-)factive predicates seem to favor local accommodation readings. If we assume that presuppositions are resolved bottom-up, i.e. presuppositions of deeper embedded triggers are resolved prior to presuppositions of higher triggers, then we might be able to explain this finding. For example, semifactive wissen ‘know’ presupposes that its clausal complement is true. But the content of its complement in (35a) depends (assuming bottom-up resolution) on the resolution of the presupposition of sollenREP. If the presupposition of sollenREP(p) were accommodated globally, the complement of wissen and hence a presupposition of the sentence would be p, as shown in (35b). But this would render the contribution of sollenREP superfluous. By contrast, if the presupposition of sollenREP is accommodated locally, we get the sensible interpretation in (35c): “It is said that Bea is in Oslo and Anna knows that”. (35) a. Anna weiß dass Bea in Oslo sein soll. Anna knows that Bea in Oslo be should b. [a b o x: A.(a), B.(b), O.(o), know(a,[: in(b,o)]), ¨(x,[: in(b,o)]), in(b,o)] c. [a b o x: A.(a), B.(b), O.(o), know(a,[y: ¨(y,[: in(b,o)])]), ¨(x,[: in(b,o)])]

4. Conclusion This paper pursued two main goals. The first was to argue for a semantically narrow but morphologically broad notion of linguistic evidentiality (in Section 2). The second was to highlight the importance of investigating evidentials in embedded contexts to uncover some of their central semantic properties. The latter point was illustrated by an investigation of the German modal verb sollen ‘should’ in its reportative use. It has been shown that sollenREP can be embedded in complement clauses of at least three classes of embedding predicates: communication predicates, (semi-)factive predicates and certain negative (denial/doubt) predicates. Embedded occurrences of sollenREP can have one of three readings that have been labeled A (assertive), G (global) and C (concord). The availability of G and C readings are problematic for standard accounts of sollenREP and necessitate a



Embedded evidentials in German

181

more fine-grained analysis. In Section 3.4, two proposals have been considered that are capable of deriving the additional readings. An interesting direction for future research is the investigation of restrictions on and semantic effects of embedding other types of evidentials in German, in particular the paradigm of German inferential evidentials identified by Diewald and Smirnova forth. The systematic cross-linguistic investigation of embedded occurrences of evidentials has the potential to reshape major theories of evidentiality that so far tended to exclusively focus on describing the semantic and pragmatic effects of evidentials in (mostly declarative) root clauses.

Notes

1.

2.

3. 4. 5. 6.

I wish to thank Gabriele Diewald and Elena Smirnova for organizing this inspiring workshop. Furthermore I'm indebted to Hans-Martin Gärtner, Manfred Krifka, Rainer Ludwig, Jakob Maché, Fabienne Salfner, Uli Sauerland, Magdalena Schwager, Frank Sode and Hubert Truckenbrodt for stimulating discussions on the topic of my contribution. The research for this paper was funded by the project CHLaSC in the FP6 Pathfinder Initiative What it means to be human of the European Commission. Speas (2004) explicitly argues that the set of evidence types available for grammaticalization is rather small. Following her account, we could posit the universal set I = {Personal Experience, Direct Evidence, Indirect Evidence, Hearsay}. However, it’s not that clear that languages only make use of four distinct evidential categories. Moreover, we will allow for lexical evidentials that can encode many more fine-grained sources of information. Simons (2007) introduces the useful notion of the ‘main point of utterance’ to characterize parenthetical uses: “the main point of an utterance U of a declarative sentence S is the proposition p, communicated by U, which renders U relevant” (p.1035). In examples “where the embedded clause has main point status, the main clause predicate appears to be functioning as a kind of evidential” (p.1036). A proposition is backgrounded in our sense if it is not the main point of utterance in the sense of Simons. This section is based on Schenner 2008. The type of conditional clause may influence the preferred reading; cf. the distinction between central and peripheral adverbial clauses in Haegeman 2006. For the IDS corpora (DeReKo) cf. http://www.ids-mannheim.de/kl/ projekte/korpora/, for the DWDS corpora cf. http://www.dwds.de/. If ‘CTP(p)’ entails ‘¨(p)’, the G and C readings coincide. However, C readings cannot be reduced to G readings in general. The C reading of the following example (Uli Sauerland, p.c.) does not entail that somebody claimed or wrote

182

Mathias Schenner that the princess is a fraud: Keine Zeitung hat geschrieben, dass die Prinzessin eine Betrügerin sein soll ‘No newspaper wrote that the princess is a fraud’.

References Aikhenvald, Alexandra Y. 2004 Evidentiality. Oxford: Oxford University Press. Asher, Nicholas 2000 Truth conditional discourse semantics for parentheticals. Journal of Semantics 17: 31-50. Bentzen, Kristine, et al. 2007 The Tromsø guide to the force behind V2. Working Papers in Scandinavian Syntax 79: 93-118. Chafe, Wallace 1986 Evidentiality in English conversation and academic writing. In Evidentiality: The linguistic coding of epistemology. Chafe, Wallace, and Johanna Nichols (ed.), 261-272. Norwood: Ablex. Chung, Kyung-Sook 2005 Space in tense: The interaction of tense, aspect, evidentiality, and speech act in Korean. Doctoral Dissertation, Simon Fraser University, Burnaby, British Columbia. Davis, Christopher, Christopher Potts, and Margaret Speas 2007 The pragmatic values of evidential sentences. In Proceedings of SALT 17. Masayuki Gibson and Tova Friedman (eds.). Ithaca, NY: CLC Publications. Diewald, Gabriele, and Elena Smirnova forth. 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, Frank Brisard, Tanja Mortelmans, and Walter de Mulder (eds.). Antwerp: University of Antwerp. Ehrich, Veronika 2001 Was ‘nicht müssen’ und ‘nicht können’ (nicht) bedeuten können: Zum Skopus der Negation bei den Modalverben des Deutschen. In Müller, Reimar and Marga Reis (eds.), 149–176. Faller, Martina 2002 Semantics and pragmatics of evidentials in Cuzco Quechua. Doctoral Dissertation, Stanford University. 2006 Evidentiality below and above speech acts. http://semanticsarchive.net/Archive/GZiZjBhO/Fallerevidentiality.above.below.pdf.



Embedded evidentials in German

183

von Fintel, Kai, and Anthony S. Gillies 2007 An opinionated guide to epistemic modality. To appear in Oxford Studies in Epistemology, Vol. 2, Tamar Szabó Gendler and John Hawthorne (eds.). Garrett, Edward John 2001 Evidentiality and assertion in Tibetan. Doctoral Dissertation, UCLA, Los Angeles. Geurts, Bart 1999 Presuppositions and Pronouns. Amsterdam: Elsevier. de Haan, Ferdinand 2001 The relation between modality and evidentiality. In Müller, Reimar and Marga Reis (eds.), 201-216. Haegeman, Liliane 2006 Argument fronting in English, Romance CLLD and the left periphery. In Cross-Linguistic Research in Syntax and Semantics: Negation, Tense and Clausal Architecture, Zanuttini R., H. Campos, E. Herburger and P. Portner (eds.), 27-52. Washington: Georgetown University Press. Hengeveld, Kees 2006 Evidentiality and reportativity in Functional Discourse Grammar. Presentation at TAM TAM Workshop, Radboud University Nijmegen. Ifantidou, Elly 2001 Evidentials and relevance. Amsterdam/Philadelphia: Benjamins. Izvorski, Roumyana 1997 The present perfect as an epistemic modal. Proceedings of SALT 7: 222-239. Kamp, Hans, and Uwe Reyle 1993 From Discourse to Logic. Dordrecht: Kluwer. Kratzer, Angelika 1991 Modality. In Semantik: Ein internationales Handbuch der zeitgenössischen Forschung (HSK 6), Arnim von Stechow and Dieter Wunderlich (eds.), 639–650. Berlin/New York: de Gruyter. Lazard, Gilbert 1999 Mirativity, evidentiality, mediativity, or other? Linguistic Typology 3: 91-110. Letnes, Ole 1997 ‘Sollen’ als Indikator fuer Redewiedergabe. In Aspekte der Modalität im Deutschen – auch in kontrastiver Sicht, Studien zu Deutsch als Fremdsprache 3, F. Debus and Oddleif Leirbukt (eds.), 119-134. Hildesheim: Olms. Matthewson, Lisa, Henry Davis, and Hotze Rullmann 2007 Evidentials as epistemic modals: Evidence from St’át’imcets, Ms., University of British Columbia.

184

Mathias Schenner

McCready, Eric, and Norry Ogata 2007 Evidentiality, modality and probability. Linguistics and Philosophy 30: 147-206. Mithun, Marianne 1999 The languages of native North America. Cambridge: Cambridge University Press. Mortelmans, Tanja 2000 On the evidential nature of the epistemic use of the German modals ‘müssen’ and ‘sollen’. In Modal verbs in Germanic and Romance Languages, Belgian Journal of Linguistics 14, Johan van der Auwera and Patrick Dendale (eds.), 131-148. Amsterdam/Philadelphia: Benjamins. Müller, Reimar, and Marga Reis (eds.) 2001 Modalität und Modalverben im Deutschen. Hamburg: Buske. Nuyts, Jan 2001 Epistemic modality, language, and conceptualization: A cognitivepragmatic perspective. Amsterdam/Philadelphia: Benjamins. Palmer, Frank R. 1986 Mood and modality. Cambridge: Cambridge University Press. Papafragou, Anna 2006 Epistemic Modality and Truth Conditions, Lingua 116: 1688-1702. Potts, Christopher 2005 The Logic of Conventional Implicatures. Oxford: Oxford University Press. Rooryck, Johan 2001 Evidentiality. Part I. GLOT 5: 125-133. Sauerland, Uli, and Mathias Schenner 2007 Embedded Evidentials in Bulgarian. In Proceedings of Sinn und Bedeutung 11, E. Puig-Waldmüller (ed.), 495-509. Barcelona: Universitat Pompeu Fabra. Schenner, Mathias 2008 Double face evidentials in German: Reportative ‘sollen’ and ‘wollen’ in embedded contexts. In Proceedings of Sinn und Bedeutung 12, Atle Grønn (ed.), 552–566. Oslo: University of Oslo. 2009 Evidentials in Complex Sentences: Foundational Issues and Data from Turkish and German. To appear in Evidentiality, R.-M. Dechaine, et al. (eds.), UBC Working Papers. Simons, Mandy 2007 Observations on embedding verbs, evidentiality, and presupposition. Lingua 117: 1034-1056. Speas, Margaret 2004 Evidentiality, logophoricity and the syntactic representation of pragmatic features. Lingua 114: 255-276.



Embedded evidentials in German

185

Stenzel, Kristine Sue 2004 A reference grammar of Wanano. Doctoral Dissertation, University of Colorado, Colorado. Willet, Thomas L. 1988 A cross-linguistic survey of the grammaticalization of evidentiality. Studies in Language 12: 51-97.

Embedding indirective (evidential) utterances in Turkish Hatice Coúkun

Turkish subordinations are mostly formed by morphological (or synthetic) means. This so-called canonical subordination uses non-finite units. Beside synthetic subordination in Turkish, there are marginal alternative subordination strategies, either asyndetic or using conjunctions or interrogative pronouns. Among other tasks, such analytical subordination strategies license the possibility of embedding evidentially marked clauses. This paper focuses on the embeddable evidential clause types in Turkish and the semantic and pragmatic properties of the predicates they select.1

1. Introduction Recent cross-linguistical typological studies based on Turkic languages, Eastern Pomo, Abkhaz and Baniwa treat the embedding of evidentiality as a phenomenon which is not licensed in dependent clauses (cf. Aikhenvald 2004: 253). Turkish possesses a grammatical means for expressing evidentials (the perfect inflection suffix -mIú for verbal clauses and the clitic form -(y)mIú2 for nominal ones), which do not primarily express evaluation or assessment of the truth of the utterance. They state the existence of a source of evidence for a propositional content. This propositional content may be acknowledged through hearsay, inference or perception. On hearing from Ayúe or from someone else that she is pregnant: (1) Ayúe hamile-ymiú. A. pregnant-INDR.CLIT ‘I heard that Ayúe is pregnant.’ On finding that one’s glasses are not in one’s bag or pocket: yan-ım-a al-ma-mıú-ım. (2) Gözlü÷-üm-ü eyeglass-1SG-ACC side-1SG-DAT take-NEG-INDR/PERF-1SG ‘It turns out I haven’t brought my glasses with me.’

188

Hatice Coúkun

After tasting some coffee: úekerli-ymiú. (3) Kahve çok coffee very sugary-INDR.CLIT ‘As I tasted the coffee is very sugary.’ It has often been argued that the evidential sentence types in Turkic languages are limited to asserted sentences, i.e. main clauses with a stated, contradictable content (cf. Johanson 2000: 61, 2006: 81). It is true that evidentially marked propositions do not appear as embedded clauses that are integrated into sentence, such as non-finite embedded clauses. Such are relative clauses and nominalized complement clauses or non-finite adverbial clauses which are marked by converbs, participles etc. But unlike nonfinite subordinate clauses, finite subordinate clauses are similar to main clauses in that they are asserted. Johanson claims that the embedded clause in (4) illustrates purely postterminal value, but we will provide some data to show that finite subordinate clauses may occur with evidential content as well. (4)

öl-müú>

bekle-di-m. Hearing > Touch > {Smell, Taste}



Evidentiality, polysemy, and the verbs of perception

251

“A verb having a basic meaning belonging to a sensory modality higher (to the left) in the hierarchy,” he explains, “can get an extended meaning that covers some (or all) of the sense modalities lower in the hierarchy” (Viberg 1983: 136-137). Hence verbs of visual perception can be expected to exhibit a higher degree of polysemy and/or generalization than verbs of olfactory perception. This is no surprise when one considers that the bulk of scholarship on perception verbs is devoted mostly to visual perception (cf. Sweetser 1990: 23-48, Schröder 1995, van der Does and van Lambalgen 2000, Harm 2000, and de Haan forth.). But considering that vision provides us with our most global and immediate access to the world around us (whereas a modality such as smell is not as global nor as reliable), it makes perfect sense that the verbs of visual perception should find themselves at the top of this hierarchy. It is necessary to draw a two-way distinction between subject-oriented and object-oriented verbs of perception (Viberg 1983, Harm 2000), for we will see later how this distinction plays into the expression of evidential meaning. Subject-oriented perception verbs (called “experiencer-based” by Viberg) are those verbs whose grammatical subject is the perceiver and they emphasize the perceiver’s role in the act of perception. They are transitive verbs, and they can be further sub-divided into agentive and experiencer perception verbs. The subject-oriented agentive perception verbs signify an intended act of perception: (2) a. b. (3) a. b.

Karen listened to the music. Karen hat die Musik angehört. Karen smelled the iris with delight. Karen hat die Iris mit Vergnügen gerochen.

So in (2) and (3), Karen intends to listen to the music and she intentionally smells the iris. On the other hand, subject-oriented experiencer perception verbs indicate no such volition; instead, they merely describe a nonintended act of perception: (4) a. b. (5) a. b.

Karen heard the music. Karen hat die Musik gehört. Karen tasted the garlic in the soup. Karen hat den Knoblauch in der Suppe geschmeckt.

So here in (4) and (5), Karen does not intend or go out of her way to auditorily perceive the music or to gustatorily perceive the garlic in her soup;

252

Richard Jason Whitt

they are simply acts of perception that she naturally experiences without any volition on her part. But one thing that (2) through (5) have in common is that Karen is both the grammatical subject of the clause and the agent – whether intended or not – of perception. Some examples of subjectoriented perception verbs include see (experiencer), look (agentive), hear (experiencer), listen (agentive), feel (agentive or experiencer), smell (agentive or experiencer), and smell (agentive or experiencer) in English; sehen ‘see’ (experiencer), ansehen ‘look at’ (agentive), hören ‘hear’ (experiencer), anhören ‘listen to’ (agentive), fühlen ‘feel’ (agentive or experiencer), schmecken ‘taste’ (agentive or experiencer), and riechen ‘smell’ (agentive or experiencer) in German. As can be seen, visual and auditory perception, the two modalities at the top of the hierarchy, have separate lexical items in both English and German for subject-oriented agentive and experiencer perception verbs, whereas the other three modalities each share a single lexical item that can signify both types of perception. The object of perception, rather than the perceiver himself, is the grammatical subject of object-oriented perception verbs (called “sourcebased” by Viberg), and the agent of perception is sometimes wholly absent from the clause. These verbs are intransitive. When using an objectoriented perception verbs, speakers make an assessment concerning the state of the object of perception, and these verbs are often used evidentially: (6) a. b. (7) a. b.

Karen looks healthy. Karen sieht gesund aus. The cake tastes good. Der Kuchen schmeckt gut.

The speaker reports on what is perceived here, and neither Karen nor the cake are perceivers. In (6), the speaker infers via visual perception that Karen is healthy (this is evidential). In (7), the cake is deemed via gustatory perception to be of good quality (not evidential). If the speaker is not the agent of perception and wishes to report on someone else’s experience, a to-clause in English or a dative object in German may sometimes be used: (8) a. The cake tastes good to the baker. b. Der Kuchen schmeckt dem Bäcker gut. And similarly, if the speaker wishes to indicate that the act of perception is indeed theirs and not someone else’s, to me may be used in English and mir can be employed in German. Examples of object-oriented perception



Evidentiality, polysemy, and the verbs of perception

253

verbs include look, sound, feel, taste, and smell in English; aussehen ‘look, appear,’ (sich) anhören ‘sound,’ (sich) anfühlen ‘feel,’ schmecken ‘taste,’ and riechen ‘smell’ in German. Some verbs that are lower in the hierarchy function both as subject-oriented and object-oriented perception verbs in English, while German employs the derivational prefixes such as an- and aus- or a reflexive pronoun to make the distinction between subjectoriented and object-oriented verbs.

2.2. Polysemy The issue of polysemy is inevitable and unavoidable when one considers the rich semantics of perception verbs. Merely a brief glance at respective entries in the Oxford English Dictionary (OED) or the Deutsches Wörterbuch (DWB) reveal how many meanings the verbs of perception can signify, especially when it comes to the verbs of visual perception. The “core” meanings of the perception verbs are, of course, physical in nature, but there is no shortage of metonymic or non-physical (i.e. internal, metaphorical) extensions to the meanings of these verbs (Sweetser 1990: 23-48; Harm 2000). The polysemy expressed by verbs of each sensory modality is outlined below.

2.2.1. Visual Perception Physical perception with the eyes is the prototypical meaning one finds expressed by verbs such as see or sehen. But these verbs can also be extended metaphorically to cover internal states of knowledge or intellection (Sweetser 1990: 32-34; Harm 2000: 127-161). One is certainly hardpressed to find the literal meaning of vision in expressions such as I see what you mean or Ich sehe, was du meinst; the meaning here is one of internal understanding, not external vision. The focusing of mental attention on an object or state of affairs can also be indicated by verbs of visual perception, as in to look ahead to or vorausschauen, where the mental focus is on some event in the future. Sweetser (1990: 33) describes this metaphorical extension as the expression of mental ‘vision’. These extensions should come as no surprise, for there is a cross-linguistic affinity between vision and knowledge: e.g. Lat. videre is cognate with Engl. wise and Germ. wissen.

254

Richard Jason Whitt

2.2.2. Auditory Perception Because verbs of auditory perception can also point to the content of what is heard – rather than the physical act of hearing itself – they display a certain degree of (metonymic) polysemy as well. This becomes especially apparent in cases of hearsay evidentiality: (9) a. I heard that Karen isn’t coming to the party. b. Ich habe gehört, dass Karen nicht zur Party kommt. As we see, the object of the verb of auditory perception is the proposition Karen isn’t coming to the party, i.e. what is heard. Of course, this information was acquired via auditory perception, so the literal meaning of hear and hören is not completely absent either. A bit more abstract is when meanings of auditory perception become extended to cover obedience, as in Listen to your mother or Hör auf deine Mutter. But seeing that auditory perception is the most likely way one will find out what is expected of him, this meaning extension (hearing > obeying) is not unexpected. 2.2.3. Tactile Perception In Modern English and German, the main verbs of tactile perception (i.e. feel and fühlen) denote forms of internal perception (emotion, intuition, inference) just as often as they signify acts of external perception. Harm (2000: 194-202) also notes that, along with the verbs of visual perception, the verbs of tactile perception can indicating understanding, yet another internal process, as in e.g. begreifen ‘comprehend, understand’ (< Middle High German grîfan ‘feel, touch’; cf. English grasp) or the English colloquial expressions I feel your pain ‘I understand and empathize with your suffering’ or simply I feel you ‘I understand what you mean.’

2.2.4. Olfactory and Gustatory Perception These two sensory modalities fail to evince the rich array of polysemy expressed by the verbs of other sensory modalities, but they are not totally devoid of metaphorical extensions to their core meanings either. Consider the evidential utterances I’m smelling a rat or Das riecht nach Skandal ‘That smells like a scandal.’ There is clearly no act of olfaction expressed



Evidentiality, polysemy, and the verbs of perception

255

in either of these statements; instead, the verb of olfaction is used to signify that the speaker has a general feeling that something is not in order with the current situation (embodied in the form of a metaphorical rodent) or that a scandal exists. And sometimes verbs of trying or choosing are derived from verbs of gustation, as in the German kosten ‘to sample, taste,’ which involves the sampling of food items. On a more general note, the verbs of olfactory and gustatory perception (as well as tactile perception) can function as both subject-oriented and object-oriented verbs. When taste, for example, is used as a subjectoriented verb, it denotes the physical act of someone’s gustatory experience (whether this experience is intended or not depends on whether the verb is being used in its agentive or experiencer sense). But when used as an object-oriented verb, taste refers to the quality of the object being tasted (e.g. I tasted the soup vs. The soup tasted spicy).

2.3. Evidential Perception Verbs So how is all this polysemy germane to a discussion of evidentiality and perception verbs? For one, a single perception verb can express a variety of evidential meanings, and these meanings generally correspond to meanings found outside of the evidential domain as well. Verbs of visual perception, for example, can point to literal vision as the source of information, but they can also indicate that understanding or knowledge forms the basis of speaker knowledge. Secondly, certain evidential meanings tend to be bound to specific construction types; yet even within a single construction type, different types of evidence may be expressed, so we can speak of constructional polysemy here as well. In order to pinpoint what exactly constitutes an evidential perception verb, it is necessary to employ a criterion that has been recognized since Jakobson ([1957] 1971): deixis. Evidentiality is, by nature, deictic, so any evidential use of a perception verb must include a deictic meaning in addition to the regular denotation of physical perception (or any metonymic or metaphorical extension thereof).3 Consider (10) and (11): (10) a. b. (11) a. b.

I see the house. Ich sehe das Haus. I see the house burning. Ich sehe das Haus brennen.

256

Richard Jason Whitt

In (10), there is only one proposition: I see the house. But in (11), there are two propositions: one, the house is burning, and two, I see the house burning. It is this second proposition that carries evidential meaning, for it is here where the act of visual perception is used to describe the speaker’s relationship with the first proposition, i.e. the house can be reported to be in a state of burning because the speaker has visual evidence to support this claim. And because evidentiality is deictic, the perceptual evidence always lies with the speaker, even when – as in the case of object-oriented perception verbs – the speaker (the agent of perception) is not syntactically expressed: (12) a. Karen looks sick. b. Karen sieht krank aus. There may not be two explicit propositions here, but there are certainly two layers of meaning: one, the speaker has visually perceived Karen, and two, the speaker infers – based on this perception – that Karen is sick. The visual evidence (Karen’s appearance) provides the basis for the speaker’s claim of Karen’s health. Examples (10) through (12) also exhibit a tendency found among evidential perception verbs: constructions involving subjectoriented perception verbs can only be evidential with a first-person subject, whereas the object-oriented perception verbs allow evidential readings when second- and third-person grammatical subjects are present.4

3. The Data In order to examine exactly when and how evidential perception verbs are used in English and German, I conducted a corpus study that covers both the Early Modern and Modern periods. For English, I consulted the Early Modern section of the Helsinki Corpus, which covers the years 1500-1710 and contains a total of 551,000 tokens (the entire Helsinki Corpus contains 1,979,420 tokens). I also used the ARCHER (A Representative Corpus of Historical English Registers) Corpus, which covers the years 1650-1990 and contains 1,789,309 tokens. For German, I examined the Bonner Frühneuhochdeutschkorpus ‘Bonner Corpus of Early New High German,’ whose texts date from ca. 1350 to 1699 and whose total token count is roughly 608,000. Because there is not yet a comprehensive corpus that covers the eighteenth and nineteenth centuries,5 I consulted two single-author corpora that I believe satisfactorily



Evidentiality, polysemy, and the verbs of perception

257

cover this time period with a variety of text types: the Kant Corpus and the Goethe Corpus. The Kant Corpus contains 3,338,068 tokens from both the philosophical writings and personal correspondences (Briefwechsel) of German philosopher Immanuel Kant (1724-1804). The Goethe Corpus contains 1,400,000 tokens from twenty-nine selected writings of various text types by the German author Johann Wolfgang von Goethe (17491832). And for the twentieth century, I consulted the Digitales Wörterbuch der deutschen Sprache ‘Digital Dictionary of the German Language’ (DWDS), which contains roughly 100,000,000 tokens of spoken and written twentieth century German. Given the large size of this corpus, I took a random sample comparable in size to the amount of data found in other corpora where relevant.

4. A Typology of Evidential Perception Verbs After the criteria for an evidential verb of perception (discussed in Section 2.4) were established, a search of the corpora discussed in Section 3 was conducted, and the following construction types were found to evince evidential meaning: Type I: Perception Verb (PV) + Finite Complementizer Clause (FCC)

(13) Now I hear you were with the old man later on last night. (ARCHER: 1960vidl.d8a) (14) Do es nue liecht ward, do sach wir, daz dy. herren zu einander giengen in ain haws. (Bonner Corpus: Text 113, Helene Kottanerin, Denkwürdigkeiten (1445-1452), S. 23) ‘When it now became light, then we saw that the lords went to one another in a house.’ As can be seen, the complementizer (that in English, dass in German) is sometimes omitted in the construction, as in (13).6 This construction is limited to subject-oriented perception verbs. Type II: PV + WH-Complementizer Clause

(15) Thence I to Westminster-hall and there hear how they talk against the present management of things, and against Sir W. Coventry for his bringing in of new commanders and casting out the old

258

(16)

Richard Jason Whitt

seamen… (Helsinki Corpus: CEDIAR3, Samuel Pepys, The Diary of Samuel Pepys (1666-1667), P VIII, 317-318) hier sehen wir abermals, wie sich die Netzhaut durch eine Sukzession von Schwingungen gegen den gewaltsamen äußern Eindruck nach und nach wieder herstellt. (Goethe Corpus: Zur Farbenlehre - Didaktischer Teil (1793), Hamburger Ausgabe, Band 13, S. 338) ‘Here we see once again how the retina recuperates from a violent external impressions little by little through a succession of oscillations.’

Instead of that or dass, this construction features a wh-word as the complemtnzier of the clause for which the speaker indicates there is evidence.7 But as with Type I, the construction is also restricted to subject-oriented perception verbs. Type III: PV + Direct Object (DO) + Non-Finite Verb (NFV)

(17) And Iohn bare record saying, I saw the Spirit descending from heaven, like a Doue, and it abode vpon him. (Helsinki Corpus: CNTEST2, The New Testament (1611), I, 20) (18) Wir sehen auch diese Theorie in der Praxis hinreichend bestätigt. (Kant Corpus: AA VIII, Über den Gemeinspruch (1793), S. 303) ‘We even see this theory adequately confirmed in practice.’ Both English and German allow the use of infinitives and past participles as NFVs here, but only English allows the use of present participles to function as NFVs as well. Only subject-oriented perception verbs are found in this construction. Type IV: PV + Prepositional Phrase (PP)

(19) Well, Thomas, why? Make me understand - because I'll tell you now, from where I stand, this looks like cowardice! (ARCHER Corpus: ) (20) Oder wenn der Sohn, als der Vater das Zusammensein mit der blühenden Schwiegertochter etwas allzu begeistert schildert, dem Alten zuruft: „Du -! Das riecht nach Inzest! - Nun ja, dies ist eben ein Theaterskandal in Paris, und der Himmel weiß, daß wir an andere gewöhnt sind. (DWDS: Paul Block, „Das Grab unter dem



Evidentiality, polysemy, and the verbs of perception

259

Triumphbogen”, in: Berliner Tageblatt (Morgen-Ausgabe), S. 2-3, (11.03.1924)) ‘Or if the son, when the father somewhat all too surprised depicts the encounter with the blossoming daughter-in-law, calls out to his father: “You! That smells like incest! Well, this is just a theater scandal in Paris, and heaven knows that we’re used to others like this.”’ Generally, there are only a limited number of prepositions that can be used with any particular perception verb, and one only finds object-oriented perception verbs occurring in this construction. Type V: PV + Adjective (ADJ)

(21) She was looking remarkably well; her very regular, very pretty features, having the bloom and freshness of youth restored by the fine wind which had been blowing on her complexion, and by the animation of eye which it had also produced. (ARCHER Corpus: ) (22) Mutter Aurelie schlug mich immer auf die Finger, wenn ich nach der Karavine griff, der Vater sah so bös aus, ich dachte, er würde mich schlagen. (Goethe Corpus: Wilhelm Meisters Lehrjahre (1795-1796), Hamburger Ausgabe, Band 7, S. 604) ‘Mother Aurelie always hit me on the finger if I reached for the Karvine, the Father looked so angry that I thought he would hit me.’ As with the Type IV construction, this construction is limited to objectoriented perception verbs as well. Type VI: PV + Conjunction (CONJ) + Clause (C)

(23) Mary is a very warm-hearted, since [sic] girl, whom I have always liked, & I doubt not will make him an excellent wife… I feel as though I am very glad he is going to marry… (ARCHER Corpus: ) (24) wenn wir mit den Gütern in Ordnung sind, mußt du gleich mit nach Hause, denn es sieht doch aus, als wenn du mit einiger Vernunft in die menschlichen Unternehmungen eingreifen könntest. (Goethe

260

Richard Jason Whitt

Corpus: Wilhelm Meisters Lehrjahre (1795-1796), Hamburger Ausgabe, Band 7, S. 501) ‘After we have taken care of the properties, you must immediately come home with us, for it looks as if you could intervene in human affairs with some rationality.’ The conjunction involved here is usually a phrasal conjunction such as English as if or as though, or the German als ob or als wenn, although single-word conjunctions can appear as well. Similar to Types IV and V, construction Type VI is also limited to object-oriented perception verbs. Type VII: PV + (Infinitive Copula (IC)) + ADJ or Noun (N) or ADJ + N

(25) The Captain on deck and in a jolly mood pointed out to us an iceberg in the distance. It looked, through the glass, to be about 30 feet long and 50 or 60 feet high. (ARCHER Corpus: )  This construction is found only in English and only allows the use of object-oriented perception verbs. When only an adjectival complement is present, the use of the infinitive copula is obligatory (thus differentiating this construction from Type V). But if there is a nominal or an adjectival + nominal complement, then the use of the infinitive copula is optional. Type VIII: Parenthetical Construction

(26) Your sister-in-law is giving a luncheon to which I hear my parent & aunt are invited Saturday in honor of my cousin who is getting married. (ARCHER Corpus: ) (27) Löbell - wiederholt; Ich - halte diese Aktennotiz im vollsten Umfange – aufrecht und weise den unerhörten Vorwurf -, den Sie (zu Erzberger) hier am Sonnabend gemacht haben und den Sie, wie ich sehe, heute wiederholen, mit der allergrößten Entschiedenheit zurück. (DWDS: “Deutscher Reichstag”, in: Berliner Tageblatt (Morgen-Ausgabe) 05.03.1907, S. 20-21) ‘Löbell repeats; I stick to what’s in this memo and dismiss the unspoken allegation, which you (to Erzberger) made on Saturday and which you, as I see, repeat again today most decisively.’ 



Evidentiality, polysemy, and the verbs of perception

261

Unlike construction Types I through VII, there is no overarching structure to parenthetical constructions involving evidential verbs of perception, other than the fact that these parentheticals “interrupt” the linear order of the clause in which they occur.8 As far as the data in the corpora are concerned, only subject-oriented perception verbs occur in parenthetical constructions, although the use of object-oriented perception verbs in such constructions is not inconceivable. Type IX: Perception Verb External to the Clause

(28) That night late his body was found in a ditch, about a mile out of town, near St. Pancras church. His sword was thrust through him, but no blood was on his clothes or about him. His shoes were clean, his money was in his pocket: but nothing was about his neck, and a mark was all round it, an inch broad, which shewed he was strangled. His breast was likewise all over marked with bruises, and his neck was broken. All this I saw; for Lloyd and I went to view his body. (Helsinki Corpus: CEH- IST3A,Gilbert Burnet, Burnet’s History of My Own Time (Part I: The Reign of Charles II) (1640-1710), 1, II, p. 164) (29) Hieraus sehen wir: da der Körper sogar in der Natur nicht deswegen unelastisch sei, weil seine Theile eingedrückt werden, sondern nur deswegen, weil sie sich nicht mit eben dem Grade Kraft wieder herstellen, mit welchem sie eingedrückt worden. (Kant Corpus: AA I, Gedanken von der wahren Schätzung der lebendigen Kräfte und Beurtheilung der Beweise… (1747), S. 071) ‘From this we see: since the body is therefore not even inelastic in nature because its parts are pushed in, but rather only because they don’t re store themselves with the same force with which they were pushed in.’ Here, the evidential perception verb occurs outside of the clause for which the speaker indicates there is evidence, and this “evidentialized” clause tends to be indicated by some sort of phoric marker such as a pronoun or demonstrative. Other time, the relationship is more asyndetic. Only subject-oriented perception verbs were found to occur in such constructions.

262

Richard Jason Whitt

5. A Case Study: see and sehen Since the verbs of visual perception attest the highest degree of polysemy, they best illustrate how the polysemy discussed in Section 2.4 transfers into the evidential domain, and how certain evidential meanings tend to be bound to specific construction types. We will focus only on two subjectoriented perception verbs here: English see and German sehen.

5.1. Quantitative Results 5.1.1. English see The perception verb see is, by far, the verb of perception attested most often in the examined corpora, in terms of both variety of evidential meanings and overall occurrence. In the Helsinki Corpus (Early Modern section), there are 1,168 attestations of the verb see, 90 (7.7%) being evidential markers. In the ARCHER Corpus, see occurs a total of 3,119 times. And of these, 280 (9%) are evidential. Thus we can see that evidential see occurs at relatively the same frequency in both of the corpora, and hence throughout the entire Modern period of the English language. Table 1: Occurrences of evidential see in English language corpora CONSTRUCTION TYPE I with complementizer without complementizer seeing construction II III infinitive present participle past participle VIII IX TOTAL

Helsinki corpus 50 (55.6%) 10 16 24 2 (2.2%) 29 (32.2%) 17 6 6 5 (5.6%) 4 (4.4%) 90

ARCHER corpus 102 (36.4%) 42 40 20 17 (6.1%) 137 (48.9%) 34 73 30 12 (4.3%) 12 (4.3%) 280

Construction Types I and III are most prevalent in both corpora, although Type I occurs more often in the Helsinki Corpus, and Type III occurs more



Evidentiality, polysemy, and the verbs of perception

263

often in the ARCHER Corpus. Types II, VIII, and IX occur at a far lesser frequency in both corpora, but interestingly, parenthetical and external constructions occur at almost the same frequency in the Helsinki Corpus, and they do occur at the same frequency in the ARCHER Corpus. There is a greater disparity among the uses of Type II in both corpora, with the ARCHER Corpora showing 3.9% higher frequency of use.

5.1.2. German sehen Like its English counterpart, German sehen is the most widely attested perception verb in the examined corpora. There are 852 total attestations of sehen in the Bonner Corpus of Early New High German, 30 (3.5%) being evidential. In the Kant Corpus, 68 (3.9%) of the 1,756 cases of sehen are evidential. 3,555 attestations of sehen can be found in the Goethe Corpus, and of these, 274 (7.7%) express evidential meaning. Finally in the DWDS, there is a surprisingly low percentage of evidential attestations. Out of a random sample of 1,500 attestations, sehen was used evidentially in only 19 (1.3%) of these cases. There is no apparent reason why this should be the case, considering all the uses attested in earlier corpora can still be used on contemporary German. In any case, an explanation for this discrepancy lies beyond the confines of the present study. Table 2 presents us with the breakdown of construction types in the German language corpora. Table 2: Occurrences of evidential sehen in German language corpora CONSTRUCTION TYPE I with complementizer without compl. II III infinitive past participle V VIII IX TOTAL

BONNER CORPUS 18 (60%) 14 4 7 (23.3%) 4 (13.3%) 3 1 -- -- 1 (3.3%) 30

KANT GOETHE CORPUS CORPUS 39 (57.4%) 73 (26.6%) 38 61 1 12 2 (2.9%) 5 (1.8%) 20 (29.4%) 181 (66.1%) 4 113 16 68 1 (1.5%) -3 (4.4%) 13 (4.7%) 3 (4.4%) 2 (0.7%) 68 274

DWDS 9 (47.4%) 8 1 3 (15.8%) 4 (21.1%) 1 3 -2 (10.5%) 1 (5.3%) 19

264

Richard Jason Whitt

Similar to the English data, Construction Types I and III are most widely attested throughout the corpora, although – as with English – Type I enjoys greater frequency in some corpora (Bonner, Kant, DWDS), while Type III is more prevalent in others (Goethe). And similar to English, all other construction types occur at a much lesser frequency than Types I and III (although in the Bonner Corpus, Type II occurs more often than Type III). There is one anomaly in the German language corpora, particularly the Kant Corpus: the appearance of one instance of the Type V construction, which is generally restricted to object-oriented perception verbs. This will be addressed later (see Section 5.2.4).

5.2. Qualitative Analysis We now turn to an examination of the relevant construction types and the evidential meaning(s) conveyed within each of these types. In general, English see and German sehen display a remarkable similarity with on another in terms of what type(s) of evidence is/are expressed by each construction.9

5.2.1. Literal visual perception is one type of evidence that can be designated by see or sehen when it is followed by a finite complementizer clause: (30) (31)

We saw that the crocodile moved only his upper jaw. (ARCHER: ) nun sah ich erst, daß unten quervor ein ziemlich langes Bänkchen stand, worauf eine Mandoline lag. (Goethe Corpus: Dichtung und Wahrheit (1811-1833), Hamburger Ausgabe, Band 9, S. 58) ‘It was then that I saw that down below a rather long bench stood diagonally in front of me, on which a mandolin was lying.’

So in (31), for example, visual perception allows the speaker to pinpoint and indicate the location of the mandolin. More common in this construction type, however, is the indication by the speaker (or writer) that inference has driven him/her to arrive at a certain conclusion. This inference is often based on observation, which of course involves some degree of visual



Evidentiality, polysemy, and the verbs of perception

265

perception, so the literal meaning of see or sehen is not wholly absent when inference is being indicated: (32)

(33)

I see you are dying with curiosity to know what has excited my anger, which I consider both inquisitive and impertinent. (ARCHER: ) Wir sehen, daß die Lebenshaltung des Volkes eine bessere geworden ist, und der weitere Ausbau des Eisenbahnnetzes wird die wirtschaftliche Entwicklung auch ferner günstig beeinflussen. (DWDS: “Landtagsverhandlungen”, in: Vossische Zeitung (AbendAusgabe) (06.03.1907), S. 4) ‘We see that the people’s standard of living has become better, and moreover, the further expansion of the railway system will influence economic development positively.’

Neither the addressee’s curiosity in (32) nor the standard of living in (33) can be visually perceived directly, but some sort of action or state of affairs that is indicative of these conditions can be visually perceived, and it is this act of perception that drives the inference made in both these examples. Of a more abstract nature is the metaphorical use of see or sehen to indicate knowledge or understanding. In these cases, traces of literal perception are more difficult to pinpoint than in (32) or (33): (34)

(35)

I now see plainly that Men cannot arrive at a full Satisfaction by Riches, nor at Power by enjoying Principalities or Kingdoms, nor at Esteem and Reverence by the Accession of Dignities, nor at Nobility by Glory, nor at true Joy by carnal Pleasures. (Helsinki Corpus: CEBOETH3, Richard Preston (trans.), Boethius (1695), p. 124) Sulzers Theorie war mir wegen ihrer falschen Grundmaxime immer verhaßt, und nun sah ich, daß dieses Werk noch viel mehr enthielt, als die Leute brauchen. (Goethe Corpus: Italienische Reise (1786-1788), Hamburger Ausgabe, Band 11, S. 207) ‘Sulzer’s theory was always abhorrent to me because of its incorrect fundamental maxims, and now I saw that this work contains much more than the people need.’

Of course, for either Boethius or Goethe to have arrived at this understanding, they must have observed or witnessed (i.e. visually perceived) events that shaped their epistemology, so even here, one cannot wholly exclude

266

Richard Jason Whitt

literal perception. After all, it is the source domain of the SEEING IS KNOWING metaphor (Sweetser 1990: 32-34; cf. Lakoff and Johnson 2003), for one knows what one has seen. Although the complementizer is present in most of the above examples, sometimes – as in (32) – it can be absent. There have been suggestions that this may be an indication of grammaticalization (see Note 6), although as far as evidential meaning is concerned, the presence or absence of this complementizer does not appear to affect the type of meaning expressed in any perceivable way. There is, however, one constructional sub-type included here that follows the structure but behaves rather differently than any other evidential construction. This is the grammaticalized use of the English present participle seeing, in which this participle behaves more like a causal conjunction (e.g. because) than an actual perception verb because the speaker/writer takes whatever propositional content falls within the scope of seeing for granted, i.e. it belongs to the domain of knowledge; this, in turn, is then described as the cause or motivation for the apodosis: (36)

(37)

Than begane he to water hys plantes, sayenge unto me, “Syr, I thowghte once never to have seene yow agayne; yow are grownowte of my knolledge; and, seynge that yt ys the wyll of God that yow shold not dye by ther crwelty, I truste that your blud shal never be requyryd at my handys.” (Helsinki Cor-pus: CEAUTO1, Thomas Mowntayne, The Autobiography of Thomas Mowntayne (1500-1570), p. 213) Seeing that diplomatic relations with Japan have been broken off, you can prevent, if you consider it necessary, any Japanese descent on Korea. (ARCHER Corpus: )

Because seeing indicates that the speaker is in possession of the knowledge, i.e. evidence, that allows the content of the protasis to be a given, this type of construction is included in the category, even though it is notably different than uses not involving subjectless present participles. But like other cases of FCCs, the complementizer need not appear in the constructions. Whatever the implications of this omission may be (a higher degree of grammaticalization?) lie beyond the confines of this study.



Evidentiality, polysemy, and the verbs of perception

267

5.2.2. The same types of evidential meanings expressed in the construction are found in this construction as well. Only here, the exact nature of the proposition is often vaguer because the complementizer is a whword that lacks specificity. Consider (38) and (39), for example, where literal vision is indicated as evidence: (38)

(39)

We were there at ye time of ye fair & so saw what there chief merchantdise consisted of - namely, flax, hemp, sheep, horses, mules, & such like, & pistols also which are there made & sold in great quantities, it being a place for that sort of work ye most remarquable in France. (ARCHER Corpus: ) Ich habe gesehen, wie der Großfürst Sergius von Rußland getötet wurde… (“Feuilleton,” in: Vossische Zeitung (Abend-Ausgabe) 03.03.1905, S. 9) ‘I saw how the grand duke Sergius of Russia was killed . . .’

So in (39), although we know the writer saw that Sergius of Russia was killed, we do not know exactly how (wie) he was killed. The case is similar in (38), where we do not learn immediately what constituted the merchandise, although this information is provided immediately after the complementizer clause. The case is similar when a general sense of observation – which includes varying degrees of literal visual perception – serves as evidence: (40)

(41)

I was delighted to see how rapidly my proposition was accepted, and we made a hasty breakfast, first sending in some of our food to the other party. (ARCHER Corpus: ) Lasset uns bei diesem heiligen Kunstwerk, der Wohlthat, durch die unser Geschlecht ein Menschengeschlecht ward, mit dankbaren Blicken verweilen, mit Verwunderung, weil wir sehen, welche neue Organisation von Kräften in der aufrechten Gestalt der Menschheit anfange, und wie allein durch sie der Mensch ein Mensch ward! (Kant Corpus: AA VII, Recensionen von J.G.Herders Ideen zur Philosophie der Geschichte der Menschheit. Theil 1 & 2 (1785), S. 48) ‘Let us linger by this holy work of art with thankful looks and amazement, the boon through which our species became the human race, be cause we see which new organization of powers begins in

268

Richard Jason Whitt

the upright form of humanity, and how only through it the person became a human!’ So in (40), we know the writer’s proposition was accepted rapidly (but we don’t know how rapidly), and (41), we do not know which “new organization of powers” (neue Organisation von Kräften) or how through these powers “the person became a human” (der Mensch ein Mensch ward), but we do know that Kant claims to have observed these events occur; the whcomplementizers, which contain part of the propositional content, simply cause their respective clauses to lack specificity. The case is the same when see or sehen have metaphorical denotations of knowledge and understanding: (42)

(43)

When I was with him I was always puzzled and uneasy, and always wondering why on earth he had ever married Dottie or rather how Dottie had ever arranged it, but I could see why he liked the family publishing business. It was a sort of ivory tower for him - a plaything with which he could do what he wanted without worrying over the financial angle - and books with the Peale imprint always had artistic distinction. (ARCHER Corpus: ) vnnd damit wir widerumb sehen/wie von der Sichtbarn Catholischen Kirchen diß zu verstehen sey/zeugt der Heyl. (Bonner Corpus: Text 157, Johan Rosenthal, Außfuehrliche Widerhol- und Vermehrung der kuertzen Bedencken vom bestaendigen Baw auff den Felsen vnd nicht auff den Sand (1653), S. 15) ‘And the Lord testifies so that we see again how this is to be understood by the visible Catholic Church.’

In (42), the writer indicates that Henry Peale (the antecedent of he) liked the publishing business and that he understood this to be true. What we do not know is why this was the case, even though the author indicates he also understands the reason behind Peale’s affinity with publishing; he simply does not specify this fact. And in (43), exactly how (wie) the Lord’s testimony clarifies what is to be understood is not indicated within the complementizer clause; that information is left elsewhere in the discourse.



Evidentiality, polysemy, and the verbs of perception

269

5.2.3. If there is any construction in which literal vision will be indicated as evidence, it is in the construction, which almost solely is a marker of visual evidentiality when either see or sehen is the PV. Whether the NFV is an infinitive, a past participle, or a present participle (in English) makes no difference as to what type of evidential meaning is expressed; rather, the NFV is used to aspectually distinguish the action or state for which there is evidence. Consider the following: (44) (45)

(46)

(47)

(48)

I saw her pass through the room where we sat before dinner. (ARCHER Corpus: ) And Iohn bare record saying, I saw the Spirit descending from heaven, like a Doue, and it abode vpon him. (Helsinki Corpus: CNTEST2, The New Testament (1611), I, 20) Vnd do wart ich aber verzucket, vnd sihe bi mir ston einen gar grosen man… (Bonner Corpus: Text 231, Rulmann Merswin, Des Gottesfreundes im Oberland (1352), S. 11) ‘And then I was thrilled and saw a very large man standing next to me.’ Sdeath! I have heard Sense run down, and seen Idiotism, downright Idiotism triumph so often, that I cou'd almost think of Wit and Folly as Mr. Hobbes does of Moral Good and Evil, that there are no such Things. (ARCHER Corpus: ) Wir sehen eine neuerliche Steigerung der Provisionen (um rund 400000 M-) und der Einnahmen auf Zinsen- und Wechselkonto (um 1 313 000 M) hervortreten. (DWDS: [“Der Aufschwung des . . .”], in: Vossische Zeitung (Morgen-Ausgabe) (03.03.1912), S. 10) ‘We see a recent increase of provisions (about 400,000 M) and the revenues from interest- and acceptance accounts (at 1,313,000 M) emerging.’

In (44) through (46), vision provides the evidence for the propositions. But in English examples (44) and (45), a further aspectual distinction is made: the infinitive in (44) provides a non-progressive reading of the event, whereas the present participle in (45) mandates a progressive interpretation of the Spirit’s descent. In (46), the infinitive is the only possible verb form because the use of a present participle in such a context in German would be ungrammatical. So whether there is to be an additional aspectual reading (or whether such a distinction in German is relevant at all) is left to

270

Richard Jason Whitt

context and reader/hearer interpretation. In (47) and (48), there is not so much a sense of direct vision as there is of observation, although in both these instances, it should be clear that vision would form at least part of the basis of observation. Hence in (47), idiotism itself cannot be visually perceived, but acts of idiotism can, and it is these acts which the writer has seen/observed that draws him to the conclusion that idiotism triumphs “so often.” When a past participle appears as the NFV, two things are indicated: the resultant state of a previous action (which seems to be more in focus in (49) and (50)), or the passive framing of an event, i.e. the agent of action being omitted from the clause and patient of action appearing as the direct object instead (emphasized in (51) and (52)): (49)

(50)

(51)

(52)

Looking from where I now stood I saw on every side a thick wood pierced with avenues in a more or less perfect state. (ARCHER Corpus: 1795twin.j4b>) Als ich eines Tages die Straße betrat, sah ich an der nahen Station der Untergrundbahn eine Menschenmenge versammelt. (DWDS: [P.B.], “Das Jahr in Paris”, in: Berliner Tageblatt (AbendAusgabe) (03.03.1908), S. 1-3) ‘One day when I entered the street, I saw a crowd of people assembled at the nearby subway station.’ For when many things are spoken of before in scripture, whereof we see first one thing accomplished, and then another, and so a third, perceiue wee not plainely, that God doeth nothing else but lead vs along by the hand, til he haue setled vs vpon the rocke of an assured hope, that no one iote or title of his word shall passe till all be fulfilled? (Helsinki Corpus: CESERM2A, Richard Hooker, Two Sermons upon Part of S. Judes Epistle (1614), p. 8) Wir sehen auch diese Theorie in der Praxis hinreichend bestätigt. (Kant Corpus: AA VIII, Über den Gemeinspruch (1793), S. 303) ‘We even see this theory adequately confirmed in practice.’

The crowd of people are already assembled at the subway station in (50), and the writer indicates that he sees this current state which is the result of a previous assembling of people. In (52), on the other hand, the focus appears to be more on Kant’s observation that the theory has been confirmed, but he does not indicate exactly who is responsible for the theory’s confirmation. There is indeed a resultant state indicated here as well, but there is also an additional sense of passivity that is lacking in (50). The same can



Evidentiality, polysemy, and the verbs of perception

271

be said of (49) and (51): the focus of (49) appears to be more of a previous action resulting in avenues throughout the woods (although who is responsible for “piercing” these woods is not indicated), whereas in (51), the actual acts of accomplishments without regard to the agents of these acts are in focus. There is one other type of evidence indicated in the construction, although it occurs far less frequently than that of literal vision or general observation involving vision: the presence of an internal emotive state. Here, the speaker/writer indicates that they base their knowledge on some sort of internal compulsion rather than on external visual perception: (53)

…und je lebendiger irgend ein Wissen in uns wird, desto mehr sehen wir uns getrieben, es in seinem Zusammenhange aufwärts und abwärts zu verfolgen. (Goethe Corpus: Tag- und Jahreshefte (?), Hamburger Ausgabe, Band 10, S. 498) ‘and the more lively any knowledge becomes in us, the more we see our selves compelled to pursue it upwards and downwards in its coherency.’

Clearly the compulsion of knowledge to engage in some sort of intellectual activity is not something to be perceived in the realm of vision, so here, Goethe uses sehen to indicate that his perception of this compulsion is internal (intuitive or emotive) rather than external. 5.2.4. There is one instance in the Kant Corpus of sehen taking an adjectival complement, something generally reserved for object-oriented perception verbs. Here, the evidential meaning is similar to (53), in that emotions or intuition – rather than external vision – are indicated: (54)

Dadurch sehen wir uns in den geheimsten Beweggründen abhängig von der Regel des allgemeinen Willens, und es entspringt daraus in der Welt aller denkenden Naturen eine moralische Einheit und systematische Verfassung nach bloß geistigen Gesetzen. (Kant Corpus: AA II, Träume eines Geistersehers (1766), S. 335) ‘Thus we see ourselves in the most secret motivations dependent on the rule of the general will, and a moral unity and systematic

272

Richard Jason Whitt

constitution according to mere spiritual laws arises from this in the world of all thinking beings.’ Dependency, something obviously not detectable by vision, is the perceived state here, and Kant indicates via a metaphorical use of sehen that he considers himself to be in this state. His use of the plural pronoun wir indicates that Kant believes his readers to share his views and find themselves in this state of dependency as well.

5.2.5. When see and sehen occur in parenthetical constructions, they are never solely markers of direct visual perception. They can, however, indicate general observation, where vision may or may not be involved: (55)

(56)

They sometimes, I see, use the word captivate thus: “Five or six ships captivated,” “His whole army captivated.” (ARCHER Corpus: ) „was aber fangen wir mit Felix an?“ fragte Makarie, „welcher, wie ich sehe, mit der Betrachtung jener Bilder schon fertig ist und einige Ungeduld merken läßt“. (Goethe Corpus: Wilhelm Meisters Wanderjahre (1829), Hamburger Ausgabe, Band 8, S. 117) ‘“But what are we going to do with Felix?,” asked Makarie, “who, as I see, is already finished observing those pictures and is showing some impatience.”’

In (56), it does seem that visual perception informs Makarie’s observation of Felix’s behavior. In (55), vision is involved as well, insofar as the writer is indicating what he as read, an act that requires visual perception. There can also be an added connotation of inference besides the indication of observation: (57)

This course we see hath been very effectual in a short time, with some more ripe witted children, but othres of a slower apprehension (as the most and best commonly are) have been thus learning a whole year together… (Helsinki Corpus: CEEDUC3B, Charles Hoole, A New Discovery of the Old Art of Teaching Schoole (1660), p. 4)



Evidentiality, polysemy, and the verbs of perception

273

Indeed, to deem whether a certain method of instruction as been effectual requires both an act of observation and subsequent inference. Finally, parenthetical constructions are capable of indicating internal states of knowledge and understanding metaphorically: (58)

Und hierin hat also, wie wir sehen, die Mathematik einen Vorzug vor der Philosophie, daß die Erkenntnisse der erstern intuitive, die der letztern hingegen nur discursive Erkenntnisse sind. (Kant Corpus: AA IX, Logik (1800), S. 23) ‘And here mathematics has a priority over philosophy, as we see, for the insights of the former are intuitive, while those of the latter are, on the other hand, only discursive realizations.’

 Here, Kant is simply putting forward the argument that mathematics has a priority over philosophy, and he employs the parenthetical construction involving sehen to indicate that this argument should be self-evident (i.e. understood) and to point out the inherent nature of both disciplines (i.e. aspects belonging to the realm of knowledge).

5.2.6. When see or sehen occur external to the clause for which they indicate there is evidence, a variety of evidential meanings can be expressed, including direct visual perception: (59)

Oons, Madam, they're broke in to the House with Fire and Sword, I saw them, heard them, they'll be here this Minute. (Helsinki Corpus: CEPLAY3B, George Farquhar, The Beaux Stratagem (1707), p. 60)

 A general sense of observation – where visual perception plays a prominent role – can also be indicated: (60)

An inquisitiue man is a pratler: so vpon the like reason, a credulous man is a deceiuer: as we see it in fame, that hee that will easily beleeue rumors, will as easily augment rumors, and adde somewhat to them of his owne, which (Tacitus) wisely noteth, when he sayth: (Fingunt simul creduntq); so great an affinitie hath fiction and beleefe. (Helsinki Corpus: CEEDUC2B, Francis Bacon,

274

(61)

Richard Jason Whitt

The Twoo Bookes of the Proficience and Advancement of Learning (1605), 21V) es ist, was ich zum voraus wußte, nun aber mit Augen sah: er hat eben immer gemacht, was andere zu machen wünschten, und ich möchte jetzt nichts darüber sagen, als daß es von ihm ist. (Goethe Corpus: Italie- nische Reise (1786-1788), Hamburger Ausgabe, Band 11, S. 103) ‘It is what I already knew but now saw with my own eyes: he just always did what others wished to do, and because it is from him, I wouldn’t like to say anything about that now.’

 Visual perception is clearly the basis for observation in (61), and it can be argued to be present in (60) as well: whether fame proves that “an inquisitiue man is a pratler” or a “credulous man is a deceiuer” can only be determined from the acts that these men engage in, acts that are perceived primarily through vision. An additional reading of inference can be included when observation is indicated: (62)

(63)

I expect every hour the definitive sentence that must separate us for ever - I am sure you feel for me - I see you do - pardon these tears! (ARCHER Corpus: ) „Die ist gesund und munter. – “Ja, das sah ich, als du vorbeirudertest. (DWDS: Ottomar Enking, “Die Darnekower”, in: Berliner Tageblatt (Morgen-Ausgabe) (08.02.1906), S. 5) ‘“She is healthy and happy.” Yes, I saw that when you paddled by.’

 In both (62) and (63), it should be clear that visual perception forms the basis of the speaker’s inference: face-to-face contact in (62), and more general observation of personal demeanor in (63).

5.2.7. Summary of see and sehen We have seen that the subject-oriented verbs of visual perception in English and German can express a remarkable variety of evidential meaning ranging from literal vision to metaphorical extensions of knowledge, understanding, and even internal emotion. And these meanings tend to be bound to specific construction types. In both constructions involving complementizer clauses ( and ), for example, a



Evidentiality, polysemy, and the verbs of perception

275

great variety of evidential meaning involving both direct and indirect evidence was found: literal perception, general observation, inference based on observation, knowledge, and understanding. Direct vision was almost the sole type of evidential meaning expressed in the constructions, on the other hand. And in the lone construction found in the Kant Corpus, internal states of intuition and/or emotion were indicated. General observation (where literal vision might play a role) and metaphorical extensions of knowledge and understanding were indicated in parenthetical constructions, while literal vision always seemed to play a role in external constructions, whether a single act was witnessed or a general state of affairs was observed.

6. Conclusion I hope to have provided a general overview of how verbs of perception in English and German can express a variety of evidential meanings, a natural consequence of their inherent polysemy. It has also been shown that certain evidential meanings are restricted to specific construction types, although some types of evidential meaning do indeed appear in multiple constructions. Only the most salient similarities between English and German have been discussed here, and a number of issues are left unaddressed: how the object-oriented perception verbs also serve as evidential markers, the consequence of coupling a modal verb or cognitive verb (think, denken, etc.) with an evidential perception verb, the effects of negation, and the role of the other four sensory modalities could not be addressed here due to considerations of space. But at the least, it should be clear how fundamental perception is in shaping speaker epistemology, and it is also worth considering the role perception plays in markers of evidentiality that are not overtly indicators of sensory perception.

Notes 1.

Support for this research was made possible by the Fulbright Program and the UC Berkeley Graduate Division. In addition, I would like to thank Gabriele Diewald, Elena Smirnova, Thomas Shannon, Irmengard Rauch, and Eve Sweetser for assistance during various stages of this project. Any errors are, of course, mine alone.

276 2.

3.

4.

5.

6.

7. 8. 9.

Richard Jason Whitt All of the distinctions made in perception verb typology are based on the prototypical meaning of each perception verb and not on any possible metaphorical or metonymic extensions. The deictic nature of evidentiality is best summed up by Joseph (2003: 308): “[evidential] 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.” I wish to emphasize that this is a tendency of evidential perception verbs, not a hard and fast rule. I did find a few counter-examples to this in my corpus study, but these were far and few between, so this rule holds in virtually all cases. The GerManC (German Manchester Corpus) project attempts to fill this gap with representative data from 1650 through 1800. Currently, only excerpts of newspapers are available, but plans to expand this corpus to a large variety of text types are underway. Some, e.g. Thompson and Mulac 1991 and Brinton 1996, have suggested that the omission of the complementizer in similar constructions involving cognitive verbs (guess, think, etc.) in English is part of a grammaticalization process that results in epistemic parenthetical constructions. They base their argument mainly on semantic and pragmatic criteria of spoken data. Their arguments do not seem to hold for the verbs of perception encountered in this study, although admittedly, more data would be necessary before one could argue one way or another if their conclusions are also applicable to perception verbs. I would like to thank Maurice Vliegen for bringing this construction type to my attention. For more elaborate discussions of parenthetical constructions, see essays in Dehé and Kavalova 2007. I initially intended to provide quantitative data on the particular evidential meanings found within each construction type. However, such precise analysis proved impossible to apply consistently, for there are sometimes shades of multiple meanings signified in one singular instance. Therefore, I only provide an overview of the varying evidential meanings found in each construction type.

References Aikhenvald, Alexandra 2004 Evidentiality. Oxford: Oxford University Press. Brinton, Laurel J. 1996 Pragmatic Markers in English: Grammaticalization and Discourse Functions. Berlin/New York: Mouton de Gruyter.



Evidentiality, polysemy, and the verbs of perception

277

Chafe, Wallace 1986 Evidentiality in English conversation and academic writing. In Evidentiality: The Linguistic Coding of Epistemology, Wallace Chafe and Johanna Nichols (eds.), 261-272. Norwood, NJ: Ablex Publ. Dehé, Nicole, and Yordanka Kavalova (eds.) 2007 Parentheticals. Amsterdam/Philadelphia: John Benjamins. Diewald, Gabriele 2004 Faktizität und Evidentialität: Semantische Differenzierung bei den Modal- und Modalitätsverben im Deutschen In Tempus/Temporalität und Modus/Modalität im Sprachvergleich, Oddleif Leirbukt (ed.), 231-258. Tübingen: Stauffenburg. Does, Jaap van der, and Michiel van Lambalgen 2000 A logic of vision. Linguistics and Philosophy: 1-92. Gisborne, Nikolas 1998 The attributary structure, evidential meaning, and the semantics of English SOUND-class verbs. University College of London Working Papers in Linguistics 10: 1-25. Gisborne, Nikolas, and Jasper Holmes 2007 A history of English evidential verbs of appearance. English Language and Linguistics 11 (1): 1-29. Haan, Ferdinand de forth. Visual evidentiality and its origins. Diachronica. Harm, Volker 2000 Regularitäten des semantischen Wandels bei Wahrnehmungsverben des Deutschen. Stuttgart: Steiner. Jakobson, Roman 1971[1957] Shifters, Verbal Categories, and the Russian Verb. Roman Jakobson: Selected Writings. Vol. II: Word and Language. The Hague: Mouton de Gruyter, 130-147. Joseph, Brian D. 2003 Evidentials: Summation, questions, prospects. In Studies in Evidentiality. Alexandra Aikhenvald and R. M. W. Dixon (eds.), 307-327. Amsterdam/Philadelphia: John Benjamins. Lakoff, George, and Mark Johnson 2003 Metaphors We Live By. With a new Afterword. Chicago/London: The University of Chicago Press. Mortelmans, Tanja 2000 On the ‘evidential’ nature of the ‘epistemic’ use of the German modals müssen and sollen. Belgian Journal of Linguistics 14: 131-148. Peirce, Charles Sanders 1998 Essential Peirce: Selected Philosophical Writings. Vol. 2 (18931913). Eds. Nathan Houser and Christian J. W. Kloesel. Bloomington: Indiana University Press.

278

Richard Jason Whitt

Schröder, Jochen 1995 Zur Verben der visuellen Wahrnehmung (Kerngruppe). In Deutsch als Fremdsprache: An den Quellen eines Faches. Festschrift für Gerhard Helbig zum 65. Geburstag, Heidrun Popp (ed.), 317-325. München: Iudicum. Smirnova, Elena 2006 Die Entwicklung der Konstruktion würde + Infinitiv im Deutschen: Eine funktional-semantische Analyse unter besonderer Berücksichtigung sprachhistorischer Aspekte. Berlin/New York: de Gruyter. Sweetser, Eve 1990 From Etymology to Pragmatics: Metaphorical and Cultural Aspects of Semantic Structure. Cambridge: Cambridge University Press. Thompson, Sandra A., and Anthony Mulac 1991 A quantiative perspective on the grammaticization of epistemic parentheticals in English. In Approaches to Grammaticalization. Vol. II: Focus on Types of Grammatical Markers, Elizabeth Closs Traugott and Bernd Heine (eds.), 313-329. Amsterdam/Philadelphia: John Benjamins. Traugott, Elizabeth Closs 1989 On the rise of epistemic meaning in English: An example of subjectification in semantic change. Language 65: 31-55. 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, 185-210. Berlin/New York: de Gruyter. Viberg, Åke 1983 The verbs of perception: A typological study. Linguistics 21: 123162. Whitt, Richard J. 2008 Evidentiality and perception verbs in English and German: A corpusbased analysis from the Early Modern period to the present. Ph.D. diss., Department of German, The University of California at Berkeley.

Evidential markers in French scientific writing: the case of the French verb voir Francis Grossmann and Agnès Tutin The present paper describes a study of the evidential functions of the French verb voir (‘see’) in the scientific genre, based on a corpus compiled from writings in the fields of linguistics and economics. Our objective was to answer two questions: (a) What are the main functions of this marker in this genre and how are these functions made manifest (meaning and use of voir)? (b) Do different fields have specific ways of using evidential markers? A brief presentation of the evidential functions of the lexicon of perception is followed by a review of the literature on the evidential meaning of voir. In order to explore this linguistic phenomenon, we compiled a large corpus of scientific writings and annotated the occurrences of voir. This approach enabled us to address the linguistic properties of voir as a statement marker and to analyze its other evidential function, that of a reference marker. We also compared the evidential functions of voir in linguistics and economics writings using a corpus drawn from these two fields.

1. Introduction In evidential typologies, visual evidence1 is considered to be one of the prime sources of information. One somewhat paradoxical result of this is that certain languages with evidential systems do not have linguistic markers to indicate visual access to information – the default mode of access – whereas modes such as inference or hearsay are marked linguistically. Aikhenvald 2004 found that in languages such as French and English that do not possess evidential markers in the strict sense of the word, the notions of evidentiality are sometimes used abusively; that is to say, evidential markers are grammaticalized. Excluding the specific lexicon used in evidentiality research, she did however recognize evidential strategies that are used to add evidential semantic extensions to linguistic tools with other main functions (modal markers, reported speech markers, etc.). This restrictive vision can be nuanced by arguing that, in the same way that it is useful to study the lexical methods some languages use to express aspect (even if only to observe their interactions with the values of inflectional

280

Francis Grossmann and Agnès Tutin

classes), it may also be interesting to study how these languages mark evidentiality, including through lexical means; especially as languages with evidential systems have often derived these grammatical tools from lexical markers, such as the verbs of perception. A tendency to grammaticalize lexical markers can be seen in the French language’s lexicon of visual perception, particularly with respect to the use of vu ‘seen’. As early as the Middle Ages, vu was being used as a preposition with the meaning of eu égard à ‘considering’, and, in the 17th century, vu began to be used in legal documents to signify après avoir examine ‘after having examined’. In addition, the conjunctive locution vu que ‘given’ appeared in the 14th century and it is still used today in some areas. It is true that this tendency to grammaticalize certain uses derived from voir ‘see’2 does not necessarily mean these uses will become part of an evidential system, as this grammaticalization is limited, for example, when voir is employed as an auxiliary to partially erase its visual aspect. Our working hypothesis is that in certain genres in French the evidential characteristic of voir is usually expressed discursively. We postulate that is particularly true in the sciences, a genre that employs very specific forms in certain uses. As a result, we felt it would be interesting to examine differences in the discursive use of evidential markers in two fields: economics and linguistics. Finally, we wanted to show how difficult it is, in certain types of written discourse, to clearly differentiate between markers used to indicate visual sources, inferences or intellectual arguments based on statements made by other parties. After briefly explaining how voir functions in French and the framework for our study (Section 2), we present the method used to analyze our corpus (Section 3) and describe the two types of evidential usage of voir (Sections 3 and 4). We conclude by contrasting the distribution of these two types of usage in the fields of economics and linguistics.

2. Voir and the lexicon of perception in scientific genres 2.1. Perceptual paradigm Voir is used in a wide variety of forms in scientific writings. Table 1 lists some of these lexical forms. The lexical field of voir can be divided into “subjective” verbs of perception (the agent is the subject and the object “seen” is the direct object) and “objective” verbs for which the subject is the object “seen” (see Whitt 2008). However, this field also contains other syntactic categories, such as adverbs or adverbial expressions, e.g. appa-



Evidential markers in French scientific writing

281

remment ‘apparently’ and à première vue ‘at first sight’, prepositional forms, e.g. au vu de ces résultats ‘given these results’, nouns, e.g. observation ‘observation’, coup d’oeil ‘quick look’ and adjectives, e.g. visible ‘visible’, observable ‘observable’. Linguistic expressions within the lexical field of visual perception often have a hedge function, with certain verbs, e.g. apparaître ‘appear’ and adverbials, e.g. visiblement ‘visibly’, apparemment ‘apparently’ being used by authors to dilute their responsibility (NØlke 1994). However, these verbs and adverbials can also have a direct evidential function (Dendale 1994). As we will see below, this hedging role does not concern the verb voir or other expressions that have a more directly evidential function, such as au vu de ‘given’ and à la lumière de ‘in the light of’. The role of these lexical elements is more to show that a “fact speaks for itself”; even if the access to this factual knowledge is rarely through purely visible perception. Table 1: The lexical field of visual perception in scientific writings “Subjective” verbs voir ‘see’ observer ‘observe’ apercevoir, s’apercevoir ‘perceive’ regarder ‘look (at)’ discerner, détecter ‘discern, detect’ examiner ‘examine’ repérer ‘recognize’

“Objective” verbs apparaître/ la figure apparaît … ‘appear/the figure appears…’ sembler ‘seem’ révéler ‘show’ dévoiler ‘reveal’ se révèle Adj.) ‘reveal’ Adj

Adverbs visiblement ‘visibly’ apparemment ‘apparently’ à première vue ‘at first sight’ en apparence ‘apparently’

Prepositions au vu de (de ces résultats, etc.) ‘given these results’ à la lumière de ‘in the light of’

Nouns

Adjectives

observation (les premières observations montrent que ) ‘observation the first observations show’ coup d’œil ‘glimpse’

visible ‘visible’ discernable ‘discernable’ apparent ‘apparent’ observable ‘observable’

We should also mention the fact that certain lexemes in the field of voir are used to contrast appearance and reality, or, more exactly, to contrast a spontaneous but erroneous visual analysis and a deeper analysis that is contrary to appearances. A première vue ‘at first sight’ and apparemment ‘apparently’ are often accompanied by an adjective such as surprenant ‘surprising’, contre-intuitif ‘counter-intuitive’ or paradoxal ‘paradoxical’,

282

Francis Grossmann and Agnès Tutin

as in (1). Here, the analysis favoured by the author is introduced by en fait ‘in fact’3: (1)

En revanche, l’influence du revenu du conjoint est conforme à celle qui ressort de travaux analogues, mais à première vue plus surprenante. On s’attendrait à ce qu’un revenu du conjoint plus élevé décourage l’activité du fait de la progressivité du système fiscal: les prélèvements sur un même salaire féminin sont d’autant plus élevés que le salaire du mari est conséquent, puisque la tranche marginale d’imposition est plus grande, ce qui n’est pas le cas sur le tableau 1. En fait, le profil est perturbé par la forte homogamie sociale qui conduit les femmes diplômées à vivre avec des conjoints, eux aussi diplômés. ‘On the other hand, the influence of the partner’s revenue is similar to that found in other studies but, at first sight, more surprising. One would expect a higher partner revenue to discourage the activity because of the progressiveness of the tax system: the amount of tax paid on the wife’s salary is higher when the husbands salary is high, because the marginal tax burden is larger, which is not the case in table 1. In fact, the profile is skewed by the strength of social homogamy, which means that highly qualified women tend to live with highly qualified partners’. (KIAP economics corpus)

2.2. Linguistic functioning of the verb voir In his Dictionnaire historique de la langue française, Alain Rey (2004: 4107) notes that the Latin verb videre, from which voir is derived, comes from the Indo-European root °weid- which indicates vision, although primarily with respect to knowledge and only secondarily with respect to perception by sight. The Latin verb, like its French successor, whether it is used intransitively or with an accusative, signifies both perceived by sight, witness and notice4. This co-existence of the perceptive and intellectual meanings of voir is the source of its uses within the scientific genre and are the subject of the present study. Numerous studies of the different uses of voir have been carried out, either as part of more general studies of verbs of perception, or separately (see e.g. Bat-Zeev Shyldkrot, 1981, 1984, 1989, 1997, Picoche, 1986,



Evidential markers in French scientific writing

283

1993, Willems, 1983, 2000a, 2000b, Franckel and Lebaud, 1990, Labelle 1996, Naukkarinen, 1997, Leeman and Sakhokia Giraud 2001, Grezka, 2006). These studies have used different theoretical frameworks to investigate the principles explaining the numerous uses of voir or the semantic variety of these uses. Franckel and Lebaud 1990, who applied Culioli’s approach, showed the importance of the speaker’s stance in interpreting voir, a point we feel is particularly important. Grezka (2006: 61), who took an object-class approach, highlighted the importance of the second argument for voir, as the first argument is always an animate NP. Discussions have frequently looked at dichotomies, for example, between direct perception (=“physiological”) and indirect perception (=“cognitive”), or between active perception (the perceptive act is controlled by the subject) and passive perception (perception is an experience that the subject undergoes)5. The roles of the co-text and of complements have been invoked to explain the passage from a passive meaning to an active meaning for prototypical “passive” verbs of perception, such as voir and entendre. Grezka (2006: 53) noted that the co-text may govern the selection of either a passive verb of perception (voir, entendre) or an active verb of perception (regarder, écouter). For example, in order to specify an object that demands intellectual participation, an active form of voir is selected, as in As-tu vu un film hier soir ‘Did you watch a film yesterday evening?’. For many researchers, (e.g. Le Goffic 1993: 250, quoted in Grezca, 2006), when voir introduces a complement clause it takes on a more cognitive meaning and signals access to knowledge. This proposition is irrefutable, although it should be realized that in many cases the cognitive dimension of voir does not entirely override its perceptual dimension. In fact, access to a piece of information is often provided by a visual clue: (2)

Mais en regagnant sa chambre, elle a vu que la porte du balcon était entrouverte. ‘But, when she got back to her room she saw that the balcony door was ajar’. (B. Friot)

This example clearly shows that when voir acquires an evidential value (i.e. when it is used to indicate an observed fact), this always implies a double predicative dimension and therefore two processes: the process expressed by the subject of voir introduced by the observer and the process that translates the observed fact, which can be carried out syntactically in different ways. Without this double structure, voir would be a simple perceptual marker and would not have any evidential value. A sentence such

284

Francis Grossmann and Agnès Tutin

as je vois un arbre ‘I see a tree appears’ to be a perceptual statement of fact and not a way of indicating the source or status of the information. This explains why, syntactically, the use of an evidential often involves either completive or infinitive propositions that can be regulated by voir, or other means of syntactically signalling a predication. The literature shows that an infinitive proposition favours a perceptual interpretation, whereas a completive construction leads to a more cognitive interpretation. This can be illustrated by contrasting the following two sentences: (3)

Je vois Paul partir vs Je vois que Paul part. ‘I see Paul leave vs I see that Paul is leaving’.

Both cases contain a double predicative structure: the structure associated with the observer who “sees”, and the structure that corresponds to the observed fact, but the completive construction, which more strongly reinforces the two predications, seems to emphasize the observer’s point of view. Hence, the completive construction appears to be a better candidate for use as an evidential marker. However, other structures can also be used. Thus, in a sentence such as: (4)

… On voit ici l’importance de R. ‘We see here the importance of R.’ (Revue de statistique appliquée, 1962, vol. X, 1)

the evidential structure is expressed using an abstract predicative noun, rather than a completive. Our corpus study identified other syntactic constructions with evidential value in scientific writings.

2.3. The framework used to study voir in scientific genres Although our study is based on the above facts, our approach was much more specific, as our aim was not to explain all the uses of voir, or its polysemy; rather, it was to examine the uses of voir in scientific writings and to try and identify criteria that can be used to identify the evidential role of these uses. We postulate that voir, when it forms part of an evidential structure in a scientific text, can be subject to three independent levels of analysis: lexical-semantic, enunciative and rhetorical-pragmatic. Our lexical-semantic analysis used a similar approach to Fillmore’s frame semantics. We believe that the evidential use of voir differs from



Evidential markers in French scientific writing

285

other uses in the nature of the frame elements involved in the valence of voir. These roles can be identified through the syntactic constructions employed and through the semantic characterisation of the arguments of the verb6. The enunciative level constitutes the polyphonic space in which the scientific demonstration occurs and indicates how the semantic frames are manifested in the enunciative interplay, most notably through the personal pronouns7. There is a complex relationship between the enunciative interplay and the lexical semantic roles linked to the frame implied by the verb. As we will see with respect to voir, the figure of the scientific author8 is represented, in principle, by the personal pronouns Nous, On, and sometimes Je, which present the demonstration to a Model Reader. Although generally implicit, this Model Reader can appear in certain injunctive forms, for example, through the use of on (e.g. pour une synthèse, on se reportera à… ‘for a summary, see…’). Lastly, the rhetorical-pragmatic level takes into account how the lexicalsemantic level is inserted into the text. We postulate that the evidential use of voir involves cognitive/rhetorical operations linked to scientific argumentation and that it frequently introduces a process of validation, thereby entering the proof system of the scientific genre. As Aikhenvald 2004 pointed out, there is no direct relationship between evidentiality and the proof or validation system. The act of specifying the source of a piece of information (evidential role) is not necessarily intended to guarantee the validity of that information. However, in certain genres of discourse, specifying the source represents a validating factor that is used argumentatively by the locutor. This is the case in scientific writing, which demands that the nature of the sources of information be specified whenever new information is added9. Specifying both the source of a piece of information and the way that information was obtained (personal work by the author, reference to work by peers) is an important element in validating research work.

3. Methodology and corpus 3.1. The corpus Our study of voir was based on a corpus of scientific writings in French, taken from articles in respected journals, PhD theses, reports and course books, in two fields within the humanities: linguistics and economics. The journal articles were taken from the KIAP corpus, which was compiled by

286

Francis Grossmann and Agnès Tutin

Kjersti Fløttum and his team (see Fløttum et al. 2006) for a detailed study of auctorial presence. The items from the KIAP corpus were supplemented by items from PhD theses, reports and course books. Table 2 outlines the composition of the corpus, which contains just over 1.3 million words. Table 2. Corpus of scientific writings used in the study of the verb voir Scientific articles (corpus KIAP) Theses, reports, course books Total

Linguistics 285 881 words

Economics 374 516 words

364 812 words 650 693 words

286 653 words 661 169 words

The corpus used was sufficiently large and varied for our purposes, which were to identify the main semantic and pragmatic functions of the verb voir, and to determine its syntactic characteristics.

3.2. Productivity of the verb voir The verb voir is one of the most frequently occurring verbs in our corpus of scientific writings. Table 3. The most frequently occurring verbs in the corpus of scientific writings Verb Être ‘be’ Avoir ‘have’ Pouvoir ‘be able to’ Faire ‘do, make’ Permettre ‘allow’ Devoir ‘have to’ Mettre ‘put’ Considérer ‘consider’ Voir ‘see’ Montrer ‘show’ Apparaître ‘appear’ Prendre ‘take’ Donner ‘give’ Utiliser ‘use’ Correspondre ‘correspond’

Economics 14709 4494 2557 783 1100 732 608 656 519 637 286 426 426 541 389

Linguistics 15518 4703 3573 1400 864 638 680 618 734 379 706 556 536 414 556

Total 30227 9197 6130 2183 1964 1370 1288 1274 1253 1016 992 982 962 955 945



Evidential markers in French scientific writing

287

As can be seen from Table 3, for economics and linguistics combined, voir is the ninth most frequently used verb and the third “full” verb (after permettre ‘allow’ and considerer ‘consider’), if auxiliary verbs, supporting verbs and modal verbs are excluded. In linguistics, it is the second “full” verb, whereas in economics it occupies fifth place, behind permettre ‘allow’, considerer ‘consider’, montrer ‘show’ and utiliser ‘use’. Voir can therefore be considered representative of scientific writings. However, as we saw above, voir has a wide variety of uses that need to be described in more detail. The corpus was annotated for this purpose.

3.3. Annotating the different uses of voir in the corpus As was noted in Section 2, voir is a polysemic verb that is not easy to analyze and that has resisted even the best efforts to identify a common semantic core for all its different meanings (see Picoche 1986, 1993; Franckel and Lebaud 1990). For the present study we adopted a pragmatic approach, as our main objective was to isolate the evidential uses of voir. This led us to identify the main senses, which represent only a subset of the uses described in reference works (for example, in the Petit Robert dictionary or in the Trésor de la Langue Française). As far as possible, we associated the main senses with precise argumental structures, in order to facilitate their analysis and to annotate them in the corpus. Five main senses were identified in the corpus and the different occurrences of these senses in the corpus were labelled using XML annotation. Voir as a statement marker, evidential, with prototypical syntactic structure: NP (hum) voit (that Sentence) | NP (abs) This use, which is very common in scientific writings, can be considered evidential. The origin of a fact, here introduced in the completive (‘visage’ a un fort degré d’affinité avec la majorité de ces cliques ‘‘face’ has a high degree of affinity with the majority of these cliques’, becomes obvious to the author and the reader (on) through simple observation, often of a visual support (e.g. a table). (5)

On peut voir dans ce tableau que visage a un fort degré d’affinité avec la majorité de ces cliques. ‘The table shows that ‘face’ has a high degree of affinity with the majority of these cliques’). (Linguistics corpus).

288

Francis Grossmann and Agnès Tutin

This sense is similar to the one described by Whitt 2008 as, ‘an inference based on observation’. We have also used this label for slightly different senses, including the one labelled ‘knowledge and understanding’ by Whitt. In most cases, in a corpus, it is extremely difficult to separate these two uses10. Voir as a reference marker, evidential, with syntactic structure: Sentence (voir NP or voir Adv). Another important function of voir – which we believe is specific to scientific writings – is its use to indicate intratextual or intertextual references. In this context, voir, used in the infinitive, is often placed in parentheses and followed by one or more intra- or intertextual references. For example: (6)

(7)

Les auteurs qui traitent (...) d'à travers et au travers de abordent systématiquement ces prépositions à partir de leurs emplois spatiaux, de nature intrinsèquement dynamique (voir, par exemple, Borillo 1998: 85, Spang-Hanssen 1963: 231-233) ‘Authors who study “à travers” and “au travers de” systematically approach these propositions from a spatial perspective, which is intrinsically dynamic (see, for example, Borillo 1998: 85, SpangHanssen 1963: 231-233’) (Linguistics corpus). On vérifie, aisément, que la fonction valeur V(xt-1, Rt) est solution de l'équation de Bellman (voir annexe E) ‘It is easy to check that the value function V(xt-1, Rt) is the solution to the Bellman equation (see appendix E)’. (Economics corpus)

Section 3 presents a detailed analysis of how these two evidential values of voir function. Voir in the sense of ‘to examine’ (non evidential) with syntactic structure: NP (hum) voir NP (- hum). This sense, which is quite common in scientific writings, is not an evidential use, as the following examples show: (8)

Voyons les sèmes que le contexte du programme des Verts nous permet d’extraire à propos de ces deux courants synchroniques dans nos pays industrialises ‘Let us look at the semantic features that the context of the Green Party’s program allows us to extract with respect to these two synchronous currents in our industrial countries’. (Linguistics corpus)



(9)

Evidential markers in French scientific writing

289

Notre objectif est maintenant de déterminer les prix d’équilibre. Nous allons procéder en deux temps. Nous verrons tout d’abord l’équilibre instantané ‘Our objective now is to determine equilibrium prices. We will do this in two stages. First we will look at the instantaneous equilibrium.’(Economics corpus)

The element introduced by voir, an SN, is not a fact. It is often an object analyzed by the author during the demonstration or the study. Voir as an opinion and judgement verb (non evidential) with syntactic structures: NP(hum) voir NP(abs) as NP(abs) or NP(hum) voir NP(abs) PrepLoc NP(abs). In these senses (see examples below), voir can be considered a verb of opinion or judgement, and it does not have an evidential use. (10)

(11)

Les cotisations du système par répartition sont souvent vues comme des prélèvements obligatoires sans contrepartie ‘Contributions paid into a redistribution system are often seen as obligatory deductions that do not give any return’). (Economics corpus). ...l'on voit dans l'énonciation une sorte de prolongement du structuralisme… ‘…we can see in the utterance a sort of extension of structuralism’ (Linguistics corpus)

Other non-evidential uses were found in our corpus, but much less frequently. These include uses that are totally specific to the verb voir in its role as a semi-auxiliary of the passive, as in the following example: (12)

La théorie des opérations énonciatives se voit intimement associée à une définition du langage “comme une activité qui modifie une situation en faisant reconnaître à autrui une intention pragmatique” ‘The theory of enunciative operations is intimately associated with a definition of language “as an activity that modifies a situation by acknowledging the pragmatic intention of a co-locutor”’. (Linguistics corpus)

290

Francis Grossmann and Agnès Tutin

It is interesting to note that the corpus did not contain any examples of direct perceptual usage. Chafe (1986) recorded a similar finding for English.  3.4. Distribution of the uses of voir in the corpus By identifying values and annotating the corpus it was possible to determine the most common uses of voir in our corpus. As Figure 1 shows, evidential uses account for an overwhelming majority of occurrences (75%).





Figure 1. Distribution of the different uses of voir in the corpus of scientific writings (number of occurrences)

In addition, statement and reference markers show interesting syntactic and enunciative specificities that we will examine in the following two sections.

4. Voir as a statement marker In these uses, the origin of the stated fact is not directly perceptual because the author is presenting a proposition. As in the perceptual sense, the syntactic subject can be considered non-active as it receives the information, but does not really “create” it. Thus voir differs from verbs such as déduire ‘deduce’ or conclure ‘conclude’, which bring into play more complex mental operations. However, in some examples, voir is not only constative; it also includes a notion of demonstration. This is notably the case when the verb is used in the present perfect tense, as in the following example:



(13)

Evidential markers in French scientific writing

291

On a vu que couper correspond à des procès très différents dans ses différents emplois lexicaux. ‘We have seen that, in its different lexical uses, to cut refers to very different processes.’ (Linguistics corpus)

It is difficult to unequivocally interpret this example out of context, but it is probable that the author has established this fact, rather than merely stating it in an earlier section of the text describing several lexical uses of the verb couper ‘cut’. The use of voir, rather than montrer ‘show’ or établir ‘establish’, gives the text greater objectivity, in the true sense of the word: the facts are self evident and the responsibility of the author – which would appear greater if he/she had used terms such as on a montré ‘we have shown’, or on a établi ‘we have established’ – is pushed to the background. Moreover, in scientific writings, voir as a statement marker occurs in specific syntactic contexts and has specific enunciative properties.

4.1. The subject of voir as a statement marker: inclusion of the “reader as witness” The subject of voir is nearly always on (‘one’) or nous (‘we’) (90.5% of occurrences in our corpus11), as in the following example: (14)

Comme nous l'avons vu précédemment, la définition et le choix d'un concept pour l'équité posent de redoutables problèmes ‘As we saw earlier, the definition and the choice of a concept for equity is extremely problematical.’ (Economics corpus)

The personal pronouns nous and on are not used merely out of protocol. They cannot be replaced by a first person singular pronoun, as they can in the case of other verbs, such as montrer or démontrer: (15)

(16)

*J’ai vu précédemment que couper correspond à des procès très différents ‘I saw earlier that to cut corresponds to very different processes’ J’ai démontré/montré précédemment que couper correspond à des procès très différents... ‘I showed earlier that to cut corresponds to very different processes…’

292

Francis Grossmann and Agnès Tutin

The fact that it is impossible to replace nous/on by je shows that the author is not the originator of the fact described (the je is not an active agent), but that the facts are, to a certain extent, self-evident. Furthermore, nous and on in these cases are not exclusive, as they include the community of peers: (17)

On admet couramment en épistémologie qu’il n’existe pas de chose en soi que le scientifique pourrait appréhender sur le mode de l’observation pure. ‘It is widely accepted in epistemology that there are no things in and of themselves that a scientist can understand through pure observation.’ (Economics corpus)

In fact, when nous and on are used with voir to form a statement marker, they are inclusive (I + You) in that they include the reader, who is called upon to witness, alongside the author, the facts described12. Here, voir, like other verbs of the same paradigm, is clearly used in a dialogical context (as it is when it is used as a reference marker), where the author draws the reader’s attention to a group of facts that support the author’s reasoning. The inclusion of the reader in verbal forms is quite rare in scientific writings, with the exception of certain imperatives13, as indications that the reader is being taken into account generally take the form of markers that guide the reader, such as metatextual elements d’une part ‘on the one hand’, pour conclure ‘to conclude’, etc.

4.2. Voir as a parenthetical Another specific characteristic of voir as a statement marker is its use as a parenthetical (on l’a vu ‘we have seen’, on le verra ‘we will see’, comme nous le voyons ‘as we can see’, etc), as in the following examples: (18)

(19)

Les métaphores vives, on l’a vu, sont sensibles au contexte, et il est nécessaire de connaître les conditions d’énonciation pour pouvoir attribuer un sens précis à l’énoncé. ‘Living metaphors, as we have seen, are context dependent, and it is necessary to know the conditions of enunciation in order to be able to attribute a meaning to the utterance.’) (Linguistics corpus) Le facteur individuel incorpore en effet les orientations du couple en matière de descendance et, comme on l’a vu, celles-ci se



Evidential markers in French scientific writing

293

traduisent probablement dans les choix du ménage en matière d’habitation ou de biens durables ... ‘The individual factor incorporates the couple’s choices in terms of having children and, as we have seen, these choices are probably reflected in their choices in terms of accommodation or consumer durables…’ (Economics corpus) These uses account for more than 20% of the occurrences in our corpus, and seem to be characteristic of scientific writings. As Rooryck (2001) showed, parentheticals are used particularly frequently with verbs expressing evidentiality. On a communicative level, the secondary predication expressed by voir reinforces its evidential function: the author provides a reminder that the facts expressed by the primary predication are justified by an observation that is made in association with the reader. In addition, the use of parentheticals helps to maintain textual coherence and the complicity with the reader by announcing information to come or by recapitulating facts that have already been demonstrated and observed. These elements help guide the reader, alongside other intertextual marks that we will look at now.

4.3. Supporting a statement: locative complements Although voir is not used as a verb of perception in scientific writings, the visual dimension is not completely absent. The author uses the text, sometimes supported by graphs and tables, to ask the reader to accept certain facts. Textual support may be implicit, in that it does not include any specific markers, but it is often given explicitly through intertextual locative complements, such as adverbs of place ici ‘here’, plus haut ‘above’, cidessous ‘below’, supra ‘above’, infra ‘below’, etc. or prepositional phrases that indicate tables, graphs or other parts of the text to be examined (dans la section 1 ‘in section 1’, dans le tableau 2 ‘in table 2’, etc.). For example: (20)

(21)

On peut voir dans le tableau précédent que ce taux est dû essentielle-ment à la contribution des cliques… ‘The above table shows that this rate is mostly due to the contribution of the cliques…’ (Linguistics corpus) Comme nous l’avons vu dans le chapitre 2, la proximité géographique relève de la localisation dans l’espace des acteurs et de leurs liens en termes de distance …

294

Francis Grossmann and Agnès Tutin

‘As we saw in chapter 2, geographical proximity decribes the localization of the parties in space and their links in terms of distance…’ (Economics corpus) In a small number of cases the localization does not appear in a complementary clause, but in a parenthetical structure, which is a point of similarity between this use of voir and its use as a reference marker: (22)

Avant le XVIe siècle, on l’a vu (chapitre I, section 2.1.), le terme d’antonomase recouvre principalement l’antonomase du nom commun. ‘Before the 16th century, as we have seen (chapter I, section 2.1.), the term antonomasia mostly covered the antonomasia of the common noun…’ (Linguistics corpus)

Like parentheticals, intratextual locative markers improve textual coherence and allow the author to guide the reader through the demonstration. These elements help support the argument by indicating the place of observation.

4.4. Summary In scientific writings, voir (and its paradigm) as a statement marker can be described using Fillmore’s frames (2003). Table 4: Different uses of voir as a statement marker Examples On voit ainsi sur la Figure 4 qu’un changement de seuil peut faire basculer une configuration d’ambiguïté ‘Thus, we can see from Figure 4 that a change in threshold can clarify an ambiguous configuration’ Les métaphores vives, on l’a vu, sont sensibles au contexte, …

Witness On

Fact un changement de seuil peut faire basculer une configuration d’ambiguïté ‘a change in threshold can clarify an ambiguous configuration’

Localization sur la Figure 4

On

Les métaphores vives … sont sensibles au contexte, …

(not realized – the piece of scientific



Evidential markers in French scientific writing ‘Living metaphors, as we have seen, are dependent on context…’

‘Living metaphors…, are dependent on context…’

295

writing by default)

Voir brings into play the following frame elements: – A witness: this element is generally realized using an inclusive subject pronoun (nous or on) that encompasses the author(s) and the readers. – A fact: prototypically, the stated fact is expressed by a complement clause or an NP consisting of an abstract noun, but it can also appear in a principal in the case of parentheticals. – A localization: Implicitly, the localization of the statement is the piece of scientific writing. Sometimes, the localization is expressed explicitly using an intratextual locative (section, chapter, figure, linguistic example). In the following section, we show that this structure can also be applied to the use of voir as a reference marker.

5. Voir as a reference marker The second use of voir in scientific writings is as a reference marker14, an aspect that might seem out of place in a discussion of evidentiality. In fact, voir is commonly used in the infinitive as a synonym for cf (Latin confer = compare)15 to direct the reader to a source. This use, which involves a double predication, differs from the typical evidential structure (described above) in appearance only. In fact, the second predication is expressed using an imperative form of voir, as the main predication corresponds to an assertion made in a locution or a section of the text to which the reference is made. This indicating function is closely linked to the use of the infinitive with an imperative value, but voir falls into a frame that is very similar to the one described in the section on voir as a statement marker. The reference introduced by voir provides the source and validates the text preceding voir. The reader is encouraged to check, to “go and see”, the content of a source that supports the author’s affirmation. The source can be external (intertextual use of voir) or internal (intratextual use). Peritextual content (footnotes) and parentheticals are important but slightly different means of introducing elements of proof into the reasoning. More detailed studies are

296

Francis Grossmann and Agnès Tutin

needed to compare the uses of voir as a source indicator alongside other referencing modes. The introduction of the author-date system by scientific publications allows authors to refer to sources supporting affirmations by simply adding the relevant publication date in parentheses. However, this sometimes leads to difficulties in interpretation16, as the reference can be regarded either as a simple reference to an affirmation made by another author, or as an element that is considered to have been accepted by the scientific community. Similar ambiguities occur with the intertextual use of voir.

5.1. Voir as an intertextual reference marker When applied intertextually, voir is most commonly used to refer to other work that supports an affirmation: L’accroissement des inégalités face à l’emploi constitue l’un des traits majeurs de l’évolution du marché du travail dans les années 80. La progression de l’écart relatif des niveaux de chômage des actifs qualifiés et non qualifiés est une caractéristique particulière de ce phénomène inégalitaire (voir, par exemple, GLYN [1995] ‘The rise in inequalities with respect to employment is one of the main features of the evolution of the job market in recent years. The increase in the differences between the unemployment rates for qualified and unqualified workers is an important characteristic of this phenomenon of inequality (see, for example, GLYN [1995])’) (Economics corpus)

(23)

Here the validation is based on the knowledge accumulated by the scientific community in a certain field. In most of its uses, voir is accompanied by modifiers (in the above example, the modifier is par exemple ‘for example’ that reinforce the validity of the presented fact. For example, terms such as entre autres ‘among others’ or par exemple indicate that the fact presented by the author is so well known to the peer community that it is unnecessary to give a detailed list of references. The adverbs notamment ‘notably’ and en particulier ‘in particular’ are used to emphasise the importance of a specific contribution by highlighting the author’s expertise, which allows him/her to establish a hierarchy in the sources used. Intertextual reference markers, which indicate sources of evidence, belong to a secondary enunciative level. Two localizations account for an



Evidential markers in French scientific writing

297

overwhelming majority of occurrences: parentheticals (as for the use of voir as a statement marker) and, even more frequently, footnotes (Figure 2), although, as we will see below, there is a substantial difference in their frequencies of use in economics and linguistics writings. ϭϲϬ

ϭϯϲ

ϭϱϭ

ϭϰϬ ϭϮϬ ϭϬϬ ϴϬ ϲϬ ϰϬ ϮϬ

ϭϱ

Ϭ /ŶƉĂƌĞŶƚŚĞƐĞƐ

/ŶĨŽŽƚŶŽƚĞƐ

KƚŚĞƌ

 Figure 2. Occurrences of voir as a reference marker

A parenthetical statement marker using voir is placed immediately after the fact it supports. The sources given, which form the localization, affirm the stated fact. When voir is used in a footnote, it operates slightly differently. Like a parenthetical statement marker, a footnote number, placed at the end of the stated fact, may refer to a note providing an intertextual reference that supports the validity of the stated fact. However, some intertextual references introduced by voir also seem to have another, not entirely evidential function, in that they provide further information about a point that has not been sufficiently explained in the main text or they introduce complementary information, as in the following example. (24)

(In the body of the text) On ne saurait considérer comme relevant tout à fait du hasard le fait qu’à la fin du dix-neuvième siècle le philosophe Victor Egger consacre un livre à la parole intérieure, tandis que dans ses leçons sur l’aphasie Charcot met à l’ordre du jour la question du langage intérieur et que le roman contemporain promeut le monologue intérieur1 ‘We would not consider it to be pure coincidence that at the end of the 19th century the philosopher Victor Egger devoted an entire book to internal speech, while in his lectures on aphasia Charcot

298

Francis Grossmann and Agnès Tutin

raised the question of internal language and contemporary novels promoted inner dialogue1). (As a footnote) 1. Voir notamment à ce sujet la contribution de Christian Puech dans ce recueil ‘For more on this subject, see the contribution by Christian Puech in this volume’. (Linguistics corpus) Parenthetical reference markers can also be used in this way, but this is much rarer.

5.2. Voir as an intratextual reference marker When used intratextually, voir refers the reader to another part of the text: (25)

Cette distinction reprend deux des fonctions (excellence et circonstances, mais pas filiation) de l’antonomase antique (voir 1.2.1.) ‘This distinction takes into account two of the functions (excellence and circumstances, but not parentage) of ancient antonomasia (see 1.2.1)’ (Economics corpus)

Here, the locative source is the complement of voir. This obligatory complement is sometimes realised by a locative adverb that guides movement through the text (ci-dessous, ci-dessus, infra, supra17), as in the example above, or by a simple paragraph, section or part number. Unlike intertextual uses, the inclusion of voir (or cf.) is obligatory. Locative complements are comparable to those that appear with statement markers and, in general, these two uses appear very similar in all aspects other than their syntactic structures. Hence, the following statement marker can easily be transformed into a reference marker by making a few adjustments to the syntax. Use as a statement marker (original example): (26) Nous avons déjà vu à la fin de la Section 2 que l’interdiction de pratiquer la discrimination du troisième degré entre marchés à élasticités différentes a des conséquences ambiguës sur le bien-être. ‘We have already seen at the end of Section 2 that the ban on carrying out third degree discrimination between markets with different degrees of elasticity has ambiguous consequences on well being’). (Economics corpus)



Evidential markers in French scientific writing

299

Use as an intratextual reference marker (manufactured example): (27) L’interdiction de pratiquer la discrimination du troisième degré entre marchés à élasticités différentes a des conséquences ambiguës sur le bien-être (voir la fin de la Section 2). ‘The ban on carrying out third degree discrimination between markets with different degrees of elasticity has ambiguous consequences on well being (see end of Section 2)). When used as an intratextual marker, voir is often placed in parentheses but it is relatively rare for it to be used with footnotes, at least in economics texts. References to appendices containing documentary sources are also common: (28)

…On vérifie, aisément, que la fonction valeur V(xt-1, Rt) est solution de l’équation de Bellman (voir annexe E. ‘It is easy to check that the value function V(xt-1, Rt) is a solution for the Bellman equation (see appendix E)’. (Economics corpus)

Intratextual references can also be more complex, as in the case of syntactic units such as voir notre développement précédent sur… ‘see our previous discussion of…’ where the indicator refers to ideas developed in those sections of the text as well as to a localization.

5.3. Summary To conclude, most uses of voir as a reference marker are cognitively and rhetorically similar to its uses as a statement marker, hence, their structures can be described using the same frame. –

– –

The witness (the reader), who is asked to check the validity of sources, is expressed dialogically by the use of the imperative infinitive. The fact is the statement preceding the reference indicated by voir. The localization is the intra or intertextual object introduced by the marker voir. In this case, voir is obligatory, contrary to statement markers. The following examples illustrate this structure.

300

Francis Grossmann and Agnès Tutin

Table 5. Different uses of voir as a reference marker Examples

Witness

Fact

Localization

On vérifie, aisément, que la fonction valeur V(xt-1, Rt) est solution de l'équation de Bellman (voir annexe E) ‘It is easy to check that the value function V(xt-1, Rt) is a solution for the Bellman equation (see appendix E)’. (Economics) La progression de l'écart relatif des niveaux de chômage des actifs qualifiés et non qualifiés est une caractéristique particulière de ce phénomène inégalitaire (voir, par exemple, GLYN [1995]) ‘The increase in the differences between the unemployment rates for qualified and unqualified workers is an important characteristic of this phenomenon of inequality (see, for example, GLYN [1995] ‘)

= the receiver, who appears implicitly in the infinitive imperative form As above

On vérifie, aisément, que la fonction valeur V(xt-1, Rt) est solution de l'équation de Bellman ‘ It is easy to check that the value function V(xt-1, Rt) is a solution for the Bellman equation’

Appendix E

La progression de l'écart relatif des niveaux de chômage des actifs qualifiés et non qualifiés est une caractéristique particulière de ce phénomène inégalitaire ‘The increase in the differences between the unemployment rates for qualified and unqualified workers is an important characteristic of this phenomenon of inequality’.

GLYN [1995]

6. Comparison between economics and linguistics One of our working hypotheses was that the evidential function in our corpus involves cognitive operations that are related to scientific reasoning and textual rhetoric. As methods, types of scientific object examined and modes of reasoning differ between fields, we hypothesized that the evidential use of voir would also vary according to the field. Detailed examination of our corpora confirmed this hypothesis. First, voir occurred more frequently in our linguistics corpus (513 occ.) than in our economics corpus (420 occ.), for similar-sized corpora. Two main differences in the distributions of the uses of voir were also noted (Figure 3). Linguistics texts contain more statement markers, whereas economics texts contain more intertextual references. However, comparable numbers of intratextual markers are found in the two fields. These results can be



Evidential markers in French scientific writing

301

interpreted in a number of ways. The greater use of statement markers by linguists may be the result of a greater need to involve the reader in the demonstration. However, this hypothesis does not seem to be supported by the occurrence of the other markers of interaction with the reader that were identified by Fløttum et al. 2006 in the KIAP corpus18. For example, metatextual markers and imperatives occurred much more frequently in the economics corpus than in the linguistics corpus. Less importantly, the frequent use of statement markers by linguists probably results from the fact that linguistics authors often support their reasoning with examples in the text, whereas observations in economics are often based on linguistic objects, in particular numerical results.

ϱϵ͘ϱ ϰϰ

 Figure 3: Evidential uses of voir in economics and in linguistics (in %)

The greater use of intertextual reference markers in economics writings suggests that economists are more likely to refer to previously established knowledge in their field, whereas linguists tend to base their statements on observations contained within their own text. It would be tempting to think that economists have a more cumulative concept of science and a greater tendency to build on the work of their peers. However, the bibliographies of economics articles tend to be shorter than those of linguistics texts, at least in the case of research articles (see Fløttum et al. 2006: 219). Furthermore, the use of reference markers appears to be slightly different in the two fields: linguists favour parentheticals (60% of occurrences), whereas economists tend to use footnotes (66.5%), which often have other

302

Francis Grossmann and Agnès Tutin

functions that are not purely evidential. From an academic point of view, economists may have a greater need to found their work on previous research carried out within their discourse community. This hypothesis is corroborated by a study carried out on information source markers, which are much more explicit in economics than in linguistics (Garcia da Silva 2008). Nevertheless, to avoid hasty judgments and caricatural sociological interpretations, a more detailed and qualitative analysis is needed to confirm these hypotheses. Finally, although both fields appear to use comparable numbers of intratextual markers, they generally refer to different types of object. Most of the markers used in economics articles refer to graphs and tables summarizing data (Figure 1, Table 2, Graph 1), or to other sections of the text, either specifically (Section 1, Appendix D, Note 4) or generally, using “vague” adverbs of place (supra ‘above’, infra ‘below’, plus haut ‘above’, ci-dessous ‘below’, etc). In linguistics, references to tables and graphs are rare, as linguistics articles do not tend to contain such elements. On the other hand, linguistics articles do contain references to examples within the text, although cf. is more commonly used in these cases.

7. Conclusion The verb voir, analyzed here within a subset of scientific writings, is only one element, albeit a very productive one, in the evidential system used in this discursive genre. Hence, it would be valuable to integrate studies of voir, and of verbs such as observer ‘observe’, s’apercevoir ‘notice’ and examiner ‘examine’ into more general studies of the evidential role of markers referring to visual metaphors. Analysis of our corpus has revealed a recurring semantic structure (frame) containing three main elements: a witness, a fact and a localization. However, this structure also includes a certain amount of diversity, which can be described using three main criteria: ʊ

The degree of syntactic and semantic integration between voir and the stated fact. Integration is strong in completive structures indicating an observation (on a vu que ‘we have seen that’, etc.). In parentheticals the degree of integration is lower. There is also a third level, where voir appears in a separated structure within parentheses, in which the degree of integration is even lower. The syntactic and scriptural dis-



ʊ

ʊ

Evidential markers in French scientific writing

303

tance from the fact is even greater when voir is used peritextually, e.g., in note19; The type of localization: implicit or explicit, intratextual or intertextual. Implicit localizations (i.e., not referring to a precise location within the same text or in a precisely identified external source) and parentheticals without explicit localizations often reinforce the rhetorical or cohesive function. Localizations are always intratextual in the case of voir as a statement marker (On voit dans Duchmoll 2003 que… ‘It can be seen from Duchmoll 2003 that…’). In the case of voir as a reference marker, a localization must be given, as the localization associated with voir is an essential complement and not a modifier; The status of the witness, who may have a different role in scientific reasoning. The witness is first a Being in discourse, who is associated with the co-observation of the facts presented. The dialogical dimension is central and aims to involve the reader in the demonstration. Thus, the reader is called upon to witness, either through a statement with which s/he is associated by the inclusive personal pronoun nous, or through an explicit invitation. In this second function - apparently more modest but nevertheless fundamental - the witness seems to be a mere guarantor of the information given.

So what is the exact status, in evidential terms, of the verb voir? Our data support the idea of a very strong ambivalence between the verb’s perceptual meaning, which although secondary remains present, and its cognitive meaning, through which it indicates the inference. In examples such as en se reportant à la figure 3b, on voit que l'espace des phases est défini par deux variables ‘looking at figure 3b, we can see that the phase space is defined by two variables’ (Naukkarinen 1997: 123) voir is used as an inference marker. When used as an intratextual marker (as well as in numerous statement uses) voir allows an author to validate a statement by inviting the reader to look at data included in other parts of the text, thereby directing the “textual circulation. When used as an intertextual reference marker, voir directs the reader to supporting evidence provided by the writings of third parties, and thus indicates that the ‘proof’ is accepted by the relevant discourse community. In both uses, the reasoning role is inextricably entwined with the evidential function. Finally, our comparison between the fields of economics and linguistics supports our hypothesis that the traditions governing the use of evidential markers in scientific writings differ between fields, even for fields within the Humanities.

304

Francis Grossmann and Agnès Tutin

Notes 1. 2. 3. 4. 5. 6.

7.

8.

9. 10.

11.

12.

13.

Note that the terms evidence and evidentiality are derived from the Latin word videre, meaning to see. See Chocheyras 1968 on this subject. For more on the use of en fait as an evidential marker, see Grossmann and Wirth 2007. Leeman and Sakhokia Giraud (2001: 59) also highlight the intellectual and perceptive values of voir in French, linking them to etymology. For a summary, see in particular, Grezka 2006. We present these roles at the end of the corpus analysis. In Fillmore’s framework semantics, the predicative meaning is represented by prototypical situations or scenarios (frames), which can be divided into different levels of organization. These frames are seen as places of articulation between human experience and/or the cognitive dimension and the lexical coding. The semantic roles (or frame elements) correspond to the arguments of the predicates, verbal or nominal predicates (Fillmore et al. 2003). Fillmore 1982 identified two types of frame: cognitive frames and interactional frames. Interactional frames aim to represent the interaction between the locutor and the allocutor, or between the author and the reader. From this idea we have retained the need to bring together cognitive frames (and the elements associated with them) and the enunciative and textual perspective. In order to avoid any ambiguity, it must be remembered that the notion of author comprises several components: the author as an empirical being, the institutional author, as defined in law and, finally, the figure of the author as constructed at the enunciative and textual rhetoric levels. The latter is the component that interests us here. Chafe 1986 stressed the fact that certain well-established written genres have, over time, codified their use of very specific forms to mark evidentiality. For example, both values are possible in the phrase on voit bien dans ce schéma que ‘this diagram clearly shows: recognition is combined with a deduction’. In most other cases, when voir is used as a statement marker, it is used in an impersonal form: Il est surprenant de voir que ce type de compétence est également produit dans l’entreprise ‘It is surprising to see that this type of skill is also produced within the company’. (Economics corpus) In their study of the pronoun on in the KIAP corpus, Fløttum, Dahl, and Kinn 2006 also showed that the inclusive use of on is particularly linked to verbs of perception. The reader’s attention can be solicited through the use of certain first person plural imperatives, as well as through pronominal and verbal forms (see Fløttum et al. 2006), as in the following examples: Enfin, examinons une dernière donnée ‘Finally, let us look at one last piece of data’. (Linguistics cor-



14. 15.

16.

17. 18.

19. 20.

Evidential markers in French scientific writing

305

pus); Voyons brièvement quelques examples ‘Let us take a brief look at a few examples’. (Linguistics corpus) It is interesting to note here that the verbs in the examples also belong to the voir paradigm, even if they do not have an evidential function. The Framenet project has catalogued the equivalent usage of see in English, using a frame called Source_of_information, which is defined as: “In a text, a Source_of_information is given that provides a reader of the text with further Information relevant to the text. In this frame the author and reader are completely deprofiled, with the Source_of_information made salient”. Nevertheless, this deprofiling of the enunciative roles remains relative, as the role of the jussive infinitive can be interpreted as an invitation to the reader. Although the two forms are often presented as equivalents (PhD students are generally advised to choose one of the two systems), semantic motivation, which is slightly different, can explain some nuances of usage. Cf. (abbreviation of the Latin confer (‘compare’) can also be used to introduce other elements of the same order as those already introduced; whereas voir emphasizes more clearly the desire to document the source of an affirmation. In practice, however, the two markers are often used interchangeably. On this question, see Grossmann 2002. We did not find passim (çà et là ‘here and there’) in the corpus, which allows us, according to the TLFI, “to avoid mentioning the exact localization of passages with a bearing on a given subject when a reference is given”. Our corpus contains the French economics and linguistics articles from the KIAP corpus. Of course, this does not prevent the information provided in a footnote from being capital, as, according to tradition, the text provides the arguments but the footnote provides the proof by giving the sources and a commentary that makes the text clearer. For a history of footnotes, see Grafton 1997.

References Aikhenwald, Alexandra Y. 2004 Evidentiality. Oxford: Oxford University Press. Bat–Zeev Shyldkrot, Hava 1981 A propos de la forme passive se voir +Vinf. Folia linguistica XV (34): 387–407. 1984 La concurrence entre la proposition conjonctive et voir + proposition infinitive. The French Rewiew, LVIII (2): 202–215. 1989 Les verbes de perception: étude sémantique. In Actes du XVIIème Congrès International de Linguistique et Philologie Romanes, 282– 294. Tome 4. Tübingen: Max Niemeyer Verlag.

306

Francis Grossmann and Agnès Tutin

La grammaticalisation des auxiliaires; le cas de voir. Scolia 10: 205224. Chafe, Wallace 1986 Evidentiality in English conversation and academic writing. In Evidentiality: the Linguistic Coding of Epistemology, Wallace Chafe and Johanna Nichols (eds.), 261-272. Norwood-New Jersey: Ablex Publishing. Chafe, Wallace, and Johanna Nichols (eds.) 1986 Evidentiality: The linguistic Coding of Epistemology. Advances in Discourse Processes. Norwood-New Jersey: Ablex. Chocheyras, Jacques 1968 Un nouvel outil grammatical en français moderne: le verbe voir. Le Français Moderne 36 (3): 219-225. Dendale, Patrick 1994 Devoir épistémique, marqueur modal ou évidentiel? Langue Française 102: 24-40. Dendale, Patrick, and Liliane Tamsmowski (eds.) 1994 Les sources du savoir et leurs marques linguistiques. Langue Française 102. Fillmore, Charles J. 1982 Frame semantics. In Linguistics in the Morning Calm, Linguistic Society of Korea (ed), 111-137. Seoul: Hanshin Publishing Co. 1985 Frames and the semantics of understanding. Quaderni di Semantica 6 (2): 222-254. Fillmore, Charles J., Chris Johnson, and Miriam R.L. Petruck 2003 Background to FrameNet. In Special Issue of the International Journal of Lexicography on FrameNet and Frame Semantics, Thierry, Fontenelle (ed.), 235-250. Oxford: Oxford University Press. Fillmore, Charles, J., and Sue Atkins 1992 Towards a frame-based organization of the lexicon: The semantics of RIK and its neighbors. In Frames, Fields, and Contrast: New Essays in Semantics and Lexical Organization, Adrienne Lehrer, Eva Feder Kittay, Richard Lehrer (eds.), 75-102. Hillsdale: Lawrence Erlbaum Associates. Fløttum, Kjersti, Trine Dahl, and Torodd Kinn (eds.) 2006 Academic Voices – across languages and disciplines. Amsterdam/ Philadelphia: John Benjamins. Franckel, Jean-Jacques, and DanielLebaud 1990 Les figures du sujet, A propos des verbes de perception, sentiment, connaissance. Paris: Ophrys. Garcia da Silva, Pedro 2008 Etudes des marques de la filiation dans les écrits scientifiques. Mémoire de master. Grenoble: Université Stendhal, Grenoble 3. 1997



Evidential markers in French scientific writing

307

Grafton, Anthony 1997 The Footnote: A Curious History. Cambridge: Harvard Univ. Press. Grezka, Aude 2006 Etude du lexique de la perception: bilan et perspectives. Suvremena lingvistika (‘Contemporary Linguistics’) 61: 45-67. Grossmann, Francis 2002 Les modes de référence à autrui: l’exemple de la revue ‘Langages’. Faits de langue 19: 255-262. Grossmann, Francis, and FrançoiseWirth 2007 Marking Evidentiality in Scientific Papers: The Case of Expectation Markers. In Language and Discipline Perspectives on Academic Discourse, Kjersti Fløttum (ed), 202-218. Newcastle, UK: Cambridge Scholars Publishing. Labelle, Marie 1996 Remarques sur les verbes de perception et la sous catégorisation, Recherches linguistiques de Vincennes 25: 83-106. Leeman, Danielle, and Madona Sakhokia Giraud 2001 Point de vue culiolien sur le verbe voir dans “Les Verbes français”, Langue française 153: 58-73. Le Goffic, Pierre 1993 Grammaire de la phrase française. Paris: Hachette. Naukkarinen, Oili 1997 Etude de quelques verbes de perception du français. Helsinki: Yliopistopaino. Nølke, Henning 1994 La dilution linguistique des responsabilité. Essai de description linguistique des marqueurs évidentiels il semble que et il paraît que, Langue française 102: 84-94. Picoche, Jacqueline 1986 Structures sémantiques du lexique français. Paris: Nathan. 1993 Didactique du vocabulaire français. Paris: Nathan. Rey, Alain 2004 Reprint. Dictionnaire historique de la langue française, Paris, Dictionnaires Le Robert. Original edition, Paris: Dictionnaires Le Robert, 1992. Rooryck, Johan 2001 Evidentiality. Part 1. Glot International 5 (4): 125-133. Whitt, Richard, J. 2008 Evidentiality, polysemy, and perception verbs: a corpus-based analysis of English and German. Hand-out of the conference “The linguistic realization of evidentiality in European Languages”, Bamberg, 27–28 February 2008.

An interactional approach to epistemic and evidential adverbs in Spanish conversation Bert Cornillie

This paper deals with the discourse functions of epistemic and evidential adverbs in Spanish conversation. It is shown that these adverbs not only vary in mood and relative frequency, but also have different roles in the organization of the turntaking process. By doing so, they go beyond the traditional epistemic and evidential qualifications mentioned in the existing literature. A lo mejor and igual ‘perhaps’ are shown to invite the interlocutor(s) to confirm or reject the view of the state of affairs (or part of it) presented, that is, the adverbs play an important role in the turn-taking process. The epistemic adverb quizá, by contrast, lacks the interactional functions typical of a lo mejor. Finally, the analysis indicates that speakers use evidential adverbs such as evidentemente to retain the turn, i.e. to keep the word.

1. Introduction Epistemic and evidential adverbs are generally defined as invariable expressions that qualify (parts of) a proposition in terms of the degree of likelihood that the event expressed may be real (in the case of epistemic markers), or in terms of the reference to the information that leads to a proposition (in the case of evidentials). Although this certainly holds for the bulk of the adverbs in monologic registers such as fiction or non-fiction prose, a closer look at their use in a conversational context points to several other functions in addition to qualifying the proposition (cf. Clift 2006). This paper is concerned with disentangling the discourse functions of epistemic and evidential adverbs. Spanish has a lot of epistemic and evidential adverbs at its disposal. First, at least seven ‘adverbs’ express ‘maybe’ or ‘perhaps’: a lo mejor, igual, quizá, tal vez, acaso, posiblemente and probablemente. Some authors prefer to use the term ‘adverbial phrase (or, in Spanish, locución adverbial) for epistemic a lo mejor. Second, several adverbs (and adverbial phrases) express a specific mode of knowing: aparentemente (also al parecer, por lo que parece), evidentemente, obviamente, supuestamente, vi-

310

Bert Cornillie

siblemente (also por lo visto).1 Note that these adverbs are to be differentiated from discourse markers such as claro, por supuesto and desde luego meaning ‘of course’, which necesessarily interact with the preceding context and do not refer to a specific mode of knowing. In this paper, I will only account for the first four epistemic adverbs and the four evidential adverbs. The paper is structured as follows. I first present some preliminary remarks (Section 2) before presenting previous accounts of the adverbs and the central hypotheses of this paper (Section 3). Then we go on to discuss the data both from a quantitative and a qualitative point of view (Sections 4 and 5). In the final section I present a series of conclusions as well as prospects for further research.

2. Preliminary remarks The relation between epistemic modality and evidentiality is a heavily debated topic in the functional-linguistic literature (cf. Dendale and Tasmowski 2001; Plungian 2001; Cornillie 2007). For some authors, an epistemic value is necessarily implicated in the reference to knowledge encoded in evidential expressions, whereas epistemic expressions sometimes lack such a reference to knowledge (Plungian 2001). Other authors argue that epistemic modality and evidentiality are two different but related categories with specific functions (Nuyts 2008; Cornillie 2009). Chafe 1986 holds an alternative view in that he establishes a broad category of evidentials which includes both epistemic and purely evidential expressions. Although the relation between epistemic modality and evidentiality is important, in this paper I will focus on what epistemic and evidential adverbs do in discourse, rather than deal with the semantic differences between them. For the purpose of this paper, it is worth recalling Chafe’s 1986 comparison of spoken and written language as far as the use of evidentials is concerned. Chafe (1986: 262) mentioned that his samples of conversational English and academic writing “showed approximately the same proportion of evidential markers to the total number of words. […] Differences appeared, not so much in the frequency of evidentials overall, as in the frequency of specific kinds of evidentials”. Hence, some evidentials will more frequently appear in written communication, whereas other ones will show up more often in spoken interaction. If we find different adverbs in speech and writing, the question arises whether a cognitive-functional motivation



An interactional approach to epistemic and evidential adverbs

311

for the different use of the adverbs can be found in the differences between speaking and writing. Since the two registers differ in many respects, this may be the case. First, a writer usually has much more time to produce, i.e. edit, a text than a speaker can use for producing utterances in interaction. Writers can read their texts over and make local changes so as to have the best final result. Since authors often tend to present the most nuanced view, they have recourse to specific epistemic and evidential expressions involving specific modes of knowing. Register-related preferences may also give rise to interesting alternations with the same verb. In previous research I have accounted for (i) the tendency of using evidential auxiliaries in written Spanish (e.g. parecer ‘to seem’ + infinitive) to the detriment of other forms with a finite que-clause and for (ii) the fact that the auxiliary + infinitive does not show up in spoken language. The answer lies in the observation that the qualification expressed by the auxiliary is based on inferences from reasoning only, whereas the parece que-construction can also involve inferences from direct evidence (Cornillie 2007a,b). Secondly, speakers are in direct, face-to-face interaction with the interlocutors or co-participants. It can be expected that the involvement in a social activity causes speakers to pay more attention to speaker-hearer related experiences. As a matter of fact, this leads me to hypothesize that evidentials used in conversation refer more frequently to evidence shared with co-participants. The preceding observations entail to the following questions. (i) with regard to written language, does more time to deliberate and to edit mean more inferences on reasoning and less circumstantial inferences? (ii) with regard to spoken language, does direct experience in interaction lead to more circumstantial inferences? Due to space restrictions, in this paper I will only address the second question.

3. Previous accounts and hypotheses Since the publication of Chafe’s seminal paper on evidentials, back in 1986, the existing literature has not satisfactorily dealt with register-related topics in epistemic modality and evidentiality. Moreover, in-depth studies of modality in conversation are rare. In this respect, the field of Spanish linguistics is no exception. The New Reference Grammar of Modern Spanish mentions that a lo mejor is “confined to spoken language or informal styles” (Butt and Benjamin 1994: 244) but does not specify whether quizá or tal vez are more frequently used in one or the other register. Moreover,

312

Bert Cornillie

no specific interactional functions are attributed to the individual adverbs. The Gramática Descriptiva de la Lengua Española (1999) only dwells on the modal adverbs briefly, comparing them with other adverbs. No attention is paid to their use in day-to-day Spanish. Whereas for English conversation there is Kärkkäinen’s 2003 book on the expression of epistemic stance, the field of Spanish conversation studies does not have any publication that comes close to it. In line with Kärkkäinen 2003, Fernández Sanmartín 2006 presents a comprehensive inventory of the epistemic and the evidential expressions that appear in the informal conversations compiled by the Val.es.Co group. It is a first step towards a better understanding of the use of different expression types in Spanish conversation, although does not deal with the dialogic context in which the expressions are used. Hence, a clearly interactional approach to Spanish epistemic and evidential expressions is still needed. Now, if the interaction has not been a central topic, what are then the approaches adopted for the study of Spanish modal adverbs? Many authors discuss the mood alternation with some of the adverbs, although some authors observe that with epistemic adverbs a different mood selection does not alter the likelihood expressed (Branza and Delbecque 2008). Most grammars point out that a lo mejor and igual combine exclusively with the indicative mood whereas, when they precede the verb, tal vez and quizá usually combine with the subjunctive mood. Yet, no satisfactory account is given for this mood distribution. In his Gramática Comunicativa del español, Matte Bon (2000: 257-258) distinguishes between the subjunctive and the indicative in terms of the “hipótesis remáticas” (new information) and “hipótesis temáticas” (given information) the adverbs introduce. The adverbs quizá(s) and tal vez are said to qualify given information, whereas a lo mejor would only introduce rhematic hypothesis or new information. In (1), Speaker B takes the turn from speaker A and attempts to explain the situation given by speaker A. (1) A. Nada... Imposible: está comunicando sin parar. (Matte Bon 1999: 257) ‘nothing… Impossible: they do not stop talking’ B. A lo mejor tienen el teléfono estropeado. ‘Perhaps their telephone is broken’ When speakers use quizá and tal vez with a finite verb in the subjunctive mood, as in (2), they formulate a hypothesis based on thematic or given



An interactional approach to epistemic and evidential adverbs

313

information. Hence, the information was previously mentioned to the speaker but was not communicated to the co-participant before. (2)

A. ¿Sabes? Quizá cambie de trabajo. (Matte Bon 1999:258) ‘You know (already)? Maybe (s)he changes job’

Quizá(s) and tal vez can combine with a rhematic hypothesis, but, according to Matte Bon, do never have the same rhematic force as a lo mejor. By using this adverb, the speaker “gives the impression of presenting hypotheses that come up at the moment of formulating them” (Matte Bon 1999: 258).2 As it will be shown in Section 5, a lo mejor does more: the speaker not only presents a hypothesis but also attempts to know what the coparticipants think about it. Moreover, a lo mejor functions in (1) as an epistemic marker with an inferential dimension, somewhat similar to inferential deber/must. By contrast, quizá in (2) is not used with an interactional purpose, nor does it have an inferential dimension. Unlike the epsitemic markers, evidential adverbs do not witness mood variation (cf Haßler 2004). The indicative is the standard mood with these expressions. Since this paper combines the study of epistemic and evidential adverbs, it is worthwhile replacing a lo mejor and quizá by the evidential adverbs under examination here. In (3) a lo mejor can be readily substituted by aparentemente ‘apparently’, obviamente ‘obviously’ and evidentemente ‘evidently’, but conflicts with supuestamente ‘supposedly’. (3) A. Nada... Imposible: está comunicando sin parar. ‘nothing… Impossible: they do not stop talking’ B. ……………… tienen el teléfono estropeado. ‘……………….... their telephone is broken’ - Aparentemente (inferential - circumstantial) - ??Supuestamente (reported - knowledge) - Obviamente (inferential - knowledge) - Evidentemente (inferential - knowledge) The fact that aparentemente fits in the open slot in (3) can be explained by the new situation the speakers are confronted with. Speaker A and Speaker B did not expect to be unable to call somebody. The inference is made on the basis of their observation that they cannot call. When obviamente and evidentemente are used, Speaker B refers to a situation which (s)he has already experienced on another occasion or which (s)he is familiar with through other means. Supuestamente, by contrast, excludes personal expe-

314

Bert Cornillie

rience and reports the observation made by someone else. It is clear that such a reading clashes with the pragmatics of the context. In sum, the test with the four evidential adverbs points to variation in the mode of knowing, which can be inferential or reported, and the type of evidence, which can be circumstantial, i.e. direct evidence, or knowledge, i.e. indirect evidence. In (4), none of the four evidential adverbs fits in the slot occupied by the epistemic adverb quizá. The first person singular of the verb cambiar may also clash with the evidential qualification following the question ¿sabes? ‘You know?’ (4) A. ¿Sabes? …………. cambia de trabajo. ‘You know?..............(s)he changes job’ - ??Aparentemente (reported - circumstantial) - ??Supuestamente (reported - knowledge) - *Obviamente (inferential - knowledge) - *Evidentemente (inferential - knowledge) The test in (4) shows two things: (i) the discourse context of epistemic marker without an inferential dimension does not allow for evidential expressions. It tells us that evidentiality and epistemic modality are two different functional categories; (ii) there is a conflict between introducing new information to co-participants and using adverbs that hint at knowledge shared with them. With regard to (ii), there is a slight difference between aparentemente-supuestamente and obviamente-evidentemente. The former can be used if the speaker frames the information as previously unknown to him/her, whereas the latter do not allow for such a reading. Whereas the mood alternation in terms of information structure is quite convincing, an interactional approach will have to disentangle the discourse function of adverbs in a conversational context. In a dialogic setting, information structure can be broadenend with the criterion of shared or personal information. Hence, the difference between the subjunctive and the indicative mood should not only be seen in terms of theme-rheme but also be examined from the point of view of the speaker’s on-line planning to ensure a reaction or a response from the co-participants (Kärkkäinen 2003: 83). The two following hypotheses can be derived from the preceding observations: Hypothesis 1: the epistemic adverbs a lo mejor and igual present a hypothesis to be confirmed or refuted by the interlocutor, whereas quizá and tal vez stick to the speaker’s epistemic qualification of the proposition ra-



An interactional approach to epistemic and evidential adverbs

315

ther than engaging in interactional strategies. This opposition could be seen as a functional elaboration of the theme-rheme opposition. Hypothesis 2: the evidential adverbs aparentemente and supuestamente emphasize the evidence, but differ as to the type of conclusion (new or reported) they can encode. The former adverb is expected to perform both readings, whereas the latter only can be reported. Obviamente and evidentemente, by contrast, focus the attention on the interlocutor or coparticipant. Not the new or the reported conclusion but the on-line planning (reasoning) behind is emphasized. Before I go on to test and subsequently verify these hypotheses in the qualitative analysis in Section 5, I will first present a quantitative overview of the relative frequency of the adverbs in different corpora.

4. A quantitative view of the epistemic and evidential adverbs In this section I present figures that illustrate the frequency of the adverbs in different corpora. In order to roughly distinguish between registers, we will start with the Corpus del español, which is a large on-line corpus containing more Latin American than peninsular Spanish. Its section of contemporary Spanish consists of the three subsections: spoken language, literature and texts. Table 1. Frequency distribution of the epistemic adverbs % Spoken % Literature % Text %

a lo mejor 1030 31.94 443 13.98 33 3.10

quizá(s) 1004 31.13 967 30.52 622 58.35

tal vez 1191 36.93 1758 55.49 411 38.56

3225 100 3168 100 1066 100

Table 1 starts from the following question: which of the three adverbs is preferred in the spoken register, in literature and in non-fiction prose, respectively? Note that igual is not taken into account due to technical problems.3 It can be observed that both in fiction (literature) and in non fiction prose (text) there is a clear preference for one of the three adverbs. Fiction authors seem to have more often recourse to tal vez (55.49%), whereas, in non-fiction prose, quizá(s) is most frequently used (58.35%). In the latter

316

Bert Cornillie

case, a lo mejor can hardly be considered a real option (3.10%). Surprisingly, in the spoken subcorpus the three epistemic adverbs are equally frequent, with tal vez being slightly more frequent. Table 2. Frequency distribution of the epistemic adverbs a lo mejor % Quizá(s) % tal vez %

Text 33 2.19 622 23.99 411 12.23

Literature 443 29.42 967 37.29 1758 52.32

spoken 1030 68.39 1004 38.72 1191 35.45

1506 100 2593 100 3360 100

Since the register corpora all comprise approximatively 5.150.000 words, we can also look at how an adverb is distributed over the various registers. The figures in Table 2 allow us to check the degree of entrenchment of an adverb in a certain register. For instance, whereas Table 1 shows more or less the same frequency in the spoken subcorpus, Table 2 indicates that only a lo mejor has a prototypical use in spoken language. Tal vez, by contrast, confirms its high frequency in literature observed in Table 1. Finally, despite the fact that quizá(s) is the speaker’s first choice in texts, the adverb is more often found in literature and in spoken language. Let us now examine along the same lines the frequency distribution of the four evidential adverbs under examination. First of all, it should be observed that these adverbs are far less frequent than the epistemic ones. Table 3 shows that the opposition between aparentemente – supuestamente and evidentemente – obviamente is reflected in their respective frequency in non-fiction prose. In fiction prose, aparentemente and evidentemente are the preferred evidential adverbs. Finally, evidentemente is by far the most frequent evidential adverb in spoken Spanish. Table 3. Frequency distribution of the evidential adverbs Press literature Spoken

Aparentemente 174 33.14 157 48.01 111 11.81

Supuestamente 162 30.86 35 10.70 60 6.38

Evidentemente Obviamente 97 92 18.48 17.52 99 36 30.28 11.01 596 173 63.40 18.40



An interactional approach to epistemic and evidential adverbs

317

The frequency distribution in Table 4 confirms the tendencies observed in Table 3: aparentemente and supuestamente are more frequent in the written register, whereas evidentemente and obviamente are much more frequent in the spoken corpus. Table 4. Frequency distribution of the evidential adverbs Aparentemente % Supuestamente % Evidentemente % Obviamente %

Press 174 39.37 162 63.04 97 12.25 92 30.56

Literature 157 35.52 35 13.62 99 12.50 36 11.96

Spoken 111 25.11 60 23.35 596 75.25 173 57.48

Total 442 100 257 100 792 100 301 100

 Given that the written genres differ among them, it may be expected that considerable variation can be witnessed within the spoken genre. Indeed, the broad term “Spoken language” covers a whole range of spoken subregisters such as political debates, telephone calls, interviews and conversations. If we really want to study interactional contexts, we have to base our analysis upon purely conversational data. As for spoken Spanish, there are at least two high quality conversational corpora from Spain, respectively from the University of Valencia and the Autonomous University of Madrid. The so-called Corpus del Habla Culta is a third option, albeit less suitable for interactional analysis because of the interview format. Interviews cannot be considered an equivalent of spontaneous speech. – – –

Corpus de conversaciones coloquiales. Val.es.Co (Valencia). +/- 100.000 words Corpus oral del castellano. Conversation section. UAM (Castilia). 252.349 words Corpus del habla culta (Castilia: 138.539 words/America: 841.644 words)

Let us look at the frequency distribution of the epistemic verbs in these different corpora. Several observations can be made: (i) the informal Val.es.co. corpus has the highest percentage of a lo mejor; (ii) igual is the second most frequent epistemic adverb in informal peninsular Spanish; (iii)

318

Bert Cornillie

quizá(s) is far more frequent in the peninsular part of the Habla Culta Corpus; (iv) the American part of the Habla Culta Corpus has the highest percentage of tal vez, which is almost absent in the corpora of peninsular Spanish. Table 5. Epistemic adverbs in different corpora Val.es.Co (Valencia) +/-100.000 words % UAM (Castilia) 252.349 words % Habla Culta (Castilia) 138.539 words % Habla Culta (América) 841.644 words %

a lo mejor

Quizá(s)

tal vez

Igual

Total

93 79.49

4 3.42

0 -

20 17.09

117 100

131 69.31

14 7.41

1 0.53

43 22.75

189 100

57 50

56 49.12

1 0.88

0

114 100

116 25.84

166 36.97

167 37.19

0 -

449 100

Table 6. Evidential adverbs in different corpora

Val.es.co (Valencia) +/-100.000 words % UAM (Castilia) 252.349 words % Habla Culta (Castilia) 138.539 words % Habla Culta(América) 841.644 words %

aparentemente

supuestamente

evidentemente

obviamente

0 -

0 -

3 100

0 -

3 100

2 8.70

0 -

19 82.61

2 8.70

23 100

0 -

0 -

2 100

0 -

2 100

20 25.32

10 12.66

41 51.90

8 10.13

79 100

Table 6 shows that also in conversation evidential adverbs are less frequent than the epistemic ones. In all of the four corpora, evidentemente is the



An interactional approach to epistemic and evidential adverbs

319

most frequent evidential adverb. Somewhat surprisingly, several adverbs do not show up in the peninsular conversation corpora, whereas they do in the Latin American interview corpus. In sum, a lo mejor is by far the most common epistemic adverb in informal and interactional contexts, igual being the second option. As for the evidential adverbs, evidentemente is clearly the preferred expression. It is striking that several adverbs are really marginal in conversation. The contrast with their frequency in interviews corroborates the assumption that interviews differ from spontaneous talk-in-interaction. There is the guidance by the interviewer, and the interviewee usually attempts to elaborate larger sequences of discourse. In other words, the interviewee is keen to retain the turn.

5. A qualitative view of the adverbs In this section we look at the discourse functions of the epistemic and evidential adverbs attested in the Val.es.Co corpus. I will focus on the contrast between a lo mejor and quizá, and also discuss the case of evidentemente. The aim is to present a functional account of the differences in frequency distribution. The main claim is that people use most those adverbs that serve interactional strategies best (a free interpretation of DuBois 1985). The discourse contexts that surround a lo mejor allow us to differentiate between its various discourse functions. The analysis below will indicate that the adverbs under examination do much more than conveying epistemic readings. That is, they go beyond the traditional focus on the evaluation of the likelihood. Table 7. Different readings of a lo mejor Discourse functions Alignment (forward-type) Alignment (backward-type) Attenuation Situational Purely epistemic Intentional Approximative

Examples 54 9 8 7 6 5 4 93

% 58.06 9.68 8.60 7.53 6.45 5.38 4.30 100

320

Bert Cornillie

From Table 7 we clearly observe that with a lo mejor speakers combine epistemic qualifications and alignment strategies with the co-participant. Example (5) illustrates such a context. (5) 1 2 3 4 5 6 7 8 9 10 11

Fragment 1 (Val.es.Co 2002: 117 – lines 1456-1457): G: la primeraĻ quitar el contacto ‘the first [question] take off the contact’ E: la primeraĻ quitar el contacto en teoría//[o sea pues] ‘the first take off the contact in theory so well’ L: [pues/no lo entiendo] ‘well I don’t understand’ G: [no no no no no o s(e)a] ‘No no no no no actually’ en teoría y en práctica//o sea§ ‘in theory and in practice actually’ E: [pero a lo mejor no es lo primero que haces] ‘but perhaps this is not the first thing you do’ L: [((a lo mejor explota))] ‘perhaps it explodes’ G: ¡normal! [que puede explotar el cocheĻ puede arderĻ=] ‘normal the car may explode may burn’ Ǽ: [sí/llevas razón tú] ‘yes you are right’ G: = si no quitas al contacto la batería estáa– está funcionando§ ‘if you don’t take off the contact the battery is is working’ E: ((y va y hace pumba)) And goes and does ‘boing’

In (5) speaker E suggests a hypothesis to speaker G which (s)he wants to be taken into account. Interestingly, the epistemic value does not correspond to doubt about the proposition that follows, but should rather be seen as a reply to speaker G. Furthermore, Speaker L elaborates the hypothesis of speaker E and directs it to speaker G, who finally takes the turn back and answers affirmatively. Hence, the two suggestions introduced by a lo mejor provoke the co-participant (i.e. speaker G) to take the turn. As a matter of fact, using a lo mejor can be considered a discourse strategy of the speaker to achieve alignment with the co-participant. This phenomenon can be termed forward-type of alignment (DuBois 2007). Let us now discuss the other functions of a lo mejor. In addition to a forward-type of alignment, the adverb also serves to confirm alignment that



An interactional approach to epistemic and evidential adverbs

321

is already achieved in the previous turn. This use is called the backwardtype of alignment. (6) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

Fragment 2 (Val.es.Co 2002: 100-101) Ǽ: § [pero yo por ejemplo] tengo problemas de cultura/ ‘but I for example I have cultural problems’ que sí que entiendo que– no debes ser así pero lo soy/ ‘which oh yes I understand you shouldn’t be so but I am’ no sé por qué/yo por ejemplo– a mí me repugna laa ‘I don’t know why I for example I disgust’ homosexualidad/yo estoy totalmente en contra// ‘homosexuality I am completely against’ [pero no (( ))=] ‘but not’ L: [¿pero por qué?] ‘But why’ E: = no lo sé/no sé por qué/nunca lo he descubierto// ‘I don’t know I don’t know why I never discovered it’ de verdadĻ hay ahí una diferencia entre la mente de ((un ‘really there is a difference between the mind of a’ sexólogo)) y de un homosexualĹ///hombre/ ‘sexologist and a homosexual you know’ casos de homosexualidad concretos no conozco ninguno ‘concrete cases of homosexuality I don’t know any’ va(le)– yo qué séĻ los aparentes eso que– uno que– ‘okay – dunno those who manifest the thing that’ entonces[de tío a tío eso a míĺ] ‘well from guy to guy for me’ G: [eso– e– e– es por eso] ‘This is it is therefore’ porque no has conocío ningún caso§ ‘because you haven’t known any case’ E: es por eso a lo mejorĻ ‘it is because of this perhaps’ porque imagínate que si yo tuvieraĺ§ ‘because imagine that if I had’ L: si estuvieras con un tío de esosŧ ‘if you were with such a guy’ E: yo creo que ahora– a lo mejor si co– si I think that now perhaps if I if

322

19 20 21 22 23 24

25 26 27 28 29 30 31 32 33

Bert Cornillie

[a lo mejor conozco=] ‘perhaps I know’ L: [igual no me aaa] ‘perhaps it doesn’t --- me’ E: = (( )) y él– a lo mejor digo ¡joder!//pero de verdad– ‘and he perhaps I say waw/jeezes but really’ pero yo de– de entrada lo rechazo§ ‘but I from the beginning I reject it’ G: § yo cuando– cuando ibaa§ ‘I when when I went’ E: § yo qué séĻ me da mucho§ ‘dunno it causes me a lot’ G: § cuando iba al institutoĹ̓ yo ‘when I went to the secondary school, I’ [tenía unn– un amigo=] ‘I had a friend’ E: [o sea/no sé] ‘well dunno’ G: = que era– que me llevaba cantidad de bien con él ¿no? ‘who was, whith whom I was on very good terms, you know’ íbamos siempre tres juntos ¿no? y uno de ellos eraa//uno ‘the three of us were always together and one of them was one’ de los otros dos era marica//maricón§ ‘one of the two others was a fag... gay’ E: § ¿pero lo reconocía él como tal?§ ‘but did he recognize it as such’ G: § lo reconocía él/peroo él no se comportaba– no se ‘he recognized it but he did not behave’ comportaba con– con los amigos que tenía dee– ni se pasaba ‘he did not behave with the friends he had nor did he” ni se [comportaba mal=] ‘nor did he misbehave ‘ E: [ya ya ya] ‘I see I see’

In line 15 of example (6), speaker E confirms speaker G’s hypothesis cautiously by means of a lo mejor. The primary focus is on the alignment between speakers, though an attenuative dimension can also be observed. Interestingly, there are contexts in which we find attenuation without clear alignment strategy. In line 18, for example, a lo mejor precedes the condi-



An interactional approach to epistemic and evidential adverbs

323

tional clause introduced by the si conjunction, and thus weakens or attenuates the condition. Speaker E wants to keep her freedom and, hence, does not want to be held responsible for the consequences of her conditional reasoning. This use is in line with other face-saving strategies. In line 18, the repeated a lo mejor is part of the conditional protasis. In this context the adverb has a situational use, in that it has scope over a potential situation sketched in this first part of the conditional. In the main clause (apodosis) we also find a lo mejor. Since in this context the adverb combines with a verb in first person, it points to the speaker’s intentions, hence the term intentional use. A lo mejor is followed by is a longer pause which indicates that the co-participant can take the turn. Since (s)he does not take it, the speaker goes on. Both the situational and the intentional dimension focus on the co-participant in that the speaker expressed that (s)he takes into account the co-participant’s opinion and feelings. An approximative use of a lo mejor before numbers is also observed in the corpus, as exemplified in (7). (7) 1 2 3 4 5 6 7 8 9 10

Fragment 3 (Val.es.Co 2002:128 - line 231) A: [pero e–] pero es que//es que es el gasto que lleva ‘but but you know the expenses which brings about’ el ascensor///es la cantidad de luz quee/gasta§ ‘the elevator is the amount of electricity that it consumes’ M: § sí yes A: luego el engrase/luego las ave[rías] ‘then the grease and then the damages’ M: [no/eso es] una cosa que se tiene ‘right, this is a thing that one has’ que– que– que/alimentarĹ/se tiene que– que/que cuidarĹ ‘to to to nourish one has to to to care’ /claro eso es [(( ))] ‘of course that’s right’ A: [pero si] ahora pagamos mil pesetas de luz de ‘but if we now pay thousand pesetas for the electricity of’ escalera luego tendrían que ser a lo mejor DIEZ MIL/// ‘the stairs later on it must be perhaps ten thousand’ [e– es un gasto enorme] ‘this is an enormous cost’

324

Bert Cornillie

Finally, note that the examples of pure epistemic modality lack any interactional dimension. The function of a lo mejor in these contexts is the judgment of the likelihood. On the other hand, uses as in (8) could also be seen as part of a personal dialogue. Evidence for this can be found in the verb decir ‘say’. (8) 1 2 3 4 5 6 7

8 9

Fragment 4 (Val.es.Co 2002: 83- line 48) G: hombreĻ […] he llegao y he llamao all– ahí al– al veinticinco ‘Listen I arrived and knocked a…there…. at the 25’ no– no abría nadie ¿no? al veinticinco§ ‘No nobody opened..no. .. at 25’ L: § ¿pero sabías que era el veinticinco? ‘But did you know that it was the 25’ G: sí/yo sabía que eraaĹ/este pisoĹ… yy bueno pues… por lo que ‘Yes, I knew that it was.. this apartment and well from what I’ me acuerdo yo de orientación y tal/sabía que más o menos ‘I remember about the direction and so I knew more or less that’ era… aquí ¿no?… y he llamao y como no abría nadie yo digo ‘It was here I knocked but as nobody opened I say’ a lo mejor no es aquí.. y he llamao ahí al la(d)o y tampoco estaban ‘perhaps it is not here and I rang at the house next to, and there weren’t at home either’ E: ¿sí?/ ‘Oh yes?’ G: pues vaya ‘A pity’

Let us now turn to quizá(s). For reasons of space, the examples of quizá(s) will not receive the same attention as those of a lo mejor. One clear difference we can mention from the outset is that quizá does not show up at the end of the turn, that is, it does not invite the co-participant to take the turn. This adverb seems to serve to modalize the speaker’s own discourse without aiming at interventions of the co-participant. Hence, quizá is not primarily directed at alignment, as shown in (9). (9) B: A:

Fragment 5 (Val.es.Co 2002: 337- line 63) § ¿pero podéis ampliar la memoria realχ en corto plazo? ‘But can you amplify the real memory in short term’ yo lo llevo pidiendo hace un año [(RISAS) (( ))] I am asking it since a year [laughter]’



B:

A: B:

An interactional approach to epistemic and evidential adverbs

325

[(( ))] bueno/pues a ver si podemoos///hacer algo aquí/por lo menos transitoriamente/¿no? ‘Well, well let’s see if we can do something here at least as a transition /okay?’ paraa/poder resolver un poco el programa porque quizáa/si estos ee– trabajos de Teseó corto no se les hiciera ‘to be able to solve a bit the program because maybe if to these eh’ ‘works of Teseó corto weren’t applied’ suapaut//pues noo– no habría tanto problema ¿no? ‘swap-out well no -- there wouldn’t be such a problem, right?’ por ejemplo ‘For example’ o sea que la degradación es demasiado fuerte debido a esto ¿no? aa– a que§ ‘Let’s say that the degradation is too strong due to that, isn’t ah that’

As example (10) confirms, quizá appears in longer turns than a lo mejor. By means of this adverb the speaker suggests an idea which (s)he can refute or attenuate subsequently. Not quizá(s) but other elements serve to organize turn completion: intonation breaks, exclamative utterances and tag questions give the co-participant the chance to intervene. (10) D:

A: D: A:

Fragment 6 (Val.es.Co 2002: 363- line 534) = tu familia/dado que tú eres experta en esos temas les irá todo muy bien ¿no? ‘your family /given that you are an expert in these topics they will do well, won’t they’ bueno/mira [verás– verás la cuestión familiar=] ‘Well /look you see – you see the family question’ [¿o a la familia no se puede enseñar?] ‘Or can’t the family be taught?’ = no/o sea/por ejemplo yoχ lo que está claro y hoy lo comentaba con otra amigaχ yo no tengo que convencer a nadieχ/o sea/el quee– yo lo que puedo explicar es algoχ y la persona pues recapacita y si le vaχ oo parece– le parece posible yy aceptableχ pues vale/entonces tengo en mi casa doss chicoss de ingeniería/con lo cual estos son más escépticos ¿no? o sea decir/tal/PERO por ejemplo pues/quizá porque me ven y soy su madre

326

B: C:

Bert Cornillie

‘more sceptical aren’t they?that say like BUT for example well maybe because they see me and I am their mother’ pues entranω no sé cómo decirte ¿no? mi hija no… mi hija sí (( )) ‘they come in I don’t know how to tell you, you know my girl don’t you.. my girl yes’ tranquila es– es mujer y entonces esa cosa pues le apetece/y mi marido me respeta un montónχ/se le ponen así unos ojitos viendo la cantidad de personas que vienen y luego le dicen ¡hay que ver tu mujer! y aquel diciendo ¡vaya tela! o sea [(RISAS)] [(RISAS)] [(RISAS)] y yo sin saberlo

Finally, we will examine the evidential adverbs in their interactional contexts. Despite the fact that they are very infrequent in the Val.es.Co-corpus, the three examples of evidentemente give us a good idea of how they function in conversation. By means of evidentemente the speaker is clearly engaged in keeping alignment with the co-participant: the speaker indicates that (s)he is aware of the co-participants potential objections. Thus, the adverb preempts the co-participant’s next turn. The two examples of evidentemente in example (11) illustrate this pattern. Interestingly, the intonation break that follows the adverb is a starting point for elaborating the main argument with which the speaker tries to convince the co-participant. Hence, the break not only does not lead to turntaking, the reverse is true: it gives the speaker the occasion to keep the turn. (11) A: B: A

C: A: B:

Fragment 7 (Val.es.Co 2002:128 - line 486) § es una decisión de administración/pura y simple ‘it is a administrative decision pure and simple’ pero no está basada en ningún estudio previoo§ ‘but is not based in any previous study’ § noω evidentemente/sí que está basado en una situación anterior donde solamente había uno [y entonces=] ‘No obviously (yes) it is based in a situation in which there was only one and then’ [y entonces] ‘and then’ = había un colapso total ‘There was a collapse’ mm



A:

An interactional approach to epistemic and evidential adverbs

327

entonces el problema quee nosotros hemos detectado/es un problema que hemos detectao evidentemente/ ‘Hence the problem that we have detected ouservelves is a problem that we have detected obviously’ entonces//lo que hicimos fue/ee/de uno pasar a tres/pero es un problema de capacidad y de prestaciones… no sé si me explico/ normalmente/si lo hiCIEses por prestaciones no tendrías tres/ tendrías cincuenta… porque el acceso es totalmente aleatorio… entonces claro/¡pf!//me explico ¿no?

Evidentemente contrasts with both a lo mejor and quizá. This evidential adverb differs from a lo mejor in that it is part of the speaker’s strategy to keep the turn, and it differs from quizá because of establishing mental contact with the co-participants. In other words, by means of evidentemente the speaker lets the co-participant know that (s)he is still developing his/her idea. After this qualitative analysis we now come back to our initial question of Section 2 concerning the type of inferences. It turns out that the evidential adverb mainly involves inferences from reasoning. Yet, the reasoning is not so much triggered by sensorial, visual or auditive stimuli but is part of the discourse planning. It can be argued that there is a semantic shift from a purely evidential reading (i.e. something that can be clearly perceived) to a discourse related reading (i.e. something that is clear according to discourse patterns). Interestingly, the other evidential adverbs do not readily allow for such a reading. It is now clear that in conversation adverbs do not readily express evidential qualifications that are based on circumstantial inferences. This confirms that conversation is very much focused on the attitudes of speech participants rather than on integrating new knowledge from the immediate physical context. Now, given the previous observations, what can we tell about the epistemic adverbs? As shown in Section (2), a lo mejor- igual are suitable for expressing an inferential dimension, whereas quizá and tal vez do not usually invite for such a reading. The fact thatthe former are more frequent in conversation than the latter leads us to conclude that the addition of a (circumstantial) inference to an epistemic evaluation facilitates the interaction. The reason why epistemic adverbs experience a shift to an evidential dimension, whereas evidential adverbs shift away from it is a topic for further research.

328

Bert Cornillie

6. Conclusions This paper has shown that epistemic and evidential adverbs deserve to be examined in terms of their discourse functions in conversation. The relative frequency in different corpora has turned out to be indicative for these functions. The quantitative corpus analysis has shown that a lo mejor is the preferred epistemic adverb in spoken language, in general, and in conversation, in particular. In (Latin-American) interviews (Corpus del Habla culta), we observed the relatively frequency of quizá(s) and tal vez. The evidential adverbs aparentemente and supuestamente contrast with evidentemente and obviamente in that the former are far less frequent in spoken language than the latter. In conversational Spanish a lo mejor reveals to have several functions other than being an epistemic qualification of the proposition. The organization of alignment appears to be the most prominent one in the Val.es.Co corpus. A lo mejor is not only used to introduce hypotheses with new information, but also invites the interlocutor(s) to confirm or reject the view of the state of affairs (or part of it) presented, that is, the adverb plays a role in the turntaking process. The relative frequency of igual goes along the same line. Thus, the above-presented analysis gives a functional explanation for the claim that with a lo mejor (and igual) the indicative mood correlates with new hypotheses. Interestingly, the corpus data of quizá contrast with those of a lo mejor, not so much because of the thematic hypotheses, but rather because of the longer turns in which quizá appears. The marginal frequency of quizá in conversational Spanish illustrates that it is less interactional than a lo mejor. The relative frequency of quizá(s) in interviews is due to the longer turns, in which the speaker does not invite for turntaking. In sum, we believe that the analysis presented in this paper demostrates that an interactional approach can account for the different mood and frequency distribution of the epistemic and evidential adverbs.

Notes 1.

In recent publications more attention has been paid to adverbial phrases such as por lo visto ‘apparently’ and al parecer ‘seemingly’ (González 2006; Fant 2005, 2006).



An interactional approach to epistemic and evidential adverbs

2.

In Spanish: “a lo mejor da la sensación […] de presentar hipótesis que se le ocurren al hablante en el momento mismo de formularlas.” Since this form also refers to the adjective with the meaning ‘same’, it is extremely frequent (more than 3500 examples). Yet, only a tiny minority of these many examples have a modal reading. Further research on igual is needed.

3.

329

References Brânză, Mircea. D., and Nicole Delbecque 2008 Variación modal con los adverbios de duda en español. In Studii de Lingvistică úi Filologie Romanică, A. CuniĠă, C. Lupu and L. Tasmowski (eds.), 58-71. Bucharest: Editura UniversitƗĠii Bucureúti. Briz, Antonio and Grupo Val.Es.Co. 2003 Corpus de conversaciones coloquiales. Madrid: Arco-libros. Butt, John, and Carmen Benjamin 1994 A New Reference Grammar of Modern Spanish. 2nd ed. Great Britain: Edward Arnold. Chafe, Wallace 1986 Evidentiality in English Conversation and academic writing. In Evidentiality: the Linguistic Coding of Epistemology, Wallace Chafe, and Johanna Nichols (eds.), 261-72. Norwood, NJ: Ablex. Clift, Rebecca 2006 Indexing stance: Reported speech as an interactional evidential. Journal of Sociolinguistics 10: 569–595. Cornillie, Bert 2007a Epistemic Modality and Evidentiality in Spanish (semi-)Auxiliaries. A Cognitive-functional Approach. (Applications of Cognitive Linguistics 5.) Berlin/New York: Mouton de Gruyter. 2007b On the continuum between lexical and grammatical evidentiality. Evidence from Spanish. Italian Journal of Linguistics 19 (1): 108129. 2009 Evidentiality and epistemic modality: on the close relationship of two different categories. Functions of Language 16 (1): 44-32. Dendale, Patrick, and Liliane Tasmowski 2001 Introduction: Evidentiality and related Notions. Journal of Pragmatics 33 (3): 339-348. DuBois, John W. 1985 Competing motivations. In Iconicity in Syntax, John Haiman (ed.), 343-366. Amsterdam/Philadelphia: John Benjamins. 2007 The Stance Triangle. In Stancetaking in Discourse, Englebretson, Robert (ed.), 139-182. Amsterdam/Philadelphia: John Benjamins.

330

Bert Cornillie

Fernández Sanmartín, Alba 2006 La expresión de la modalidad epistémica en el español conversacional. Ma Thesis. Department of Spanish Philology. Santiago de Compostela: Universidade de Santiago de Compostela. González Ramos, Elisa 2005 Cómo eludir responsabilidades sobre lo dicho: los signos “por lo visto” y “al parecer” (analogías y diferencias en su empleo actual). Español actual: Revista de español vivo. 84: 153-158. Haßler, Gerda 2004 El uso evidencial de adverbios modales. In Algunos problemas específicos de la descripción sintáctico-semántica, Juan Cuartero, and Gerd Wotjak (eds), 229-244. Berlin: Frank & Timme Verlag. Kärkkäinen, Elise 2003 Epistemic stance in English conversation. A description of its interactional functions, with a focus on I think. Amsterdam/Philadelphia: John Benjamins. Matte Bon, Francisco 2000 Gramática Comunicativa del español. Madrid: Edelsa. Nuyts Jan 2008 The “one-commitment-per-clause” principle and the cognitive status of qualificational categories. Linguistics 47 (1): 141-171 Plungian, Vladimir A. 2001 The Place of Evidentiality within the Universal Grammatical Space. Journal of Pragmatics 33 (3): 349–357. Torner Castells, Sergi 2005 Aspectos de la semántica de los adverbios de modo en español. Ph. D. diss. Department of Linguistics. Barcelona: Universitat Pompeu Fabra.

Revelative evidentiality in European languages: linguistic marking and its anthropological background Alexandra Kratschmer and Adriënne Heijnen

In this article we aim to present the results to date of an ongoing interdisciplinary research project on dreams and dreaming, focusing primarily on the linguistic aspects, i.e. the marking of the source and manner of acquisition of dream material in dream accounts, pointing at relevant relations to the cultural background of the speakers.1 The project is executed by a linguist (AK) and an anthropologist (AH). The objective of this article is threefold: a) empirical (as we draw a first inventory of linguistic markers used in European languages in order to present information as “dreamed”, diachronically and synchronically, helping us to seize the complex semantic structure of this domain), b) theoretical (as we wish to show that the semantic domain of evidentiality, revelative but also in general, will draw profit of the integration of a ,rather undetected aspect, namely the interaction between the human receiver and a human/divine/personified source of information) and c) methodological (as we seek to use our results on the linguistic and anthropological aspects of dream treatment to illustrate the heuristic potential of cross checking “claims” made by language structure about non linguistic matters (culture, cognition, etc.) in the relevant discipline, in our case anthropology.

1. Revelative Evidentiality We start out with a definition of our concept of revelative evidentiality (Section 1), continue with considerations regarding the heuristic value of our interdisciplinary approach (Section 2), introduce briefly our written and oral corpora (Section 3), after which we present the empirical linguistic and anthropological results of our studies to date and address possible relations between these (Section 4). In our conclusions (Section 5) we return to our previously established threefold objective, namely rendering account for the empirical, theoretical and methodological results of our studies.

332

Alexandra Kratschmer and Adriënne Heijnen

1.1. A first working definition Following Aikhenvald’s seminal monography on evidentiality in the languages of the world, “Evidentiality is a linguistic category whose primary meaning is source of information” (Aikhenvald 2004: 3). The author clearly states that this term should be used uniquely to refer to grammatical means of coding the source of information (Aikhenvald 2004: 11). In the mean time, an increasing number of scholars, especially those working on European Languages, insist on “considering evidentiality as a more general functional category whose scope includes not only grammatical but also lexical phenomena” (Squartini 2007: 1). For example, Squartini (2008: 918) aims at “exploring how an integrated account of both grammatical and lexical evidentiality can contribute to a better understanding of the whole domain” [our emphasis]. In what follows, we will use the term of evidentiality noting the semantic domain of information source as marked by linguistic means. These linguistic markers can be grammatical, lexical or even discoursive (contextual).2 We will hence speak of grammatical vs. lexical vs. discoursive evidentiality marking. Thus, the inclusion of the option of discoursive evidentiality marking even goes further in challenging the original3 understanding of the category of evidentiality than e.g. Squartini cited above.4 Within this domain, another distinction, advocated for by e.g. Squartini (2008: 918), is useful for our purpose, namely “mode of knowing” vs. “source of evidence”: the “mode of knowing” is always “dreaming” in our case, but the “source of evidence” can be linguistically represented as the dreaming subject him/herself or as a more or less unspecified source outside the subject. Note here that the dreaming subject might be distinguished from the waking subject in certain societies, especially in epistemological terms (Heijnen forth.). In languages with grammatical evidentiality marking, one can typically observe the use of different markers for different types of information sources, with different degrees of specification depending on the language. The recurrent semantic parameters are visual, non-visual sensory, inference, assumption, hearsay and quotative (Aikhenvald 2004: 63-64). Some languages do also have markers for endophoric evidentiality that “refer to the speaker’s own physiological or psychological states” (Plungian 2001: 354, as well as personal communication). As early though as in 1910, Boas inventarizes a suffix with a special evidential meaning ‘see in a dream’ in Kwakiutl. In 1986, Jacobsen suggests that this suffix could be related to a formative suffix meaning ’have a dream’ in Makah. It was actually Roman



Revelative evidentiality in European languages

333

Jakobson, who coined the term “revelative evidence” for such phenomena in 1957 (Aikhenvald 2004: 345)5. Aikhenvald discusses the grammatical marking of that kind of information source in her monography. She remarks that in different languages different types of knowledge and experience are associated with conventionalized sources. In Jarawara, Quechua, Tuyuca and Tatuyo dreams are treated on a par with directly observed reality, in Shipibo-Konibo, Yukaghir, Cree and Modern Eastern Armenian dreams are represented as being outside conscious reality and are coded as reported or non-firsthand evidential, while in Tucano and Tariana ordinary humans’ dreams are coded as non-visual evidential, but shamans’ as visual evidential (Aikhenvald 2004: 380-381). Thus, dreams are very often marked by means normally used to refer to other kinds of evidential categories. Since we interpret “evidentiality” as a semantic category, we treat information revealed in dreams as a special evidential category meriting its own attention. What interests us primarily is which linguistic means (grammatical, lexical, or discoursive) can be used in different languages to mark this semantic category, whether dream marking draws on other evidential categories (e.g. the visual one) and/or non-evidential semantic categories (e.g. epistemic modality). Our working definition of revelative evidentiality reads broadly “information linguistically marked as created inside the mind of a subject without direct input from the outside world (which is not necessarily the view of the experiencing subject, see below), e.g. dreams, visions/revelations, hallucinations, inspiration/ideas, etc.”, information revealed in a dream being a very important subcategory, in fact the category this article will focus on.

2. The interdisciplinary point 2.1. Basic methodological assumption Conventionalized and grammaticalized linguistic expressions “make claims” about cultural experience and knowledge and cognitive processes. While inferring cultural and cognitive circumstances directly from linguistic structures can lead to erroneous conclusions, we see testing “claims made by language” against anthropological and (neuro-, psycho- etc.) cognitive empirical data to be a powerful heuristic tool. The importance of this becomes obvious when considering Aikhenvald’s conclusions about dream

334

Alexandra Kratschmer and Adriënne Heijnen

theories among the Yukaghir, mentioned in Section 1.1. While Aikhenvald claims that for the Yukaghir, dreams are not equal to conscious reality, basing herself on the older work of Jochelson 1905, recent anthropological study undermines her point. Willerslev, who has conducted ethnographic fieldwork among the Yukaghir in the 1990s and 2000s, writes that, for the Yukaghir, dreaming involves a conscious experience of the ayibii6 the soul, often unfolding as sexual engagement with spiritual beings (Willerslev 2007: 175). In the ongoing project we intend to test “claims” made by language on the cultural-cognitive basis of dream experiences against anthropological data on culture related perspectives to dreams and dreaming in order to see whether there can be detected any convincing relations between language use and cultural ideas and practices, and, in the positive case, whether there is any plausible way in determinating a causal directionality from culture to language or vice versa. At the present point of the investigation there seem to emerge certain patterns, even though they lack statistical significance.

2.2. Expected benefits At a general level, an interdisciplinary investigation will further our understanding of the relation between thought, practice and language and will be able to put NeoWhorfian approaches to the thought/language problem to the test. At the linguistic level, it will further our understanding of the semantic category of evidentiality (also regarding such factors as e.g. interactivity, see below) and of its cultural and cognitive basis, besides helping to inventorize its linguistic realization options. At the anthropological level as well as that of cognitive sciences, it will further our understanding of the use of linguistic material as an important heuristic source. Methodologically, in anthropology, “what people do” often predominatess “what people say”, while anthropologists depend increasingly on semi-structured interviews due to a change of research conditions. These new settings urge for a reconsideration of the relation between language use, cognition, culture and social practice.



Revelative evidentiality in European languages

335

3. The empirical data 3.1. The linguistic corpus 3.1.1. Bible versions of Joseph’s story Our linguistic corpus is meant to cover both diachronic and synchronic data. As a diachronic corpus, we chose different versions of Joseph’s story (Old Testament, Genesis 37-44)7 in different European languages (as well as the Biblical Hebrew original) through history. This method allowed us to compare variations on a constant theme, notably a story focusing on dreams and dream accounts, both interlinguistically and language internally through time. We have to date analyzed the following versions of Joseph’s story: Biblical Hebrew (original = O, B.C., exact period heavily discussed), Classical Greek (Septuagint = SE, B.C-A.D, exact period equally discussed), Latin (Vulgata = V, 4th cent. A.D.), German (1545/modernized = G1, 1871 = G2, 1905 = G3, 1912 = G4, 1951 = G5, 1984/99 = G6, 2000 = G7), Dutch (1618/modernized = DU1, 1734 = DU2), Danish (1550 = DA1, 1647 = DA2, 1931 = DA3, 1992 = DA4, 2002 = DA5), Icelandic (1815 = IC1, 1841 = IC2), Faroese (1949/74 = FA), French (1707/44 = FR1, 1744/1881/86/1996 = FR2, 1859/1991 = FR3, 1894/1910 = FR4,), Italian (1649 = IT1, 1821 = IT2, 1925 = IT3, 1974 = IT4, 2007 = IT5) as well as Spanish (1569 = S1, 1909 = S2, “modern” = S3, 1986/97 = S4, 2005 = S5). We have used the English spoken American Standard Version (ASV) for glossing purposes if convenient.

3.1.2. Interviews with informants We have to date registered oral dream accounts and qualitative data regarding language use in dream telling, combined with linguistics tests on grammaticality, discourse compatibility and paraphrasing dream related expressions, among Icelanders, Faroese, Dutchmen, Germans as well as Italians. The main criterium for selecting informants was that the group should allow for comparison between the various languages and societies. The informants are men and women between 20 and 35 years of age, who study various disciplines at institutions of higher education in Denmark. Whereas our corpus of Icelandic dream accounts possesses a clear quantitative significance (cf. Heijnen 2005a, forth.), our Faroese, Dutch, Ger-

336

Alexandra Kratschmer and Adriënne Heijnen

man and Italian oral corpus to date consists of qualitative data collected through focus group interviews and semi-structured interviews. Even though our data is extensive enough to observe certain patterns, it still lacks quantitative statistical weight.

3.2. Anthropological methods and data collection With the help of focus group interviews and semi-structured interviews, we have collected data on the cultural perspectives and practices related to dreaming, dream telling and interpretation among Faroese, Italians, Germans, Dutchmen and Icelanders. For the Icelandic case, our data is extensive. Heijnen collected the data during fieldwork in Reykjavík and South Iceland during the period 1996 to 2008, using the methods of participant observation (including semi-structured interviews) and literature studies. Participant observation allows for collecting data on the context in which dreams are remembered and narrated. To obtain historical depth in the analysis and to reduce personal bias, Heijnen collected numerous dream accounts written from the 13th century until the present day8. Thus, the anthropological data consists of nearly 1000 Icelandic dream accounts, covering a period of almost 1000 years (Heijnen 2005a, forth.).

4. The results to date 4.1. Linguistic results 4.1.1. Issues of revelative evidentiality in Joseph’s story across time and across languages During our work with the biblical material, the following cardinal points of revelative evidentiality emerged as subject to variation as well as pattern forming: – – – – –

reference to the dream experience as such marking the beginning of the dream account marking of the dream account while unfolding it combined forms where at least two of the aspects just mentioned are fused marking the ending of the dream account



– –

Revelative evidentiality in European languages

337

the revelative element the etymology of the lexemes ‘(a) dream‘,’ to dream’

Not all of these cardinal points have to date been subject of systematic data scrutiny, therefore, we will focus here on the reference to the dream experience (4.1.1.1.), the marking of the beginning of the dream account and of its unfolding (4.1.1.2.) and finally the revelative element (4.1.1.3.).

4.1.1.1. The reference to the dream experience as such In this area, the questions which arise are those of verbal valence and of the semantic roles selected by the verb ‘to dream’ as well as those of the choice of verb in combination with the noun ‘a/the dream’. Clear diachronic and synchronic typological patterns can be observed which also can be blurred by obvious traductological issues. In our corpus, the following constructions and diachronic patterns are found: –

–

– –

–

‘to dream a dream’: this is the only documented form used in the Biblical Hebrew original9 and is, probably by strong traductorial influence, present in all our corpus-languages, except Latin (translated from Classical Greek, not from the Hebrew original) and Icelandic (on the special status of Icelandic, see below) ‘to see a dream’: this is the by far preferred construction in Classical Greek10 and the only documented construction with the noun dream in Latin; the Latin use of this construction is not pursued by the Romance Languages, nor does it exist in the Germanic Languages ‘to see something’ with reference to dreaming is documented in the Latin Vulgata ‘to dream’ (absolute use) or ‘to dream + direct object’ (typically occurring in constructions like ’I/he dreamed something’ or with completive clauses as ‘tell me what you have dreamed’, ‘listen what I have dreamed’, ‘I have dreamed that’): this construction is documented in German, Danish, Dutch, Faroese, French, Italian and Spanish remains active throughout history ‘to have a dream’: Dutch and Faroese, and Spanish show a stable (and exclusive) use in our corpus11 of ‘to have’ in combination with ‘a dream’ and especially for Spanish this construction is by far the preferred one in our corpus, compared to ‘to dream (a dream)’.

338 –

– –

Alexandra Kratschmer and Adriënne Heijnen

German and Danish on the one hand and French and Italian on the other undergo a characteristic diachronic development towards constructions suggesting increased subject participation (cf. 4.1.1.1.1) Icelandic shows an extremely stable use throughout history of a double accusative construction (cf. 4.1.1.1.2)

4.1.1.1.1. The subject participation axis In our corpus, there can be observed that earlier Bible versions of German and Danish make extensive (though not exclusive) use of constructions with the dreaming person in a dative/oblique12 case (it dreamed [to] me, a dream [nom.] dreamed [to] me): Es [nom.] hat uns [dat.] geträumet ‘it has dreamed [to] us’ (40:8, G1), Was ist das für ein Traum, der [nom.] dir [dat.] geträumet hat? ‘what kind of dream is that, that has dreamed [to] you?’ (37:10, G1), Mig [obl.] drømde ocsaa ’[to] me [it] dreamed also’ (40:16, DA1), Da drømde oss [obl.] baade en nat huer sin drøm [nom.] ’it dreamed [to] both of us one night a dream’ (41:11, DA1), which later on recede in favor of constructions with the dreaming person as a nominative subject: […] der Pharao [nom.] träumte ‘the Pharao dreamed’ (40:8, G2), Hør dog, hvad jeg [nom.] har drømt! ’listen what I have dreamed’ (37:6, DA3). Parallel to this, a representation of the dreaming subject as a nominative ‘possessor’ (‘I had a dream’) is documented from the first sources onward and remains strong throughout history: […] hatte Joseph [nom.] einmal einen Traum ‘had J. once a dream’ (37:5, G1), […] haffde Joseph [nom.] en gong en drøm ‘had J. once a dream’ (37:5, DA1). On the other hand, our corpus shows a development, concerning French and especially Italian, from a productive ‘to have a dream’ (Joseph eut un songe13 ‘J. had a dream’, 37:5, F4; Giuseppe ebbe un sogno ‘J. had a dream’, 37:5, IT314) to a more and more frequent use of ‘to make a dream’ (Nous avons fait un songe ‘we have made a dream’, 40:8, F2; Giuseppe fece un sogno ‘J. made a dream’, 37:5, IT5). In both languages, ‘to dream’ with a nominative agentive dreaming subject is documented throughout history (J’ai aussi songé ‘I have also dreamed’, 40:16, F1; sognò di nuovo ‘he dreamed again’, 41:5, IT1). Looking at these data, language use seems to make a claim that the dreaming subject, in a way consistent for several European language communities, represents itself on an “activity” or “subject participation” axis, as moving towards more subject participation/responsibility: starting as an



Revelative evidentiality in European languages

339

experiencer, (s)he becomes an owner and finally a creator of his/her dreams (see Table 1). Table 1. Diachronic development along the subject participation axis (simplified) languages experiencer >> owner >> creator German, Danish ‘It dreamed to me’ ’I had a dream’ ’I dreamed’ ‘A dream dreamed to me’ French, Italian

’I had a dream’ ’I made a dream’ ’I dreamed’

The validation by historical and present day anthropological empirical material of this “claim made by language” about a development of the cultural attitude to dreams and dreaming in these linguistic communities represents an important heuristic opportunity, which we will follow up in 4.2.2.2 and 4.2.2.3 resp.

4.1.1.1.2. The double accusative construction in Icelandic Both in our corpus, but also in the remainder of the thousand years of Icelandic language history, documented in the prolific written source material (cf. Heijnen 2005a, forth.), a strong persistence in the use of the construction ‘[It] dreamed me [accusative] a dream [accusative] [evt. for somebody/something [dative]]’: Mig [acc.] dreymdi draum [acc.] [fyrir einhverju [dat.]] is found.15 Based on different theoretical backgrounds, different syntactic interpretations can be given of this construction (Hrafnbjargarson/Tromsø, personal communication). While traditional grammar considers it as a double (accusative) object construction, Generative Grammar (e.g. Hrafnbjargarson 2004: 133) analyzes it as a construction with a human accusative subject and nonhuman accusative object. These two different interpretations make, as we see it, different, but equally intriguing claims about how the dreaming subjects experience their role in a scenario which is equally complex in both cases. According to the traditional (double object) analysis, there is an unidentified agentive entity (the implicit nominative subject of the verb), a “power”, exercising a direct influence (marked by the direct objects) on the dreaming person as well as on the dream (its content), sometimes regarding a third party (the prepositional dative object). The verb ’to dream’ shows a complex valence pattern

340

Alexandra Kratschmer and Adriënne Heijnen

with a series of actants (1 nominative, 2 accusatives) and the frequent presence of a standardized circumstant (a prepositional dative), mirroring a complex constellation of involved entities. According to the Generative Grammar analysis (human accusative subject), the dreaming person combines “agentive” (subjecthood) and patient/theme features (accusativity), “acting” and “undergoing the action” at the same time. The dream itself is (equally) under direct influence of the action/the agent.16 Also in this interpretation17, there is a complex constellation of entities involved. We will show below that this complex constellation of involved entities in dream experiences can indeed be confirmed as an important underlying metaphysical stance of Icelandic culture (cf. 4.2.1 and 4.2.2.1).

4.1.1.2. The marking of beginning of the dream account and of its unfolding Due to lack of space we have to limit ourselves to a very synoptic presentation of the linguistic phenomena regarding these two cardinal points, reserving the extensive discussion of the extremely rich data material for another occasion. As some of the following phenomena can mark both the beginning and the unfolding of the account, a collective treatment is by no means unnatural.

4.1.1.2.1. Visual evidentiality This is by far the most dominating evidential domain in our corpus. Here, a pervasive role is played by the ‘behold’-marker which goes back to the Biblical Hebrew original hinneh18 and which is subsequently found in all corpus languages except Faroese, e.g. German (siehe, as in 37:7, G2). It has to be noted though that the visuality conveyed by the ‘behold’-marker is a phenomenon that obviously arose due to a traductorial problem: most target languages lacked a comparable focus marker, and the marker with a similar textual deictic function at hand was one conveying the idea of summoning the gaze of the listener (‘behold’). The traductorial pressure to conserve this marker (but in a new disguise) seems to have been rather strong. Three of our Germanic corpus languages (namely German, Dutch and Danish) share the feature that they allow for perception verbs to be fol-



Revelative evidentiality in European languages

341

lowed by a subordinate clause introduced – instead of by ‘that’ (dass, at, dat) – by the manner conjunction ‘how’ (wie, hvordan/hvorledes, hoe dat), expressing not a manner, but the intensive perception of an action in progress or a situation under development (e.g. Duden [Drosdowsky 1984: 686], for German). While there are no German occurrences of this phenomenon in our corpus, both one Dutch and one Danish version (DA1) use the ‘how’-connector, not after a perception verb, but after ‘had a dream’, probably capable of implying a context of (visual) perception, in the same verse19 (41:1): hadde Pharao eenen droom hoe dat hy stond aan het waater, DU2. Another aspect of visuality in dream accounts is what we call tableau painting, grammatically marked by imperfective verb forms (see 5.1.1.2.4.).

4.1.1.2.2. Inferential evidentiality and epistemic modality intertwined In our corpus, it is often verbs or complex expressions of epistemic modality that are used to mark the beginning or the continuation of a dream account. In the European Languages represented in our corpus, these markers appear to carry systematically an evidential nuance inherent to the lexeme/construction (cf. Nuyts 2001: 64-72, 122-129, 205-208 on modal adverbs and adjectives, mental state predicates as well as modal auxiliaries in German and Dutch; Cappelli 2007, on English verbs of cognitive attitude) or depending on the grammatical and/or semantic context (cf. Kratschmer 2006 on the broad, context sensitive variety of evidential nuances of the Italian epistemic verbs sembrare and parere, both ‘to seem’, as well as Cornillie 2007 on Spanish parecer ‘to seem’, equally expressing different nuances depending on the syntactic context).20 Typically, the epistemic verbs and expressions found in our corpus are of that latter kind, i.e. expressing a clearly subjective judgment, sometimes combined with an element of epistemic reserve in the sense of marking that the content of the narration is not unambiguously assessable. It has to be noted that these epistemic verbs are not attested in the Hebrew original, but find their way into the text already in the Greek Septuagint (Оΐ΋Α oimen ’to believe’, ‘to seem‘). They are consequently consistently used in the Latin Vulgata (puto ’to mean’, ‘to judge’, ‘to consider as’), in the German (dünken, archaic ‘to seem’), Dutch (dunken ‘to seem’, ‘to think), Danish (tykke (archaic), syntes, both ’to mean’), Icelandic (að þykja ‘to find’, ‘to feel’, ‘to seem’, ‘to like’; að þykjast ‘to claim‘, ‘to pre-

342

Alexandra Kratschmer and Adriënne Heijnen

sume‘, ‘to profess’), Faroese (tað var (fyri) mær, sum ‘it was (to) me, as (if)’) French (sembler, roughly ‘to seem’), Italian (parere ‘to seem’) as well as in the Spanish Bible versions (parecer ‘to seem’). The evidential nuance carried by these lexemes is mostly inferential, in rare cases possibly visual.21 In a series of versions, a comparative connector is the inherent part in an expression of epistemic reserve, as in the subordinate connector following the verb ‘to seem’. This is true for Danish som ‘as’ following syntes ’to seem’ (an archaic combination, e.g. 40:10: Oc (mig siuntis) som det blef grønt22, DA2), as well as for Faroese sum ‘as’ following ‘it was [to] me’: Tað var mær, sum vit bundu bundi úti á akrinum23, 37:7, FA.

4.1.1.2.3. Alternative reality Dreams can be marked as alternative reality by various means, as e.g. a comparative connector ‘as if’.24 This is documented in the following Icelandic version of 40:9 based on þykja ‘to seem’ followed by sem ‘as (if)’ and the subordinated verb in the preterite subjunctive, a non-realis mood: Mér þótti í svefninum sem vínviður stæði [pret. subj.] fyrir framan mig25, IC1.26 The alternative reality aspect of the narrated dream can be marked in Icelandic, not only by the preterite subjunctive mood following ‘to seem’ just mentioned, but also by this mood following mig dreymdi ‘[it] dreamed [to] me’, as in mig dreymdi að vínviður væri gegnt mér27 (40:9, IC2).28 The same is true for German, where the 2nd subjunctive is used: Mir hat auch geträumet, ich trüge drei weiße Körbe auf meinem Haupt29 (40:16, G1).30

4.1.1.2.4. Narratological aspects A “typical” feature of Romance dream telling is the use of a characteristic aspectual pattern in the preterite verbal conjugation, at least following grammaticography.31 This use is opposed to conventional narrative aspect use which combines perfective (marking actions as single and concluded) and imperfective preterite (marking actions or states as ongoing at reference time). The oniric preterite consists of casting the whole account in the imperfective aspect, which is often explained by the imperfect’s potential of expressing “alternative realities”.32 It can, on the other hand, be used as a mere (non fictional) narrative marker, supposed to “immerge” the listener



Revelative evidentiality in European languages

343

into the action, typically in history writing or modern journalist style (Dardano, Trifone 1995: 354; Togeby 1982: 319).33 Interestingly enough, the use of the oniric imperfective aspect in our Romance Bible versions is not systematic at all, compare e.g. the two competing version of Spanish 37:7: […] y he aquí que mi gavilla se levantó [perfective preterite] […] (S4 + S5) vs. […] y he aquí que mi manojo se levantaba [imperfective preterite] […] (S1-3).34 One can conclude that the oniric imperfect, claimed by Romance grammaticography to be a definitorial feature of dream telling, is, as far as our corpus is concerned, regularly overridden by the “conventional” narrative aspect use.35 This question of the stability of use of the oniric imperfect has also been part of our native speaker investigation with Italian informants (see 4.1.2.3.). (Visual) evidentiality and concerns of story telling go hand in hand, when the teller makes use of the narrative strategy of painting a “tableau” for his/her public’s (inner) eye. This is done in several of our corpus Bible versions by verbal markers of imperfective aspect which represent the action “as it unfolds”, either in the form of present participles as in the original Hebrew version, in the Greek Septuagint, the Latin Vulgata as well as our Icelandic versions (blómstrandi ‘blossoming’, 40:10), as a gerund as in Spanish (estábamos atando ‘we were binding’; 37:7, S4 + S5), or as finite verbal forms as the Classical Greek imperfective aspect: (which stands in opposition to the aorist, coding perfective aspect)36, as well as the Latin imperfective preterite.37 The imperfective markers in our corpus are the only grammatical markers of (revelative) evidentiality, but they are not autonomous. They require a context marked as a dream beforehand and can therefore only be classified as combinatorial markers. The context is created by lexical elements, which obligatorily have to be members of the lexical family ‘a dream’/‘to dream’. The same is true for the above mentioned lexical markers, namely the epistemic verbs.

4.1.1.3. Interactive evidentiality Finally, concerning the aspect of dreams being experienced as divine revelations, it turns out that evidentiality can and should be regarded from an interactive perspective where information can be passed and received in a deliberate and conscious manner.

344

Alexandra Kratschmer and Adriënne Heijnen

The idea of revelation is verbalized three times in our written corpus in Gen 41. Joseph explains the divine origin and intention of the Pharao’s dreams to the Pharao: […] God has told to Pharaoh what He is about to do (Gen 41:25); […] God has shown to Pharaoh what He is about to do (Gen 41:28). The Pharao in his turn speaks about Joseph’s dream interpreting gifts equally as being divine revelations: Since God has informed you of all this […] (Gen 41:39; all ASV, all emphases ours).38 In both constellations there is a relevator (God) who reveals something (the target) to the other (the Pharao or Joseph), and in all cases the action is seen from the perspective of the agent (God). Table 2. Revelation: the interactive patterns level of interaction aspect focussed on visual perception ‘to direct the other’s gaze to the target by one’s finger’

expressions: tokens

’to point/indicate’: 11 3 σįİȚȟİȞ SE 5 zeigen G 1, 4-7 1 anzeigen G5 2 indicare IT4,5 ‘to move the target into the field of ‘to hold up/forward’: 3 perception of the other and hold it 1 hiGGîd O there’ 2 ostendere V ‘to cause the other’s perception of the ‘to make see’: 10 target’ 1 her´â O 2 sehen lassen G2,3 1 lade skue DA3 2 lade se DA2,4 1 láta sjá IC2 1 láta síggja FA 2 fair voir F1,2 ‘to remove an obstacle in order to ‘to unveil/to reveal’: 3 enable the other to perceive the target’ 1 enthüllen G7 2 afsløre DA5 ‘to arrange the naked (= non-hidden) ‘to render bare in the target in the open (= in the visual field open’: 3 of the other)’ 2 åbenbare DA3,5 1 opinbera FA



Revelative evidentiality in European languages

345

‘to shed light on the target in order to ‘to declare/to make enhance the other’s perception of it’ clear, to illuminate’: 5 1 birta IC1 1 auglýsa IC2 3 déclarer F1-3 ‘to arrange the target in a way that ‘to exhibit/display’: 15 enables the other to perceive it’ 1 vertoonen DU1,2 1 sýna IC1 1 montrer F3 4 mostrare IT1-3 8 mostrar S1-5 ‘to make the target touchable to the ‘to manifest’: 5 hand of the other’ 5 manifestare IT1,4,5 symbolic behavior symbols used ‘to give signs’: 2 2 significare IT1,3 utterance act ‘to communicate verbally (to tell)’: 1 1 fortælle DA4 content ’to deliver a message’: 20 3 verkündigen G1,4 6, 8 kundtun G1-6 1 ankündigen G7 2 verkondigen DU1,2 1 kond doen DU2 1 kundgøre DA3 1 boða IC1 1 kunngjöra IC2 2 anunciar S4,5 cognition enrichment of the other’s cognitive ‘to make know’: 21 universe 1 hôdîª` O 1 te kennen geven DU1 2 give til kende DA1,2 3 lade vide DA1,2,4, 1 láta vita FA 6 faire connaître F1-4 2 far conoscere IT2,3 5 hacer saber S1-5

346

Alexandra Kratschmer and Adriënne Heijnen

In Table 2 the verbs and expressions used in the single versions of our corpus text are listed in conceptual groups following their etymology39. The etymologies of the used verbs and expressions very clearly contain the concept of interaction. As one can se, in our corpus, the level of visual interaction is by far the dominant one (55 vs. 23/symbolic behavior vs. 21/cognition). Inside the visual group, we find three dominating subgroups: ‘to indicate’ (11), ‘to make see’ (10) and ‘to display’ (15). Inside the symbolic behaviour group, ‘to deliver a message’ is the dominating choice (20 of 23). The cognition level, despite of disposing of only one realization pattern: ‘to make know’, is nevertheless important (21). There are no absolute patterns regarding the distribution of the above conceptual options of revelativity over the different languages in different periods (e.g. typological constraints regarding interaction levels), but one can detect a marked preference for the concept ‘to deliver a message’ in the Germanic (18 of 20) and a marked preference for ‘to display’ (13 of 15) in the Romance and a certain preference for ‘to make know’ (13 of 21, thereof one in the Hebrew original) again in the Romance languages. On the basis of these data, the interactive aspect seems crucial in evidentiality matters and we wish to make a plea for an integration of interactive aspects in evidentiality studies in general, that is, in all studies on information source marking. At that point, we even want to raise the question, not at all rhetorically meant, if one can find, in the languages of the world with grammatical evidential marking in the Aikhenvaldian sense, different evidential markers for ‘overhearing something’ vs. ‘(inter-)actively being told’ or ‘ (visually) perceiving’ vs. ‘being shown’.

4.1.2. Issues of revelative evidentiality in modern European languages form the native speakers’ perspective We repeat that we possess statistically relevant material only for the Icelandic situation at this point of the project (viz. Heijnen 2005a; 2007, forth.). The data material for the other communities has to be regarded as a first insight that has to be corroborated by a far larger amount of data.



Revelative evidentiality in European languages

347

4.1.2.1. Linguistic results of participant observation among and interviews with Icelandic native speakers Our Icelandic corpus consists of approximately 80 semi structured interviews, conducted between 1996-2009 (Heijnen 2005a, forth.); a collection of interviews primarily carried out by the Icelandic ethnologist Hallfreður Örn Eiríksson40 and the earlier listed literary sources (cf. n. 4). In line with the investigations for this particular project, we also conducted two interviews with Icelandic students in Denmark in the age group 25-35. Our data show a strong continuity in the use of the construction: mig dreymdi draum ‘it dreamed me a dream’, since the 13th century41 at least. However, the in note 11 mentioned inversion of nominative, accusative and dative42 is to some degree present in our material. Our modern data demonstrates that the double accusative construction mig dreymdi draum is especially consequently used among Iceland’s older population. In addition, the mixing up of accusative and nominative occurs primarily in the perfect tense, mig hefur dreymt ‘it has been dreamed me a dream’, people using ég hef dreymt ‘I have dreamed’ instead, and in the general mann dreymir ‘it dreams one’. Thus, the application of nominative ég dreymdi ‘I dreamed’ occurs, but is rare. We have not encountered Icelandic native speakers who consequently apply nominative in the marking of dream accounts. Having analyzed numerous dream accounts, we found the following epistemic verbs: the most important being að þykja ‘to find‘. ‘to feel‘, ‘to seem‘, ‘to like’, að þykjast ‘to claim‘, ‘to presume‘, ‘to profess‘, ‘to make believe’, ‘to pretend’ and að finnast ‘to seem‘, ‘ to think, ‘to like’. Að þykja and að þykjast are especially found in (older) written material and in dream accounts of the older population. The verb að finnast is less formal and more commonly used nowadays. These epistemic verbs are often followed by a designation of place43. Our data shows that dream accounts are told in the past tense and in the “historical” present. In the latter case, the dreamer enhances the recalling of the dream by “reexperiencing” the dream, as well actualises the content for the audience. The use of the “historical” present is often related to intense dream experiences, such as nightmares. In addition, the subjunctive mood occurs frequently in the modern dream accounts (see the example in 4.2.2.1), in a similar way as we noted for the biblical references (see 4.1.1.2.3).

348

Alexandra Kratschmer and Adriënne Heijnen

4.1.2.2. Results of interviews with German native speakers We interviewed one female and two male university students from the Northern and Eastern part of Germany aged between 21 and 23 years. We started with linguistically oriented single interviews and proceeded with a focus group interview to obtain anthropological data. Two of the students had never heard the construction mir [dat.] hat geträumt (‘it has dreamed to me’) and experienced it as outright ungrammatical, the last one remembered reading the form in older texts44. For all of them ich habe geträumt (‘I have dreamed’) was the only form they claimed to use.45 They spontaneously observed that the dative construction would vehiculate the idea that the dream was coming from outside themselves, while on the contrary they felt it was their own mental product. This is clearly a case of what we would like to call “norm-violation induced transparency”, i.e. the “literal” semantic interpretation of constructions that depart form the norm. This phenomenon has its counterpart in “norm-compliance induced opacity”, i.e. that fact that conventionalized (“salient”) non-“literal” meanings are automatically accessed as whole units without their constitutional parts being analyzed each for themselves.46 That means that the fact, that the dative is experienced as expressing extra-subject dream origination does not imply that the form with the nominative experiencer is automatically and consciously used and interpreted as expressing intra-subject dream creation by the naïve language user. Inspired by the biblically documented use of the archaic epistemic verb dünken (in G1 and G4) for marking the beginning of the dream account, the subjects’ readiness to interpret the modern equivalent scheinen ‘to seem’ not as an epistemic reserve marker, but as a mere marker of revelative evidentiality, was tested, but the subjects could actually not decide on the question in a conclusive manner. Inspired by the biblical use of the visual perception verb (ich/er sah ‘I/he saw’) in dream accounts, we tested the natural interpretation of this verb in dream accounts either as a marker of revelative evidentiality or as a marker of straightforward visual perception as an ingredient to the dream action (on a par with other events occurring in the dream). Here, the subjects agreed on only accepting the latter. A phrase Ich sah, wie ich Fahrrad fuhr ‘I saw how I rode the bike’ was not interpretable as ’I dreamed that I rode the bike’ to them, but only as ‘I saw myself riding the bike (from outside, as one can experience that in a dream)’. Asked to produce an actual dream account, we registered both “historical” present (actualizing the content for the teller and the audience) and



Revelative evidentiality in European languages

349

past tense forms. One subject spontaneously produced a subordinate clause in the second subjunctive, as documented in G1 and G4 (cf. 4.1.1.2.3): ich habe geträumt, ich wäre [2nd subj.] in einem Turm ‘I have dreamed I were in a tower’, thus marking the alternative reality status of the dream. Directly questioned, all subjects accepted that use. To conclude, it can be said that the results of our interviews did not contradict the diachronic evolution documented by our Bible studies, i.e. that the verb träumen ‘to dream’ is becoming more and more agentive, meaning that it is more easily combined with a nominative agent than with a dative experiencer.

4.1.2.3. Results of interviews with Italian native speakers We interviewed three female university students aged between 21 and 31, originating from Northern, Center and Insular Italy respectively. All linguistically oriented interviews were single interviews; two of the informants gave a group interview, while one informant, for schedule reasons, also responded to the anthropological questions during her individual interview. They all came to the conclusion that the noun un sogno ‘a dream’ was absolutely incompatible with avere ‘to have’ and always required fare ‘to make’, thus confirming the diachronic tendency announced by our Bible corpus. According to two of them (the third one not being present), avere ‘to have’ was compatible though with un incubo ‘a nightmare’, or with un sogno premunitorio ‘a premonitory dream’.47 Again they spontaneously observed, that ‘to have’ would indicate that the dream was not the product of their own mind, as they felt it was and as ‘to make’ appropriately verbalized according to their view, while the premonitory dream could be seen as a message from elsewhere. Again, we have to suspect a possible “norm violation transparency”-effect here. ‘To dream about somebody’ can only be construed as sognare qualcuno ‘to dream somebody’, and one subject spontaneously produced the sentence una mia amica […] sognava se stesso correre ‘a [girl] friend of mine dreamed herself [to] run’.48 The epistemic verb sembrare ‘to seem’ is still possible as a dream account marker (it was not produced spontaneously though), but the linguistic tests made it very clear that this function is only accessible after the introduction of the dream concept by the noun sogno or the verb sognare.

350

Alexandra Kratschmer and Adriënne Heijnen

The informants were asked to give an actual dream account as well as to create a fictive dream account on the basis of a prefabricated mini-plot in key words about a dog chasing them (i.e. containing actions succeeding each other and thus potentially reflecting clearly any oniric aspect use). Both in the genuine dream accounts and the fictive ones, the subjects used mainly either present tense or perfective tense forms, with only isolated occurrences of the oniric imperfect (and only by one informant). Asked about it post hoc, they knew the concept and judged dream accounts based on the oniric imperfect as “more beautiful”, “more poetic”, “more correct”, “more story”, while the normal aspect was “more direct”, “more effective”, “quicker”, obviously meaning that it did not require a genre related elaboration effort. The informants confirmed further that the Italian conjunction come ‘how’ following vedere ‘to see’ cannot be used to emphasize the perceptual process regarding actions in progress as the Dutch, Danish and German equivalents can, but only can refer to the manner in which an action is performed (hence the lack of it in our biblical corpus). We can conclude that, also what our Italian informants are concerned, the tendencies found in our diachronic Bible corpus were at least not contradicted. This was true for the weakening of the expression avere un sogno ‘to have a dream’ in favour of fare un sogno ‘to make a dream’, as well as for the not at all unchallenged position of the oniric imperfect as an obligatory marker of dream accounts.

4.2. Anthropological results 4.2.1. Written data – The Icelandic case Supported by the written sources, listed in note 4, we argue that the idea of dreaming being a means to experience temporal and spatial realms that are hidden from the waking mind can be traced continuously from the first Icelandic sources written down in the 13th century until the present day. Christianity, foreign influence and modernisation have, of course, generated change in Icelandic society, and it should be noted that the narrative form, dream content and the actors involved in dreaming, dreamtelling and interpretation vary from period to period. However, the basic theory of dreaming being a revealing activity is persistent. In the Old Norse literature, dreaming is primarily related to revealing the fate of the dreamer or the persons in his or her direct environment. In



Revelative evidentiality in European languages

351

the 16th to 18th century, many dreams are revelations, as we know from other European Christian divine literature and provide guidance in Christian ethics, while few dream references found in the rímur (metrical romances) and folktales witness of dreaming as a gateway to hidden realms. In the literature, the realness of other realms is often proved by the dreamer’s ability to bring something from these realms to the waking world. This can be knowledge, objects (e.g. Skíðaríma in Homan 1975: 197-216) or a change of physical state, for example by becoming pregnant (e.g. Kötludraumur in ÍÞÆ 1954-1961, I: 59). Two surveys, conducted respectively in 1974 (Haraldsson 1978) and in 2006 and 2007 (Arnalds et al. 2007), show the persistence and even slight increase of an inclination towards considering dreams as revelations in present day Iceland. People were asked whether they considered themselves being berdreyminn49, the results being 36% of positive answers for 1974, 39% for 2006 and 42% for 2007 (Arnalds et al. 2007).

4.2.2. Oral data 4.2.2.1. Results of interviews, dialogues and participant observation among Icelandic native speakers Also our ethnographic data shows the strong continuing presence in modern Iceland of the theory that dreaming does not involve a withdrawal within the self, but social engagement whereby hidden knowledge is revealed (Heijnen 2005a, 2005b, in press). The following example from one of Heijnen’s keyinformants, who worked at the time of the dream as night watch at a home, illustrates the revelative role of dreams in contemporary Iceland: Mig dreymdi að ég væri50 í vinnunni og ég var að klifra upp einhvern vegg, ofan á bakinu á mér sat sú sýkursjúka, og það sem ég var að gera, var að ég var að lyfta henni upp í sykri, fyrir ofan vegginn var einhverskonar op, eða gluggi og ég þurfti að koma henni þangað inn, fyrir innan gluggann var starfsfólk, ég mann bara einn starfsmanninn og það var hún Rósa. [It dreamed me a dream that I was at work and I was climbing up a certain wall. On my back was the person with diabetes and what I was doing was, that I was raising her level of insuline. High at the wall was a kind of radiator, or window, and I needed to get her in there. On the other side of the window were staffmembers, I only remember one staff member and that was Rosa.] (Icelandic man, b. 1969, Reykjavík, 21 January 2000)

352

Alexandra Kratschmer and Adriënne Heijnen

The dream refers to a Diabetes patient. The informant said that the night he dreamed his dream, this patient got an insulin chock due to a wrong level of medication. Other dreams about this patient followed this first dream, encouraging the dreamer to ask for further medical examination. Our data contain many examples, where people follow the wellbeing of others through their dreams and gain knowledge about their own life or that of others. However, despite for persistence of the theory that dreams are revelative, Heijnen’s ethnographic studies in the period 1996-2009 demonstrate an increase in the inclusion of personal feelings and emotions in Icelanders’ dream accounts and interpretations, as well as a slight change in epistemology. Rather than experiencing the dream solely as originating from an external source, a few dreamers mentioned that they picked up unconsciously hidden hints in daily life, which were responsible for the generation of revealing dreams during the night. Against the background of Aikhenvald’s claims about Yukaghir dream theory, we argued earlier (see 2.1.) that to understand the interrelation between local attitudes and practices concerning dreaming, on the one side, and language use, on the other, as well as to support our thesis that evidentiality should be approached from an interactive perspective, it is crucial to consider notions of personhood (including ideas about the “soul”) and agency. As such, the Icelandic dream theory, which recognizes that dreams are not taking place within persons’ head but involve engagement with other realms, has radical implications for understanding the use of linguistic markers of evidentiality. While according to a mainstream Euroamerican psychological point of view, dreamers create their dreams and are the only agents in the dream, in Iceland the agents in dreams might be the dreamer, but also the dead and other beings who appear in dreams, with or without counterparts in the waking world. Telling a dream is therefore not necessarily an act of uttering personal sentiments, anxieties or perceptions, but might solely involve the communication of information (visually experienced in the dream) about the life of others. Also, as for the Yukaghir, while the dreamer might be observed passively lying asleep, his or her “soul” might be actively engaged in conscious “first-hand” experience in the dreaming realm. Thus, to conclude, our data shows a strong parallel between the persistence of the double accusative construction mig dreymdi draum ‘it dreamed me a dream’ and the continuing importance of the theory that dreams have the potential to reveal knowledge, otherwise hidden for the waking mind51.



Revelative evidentiality in European languages

353

4.2.2.2. Results of interviews with German native speakers For our subjects, dreams are products of their own psyche, processing phenomena of real life impressions and (pre)occupations. They are not revelations coming from whatever source outside themselves. Even as products of their own psyche, dreams do hardly directly influence their choices in the woken life, at most could give occasion for reflection. Nevertheless, they all knew other persons from their nearest community, most female, to whom dreams serve as guidelines or objects of speculation or preoccupation, either because of their supposed psychoanalytic value or their status of being messages from supernatural realms. At least what our few subjects is concerned, we can conclude that the linguistic and the anthropological data provided by them are highly consistent: the attitude towards dreams and dreaming as phenomena produced by the dreaming subject’s psyche seems to be perfectly mirrored by the diachronically successful verbal valency structure where the verb träumen ‘to dream’ selects for a nominative (agentive) subject representing the dreaming person.

4.2.2.3. Results of interviews with Italian native speakers As it was true for our German subjects, our Italian subjects equally saw dreams as products of their own minds. One of our subjects reported having had several premonitory dreams having come true later on. The source of these dreams was seen by her as a supernatural sphere consistent with her catholic faith, a conceptualization not infrequent in Italian society where Catholic faith is a social parameter of a certain weight. Nevertheless, she claimed to conceive herself as the creator of the actual dream, of the form it takes, while it is uniquely its content, the message that comes from a source outside her. Italy has further a strong tradition, originating from the town of Naples, of dream interpretation regarding guessing the right numbers for different lottery institutions, the so called Smorfia.52 Our subjects that had not experienced premonitory dreams did not attribute any importance to dreams, but again knew people that did, be it for psychoanalytic or religious, revelative reasons (catholic or less orthodox, as in the case of believers in the Smorfia). Also concerning our Italian subjects we could conclude that the cultural attitude towards dreams and dreaming and the linguistic realization of the reference to this phenomenon is consistent to a certain degree, taking our subject’s metaphysical subtle-

354

Alexandra Kratschmer and Adriënne Heijnen

ties into account: the dreaming person creates the dream (entirely or at least what its actual form is concerned, in case it is revealed from a supernatural source outside the dreamer), which can be said is mirrored by the loss of the construction form ‘to have a dream’ in favour of ‘to make a dream’.

5. Conclusions 5.1. Empirical conclusions As far as the linguistic data analyzed is concerned, it is no surprise that no independent grammatical markers for revelative evidentiality could be found in the European corpus languages, Classical Greek, Latin, German, Danish, Dutch, Icelandic, Faroese, French, Italian and Spanish, nor in the Biblical Hebrew part of our written corpus. The only grammatical markers found were dependent on a previous marking of the oniric sphere by a lexical element of the etymological field of dream or dreaming. These markers were ones of imperfective verbal aspect, either as finite inflection or as participial/gerund forms, representing the dream content either as belonging to a sphere of alternative reality or as an action unfolding before the eye of the teller as well as the listener (‘visual evidentiality’). Visual evidentiality in general proved to play a dominant role in European revelative evidentiality. This aspect can be marked by the use of perception verbs or conjunctions emphasizing the perceptive process (as in Dutch and Danish). As a traductorial heritage from the Hebrew original, the ’behold‘-marker equally belongs to this sphere. As typical for European languages, inferential evidential aspects are often intertwined with aspects of epistemic judgment. Epistemic reserve is a part of dream accounts in the form of verbs like ‘to seem’ or expressions as ‘it was (to me)’. What the creative source of the dream is concerned, we can distinguish between linguistic structures that seem to mark an attribution of creative power and thereby responsibility for the dream to the dreaming subject and those seeming to mark its attribution to a metaphysical realm outside the dreaming subject. While a language as Icelandic shows a diachronically rather stable pattern of the dreaming subject in oblique case position, suggesting exposure to the action and not acting, we can pin down a diachronic process towards a marking that suggests more subject participation and



Revelative evidentiality in European languages

355

thereby responsibility in languages like German, Danish, French and Italian by the use of nominative subjects and/or the creation/production verb ‘to make’. With the exception of the lexical family ‘to dream’, ‘a dream’, revelative evidentiality in European Languages is not a semantic domain that has a specialized formal apparatus of markers at its disposition, but which presents itself more like a dynamic system that has to be constructed to the purpose, building on elements of visual and inferential evidentiality, of epistemic reserve and of comparison, and in some cultures of a seeming interaction between the dreaming subject and another realm. It is furthermore absolutely depending on the explicit lexical marking of the concepts ‘a dream’ or ‘to dream’. What the anthropological background of these data is concerned, we could witness a certain consistency between dream conceptualization “suggested” by construction types in a certain language and the cultural attitude of (some of) its speakers, a consistency that by no means should be considered as self evident though. This was true e.g. for Icelandic, whose double accusative construction matches perfectly well with the Icelandic attitude to dreaming as messages received, by the dreaming person, from another realm. German and Italian, for which there could be observed a diachronic development from linguistic constructions suggesting less subject participation and hence responsibility for the dream towards constructions seemingly mirroring a higher degree of both, speakers agreed on experiencing dreams as creations of their own mind, at least what their outer form was concerned. This field of investigation deserves by all means a larger scale attention, whereas the outcome hardly can be predicted at this point.

5.2. Theoretical conclusions In so far, as revelative evidentiality is perceived, in many historical and some contemporary cultures, as a phenomenon implying a metaphysical source outside the dreaming subject to whom a message is revealed, the interactive aspect of revelative evidentiality emerged as crucial. A large inventory of metaphors or metonymies for this passing on of information between the revelator and the dreaming subject could be inventarized in our corpus. These metaphors and metonymies stem from the sphere of – again – visual perception, symbolic behavior and cognition.

356

Alexandra Kratschmer and Adriënne Heijnen

These data suggest that the interactive aspect might be important in evidentiality matters in general, and we would like to encourage typologists to follow this lead. It might be rewarding to see whether one can find, in the languages of the world with grammatical evidential marking, different evidential markers for e.g. ‘overhearing something’ vs. ‘(inter-)actively being told’ or ‘(visually) perceiving’ vs. ‘being shown’.

5.3. Methodological conclusions Based on the assumption that testing “claims” made by conventionalized and grammaticalized linguistic expressions against anthropological and, by the way, (neuro-, psycho- etc.) cognitive, empirical data represents a powerful heuristic tool, we have collected some first data regarding “claims” made by language on the cultural-cognitive basis of dream experiences and anthropological data on culture related attitudes to dreams and dreaming. We have focussed, in the first place, on Icelandic, German, and Italian. Independently of any concrete outcome, which in our case actually happened to show a certain consistency between linguistic representation and cultural attitude, we consider this method to be a fruitful guideline to both linguistic and anthropological studies. The merits of this method are twofold: furthering and preventive. On one hand, this method helps linguistics and anthropology (or cognitive sciences) to fertilize each other, taking advantage of the other discipline’s results and emulating them by their own. On the other hand, this method helps linguists not to jump to cultural or cognitive conclusions based on mere linguistic data, while it helps anthropologists not to ignore relevant linguistic clues, circumventing the occasional “blind spots” of both disciplines. On the long run, an interdisciplinary investigation will further our understanding of the relation between thought, practice and language and will be able to put NeoWhorfian approaches to the thought/language problem to the test. At the linguistic level, it has proved able to further our understanding of the semantic category of evidentiality, especially regarding the factor of interactivity.



Revelative evidentiality in European languages

357

Notes 1.

2. 3. 4. 5.

6. 7.

8.

We are deeply indebted to the following scholars for sharing their expertise on the following languages: Dan Enok Sørensen/Aarhus: Biblical Hebrew; Steffen Krogh/Aarhus: Biblical Hebrew & Early New High German; Jette Persiani/Aarhus: Classical Greek; Per Overgaard/Aarhus: Latin; Henrik Jørgensen/Aarhus: Early Modern Danish; Sten Vikner/Aarhus: Early Modern Danish & Icelandic; Gunnar Hrafn Hrafnbjargarson/Tromsø: Early Modern Danish & Icelandic; Susana Fernández/Aarhus: Spanish; Karsten Hvidtfeld/Aarhus: Modern German & Modern Danish, Peter Bakker/Aarhus: Old and Modern Dutch. All responsibility for errors is of course ours. We are equally deeply indebted to our Icelandic, Faroese, Dutch, German and Italian informants, as well as to an anonymous reviewer whose observations helped us to enhance the focus of this paper and to elaborate on some crucial issues. As we will show in the case of dream accounts. In the Aikhenvaldian sense. We thank the anonymous reviewer for pointing out to us, how far we actually stretch the understanding of this category. It was also Jakobson who, on that very same occasion, first used the term ”evidential” in the way we understand it now, as a generic term for information source. Until then, it was understood as used by Boas, namely ‘something for which there is evidence’ (Aikhenvald 2004: 13). The ayibii can be literally translated as shadow and be physically manifested in dreams (Willerslev 2007: 57). The quoted examples stem from the following different dream accounts (which is sometimes also true for our exemplification of a particular phenomenon, when different languages or even different Bible editions in the same language choose different marking strategies for a particular verse): Joseph’s direct speech account to his family about his own two dreams (37:6-7 and 37:9), the chief butler’s direct speech account to Joseph (40: 9-11), the chief baker’s direct speech account to Joseph (40: 16-17), the account of the Pharao’s two dreams as told by the narrator’s voice (41:1-7) as well as the Pharao’s own direct speech account of his two dreams to Joseph (41:17-24). Literature studies included the following Old Norse sources, containing ca. 530 references to dreams (Kelchner 1935: 3). Eddukvæði (the Poetic Edda); Landnámabók (Book of settlements); Íslendingabók (Book of Icelanders); Konungasögur (Kings’ sagas); Fornaldarsögur (Sagas of antiquity); Íslendingasögur (the Icelandic family sagas) and Sturlungasaga. For the period 14th 19th century, mainly rímur (metrical romances); vitranir (epiphanies) and folk tales were used and, for the modern society, personal collections of dreams; diaries; letters and dreamfora at radio, television and in newspapers, as well as surveys on beliefs and practices around dreaming and related phenomena

358

9. 10. 11. 12.

13.

14.

15.

16. 17.

18.

19. 20.

Alexandra Kratschmer and Adriënne Heijnen (Haraldsson 1978; Arnalds et al. 2007 and the survey Skrá 61. Draumar, fyrirburðir, spádómar of Iceland’s National Museum, 1985). Steffen Krogh/Aarhus, personal communication. Jette Larsen Persiani/Aarhus, personal communication. For different data regarding Faroese cf. note 46. The oblique case in Danish, a phonetically merged form of both prior accusative and dative, has probably to be analyzed as deriving from a historical accusative in this case, especially given the Icelandic situation, usually representing a conservative condition af Scandinavian; Steen Vikner/Aarhus, personal communication. In our Bible corpus F1-F4, we exclusively found the “old or literary” (cf. Larousse [Guilbert 1977], s.v. songe) songe ‘(a) dream’ and songer ’to dream‘, but not rêve and rêver, the unmarked forms in contemporary French (cf. Larousse 1977, s.v. rêve). The same version shows ‘to make a dream’ systematically in direct speech, reflecting the innovativity of the construction, felt probably as more informal at that time: il sogno che ho fatto ‘the dream that I have made’, 37:6, but also in 40:8 and twice in 41:11, IT3. But there have been exceptions already in Old Icelandic, with a growing tendency from the middle of the 19th century (Jónsson, Eythórsson 2005: 224f.): it is true for many verbs with human semantic roles in accusative and dative case that speakers use nominative or dative for accusative and nominative or accusative for dative, or even dative or accusative for nominative; among a longer series of concerned verbs, ‘to dream’ is one of the most stable ones though, still standing strong with its use of the dreaming subject in the accusative, and dative preferred over nominative as an alternative to accusative (Jónsson, Eythórsson 2005: 232). Note also the form mig dreymdi sjálfa mig ‘me [acc.] dreamed myself [acc.]’, ‘I dreamed about myself’. Which is completely ours; for e.g. Hrafnbjargarson/personal communication, there is no clear semantic role assignable to the human accusative subject in this or any of the other occurrences of double accusative constructions in Icelandic. Hinneh is a focus marker used in direct speech and narratives to draw the listener’s or reader’s attention to the piece of information following it. Its etymology is heavily discussed, but does as far as is known not comprise the idea of visuality (Sørensen 2003: 418-419) ASV: Pharaoh dreamed: and, behold, he stood by the river. [lit. ‘P. had a dream how he stood by the water’] Cf. also Plungian (2001: 354): “[…] an evidential supplement can always be seen in an epistemic marker”; for this author, the opposite does not automatically hold in all languages, but following him, there exist “modalized evidential systems”, viz. systems where evidential markers (grammatical, lexical, or



21.

22. 23. 24.

25. 26.

27. 28. 29. 30.

Revelative evidentiality in European languages

359

else) must code aspects of epistemic modality at the same time. This view of Plungian’s is critized by Aikhenvald (2007, note 3), who argues that “in the same language one evidential may have en epistemic extension, and another one may not”). The inferential domain can comprise subdomains as circumstantial inferences (based on circumstantial clues), generic inferences (based on world knowledge) and conjectures (lacking any sensory evidential basis; cf. Squartini (2008: 924) on Romance languages). Concerning the exact evidential nuances of sembler/paraître in French, cf. Nølke (2001:30), of sembrare/parere in Italian, cf. Kratschmer 2006 (synchronically) as well as Squartini 2006 (diachronically), and of parecer in Spanish, cf. Cornillie (2007:118). ASV: and it was as though it budded [lit. ‘and it seemed to me as though it became green’]. ASV: for, behold, we were binding sheaves in the field [lit. ‘it was [to] me as […]’]. In a similar way as regarding the ‘behold’-marker, this could equally be a traductorial issue: the original Hebrew form in the relevant verses is the conjunction k, with the possible readings ‘as if’, but also purely temporal ‘when’. Clines (1998: 348) gives as an example specifically Gen 40:10 and translates it into and when it budded. Note that the ASV writes and it was as though it budded. The Septuagint shows hosper, which can have the same two readings (comparative and temporal) as k and therefore did not cause problems of translation. The choice of the comparative reading ‘as if’ in certain translated versions could therefore be due to inaccurate translation. ASV: In my dream, behold, a vine was before me [lit. ‘It seemed [to] me in the sleep that […]’]. This example combines the alternative reality reading with an epistemic verb. This observation together with the fact that parallel contexts also can be presented as contexts of mere epistemic reserve (e.g. 40:10: Oc (mig siuntis) som det blef grønt, DA2), confirms, as we see it, the parallel analysis, proposed in Kratschmer (in press), of contexts of epistemic reserve and of alternative reality comparison (illustrated by two possible readings of the Italian verbs sembrare/parere ‘to seem’), attributing their difference to a mere degree of epistemic force (‘epistemic factor’): compare ‘the computer seems broken’ (epistemic reserve) vs. ‘the computer seems sentient’ (alternative reality comparison). ASV: In my dream, behold, a vine was before me [lit. ‘it dreamed me that a vine were before me’].. This use is also documented in our modern oral corpus, see 4.2.2.1. below. ASV: […] I also was in my dream, and, behold, three baskets of white bread were on my head. This use is equally documented in our modern oral copurs, see 4.1.1.2. below.

360

Alexandra Kratschmer and Adriënne Heijnen

31. E.g. Dardano and Trifone (1995: 354) for Italian, and Fernández Ramírez (1986: 282) for Spanish. 32. Cf. Dardano and Trifone (1995: 354). The imperfective preterite is also used as a fiction marker in children’s play (Fernández Ramirez 1986: 275-276., Togeby 1982: 315), as well as in the French prostasis of the present irrealis hypothetic period: Si j’avais [imperfect preterite] de l’argent, j’acheterais [present conditional] une voiture ‘if I had money, I would buy a car.’ (Italian and Spanish use the imperfective subjunctive in this context). 33. Marking thus punctual, concluded actions: fr./sp./it. Colombe/Colombo naissait/nacía/nasceva [imp.] en/nel 1451, ‘Columbus was born in 1451’. 34. ASV: and lo, my sheaf rose up. 35. The interference of narrative conventions with evidentiality is also noted by Aikhenvald (2004: 380): “The narrative genre as a macro-convention typically overrides all our preferences. Narrative conventions thus serve to narrow down the polysemy of evidentials depending on person and other factors.” 36. Jette Larsen Persiani/Aarhus, personal communication. 37. Per Overgaard/Aarhus, personal communication. 38. As correctly pointed out by the anonymous reviewer, these contexts do equally challenge the boundaries of the category of evidentiality, namely its (presupposed?) definition as marking the source of a piece of information presented as a full proposition expressed in clausal form. Our contexts enumerated here do admittedly show pronominal combinations (‘all this’) and pro-phrases (‘what he is about to do’), hence not semantcially full propositions but rather propositional variables (nominal and clausal). Nontheless, we still consider these contexts relevant to evidentiality matters, by two reasons. The first reason is that comparable contexts exist where the same revelative verbs are followed by semantically and syntactically full clauses, as in e.g. II Kings 8:13 (Jehovah hath showed me that thou shalt be king over Syria.; ASV). The second reason is that even in languages with grammaticalized evidentiality marking, these marker do not only occur in the standard contexts of full clauses, but can be attached to pro-phrasal elements, e.g. an interrogative marker standing in for a proposition (which by definition is uncomplete due to its interrogative status) as in the following example from Kham (mentioned in Aikhenvald 2004: 248: “the question is directed by the speaker to the addressee (second person), but it ‘has its origin outside the speech situation’: the author of the question is ‘he or she’ “): karao di why REP [= reportative; note by AK + AH] ‘(He or she wants to know) why’ 39. Information source marking verbs like speech verbs are often taken into consideration in evidentiality studies only when standing e.g. in parenthetical form, not when taking a clausal complement. It appears that this is due mostly to practical reasons, the body of data simply threatening to grow to unsur-



40.

41. 42.

43.

44.

45.

46.

47.

48. 49. 50.

Revelative evidentiality in European languages

361

mountable dimensions. This, of course, is not a valid reason for not investigating these structures at all, hence our decision to include revelative verbs here. We always chose the original etymology, even though not all lexemes are etymologically transparent any more to non linguist nativespeakers, as e.g. dan. åbenbare ‘to render bare in the open’, being more or less interpreted as an abstract ‘tell a secret’, the same way as engl. reveal is no longer felt as to express ‘to remove a veil from’ by non linguist native speakers. This collection is part of the Oral Archives of the Árni Magnússon Institute of Icelandic Studies. The construction appears for example in Landnámabók (Book of Settlement) of which the oldest manuscript survived dates from the 13th century. However, it is generally thought that Landnámabók was created in the 12th century. This mixing up of nominative and accusative is recognized in Iceland as a widespread grammatical deficiency and is designated in Icelandic with the general term þolfallssýki, literally, ‘the illness of accusative’. Also inversion between nominative and dative is a recognized problem and called þágufallssýki ‘illness of dative’. The following examples illustrate this: […] að þau hjón voru nýháttuð; þótti konunni sem maður kæmi að rúmstokknum. Þessi maður biður hana koma fljótlega á fætur (ÍÞÆ, I, 1954-1961: 17). ’[…] when the couple just had gone to bed, it seemed to the woman that a man had come to the bedside. This man asks her to get up quickly’; […] svo dreymir mig nokkuð eftir, bara skamm eftir jarðförina, þá finnst mér að ég fari gegnum dimm göng. ‘[…] then it dreamed me, somewhat later, just after the funeral, it seems to me that I am going through a long dark corridor’ (quote from an interview with an Icelandic woman of 70 years old, 2001). He was unsure though whether it should not rather be an accusative: mich [acc.] hat geträumt, which gives associations to the Icelandic case variations mentioned above. As a more anecdotical note, the linguist in this collaboration (AK), remembers her own mother using mir hat geträumt still about 20 years ago, experiencing it as slightly affectated at that time, while her mother now claims never to use anything else but ich habe geträumt. Regarding the concept of saliency, cf. e.g. Giora 1997; regarding a discussion of the concept of literal/”minimal” meaning (and among others saliency), cf. e.g. Ariel 2002. A statement not all Italians want to subscribe to though (Valentina Bambini, Pisa, personal communication). Cf. the parallel Icelandic construction mentioned in note 12. Ber (‘naked’ or ‘revealed’) dreyminn (masc. nom. adjective of the verb að dreyma ‘to dream’) means to be able to dream dreams that reveal knowledge on past, present and future events; cf. the interactive evidential pattern ‘to render bare in the open’ mentioned in 5.1.1.3.

362

Alexandra Kratschmer and Adriënne Heijnen

51. Note, here, the use of the subjunctive mood (cf. 4.1.1.2.3). 52. The Faroese situation provides an interesting parallel: In Faroese, a development from the construction meg droymde ‘it dreamed me [acc.]’ to eg droymde ‘I dreamed’ can be detected (cf. Poulsen 1997 as well as Jónsson and Eythórsson 2005: 227). This linguistic change seems to run parallel with a change in attitude towards dreaming, whereby dreaming increasingly is thought to take place within people’s head. Even though, for some, dreams have the potential to reveal knowledge, the successful integration of Christian doctrines in Faroese society has defined revealing dreams as superstition (see e.g. Joensen [1975: 139-147] for information on the role of dreams in Faroese society). 53. Following this tradition dating back to the 16th century (Cosentino [2003: 1]), different dream contents are associated with different numbers, conceived as secure bets for the next lottery extraction (e.g. mulino ad acqua ‘water mill’ = 7, mulino a vento ‘wind mill’ = 77; Cosentino [2003: 333]).

References Aikhenvald, Alexandra Y. 2004 Evidentiality. Oxford: Oxford University Press. 2007 Information source and evidentiality: what can we conclude? In Evidentiality between lexicon and grammar; Mario Squartini (ed.), 209-282. Ariel, Mira 2002 The demise of a unique concept of literal meaning. Journal of Pragmatics 34: 361-402. Arnalds, Ásdís A., Ragna Benedikta Garðarsdóttir and Friðrik Jónsson 2007 Könnun á íslenskri þjóðtrú og trúarviðhorfum. Reykjavík: Félagsvísindastofnun, Háskóla Íslands. Capelli, Gloria 2007 “I reckon I know how Leonardo da Vinci must have felt ….” Epistemicity, evidentiality and English verbs of cognitive attitude. Pari: Pari Publishing. Clines, David J.A.I. (ed.) 1998 The dictionary of Classical Hebrew. Sheffield Acedemic Press. Cornillie, Bert 2007 The continuum between lexical and grammatical evidentiality: a functional analysis of Spanish parecer. In Evidentiality between lexicon and grammar; Mario Squartini (ed.), 109-128. Cosentino, M. (ed.) 2003 La vera smorfia napoletana. Sogni e numeri per vincere al lotto. Florence: Giunti.



Revelative evidentiality in European languages

363

Dardano, Maurizio, and Pietro Trifone 1995 Grammatica italiana con nozioni di linguistica. Terza edizione. Bologna: Zanichelli. Drosdowsky, Günther (ed.) et al. 1984 Duden - Grammatik der deutschen Sprache. 4. Auflage. Mannheim: Dudenverlag. Fernández Ramírez, Salvador 1986 Gramática Española. 4. El verbo y la oración. Madrid: Arco Libros. Giora, Rachel 1997 Understanding figurative and literal language: The graded salience hypothesis. Cognitive Linguistics 7: 183-206. Guilbert, Louis (éd.) Grand Larousse de la langue française. Tome sixième. Paris: Libraire Larousse. Haraldsson, Erlendur 1978 Þessa heims og annars. Könnun á dulrænni reynslu Íslendinga, trúarviðhorfum og þjóðtrú. Reykjavík: Bókaforlagið Saga. Heijnen, Adriënne 2005a Dream Sharing in Iceland. Ph.D. diss. Århus: University of Aarhus. 2005b Dreams, Darkness and Hidden Spheres. Paideuma 51: 193-207 2007 Stregkoder, sparegrise og barnedåb. Det islandske selv i entreprenørskab og navngivning, Tidsskriftet Antropologi 55: 21-35. in press Name-giving by the dead. An argument against the geneticisation of relatedness, Social Anthropology. forth. The Life of Dreams. Sharing Sensory Experience in Iceland. Princeton: Princeton University Press? Hrafnbjargarson, Gunnar Hrafn 2004 Oblique Subjects and Stylistic fronting in the History of Scandinavian and English: The Role of IP-Spec. Ph.D. diss., Århus: University of Aarhus, Aarhus. Jochelson, Vladimir Illich 1905 Essay on the grammar of the Yukaghir Language. American Anthropological Supplement 7: 369-424. Joensen, Jóan Pauli. 1975 Færøske sluppfiskere.Etnologisk undersøgelse af en erhvervsgruppes liv. Tórshavn: Føroya Fróðskaparfelag. Jónsson, Jóhannes Gísli, and Thórhallur Eythórsson 2005 Variation in subject case marking in Insular Scandinavian. Nordic Journal of Linguistics 28 (2): 223-245. Kelchner, Georgia Dunham 1935. Dreams in Old Norse literature and their affinities in folklore. Cambridge: Cambrodge University Press.

364

Alexandra Kratschmer and Adriënne Heijnen

Kratschmer, Alexandra 2006 Che te ne sembra? Semantica e pragmatica delle costruzioni italiane con sembrare/parere. In Atti del XVI Congresso dei Romanisti Scandinavi, Copenaghen e Roskilde, 24-27 agosto 2005, Michel Olsen, and Erik Zwiatek (eds.). http://www.ruc.dk/cuid/publikationer/ publikationer/XVI-SRK-Pub/MMH/MMH07-Kratschmer in press Catégorisation vs comparaison : une question de quantification épistémique. Modèle interprétatif sémantico-pragmatique modulaire des constructions italiennes avec sembrare /parere. Cahiers Chronos. Nølke, Henning 2001 La dilution linguistique des responsabilités. Essai de description des marqueurs évidentiels il semble que et il paraît que. In Le regard du locuteur 2. Pour une linguistique des traces énonciatives, Henning Nølke (ed.), 17-34. Paris: Kimé. Nuyts, Jan 2001 Epistemic Modality, Language, and Conceptualization. Amsterdam/Philadelphia: John Benjamins. Poulsen, Jóhan Hendrik W. (ed.) 1997 Føroysk orðabók. http://obg.fo/ Plungian, Vladimir A. 2001 The place of evidentiality within the universal grammatical space. Journal of pragmatics 33: 349-357. Squartini, Mario 2006 Hearsay and quotatives in the diachronic evolution of the Italian pare/sembra evidential pair. Abstract at the 39th annual meeting of the Societas Linguistica Europaea (SLE), 30th august – 2nd september 2006, Bremen. 2007a Investigating a grammatical category and its lexical correlates. In Evidentiality between lexicon and grammar; Mario Squartini (ed.), 1-6. 2007b (ed.) Evidentiality between lexicon and grammar. Italian Journal of Linguistics 19 (1). 2008 Lexical vs. Grammatical evidentiality in French and Italian. Linguistics 46 (5): 917-947. Sørensen, Dan Enok Bibelsk-hebraisk grammatik. Kopenhagen: Hovedland. Togeby, Knud 1982 Grammaire Française. Vol. II. Copenhague: Akademisk Forlag. Willerslev, Rane 2007 Soul Hunters. Hunting, Animism, and Personhood among the Siberia Yukaghirs. Berkeley: University of California Press.



Revelative evidentiality in European languages

365

Source material: Bible versions – books Danish 1550

1647 Dutch 1734

Icelandic 1841

Faroese 1949

Kong Christian the Third’s Bible translation, primariylbased on Luther’s German version. Facsimile-Edition. Det Dansk Sprog- og Litteraturselskab. 1917. Copenhagen. Bishop Svane’s revised version of Hans Povl Resen’s 1607 translation, based on the original languages. Copenhagen. Nederduytse bijbel. Dat is het oude en nieuwe Testament na de Oversetting van D.M. Lutherus met Alle desselfs gelykluydende Texten, en Inhoud der Kapittele tot Dienst van de Christelyke Gemeentes, toegedaan d'Onveranderde Augsburgsche Confessie van verscheyde Drukfouten vebetert. t' Amsterdam gedrukt voor het Lutherse Weeshuys en zijn te Bekoomen by de Eygenaar van 't Prvilegie van het Psalm boek. Amsterdam 1734. Biblía. Það er heiløg ritning. I 5ta sinni útgefin, á ný yfirskoðuð og leiðrétt, að tilhlutun ens íslenðska Biblíu félags. Viðeyar Klaustri. 1841. BÍBLIAN - HIN HEILAGA SKRIFTIN í týðing Victor Danielsens fra 1949©. Endurskoðað 1974©. Hin Føroyski Bíbliugrunnurin eigur upphavsrættindini til hesa týðingina av Bíbliuni.

Source material: Bible versions – online http://scripturetext.com [as accessed in december 2007; bibliographical indications as listed on the website, with varying degree of informativity, authors’ comments in square brackets]: Hebrew Orig. Westminster code with vowels and consonants: The BHS Hebrew Bible, Leningrad Codex has been made available through the hard work of several organizations, including the Westminster Theological Seminary, University of Pensylvania CCAT, German Bible Society, and Unbound Bible. Septuagint The Septuagint /LXX /Greek Old Testament has been made available through the hard work of several organizations, including the Univer-

366

Alexandra Kratschmer and Adriënne Heijnen

Vulgata German 1545

1871 1905 1912 1951 Danish 1931

Dutch 1618

French 1707

1744 1859 1894 Spanish 1569 1909 Mod.

sity of Pensylvania CCAT, German Bible Society, and Unbound Bible. Biblia Sacra Vulgata [translated from Greek by Hieronymus, 4th cent. AD] Luther: Made available in electronic format by Michael Bolsinger at http://www.luther-bibel-1545.de (see here for the most recent versions in text and HTML format). -- It was converted to SWORD format by Matthias and Joachim Ansorg. This text has been modernized, and the Johannine Comma added to 1 John 5. 1871 Elberfelder Bible 1905 Elberfelder Bible 1912 Lutherbibel 1951 Schlachter Bible Det Gamle Testamente af 1931: copyrighted Etext (c)1931 Det Danske Bibelselskab. -- The Old Testament of 1931: copyrighted Etext (C) by Det Danske Bibelselskab 1931. (The Danish Bible Society). The New Testament is Public Domain. Staten Vertaling: BIJBEL - DAT IS DE GANSE HEILIGE SCHRIFT -- DOOR LAST VAN DE HOOG-MOGENDE HEREN STATEN-GENERAAL DER VERENIGDE NEDERLANDEN EN VOLGENS HET BESLUIT VAN DE SYNODE NATIONAAL GEHOUDEN TE DORDRECHT IN DE JAREN 1618 EN 1619 UIT DE OORSPRONKELIJKE TALEN IN ONZE NEDERLANDSE TAAL GETROUWELIJK OVERGEZET - BEVATTENDE AL DE KANONIEKE BOEKEN VAN HET - OUDE EN NIEUWE TESTAMENT - Based on electronic edition from http://www.coas.nl/ bijbel. La Bible David Martin, 1744. Downloaded from http://desmond. oshea.free.fr/BibleMartin/Bible%20Martin%201744/Bible%20Marti n%201744.html French: Osterveld 1744. Text downloaded from www.theophilos.sk French: Darby. Text downloaded from www.theophilos.sk. Version Louis Segond 1910 (LSG) -- Texte libre de droits. Spanish: Sagradas Escrituras (1569) Spanish: Reina Valera (1909) Spanish: Modern



Revelative evidentiality in European languages 1986

2005

367

Las citas bíblicas son tomadas de La Biblia de las Américas © 1986, 1995, 1997 by The Lockman Foundation, La Habra, Calif, http://www.lockman.org. Usadas con permiso. Las citas bíblicas son tomadas Nueva Biblia de los Hispanos © 2005 by The Lockman Foundation, La Habra, Calif, http://www.lockman. org. Usadas con permiso.

Italian 1649 1649 Italian Giovanni Diodati Bibbia 1927 Italian: Riveduta Bible (1927) English [for glossing purposes only] 1960 New American Standard Bible Copyright © 1960, 1962, 1963, 1968, 1971, 1972, 1973, 1975, 1977, 1995 by The Lockman Foundation, La Habra, Calif. All rights reserved. For Permission to Quote Information visit http://www.lockman.org http://www.dbg.de/[as accessed in nov. 2007 and jan. 2008] German 1984 Bibeltext in der revidierten Fassung von 1984. Herausgegeben von der Evangelischen Kirche in Deutschland. Durchgesehene Ausgabe in neuer Rechtschreibung. © 1999 Deutsche Bibelgesellschaft Stuttgart. Die Lutherbibel wurde in den Jahren 1952-1984 überarbeitet. 2000 Gute Nachricht Bibel. Revidierte Fassung der „Bibel in heutigem Deutsch“. Durchgesehene Ausgabe in neuer Rechtschreibung. © 2000 Deutsche Bibelgesellschaft Stuttgart. http://www.bibelselskabet.dk/[as accessed in feb. 2008] Danish 1992 Bibelteksten er hentet fra den autoriserede oversættelse, © Det Danske Bibelselskab 1992. http://www.hverdagsbibel.dk/ [as accessed in feb. 2008] Danish 2002 Hverdagsbibel [Bibel in ”every day language”]: den nye, gennemreviderede udgave af hele Bibelen på Hverdagsdansk udkom den 1. november, 2007. Den kan læses online sammen med en række andre oversættelser fra International Bible Society på www.bibleserver. com. http://www.snerpa.is/net/biblia/biblia.htm [as accessed in dec. 2007] Icelandic 1815 Hið íslenzka biblíufélag. Biblían Heilög ritning. 1815. [identical to the following 1966-edition: Biblía. Það er heilög ritning. Ný þýðing úr frummálunum. Reykjavík: Á kostnað hins íslenzka biblíufélags. Prentsmiðjan Oddi H.F. 1966] http://www.laparola.net/[as accessed in jan. 2008]

368

Alexandra Kratschmer and Adriënne Heijnen

Italian 1821 1974

Italian 2008

Nuova Diodati, Revisione (1991), edizione La Buona Novella Brindisi. [modernized Diodati-version, under continuous revision] La versione C.E.I., seconda edizione (1974), a cura della Conferenza Episcopale Italiana. La Bibbia di Gerusalemme ha lo stesso testo, con l'aggiunta di alcune note prese dalla Bibbia di Gerusalemme francese. È una versione approvata dalla Chiesa cattolica romana http://www.biblegateway.com [as accessed in jan. 2008] La versione C.E.I., a cura della Conferenza Episcopale Italiana.

Subject index Adposition 65-67, 99, 109 Adverb 65f., 69, 78, 90, 92f., 98, 109, 112, 115, 117, 134, 143, 149-152, 202, 225, 234f., 282f., 292, 298, 300, 304, 311, 314-318, 320f., 325, 328330, 343 Agglutinative 7, 69, 72, 111, 118, 192 Ambiguity 8, 133, 138, 152-154, 177180, 191, 199 Anthropology 333, 336, 358 Auxiliary 67, 71, 78, 81f., 84, 138, 282, 289, 291, 313 Basque 133-155 Complementation 100, 190, 204, 217, 222, 251 Complementizer 65-67, 77, 89, 92, 99f., 107f., 112, 117, 190, 205, 210, 213215, 217, 259, 264, 268-270, 278 Conjunction 67f., 76, 98, 108, 114, 131, 155, 189, 202, 261f., 268, 325, 342, 352, 356 Construction 4, 18, 29, 40, 61, 66, 70, 75f., 82, 107, 117, 166, 176, 210, 241, 251, 257, 262, 286, 313, 339341, 357 Croatian 95, 98, 119 Danish 337, 339, 343, 356 Deixis 9, 229, 232, 239, 244 Embedding 165, 168, 171, 174f., 182, 189, 191, 207, 213, 216f. English 2, 5, 9, 24, 48, 85-87, 94, 101, 114, 133, 140, 163, 169, 191, 194,

200, 202, 251, 254f., 259f., 264f., 278, 312, 337 Epistemic modality 46, 61f., 161f., 221223, 230, 235f., 308, 312, 321, 339 Evidentiality auditory 33, 35, 91, 236 direct 26, 28f., 32, 34-37, 41, 90, 169, 222, 228, 272, 279, 309 hearsay 7, 10, 25, 29, 35f., 61-116, 134f., 138, 141, 144-146, 148150, 187, 192, 211, 233, 236, 252, 277, 330 firsthand 28, 35, 37, 331 indirect 26, 28f., 34-37, 41, 46, 69f., 74, 78, 82, 86, 91, 106, 115, 169, 191f., 272 inferential 8, 10, 23, 30, 34, 36f., 40, 48, 63, 69f., 72, 74, 77-80, 84, 86, 91, 106, 112, 117, 154f., 182, 192-195, 205, 207, 209, 222, 311f., 325, 339, 352 interactive 227, 341, 344, 349, 353 personal 26, 29-32, 36f., 45, 223, 311 quotative 29, 35, 78, 98, 100, 102, 104, 113, 192, 211, 215, 237, 330 reportive/ reportative 29, 35-37, 39, 48, 61f., 66, 69, 74, 76, 78, 80, 82-85, 91, 93, 99, 103f., 108, 112, 115, 117, 160, 166-168, 179f., 192, 194, 198, 201, 209 revelative 25, 329, 334, 341, 346, 351 sensory 33f., 37, 330 third-hand 35-36 visual 35-37, 40, 223, 231, 236, 277,

370

Subject index

331, 338f., 351 Evidentiality marking discoursive 330f. grammatical 12, 16, 19f., 30, 36, 38, 63-65, 67, 74, 78, 82, 135, 140, 146, 165, 171, 191, 221, 247, 330f., 341, 351 lexical 3-8, 18, 62, 65f., 108, 115, 133, 135, 146, 149f., 165, 277, 330, 341 Faroese 333, 335, 338, 351 French 48, 80, 154, 204, 234, 277, 283, 335f., 339, 352 German 23, 48, 68, 77, 79, 81f., 85, 88, 103, 117, 159f., 166f., 171, 182, 239, 247f., 250, 254-256, 260f., 267, 272, 333, 335, 338f., 345f., 350f. Grammaticalization 3f., 22, 63, 6, 110, 125, 264, 278 Imperfect 69, 73, 77, 237f., 240, 340f., 347 Italian 78, 80, 86, 112, 230, 333, 335f., 339f., 346f., 350f. Lexicalization 2f., 63 Metaphor 251, 253, 263, 266, 271, 353 Modal verb 3, 166, 182, 273, 282 Parenthetical 66, 92f., 105f., 136, 152, 165, 169, 174f., 178, 231, 258f., 270, 290f., 293, 295, 300 Particle 5, 7, 40, 65f., 76, 80, 92f., 96, 98f., 102, 105, 107, 111-113, 116, 135, 140, 147, 225f.

Perception verb 247-251, 256, 259, 269, 273, 338, 346, 351 Person 42f., 72, 88, 104, 149, 254, 283, 289 Polish 84, 89f., 94, 97, 103, 106, 108, 112, 117 Polysemy 84, 248, 251f., 252, 259, 273, 282 Portuguese 77, 80 Predicative 65f., 69, 88f., 112, 114, 206, 212, 214, 282 Presupposition 165, 178-181, 184 Reasoning 30, 35, 37, 192, 208, 290, 293, 298, 301, 309, 313, 320, 325 Romance 48, 67, 77, 79, 82, 87, 112, 115, 117, 221, 230, 235, 335, 340, 344, 348 Russian 25, 31, 90, 97, 99, 102, 106, 108, 110, 112, 117 Scope 63, 65, 67, 89, 92, 96-99, 105, 108, 135f., 150, 152-154, 198, 264, 321, 330 Spanish 80, 86, 94, 133-149, 153, 237f., 242, 307, 309f., 313, 315, 335, 339, 349, 351 Strategy, evidential 10, 62, 135 Subordination 188, 190f., 208, 214, 219 System, evidential: modalized 7, 48 non-modalized 7, 48 Temporality 1, 4, 239 Tense 4, 7, 19f., 22f., 42, 49, 70-74, 78, 80f., 87, 118, 139, 191, 193, 200, 203, 228, 242, 288, 344

 Turkish 8, 19, 27, 31, 63, 69f., 74, 111, 187, 189-192, 200, 202 Volition 7, 44f., 81, 87f., 166, 249

Subject index

371

Witness 28, 31, 37, 72, 102f., 136, 159, 192f., 195, 226, 263, 273, 280, 290, 292, 297, 301, 311, 352