Causal Categories in Discourse and Cognition 9783110224429, 9783110224412

Citation preview

Causal Categories in Discourse and Cognition

≥

Cognitive Linguistics Research 44

Editors Dirk Geeraerts John R. Taylor Honorary editor Rene´ Dirven Ronald W. Langacker

Mouton de Gruyter Berlin · New York

Causal Categories in Discourse and Cognition

Edited by Ted Sanders Eve Sweetser

Mouton de Gruyter Berlin · New York

Mouton de Gruyter (formerly Mouton, The Hague) is a Division of Walter de Gruyter GmbH & Co. KG, Berlin

앝 Printed on acid-free paper 앪

which falls within the guidelines of the ANSI to ensure permanence and durability.

Library of Congress Cataloging-in-Publication Data Causal categories in discourse and cognition / edited by Ted Sanders, Eve Sweetser. p. cm. ⫺ (Cognitive linguistics research ; 44) Includes bibliographical references and index. ISBN 978-3-11-022441-2 (hardcover : alk. paper) 1. Psycholinguistics. 2. Causation. 3. Causative (Linguistics) I. Sanders, Ted, 1963⫺ II. Sweetser, Eve. P37.C38 2009 4011.9⫺dc22 2009045533

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.

ISBN 978-3-11-022441-2 ISSN 1861-4132 쑔 Copyright 2009 by Walter de Gruyter GmbH & Co. KG, D-10785 Berlin All rights reserved, including those of translation into foreign languages. No part of this book may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Typesetting: ptp, Berlin Printed in Germany

Acknowledgements This book volume results from an inspiring international workshop on Causal connectives in Discourse and Cognition, which we co-organized at the Department of Linguistics of the University of California at Berkeley in April 2007. Speakers included Barbara Dancygier, Gilles Fauconnier, Jos´e Sanders and the organizers, while Dan Slobin and Len Talmy, who were invited as respondents, joined us in a lively discussion, together with many colleagues from several countries and backgrounds. Some chapters in this book volume are directly based on these papers and discussions. Others were invited later. We took care to come up with a coherent volume of contributions, all of which were reviewed anonymously by two referees and the editors. We believe this procedure has lead to a high quality book volume. We would like to thank all authors, as well as speakers and respondents at the original workshop, for contributing to the ideas presented in this book volume, and for acting as reviewers. We are especially grateful to Bonnie Howe (UC Berkeley) for additional reviewing. We gratefully acknowledge the financial support of the Netherlands Organization for Scientific research NWO, through NWO-Vici-grant 277-70-003, awarded to Ted Sanders, enabling his 6 month’s stay at Berkeley during the Spring of 2007. Our joint work was further supported by the Utrecht institute of Linguistics OTS and the University of California – Universiteit Utrecht program for collaborative research, while the April workshop was supported by the Department of Linguistics at UC Berkeley and by the University of California’s Intercampus Travel Fund. We are grateful to Dirk Geeraerts for his initiative to publish the book volume in this Mouton de Gruyter series, and to him and Birgit Sievert (Mouton) for managing a speedy and thorough printing process. Finally, we would like to thank Fang Li (Utrecht) for her editorial assistance. Ted Sanders, Eve Sweetser Utrecht/Berkeley, October 2009

Table of contents Acknowledgements

v

List of Contributors

ix

Introduction: Causality in language and cognition – what causal connectives and causal verbs reveal about the way we think Ted Sanders and Eve Sweetser

1

Causality, cognition and communication: A mental space analysis of subjectivity in causal connectives Ted Sanders, Jos´e Sanders and Eve Sweetser

19

Causal Connectives in Dutch Biblical Translations A cognitive linguistic approach Jos´e Sanders

61

Causes and consequences: Evidence from Polish, English, and Dutch Barbara Dancygier

91

Categories of subjectivity in Dutch causal connectives: a usage-based analysis Ninke Stukker, Ted Sanders and Arie Verhagen

119

Causes for causatives: the case of Dutch doen and laten Dirk Speelman and Dirk Geeraerts

173

Causal categories in discourse – Converging evidence from language use Ted Sanders and Wilbert Spooren

205

Index

247

List of Contributors

Barbara Dancygier Department of English University of British Columbia 397-1873 East Mall Vancouver, BC V6T 1Z1 Canada [email protected] http://faculty.arts.ubc.ca/ bdancygier/ Dirk Geeraerts Department of Linguistics University of Leuven, Faculty of Arts, Box 3308 Blijde Inkomststraat 21, BE-3000 Leuven, Belgium [email protected] http://wwwling.arts.kuleuven.be/ qlvl/dirkg.htm Ted Sanders Utrecht institute of Linguistics OTS Universiteit Utrecht Trans 10 NL-3512 JK Utrecht The Netherlands [email protected] http://www.let.uu.nl/ ∼ted.sanders/personal/ index.php Jos´e Sanders Centre for Language Studies Radboud University Nijmegen P.O. Box 9103 NL - 6500 HD Nijmegen The Netherlands [email protected]

Dirk Speelman Department of Linguistics University of Leuven, Faculty of Arts, Box 3308 Blijde Inkomststraat 21, BE-3000 Leuven, Belgium [email protected] http://wwwling.arts.kuleuven.be/ qlvl/dirks.htm Wilbert Spooren Dpt. of Language & Communciation VU University Amsterdam De Boelelaan 1105 NL-1081 HV Amsterdam The Netherlands [email protected] Ninke Stukker Utrecht institute of Linguistics OTS Universiteit Utrecht Trans 10 NL-3512 JK Utrecht The Netherlands [email protected] Eve Sweetser Dept. of Linguistics University of California at Berkeley Berkeley CA 94720-2650 USA [email protected] Arie Verhagen Leiden University Centre for Linguistics P.N. van Eyckhof 1 NL - 2311 BV Leiden The Netherlands [email protected] http://www.arieverhagen.nl

Introduction: Causality in language and cognition – what causal connectives and causal verbs reveal about the way we think Ted Sanders and Eve Sweetser 1. What language use tells us about causal categories All languages of the world provide their speakers with connectives to express causal relations in discourse – indeed, although no physicist has found “causation” out in the world, all humans in all cultures seem to interpret and describe the world in terms of causal relations. As in other semantic domains, the cognitive scientist and the linguist are therefore interested in how much of this causal modeling is specific to a given culture and language, and how much is characteristic of general human cognition. Causal connectives and causative auxiliaries are among the salient markers of causal construals. Speakers of English, for example, can choose between because and since or between therefore and so. How different are these from the choices made by Dutch speakers, who speak a closely related language, but (unlike English speakers) have a dedicated marker for non-volitional causality (daardoor)? On another grammatical level, speakers may use causal auxiliary verbs, such as make and let to mark causal relations expressed within one clause – but how different are these from laten (related to “let” both etymologically and semantically) and doen ‘do, make’ in Dutch? It is also well known that at a young age, children learn to have very different models of human/animate volitional causation (a person throwing an object) as opposed to inanimate object causation (a branch falling from a tree) (Boyer 1996; Sperber, Premack, and Premack 1996). This could lead us to expect that crossculturally, there will be some important contrasts between causal categories, which are part of a universal repertory potentially named by any language. Such a view would be similar to many analyses of color terms – it is not the case that every language must name every human visual color distinction, but rather that color vision provides a universal perceptual basis for the range of possible linguistic categories. Language users often systematically prefer one lexical item rather than another (even highly similar) one to express a certain type of causal relationship. Such choices could provide a window on speakers’ cognitive categorizations of causality. Studies of the linguistic categories apparent in people’s everyday lan-

2

Ted Sanders and Eve Sweetser

guage use have already produced many interesting insights into the working of the mind in other domains (see, for instance, Lakoff 1987; Lakoff and Johnson 1999). The linguistic study of the meaning and use of causal connectives and auxiliaries may reveal insights into human categorization of causality. Sweetser (1990) introduced the categories of content, epistemic and speech act use of causal conjunctions like because and since are illustrated in (1) – (3). (1)

John came back because he loved her. (i.e. the loving caused the return)

(2)

The neighbors are not at home because the lights are out. (i.e. the observation that the lights are out causes the conclusion that the neighbors are away)

(3)

Since you’re so smart, when was George Washington born? (i.e. the question is presumed to be motivated or enabled by the addressee’s claim to superior intelligence)

Similar distinctions have been dominant in many existing classifications of coherence relations – meaning relations that exist between discourse segments, e.g. Cause-Consequence relations between events as opposed to the relationship between premises or arguments in a Claim-Argument or Conclusion (Argument-Claim) relation (Sanders, Spooren, and Noordman 1992, 1993). Like Sweetser, these studies are sensitive to a contrast between content relations (sometimes also called ideational, external, or semantic relations), epistemic relations, and speech-act relations. In the first type of relation, segments are related because of their propositional content, i.e., the locutionary meaning of the segments. They describe events that cohere in the world. Epistemic relations relate speaker’s reasoning and conclusions (see [2]), and (3) illustrates speech act use. Over the last 20 years or so, the field of discourse studies has witnessed important progress in the linguistic study of connectives (see the contributions to Couper-Kuhlen and Kortmann 2000; Risselada and Spooren 1998; Spooren and Risselada 1997; Knott, Sanders, and Oberlander 2001; Sanders, Schilperoord, and Spooren 2001). By focusing on the way in which one, crucial, conceptual relation – that of Causality – is instantiated linguistically, this book volume seeks to bring this discussion one step further.

Introduction: Causality in language and cognition

3

2. Cross-linguistic comparison of causal categories To what extent does causal connective use show causal categorization? The study of the meaning and use of connectives like these, in several languages, is a central topic in this book volume. Over the last 15 years, a number of publications have dealt with the question whether connectives specialize in expressing certain types relations. In English, backward causal relations (first Consequence, then Cause) can be expressed by because, which may be used to express all interpretations, while since seems to specialize in epistemic and speech act use. Similar observations have been made for German, French and Dutch connectives, (see Pit 2003; Sanders 2005; see also contributions to special issues like Risselada and Spooren 1998; Spooren and Risselada 1997; Knott, Oberlander, and Sanders 2001), with the difference that these languages have a more differentiated repertoire of connectives than English seems to have. The same kind of differences have been observed for forward causal connectives such as English that’s why and so, Dutch daarom and daardoor and French de ce fait and c’est pourquoi on the one hand and alors and donc on the other (Pander Maat and Degand 2001; Degand and Pander Maat, 2003; Jayez and Rossari 2001; Sanders 2005). The general picture emerging from these studies is that connectives do specialize, although their semantic interrelations are more subtle than a simple one-to-one assignment from connectives to classes of coherence relations would suggest (Knott and Dale 1994; Knott and Sanders 1998; Pander Maat and Sanders 2006). Closer comparison of such related languages as English, Dutch, French and German can show clear differences in the way these languages “cut up” the domain of causality by choosing different markers for different relations. The contrast between these languages enables us to examine the constraints on conceptualization and labelling of causal relations. Therefore, the first leading question in this book volume is: What parameters of categorization shape the use of causal connectives and auxiliary verbs across languages? This question is taken up in all contributions. English and Dutch are studied in each chapter, in more or less detail, whereas Polish connectives are studied by Barbara Dancygier in chapter 3, in comparison with the other two languages.

4

Ted Sanders and Eve Sweetser

3. Characterizing the categories: subjectivity, perspective and mental spaces In recent years, we have also seen proposals to replace distinctions like content, epistemic and speech act domains by a subjectivity scale of speaker involvement (Pander Maat and Degand 2001). This scale is a continuum on which content relations such as Cause-Consequence are maximally objective, whereas epistemic relations are very subjective. Volitional causal relations such as the Reason-relation in John wanted to leave. He was tired hold an intermediate position. Some corpus evidence may be found in the distribution of Dutch and French connectives, since the notion of subjectivity, i.e., the amount of speaker involvement – to what extent is the speaker responsible for the utterance? – seems to provide an explanation for differences in meaning and use of causal connectives like Dutch daardoor ‘as a result’, daarom ‘that’s why’, and dus ‘so’ (Pander Maat and Sanders 2000, 2001). In the case of the non-volitional daardoor (see [4]), for instance, the causality is located outside of the speaker as a subject-of-consciousness. There is a minimal amount of speaker involvement. In the epistemic use of dus in (6) and the volitional use of daarom in (5), a subject-of-consciousness can be identified, either the current speaker or the actor. (4)

Er was een lawine geweest op Roger’s pass. Daardoor was de weg geblokkeerd. ‘There had been an avalanche at Roger’s pass. As a result, the road was blocked.’

(5)

Daan wilde op tijd thuis zijn. Daarom vertrok hij om 5 uur. ‘Daan wanted to be home in time. That is why he left at 5 o’clock.’

(6)

Het waren grote grijze vogels, die veel lawaai maakten. Dus het moeten wel kraanvogels geweest zijn. ‘They were large grey birds that made a lot of noise. So it must have been cranes.’

Proposals such as these illustrate the unmistakable tendency in recent textlinguistic work to use the notions of subjectification and perspective. This tendency goes back on Ducrot (1980), who already stressed the diaphonic nature of discourse (in the French tradition the seminal work by Anscombre and Ducrot 1983, and the Groupe Lambda-l should also be mentioned). Even in monologic texts, traces can be found of other “voices” – information that is not presented as fact-like, but as coming from a particular point-of-view, either the current speaker’s (subjectified information, in the terminology of J. Sanders and Spooren

Introduction: Causality in language and cognition

5

1997) or another cognizer’s (perspectivized information). Cognitive Linguistics has a large role to play in the development of this line of work, because of the key role it attributes to processes of subjectification in natural language, but also because it allows for a dynamic approach to connectives “as processing instructors”. Fauconnier’s Mental Space framework is very suitable to model this type of phenomena, as has been suggested by Dancygier and Sweetser (2000, 2005), Verhagen (2000, 2005), and Sanders and Spooren (2007). Fauconnier (1994) treats connectives as one of the so-called space-builders, that is, linguistic expressions that typically establish new Mental Spaces. Mental Spaces are mental constructs set up to interpret utterances, “structured, incremental sets [. . . ] and relations holding between them [. . . ], such that new elements can be added to them and new relations established between their elements” (Fauconnier 1994: 16). An example of a connective acting as a spacebuilder is the if-then conditional, as in If I were a millionaire, my VW would be a Rolls. An expression like if p then q sets up a new mental space H in which q holds. In other words, if I were a millionaire is the space builder and in this new space my VW from the initial space is identified with the Rolls in the new space. Dancygier and Sweetser (2005) have shown in detail how conditional constructions can be analyzed in an MST framework (see also Fauconnier 1994: chapters 3 – 4; and Sweetser 1996). Does such an approach work for causals? In chapters 2, 3 and 4 of this book volume, Mental Spaces Theory (from now on MST) figures prominently. It is used to further clarify the causal categories expressed in connective use across languages. In chapter 2, Sanders, Sanders and Sweetser discuss three Dutch causal connectives expressing forward causality daardoor, daarom and dus. As we have just shown, this part of the lexicon of Dutch language users clearly illustrates the categorical distinctions. The meaning and use of dus, daarom and daardoor were investigated in several studies over the last 15 years (see especially Degand 2001; Pander Maat and Sanders 2000, 2001; Pander Maat and Degand 2001; Stukker 2005). In most studies, text fragments were selected to form a newspaper corpus, consisting of different text types: argumentative / persuasive as well as descriptive / informative texts. The methodology usually involved three steps. First, the possible relational interpretations of fragments was determined without connectives, by examining possible and impossible paraphrases using explicit connectives (John wanted to leave. He was tired. might be paraphrased as John wanted to leave because he was tired). Then, it was investigated how often a given connective expressed a certain relation in corpus data. In the final step, it was checked whether the original connective could be substituted by another. This substitution method is a way of testing semantic intuitions (Knott and Dale 1994; Knott and Sanders 1998). The questions are:

6

Ted Sanders and Eve Sweetser

Does substitution lead to a sequence that is still acceptable? And, if acceptable, does the relational interpretation change as a result of this substitution? The findings, formulated in terms of the relations the three connectives can and actually do express, can be summarized as follows: – Daardoor can only express relations of the content non-volitional type; – Dus can express content volitional, epistemic, but not content non-volitional relations. It most often expresses epistemic relations; – Daarom can express content and epistemic relations. It most often expresses content volitional relations. Note that speech act and meta-linguistic relations are absent from this overview, because they did not appear in the corpora of written text. In addition to corpus studies, Pander Maat and Sanders (2001) presented experimental data. They found that Dutch speakers show clear patterns of preferences when asked to choose the best-fitting forward causal connective in natural discourse fragments. Dus is considered more appropriate when the distance between the speaker and actor / concluder is small, or when they are even identical, as in the first-person (I) example (7). Daarom fits better when the distance between speaker and the textual protagonist increases, as in the third-person example (8). In the case of implicit Speaker/Concluders, the distance between SOC and Speaker is smallest and the preference for dus at a maximum. Pander Maat and Sanders also concluded that a Subjectivity account rather than a Domain of use account explains the choice for dus versus daarom. Dus is not more appropriate in epistemic relations in general; it only fits better in the case of first-person SOC’s. In other words: SOC-Speaker distance seems to overrule domain differences. Moreover, the fact that dus is more appropriate in epistemic relations with implicit first-person concluders (9) than in those with explicit first-person concluders (10) can not be explained in term of domain differences. (7)

The weather-forecaster predicted that there will be 10 degrees of frost. I will dus not come for a walk.

(8)

Willem heard that there will be 10 degrees of frost. Willem will daarom not come for a walk.

(9)

Yesterday evening I did not see the lights burning in our neighbors’ house. Dus I think that they haven’t returned yet from their holiday.

(10)

Yesterday evening Alex did not see the lights burning in our neighbors’ house. Daarom he thinks that they haven’t returned yet from their holiday.

Introduction: Causality in language and cognition

7

All in all, data suggest the following organization of the Dutch causal connective lexicon. Table 1. How Dutch speakers cut up forward causality; a summary from corpus and experimental studies.

no SoC non-volitional content SoC volitional content SoC

Epistemic

large SoC-S distance (3rd person) small / no SoC-S distance (I )

DAARDOOR DAAROM (dus) DUS (daarom)

The challenging categorization picture of Table 1 is further investigated in this book. Overall, a major question is the extent to which a “cline” of subjectivity is equivalent to a spectrum of different possible mental space configurations – perhaps more equivalent than is immediately obvious, since more than one space configuration may be accessible simultaneously. For example, in (8), I will not come for a walk can be taken as both a description of a future event, and as an epistemic conclusion or decision – so (8) may not be uniquely classifiable as either only a content causal or only an epistemic causal. In chapter 2, Sanders, Sanders and Sweetser draw on Mental Spaces Theory to offer new insights into central problems of Subjectivity and Perspective, including an explanation of the dotted line between daarom and dus in the right column. More specifically, they introduce the concept of a Basic Communicative Spaces Network (from now on BCSN). The BCSN is the network of space structures which are automatically accessible to participants in any communicative event. As Sweetser (1990) proposed, there is no need to do extra cognitive work in order for communicative participants to be aware of the fact that some Content is being mentioned, by a speaker in a particular Speech Act context, and resulting from some Epistemic states of the speaker. The BCSN therefore starts with the deictic anchoring of the speech event (I-here-now – B¨uhler [1934] 1990), relates this to the content and epistemic spaces, and accounts for the relationship between Subjectivity and Domains of use in terms of blending of Mental Spaces. We seem to need at least this much analytic structure, to explain the contrasts between English or Dutch causal connectives.

8

Ted Sanders and Eve Sweetser

4. Cross-linguistic categories? Looking for causal categories in less-related languages than English and Dutch, Barbara Dancygier discusses the Polish causal connectives bo and to in chapter 3. To would be used in cases like (11), whereas many cases of bo suggest speech act causality as in (12). (11)

Because he tried hard, (to) he passed the exam.

(12)

What are you doing tonight, because (bo) there’s a good movie on?

Again, Mental Spaces Theory appears useful in modelling differences and similarities in meaning and use of connectives. Dancygier shows how the two connectives invoke constructions that both signal causal links in the discourse, but in a different order: To marks the construed result, while bo marks the construed cause. She relates her findings to the BCSN in chapter 2 and includes Verhagen’s (2005) Intersubjectivity in the discussion. She concludes that argumentative causal links play a significant role in the use of bo and to, unlike some other Polish causal connectives which are more content-based. In terms of Verhagen (2005), bo and to are both used in the intersubjective domain, with a very broad range of uses. They do not show too many restrictions on their use, although to cannot mark inferential (epistemic) conclusions when no conjunction explicitly marks its cause / antecedens-segment (p).

5. Categories in diachronic development? As Traugott pointed out (1989 and elsewhere), causal markers in English have changed over time, from double marking (Because P, therefore Q) as the Old English norm, to single backwards or forwards marking as the Modern English norm. In chapter 4 of this volume, Jos´e Sanders uses MST prominently in her approach to the diachronic development of Dutch causal connectives in successive Dutch translations of biblical narratives. She shows how the changing preferences of Dutch translators lead them not just to different grammatical and semantic choices but thereby to different rhetorical strategies with respect to causal structure. The categories of causal connectives have repeatedly been shown to be relevant in research on diachronic development. Sweetser (1990) originally introduced her three-domain distinction to cover the semantics of a number of related phenomena involving verbs of perception, modal elements, and connectives. She

Introduction: Causality in language and cognition

9

argued that, from their original content meanings, these linguistic elements have diachronically developed new meanings in the more subjective epistemic and speech-act domains. Examples of such developments in the realm of connectives have been presented by K¨onig and Traugott (1988), Traugott (1995) and Traugott and Dasher (2005). Thus, still originally meant “now as formerly” but has changed from an expression of simultaneity to one of denial of expectation. Similarly, while developed from a marker exclusively expressing simultaneity (“at the time that”) to a marker used to express contrast and concession (see [13]); German weil had the same root meaning, but developed into a causal connective. Traugott (1995: 31) considers this a case of “subjectification: meanings become increasingly based in the speaker’s subjective belief state/attitude toward the proposition”. (13)

Mary read while Bill sang. Mary liked oysters while Bill hated them. (Traugott 1995: 31)

Traugott shows how subjectification plays a significant role in the grammaticalization processes on the sentence level. Are subjectivity and subjectification also valid at the discourse level? Sanders’ analysis of Bible translations suggests that this is so – that rhetorical conventions can be more or less subjective.

6. Cross-level categories? Causal connectives and causative verbs In chapter 5, Stukker, Sanders and Verhagen return to the discussion of Subjectivity in the three Dutch causal connectives discussed earlier. They start from the “categorization hypothesis”: the idea that each one of the connectives is related to a specific conceptual model of causality. They build on earlier corpus studies, which revealed that, indeed, the majority of connectives’ natural usage contexts reflect these conceptual categories of causality more or less directly, as was shown in table 1. However, in a minority of cases, the relation of a connective to its presumed typical category of causality seems less straightforward. Should these findings be interpreted as evidence against the categorization hypothesis? Stukker et al. re-interpret the findings from previous connective studies within a usage-based framework. The “usage-based” approach to language assumes that variation is an inherent characteristic of language use, and seeks to explain occurring patterns of variation with reference to more general cog-

10

Ted Sanders and Eve Sweetser

nitive mechanisms (cf. Langacker 1987; Bybee 1985, 2006, 2007; contributions to Barlow and Kemmer 2000). In line with this framework, they propose that an interplay of conceptual and usage factors can explain why the usage of Dutch causal connectives does not always conform to abstract definitions that seem to be quite straightforward otherwise. The hypothesis is that the apparent counter-examples are actually non-typical, peripheral members of the very same conceptual category the connectives refer to in their more typical usage contexts: “prototypical usages”. The chapter focuses on one specific factor causing variation in connective use: the idea that language users categorize causal relations not on the basis of “objective reality”, but on the basis of their subjective construal of the situation (Langacker 1990 and elsewhere) and on one specific factor constraining variation: the prototypicality structure of semantic categories discussed above. Stukker et al. also mention the issue of the similarities between causal categories in connectives and in auxiliary verbs like Dutch doen ‘make’ and laten ‘let’, see Stukker, Sanders and Verhagen (2008), for a more elaborate discussion. Prototypical uses of doen and laten are illustrated in (14) and (15), taken from Stukker (2005:1). (14)

De extreme koude deed de rivieren bevriezen. ‘The extreme cold caused the rivers to freeze.’

(15)

Hij liet de soep afkoelen. ‘He let the soup cool.’

Supposed similarities in cross-level categories are explicitly challenged by Speelman and Geeraerts in Chapter 6. On the basis of Verhagen and Kemmer (1997) and Stukker (2005), they formulate the (in)direct causation hypothesis: the choice for either doen or laten is influenced by the degree of involvement of the causee. Speelman and Geeraerts derive a series of concrete predictions from this hypothesis and argue that it is crucial for the further development of the “scientific method” in linguistics, that such hypotheses are tested against a sample of “observable behavior”. In their chapter they test the predictions against a corpus of spontaneous spoken Dutch, making using of the innovative statistical method producing collocation patterns. In a stepwise logistic regression analysis, they incorporate factors that were predicted to affect the choice of doen versus laten. The choice of these factors was based on the indirect causation hypothesis. Interestingly, Speelman and Geeraerts conclude that their data falsify several predictions based on the (in)direct causation hypothesis. In stead, it seems like doen is a marked form in comparison with laten: it appears to have a more restricted and more specific range of application than laten does. The

Introduction: Causality in language and cognition

11

authors argue that it will be necessary to pursue a new hypothesis for determining the choice between doen and laten, which in turn requires further empirical testing.

7. Research methods: Converging evidence in causal categories? In chapter 7, Sanders and Spooren join Speelman and Geeraerts in their argumentation in favor of a stronger empirical bias for studies of language use. Sanders and Spooren acknowledge that several chapters in this book volume focus on the system behind meaning and use of these causal connectives in various languages, and by doing so, provide interesting insights in the organization of the lexicon of causal connectives. However, Sanders and Spooren see these studies as providing one window on conceptual categorization, but not the only one. Other windows are provided by the study of implicit and explicit causal relations in discourse, in such diverse areas as relation categorization, discourse processing and language acquisition. The authors summarize results from all these fields and conclude that Causality and Subjectivity are two cognitive principles that organize human knowledge of both cognitive construal of coherence relations and linguistic use of connectives. Notions like Causality and Subjectivity indeed help explain the system and use of causal relations and their linguistic expressions in everyday language use, in language acquisition and in discourse processing. As has probably become clear by now, this book volume provides specific attention to the adequate research methods that may be used to investigate research questions, an important discussion that is also taken up in contributions by, among others, Geeraerts, Gibbs, Sweetser and Talmy to a recent book volume Methods in Cognitive Linguistics (Gonzalez-Marquez et al. 2007). Some chapters in the current volume stick to classical linguistic “introspection”, while others report extensive corpus analyses and even psychologically oriented experimental studies. The basic notion of causality appears to be an ideal linguistic phenomenon to provide an overview of methods and, perhaps more importantly, invoke a discussion on the most adequate methodological approaches to study fundamental issues in language and cognition. Speelman and Geeraerts even present their chapter as a case study to illustrate how the scientific method can be used in linguistics. Needless to say that empirical testing of hypotheses is paramount in this approach. All kinds of data have their strengths and weaknesses. Constructed examples, for instance, only tell us what judgments are consciously available to speakers;

12

Ted Sanders and Eve Sweetser

corpus data may tell a very different story about speakers’ behavior. However, an analyst’s intuitions play an indispensible role in the formulation of hypotheses for testing – and only constructed examples may make us aware of what cannot be done with connectives (Pander Maat and Sanders 2006). The in-depth analysis of naturally occurring discourse may provide insights into the intricate interaction between semantic features and interactional conditions, but do not enable us to systematically tease out the contributions of these two factors. Corpus research may compare larger numbers of connective uses on both linguistic and contextual factors, but every corpus analyst in the field may testify to the fact that the interpretation of discourse relations may differ between several analysts. Finally, experimental research into connective effects is a superior way of supporting causal models, but the often very short texts used in these experiments have sometimes rightly been criticized for their lack of external validity (Graesser, Millis, and Zwaan 1997). From a methodological point of view, it can be concluded that the integration of cognitively plausible theories with empirical testing is the ultimate aim rather than a situation that has already been realized (Sanders and Spooren 2007). One way to realize this goal is to proceed with the thorough investigation of corpora of actual language use. Digital corpora enable researchers to do this on a larger scale than ever, and recent studies show how fruitful statistic and (partly) automatic analyses of corpora can be for the area of causal verbs and connectives, too. Speelman and Geeraerts’chapter is an outstanding example for causative verbs and for connectives Bestgen, Degand, and Spooren (2006) have shown the way. Furthermore, it is especially important to extend corpus research in the direction of spoken discourse. This challenge is clearly taken up in this book volume in the chapter by Speelman and Geeraerts and by Sanders and Spooren, but on the whole, the field of corpus-linguistic studies is still largely based on the study of written discourse. There are at least two important questions to consider: to what extent can results be generalized to spoken discourse? And what do the specific insights from the linguistic analysis of spoken discourse add to the picture we have so far? At present, we have only limited results on nonwritten connective use (Couper-Kuhlen 1996; Ford 1993; Gohl 2000). Analysis of multi-modal discourse data would presumably also allow workers to examine the role of visual cues (gaze, gesture, stance) in guiding the interpretation of causal relations; there is good evidence that gesture itself is interpretable at the different levels of content, epistemic and speech-act structure (Sweetser 1998, 2007). Integration of text-linguistic and psycholinguistic insights is a second way to realize the goal of interaction between theory and empirical testing. The

Introduction: Causality in language and cognition

13

sometimes subtle semantic-pragmatic distinctions proposed by linguists on the one hand, and the processing effects revealed by psycholinguistic research on the other hand, still need to be linked. For instance, the general processing effects of because and and have been investigated (e.g. Millis, Golding, and Barker 1995), but there is very little experimental research into the processing instructions encoded by connectives that differ in specificity (e.g. but versus although, or dus versus daardoor, see De Leeuw, Mak, and Sanders 2008). Similarly, there are a few processing studies based on linguistically sophisticated analysis of the causal categories discussed earlier in this chapter – such as the differences between content and epistemic causals. Studying on-line text processing, Traxler, Bybee and Pickering (1997) focused on the difference between content and epistemic causal relations and conclude that content relations are processed quicker than epistemic ones. Noordman and De Blijzer (2000) arrive at similar conclusions. Still, many questions remain unanswered. We believe a closer integration of theoretical and corpus-linguistic work with this type of processing studies would lead to significant further progress in the research field as a whole. In conclusion, we believe that crucial contributions of this volume are (1) demonstrating convergence of linguistic, corpus-linguistic and psycholinguistic methodologies in determining cognitive categories of causality, (2) showing how differences between even quite closely related languages (English, Dutch, Polish) can help us to elaborate the typology of levels and categories of causation represented in language, and (3) using mental spaces theory to represent a general theory of linguistic construal of causation.

References Anscombre, J.-C. & O. Ducrot 1983 L’argumentation dans la langue. Li`ege/Bruxelles: Mardaga. Barlow, M. & S. Kemmer 2000 Usage based models of language. Stanford: CSLI Publications. Bestgen, Y., L. Degand, & W. Spooren 2006 Towards automatic determination of the semantics of connectives in large newspaper corpora. Discourse Processes 41(2): 175–194. Boyer, P. 1996 Causal understandings in cultural representations: cognitive constraints on inferences from cultural input. In: Sperber, Premack & Premack (eds.), Causal Cognition: A Multidisciplinary debate, 615– 649. Oxford: Clarendon Press.

14

Ted Sanders and Eve Sweetser

B¨uhler, K. 1990 [1934] Theory of Language: the Representational Function of Language. D.F. Goodwin (trans.). Amsterdam/Philadelphia: John Benjamins Press. Bybee, J. 1985 Morphology: a study of the relation between meaning and form. Amsterdam/Philadelphia: John Benjamins. Bybee, J. 2006 From usage to grammar: The mind’s response to repetition. Language 82: 711–733. Bybee, J. 2007 Frequency of use and the organization of language. Oxford et al.: Oxford University Press. Couper-K¨uhlen, E. 1996 Intonation and clause combining in discourse: The case of because. Pragmatics 6: 389–426. Couper-Kuhlen, E. & B. Kortmann 2000 Cause, condition, concession and contrast: Cognitive and discourse perspectives. Berlin: Mouton de Gruyter. Dancygier, B. & E. Sweetser 2000 Constructions with if, since, and because: Causality, epistemic stance, and clause order. In: E. Couper-Kuhlen & B. Kortmann (eds.), Cause, condition, concession and contrast: Cognitive and discourse perspectives, 111–142. Berlin: Mouton de Gruyter. Dancygier, B. & E. Sweetser 2005 Mental Spaces in Grammar: Conditional Constructions. Cambridge: Cambridge University Press. Degand, L. 2001 Form and function of causation: A theoretical and empirical investigation of causal constructions in Dutch. Leuven: Peeters. Degand, L. & H. Pander Maat 2003 A contrastive study of Dutch and French causal connectives on the Speaker Involvement Scale. In: A. Verhagen & J. van de Weijer (eds.), Usage based approaches to Dutch, 175–199. Utrecht: LOT. De Leeuw, S., P. Mak, & T. Sanders 2008 Effects of the Dutch causal connectives ‘dus’ and ‘daardoor’ on discourse processing. Toegepaste Taalwetenschap in Artikelen. Ducrot, O. 1980 Essai d’application: MAIS – les allusions a l’´enonciation – d´elocutifs, performatifs, discours indirect. In: H. Parret (ed.), Le langage en context: e´ tudes philosophiques et linguistiques de pragmatique, 487–575. Amsterdam: Benjamins.

Introduction: Causality in language and cognition Fauconnier, G. 1994 Ford, C. E. 1993 Gohl, C. 2000

15

Mental spaces: Aspects of meaning construction in natural language, 2nd edition. Cambridge, M.A.: Bradford. Grammar in interaction: adverbial clauses in American English conversations. Cambridge: Cambridge University Press.

Causal relations in spoken discourse: Asyndetic constructions as a means for giving reasons. In: E. Couper-Kuhlen & B. Kortmann (eds.), Cause, Condition, Concession and Contrast: Cognitive and Discourse Perspectives, 83–110. Berlin: Mouton de Gruyter. Gonzalez-Marquez, M., I. Mittelberg, S. Coulson & M. J. Spivey 2007 Methods in Cognitive Linguistics. Smaterdam/Philadelphia: John Benjamins Publishing Company. Graesser, A. C., K. K. Millis & R. A. Zwaan 1997 Discourse comprehension. In: J. Spence, J. Darley & D. Foss (eds.), Annual Review of Psychology 48: 163–189. Palo Alto, CA: Annual Reviews Inc. Jayez, J. & C. Rossari 2001 The discourse-level sensitivity of consequence discourse markers in French. Cognitive Linguistics 12: 275–290. Knott, A., & R. Dale 1994 Using linguistic phenomena to motivate a set of coherence relations. Discourse Processes 18: 35–62. Knott, A., & T. Sanders 1998 The classification of coherence relations and their linguistic markers: An exploration of two languages. Journal of Pragmatics 30: 135–175. Knott, A., T. Sanders & J. Oberlander 2001 Levels of representation in discourse relations. Cognitive Linguistics 12: 197–209. K¨onig, E. & E.C. Traugott 1988 Pragmatic strengthening and semantic change: The conventionalizing of conversational implicature. In: Werner H¨ullen & Rainer Schulze (eds.), Understanding the lexicon: Meaning, sense and world knowledge in lexical semantics, 110–124. T¨ubingen: Max Niemeyer Verlag. Lakoff, G. 1987 Women, Fire and Dangerous Things. What Categories Reveal about the Mind Chicago et al.: University of Chicago Press. Lakoff, G. & M. Johnson 1999 Philosophy in the Flesh. The Embodied Mind and its Challenge to Western Thought. New York: Basic Books.

16

Ted Sanders and Eve Sweetser

Langacker, R. 1987

Foundations of Cognitive Grammar I:Theoretical Prerequisites. Stanford CA: Stanford University Press.

Langacker, R. W. 1990 Subjectification. Cognitive Linguistics 1: 5–38. Millis, K. K., J. M. Golding, & G. Barker 1995 Causal connectives increase inference generation. Discourse Processes 20: 29–49. Noordman, L. & F. De Blijzer 2000 On the processing of causal relations. In: E. Couper-Kuhlen & B. Kortmann (eds.), Cause, Condition, Concession and Contrast: Cognitive and Discourse Perspectives, 35–56. Berlin: Mouton de Gruyter. Pander Maat, H. L.W. & L. Degand 2001 Scaling causal relations and connectives in terms of speaker involvement. Cognitive linguistics 12: 211–245. Pander Maat, H. & T. Sanders 2000 Domains of use or subjectivity? The distribution of three Dutch causal connectives explained. In: E. Couper-Kuhlen & B. Kortmann (eds.), Cause, Condition, Concession and Contrast: Cognitive and Discourse Perspectives, 57–81. Berlin: Mouton de Gruyter. Pander Maat, H. L.W. & T. J. M. Sanders 2001 Subjectivity in causal connectives: an empirical study of language in use. Cognitive linguistics 12: 247–273. Pit, M. 2003 How to express yourself with a causal connective. Amsterdam/New York: Editions Rodopi B.V. Risselada, R., & W. Spooren (eds.) 1998 Discourse Markers. Special issue on discourse markers of Journal of Pragmatics 30. Sanders, T. 2005 Coherence, causality and cognitive complexity in discourse. In: M. Aurnague, M. Bras, A. Le Draoulec & L. Vieu (eds.), Proceedings/Actes SEM-05, First International Symposium on the exploration and modelling of meaning, 105–114. Toulouse: University of Toulouse-le-Mirail. Sanders, T. & H. Pander Maat 2006 Cohesion and Coherence: Linguistic approaches. In: Brown. K. et al. (eds.), Encyclopedia of Language and Linguistics, 2nd edition, Volume 2. London: Elsevier. Sanders, T., J. Schilperoord & W. Spooren 2001 Text representation: Linguistic and psycholinguistic aspects. Amsterdam: John Benjamins.

Introduction: Causality in language and cognition

17

Sanders. J., & W. Spooren 1997 Perspective, subjectivity and modality from a cognitive linguistic point of view. In: W. A. Liebert, G. Redeker, & L. Waugh (eds.), Discourse and perspective in cognitive linguistics, 85–112. Amsterdam: Benjamins. Sanders, T. & W. Spooren 2007 Discourse and text structure. In: D. Geeraerts & H. Cuyckens (eds.), Handbook of cognitive linguistics. Oxford: Oxford University Press. Sanders, T., W. Spooren & L. Noordman 1992 Toward a taxonomy of coherence relations. Discourse Processes, 15, 1, 1–35. Sanders, T., W. Spooren, & L. Noordman 1993 Coherence relations in a cognitive theory of discourse representation. Cognitive Linguistics 4: 93–133. Sperber, D., Premack & Premack 1996 Causal Cognition: A Multidisciplinary debate. Oxford: Clarendon Press. Spooren, W. & R. Risselada (eds.) 1998 Discourse Markers. Special issue of Discourse Processes 24: 1–168. Stukker, N. M. 2005 Causality Marking across Levels of Language Structure: A CognitiveSemantic Analysis of Causal Verbs and Causal Connectives in Dutch. Utrecht: LOT. Stukker, N., T. Sanders & A. Verhagen 2008 Causality in verbs and in discourse connectives. Converging evidence of cross-level parallels in Dutch linguistic categorization. Journal of Pragmatics 40: 1296–1322. Sweetser, E. 1990 From etymology to pragmatics. Metaphorical and cultural aspects of semantic structure. Cambridge: Cambridge University Press. Sweetser, E. 1996 Mental spaces and the grammar of conditional constructions. In: G. Fauconnier, & E. Sweetser (eds.), Spaces, worlds and grammar, 318– 333. Chicago: The University of Chicago Press. Sweetser, E. 1998 Regular metaphoricity in gesture: Bodily-based models of speech interaction. In Actes du 16e Congr`es International de Linguistes (CD ROM). Oxford: Elsevier. Sweetser, E. 2007 Looking at space to study mental spaces: Co-Speech gesture as a crucial data source in cognitve linguistics. In: Gonzalez-Marquez, M., I. Mittelberg, S. Coulson & M.J. Spivey (eds), Methods in Cogni-

18

Ted Sanders and Eve Sweetser tive Linguistics, 201–224. Amsterdam/Philadelphia: John Benjamins Publishing Company,

Traugott, E. C. 1989 Traugott, E. C. 1995

On the rise of epistemic meanings in English: an example of subjectification in semantic change. Language 57: 33–65.

Subjectification in grammaticalization. In: D. Stein & S. Wight (eds.), Subjectivity and subjectivisation: Linguistic perspective, 31–54. Cambridge: Cambridge University Press. Traugott, E. C. & R. B. Dasher 2005 Regularity in Semantic Change. Cambridge: Cambridge University Press. Traxler, M. J., M. D. Bybee, & M. J. Pickering 1997 Influence of connectives on language comprehension: Eye tracking evidence for incremental interpretation. The Quarterly Journal of Experimental Psychology 50A: 481–497. Verhagen, A. 2000 Concession implies causality, though in some other space. In: CouperKuhlen, E. & Kortmann, B. (eds.), Cause, condition, concession, contrast. Cognitive and Discourse Perspectives, 361–380. New York: Mouton de Gruyter. Verhagen, A. 2005 Constructions of intersubjectivity: Discourse, syntax, and cognition. Oxford: Oxford University Press. Verhagen, A. & S. Kemmer 1997 Interaction and causation: causative constructions in modern standard Dutch. Journal of Pragmatics 27: 61–82.

Causality, cognition and communication: A mental space analysis of subjectivity in causal connectives Ted Sanders, Jos´e Sanders and Eve Sweetser 1. Causal connectives, domains, subjectivity and Mental Spaces Theory 1.1.

Causal connectives: similarities and differences

For speakers of English, it is possible to express all three of the coherence relations manifested in examples (1) – (3) with the connective so. Consider these three examples. (1)

There is coffee and tea. So, what do you want to drink?

(2)

The neighbors’ lights are out. So they are not at home.

(3)

The sun was shining. So the temperature rose.

This observation suggests that these relations have something in common – indeed, we agree with many other analysts that they share a conceptual relation of causality. English speakers also have other discourse connectives at their disposal, which seem more specifically suited to express these specific relations, and no other (Halliday and Hasan 1976; Knott and Dale 1994; Knott and Sanders 1998; Sweetser 1990). Similarly, Dutch divides up the domain of causal connection more precisely, with no overall marker such as so which can cover all of (1) – (3). The prototypical connectives used to express the very same relations are shown in (1’) – (3’) for both English and Dutch. (1’)

Er is koffie en thee. Dus wat wil je drinken? ‘There is coffee and tea. So, what do you want to drink?’

(2’)

Het licht bij de buren is uit. Dus ze zijn niet thuis. ‘The neighbors’ lights are out. Therefore, they are not at home.’

(3’)

De zon scheen. Daardoor steeg de temperatuur. ‘The sun was shining. As a result, the temperature rose.’

In particular, daardoor could not possibly express the connections in (1’) and (2’); and if dus were used in example (3’), it would sound as if the speaker

20

Ted Sanders, Jos´e Sanders and Eve Sweetser

was focusing not on the causal relation between sunshine and temperature, but on causal relations at a higher epistemic or argumentative level. This observation suggests that the examples may have causality in common, but also show differences. These differences have often been characterized in terms of different domains (Sweetser 1990), semantic or pragmatic types (Sanders, Spooren and Noordman 1992; Sanders 1997) or levels of causality (for an overview see Knott, Sanders and Oberlander 2001; Sanders and Spooren 2001). In this chapter we describe both the similarities and the differences between these types of causality, starting from the causal connectives used to express them, mainly in Dutch.

1.2.

Domains of use

Sweetser (1990) has argued that a conjunction like because is used in the contentdomain when one event causes another in the described world (4), while epistemic use (5) concerns the speaker’s reasoning and (6) illustrates the speech act use. (4)

John came back because he loved her.

(5)

John loved her, because he came back.

(6)

What are you doing tonight, because there’s a good movie on.

Although because can be used across the three domains, Sweetser (1990) also suggested that some connectives specialize in one domain: English since and French puisque would be specifically used in the epistemic and speech act domains. Similarly, German denn can only be used to express epistemic relations (G¨unthner 1993; Keller 1995). Dancygier (1998) and Dancygier and Sweetser (2005) also discuss a metalinguistic domain – an example of a metalinguistic causal connection might be OK, since we’re being politically correct, her partner is coming to dinner with her (where the speaker has been reproved for using the term boyfriend). We will not be discussing these cases in detail here, since we do not yet have enough Dutch examples to develop a solid approach. The multi-domain analysis was tested empirically for Dutch. In a number of corpus studies, Dutch connectives expressing forward causality – that is, in the order “S1, CONNECTIVE S2” – were investigated, where S stands for discourse segment, which is minimally a clause. The prototypical use of these forward connectives is illustrated in (3) daardoor ‘as a result’, (2) dus ‘so’ and (7) daarom ‘that’s why’.

Causality, cognition and communication

(3)

De zon scheen. Daardoor steeg de temperatuur. ‘The sun was shining. As a result the temperature rose.’

(2)

Het licht bij de buren is uit. Dus ze zijn niet thuis. ‘The neighbors’ lights are out. So they are not at home.’

(7)

Het was een warme dag. Daarom ging Jan zwemmen. ‘It was a hot day. That’s why Jan went swimming.’

21

Daardoor ‘as a result’ in (3) expresses a simple cause-consequence relation in the content domain, (2) can only be interpreted as an epistemic conclusion that is expressed by dus ‘so/therefore’ and daarom ‘that’s why’ in (7) expresses the reason for an intentional action in S2. Several studies have shown that these connective-characteristics are robust, and vary from strong preferences to clear restrictions on the relations they can express (Pander Maat and Sanders 2000, 2001; Stukker 2005). Daardoor can only express non-volitional content relations, but these relations cannot be expressed by daarom and dus. Dus and daarom show rather gradual preferences. Dus most often expresses epistemic relations and can be used to express content volitional relations. Daarom most often expresses volitional relations, but can express epistemic relations. Taken together, these observations show how the Dutch language “cuts up” forward causality. Roughly the same mechanism accounts for backward causals, where doordat can only express non-volitional content relations, epistemic relations are often expressed by want ‘since / because’ and omdat ‘because’ has a slight preference for volitional content relations (Degand 2001; Degand and Pander Maat 2003; Pit 2003). The clearest case of this “cutting up” concerns daardoor and the corresponding backwards connector doordat. There are clear restrictions on their use. They can only express non-volitional content relations. Interestingly enough, daarom and dus can both express volitional and epistemic relations. In fact, these relations are regularly lexicalized by the same connectives: daarom and dus. The conceptual meaning these two connectives share is that they both crucially involve an animate subject, a person, whose intentionality is conceptualized as the ultimate source of the causal event, be it an act of reasoning or some “realworld” activity (Pander Maat and Sanders 2000, 2001). In terms of conceptual categories, this is a very fundamental distinction: the one between events ultimately originating from some intentional mind, versus events that originate from non-intentional causes; between causes that are crucially located in a Subject of Consciousness (from now on SoC), and those that are located in the inanimate, outside world (cf. Verhagen 1995, 2005; Stein and Wright 1995).

22

1.3.

Ted Sanders, Jos´e Sanders and Eve Sweetser

Subjectivity and Mental Spaces

The notion of Subjectivity (developed from Langacker 1990) helps us to express these contrasts more precisely: The degree of subjectivity is the distance between the current speaker and the SoC involved in the construction of the causal relation; the smaller this distance, the more subjective the relation. Under this approach, daardoor (see [3]) and doordat express objectivity: the speaker is not involved in the construction of the causal relations between the events. In fact, it seems like there is no SoC at all. The speaker merely reports events in the world that are causally related. In epistemic dus/want-relations (see [2]), the SOC is heavily involved and is often identical to the speaker: the speaker construes the relation, even though she is usually not mentioned explicitly in the discourse. That is, epistemic relations come in the form of (2): The neighbors’ lights are out. So they are not at home rather than in the form The neighbors’lights are out. So I am sure they are not at home. Finally, in volitional causal relations (see [7]), there is an explicitly verbalized SOC who acts and is responsible for the causal relation, but the speaker is not involved in the construal. This characterization in terms of Subjectivity is corroborated by corpus analyses and experimental studies on language users’ preferences (Pander Maat and Sanders 2000, 2001; Pander Maat and Degand 2001; Pit 2003; Stukker 2005; Stukker, Sanders and Verhagen 2008). Hence, in terms of Langacker’s (1990) notion of Subjectivity, we could argue that, when there is no SoC present, in the case of a non-volitional causal relation, the ground (speech event, its participants, and its immediate circumstances) may be entirely external to the semantics of the utterance. The ground can also be included in the scope of predication as an off-stage, unprofiled reference point: yesterday, tomorrow etc. In volitional and epistemic cases, there is a SoC present and this will be more or less clear by (implicit) evaluation by the speaker: probably, is likely to, must be: SoC present. Finally, the ground may be on stage: So I think they are not at home. In the latter case, the ground is in a sense objectified: that is, made part of the situation referred to in the utterance. The latter cases are of another category; then the speaker is made explicit and foregrounded so that it is comparable to looking at another actor. In this chapter, we use Mental Spaces Theory (Fauconnier 1985, 1994; Sweetser and Fauconnier 1996) to model the similarities and differences in the meaning and use of the Dutch causal connectives. Why Mental Spaces Theory (from now on MST)? A first reason is that we strive for a cognitively plausible account. Both in linguistic and psycholinguistic approaches to discourse, connectives are considered as linguistic signals of coherence, or as operating instructions for interpretation: They instruct the interlocutor to relate the content of

Causality, cognition and communication

23

the connected segments in a specific type of relationship (Sanders and Spooren 2001, 2007). This view on connectives is cognitively plausible and in fact fits in with what is known about the role of causal coherence markers during discourse processing (Millis and Just 1994; Noordman and Vonk 1997; Cozijn 2000; Kamalski 2007; Mulder 2008). MST seems to be particularly compatible to this conceptualization of connectives. In MST, connectives are often treated either as elements that block certain inferences (such as but) or as space-builders, i.e. linguistic expressions that typically establish new mental spaces, such as if-then conditionals. As a theoretical framework, MST seems compatible to findings in research on discourse processing. A second reason for using MST is that this model has proven to be descriptively adequate for linguistic items that are related to causal connectives. In recent years, cognitive linguists have shown how MST can be used fruitfully to clarify the meaning and use of conditionals (Sweetser 1996; Dancygier 1998; Dancygier and Sweetser 1997, 2000, 2005) and other connectives (Verhagen 2005). Interestingly enough, these approaches seem compatible with Sweetser’s original multi-domain theory. Different if -conditionals, for instance, have been shown to correspond to content, speech act and epistemic domains and the MST framework is used to describe these differences. Therefore, we will set out to develop a similar integrative approach for causal connectives: are Subjectivity accounts compatible to domain theory? What is the difference between referring to mental spaces, as opposed to domains (as Sweetser 1990 did)? First, a mental spaces analysis allows for an integrated approach: both explicitly built mental spaces (She said. . . , He thought. . . ) and implicit domains such as the speaker’s reasoning processes can be treated as mental spaces. The difference, as we shall argue below, lies in how the spaces relate to the communicative situation, and whether they are explicitly mentioned. Finally, there are some unresolved issues in current Subjectivity accounts, which may be clarified in an MST-approach. A crucial one is related to perspective. For instance, while describing the difference between the causality in (2) and (7) – repeated here for sake of clarity – , it is tempting to say that the speaker is responsible for the causality in (2), whereas Jan is responsible for the causality in (7). But what happens when these examples are changed slightly, as in (2a) – where the verb tense is changed from present to past – and (7a) – where changed the perspective from Third Person Jan to First Person I ? (2)

Het licht bij de buren is uit. Dus ze zijn niet thuis. ‘The neighbors’ lights are out. So they are not at home.’

(7)

Het was een warme dag. Daarom ging Jan zwemmen. ‘It was a hot day. That’s why Jan went swimming.’

24

Ted Sanders, Jos´e Sanders and Eve Sweetser

(2a)

Het licht bij de buren was uit. Dus ze waren niet thuis. ‘The neighbors’ lights were out. So they were not at home.’

(7a)

Het was een warme dag. Daarom ging ik zwemmen. ‘It was a hot day. That’s why I went swimming.’

About (2a), we can now ask again: Who is responsible for the causality? Who is the SoC construing the causal relation here? Is it still the speaker (as in [2])? And has (7a) suddenly become an epistemic relation, simply as a result of a change in perspective from third to first person (compare [7a] with the epistemic example [2])? This type of questions are hardly addressed in the literature, where usually clear-cut cases like (1) – (3) are discussed. In order to answer these questions, integration is needed between multidomains theory and models of subjectivity and discourse perspective. MST has already been fruitfully used to describe issues of perspective in narrative discourse, such as the free indirect speech that we are witnessing in (2a), see J. Sanders and Redeker (1996). In the following, we investigate the relationship between domains of use, subjectivity and perspective, making use of Mental Space Theory (MST). We will argue that it is useful to think of subjectivity within this framework in terms of distance from the speaker’s internal mental spaces (or the speaker as SoC), in the mental space network. We will develop a Basic Communicative Space Network to account for the representation of causal coherence between clauses. We will show how these insights from MST indeed allow us to develop an integrative theory of causal connectives, which illuminates the relationships between connective domains, subjectivity and perspective. Hence, this chapter investigates to what extent MST can help uncovering the system behind the meaning and use of causal connectives. The meaning and use of these connectives is studied from the point of view of linguistic categorization. We focus on Dutch speakers who categorize causally related events in the order cause-consequence, by expressing them with the connectives like daardoor, daarom and dus. We propose an analysis of these connectives, clarifying their similarities and differences. In reverse, this analysis is likely to have implications for theories of connective categorization.

Causality, cognition and communication

25

2. Towards a Basic Communicative Spaces Network 2.1.

Developing an integrative account: Principles of the Basic Communicative Spaces Network

Mental Spaces Theory offers us a theoretical and descriptive apparatus which readily permits an analyst of causal markers to do justice to the complex data. One important initial distinction is that between setting up a mental space, evoking an accessible mental space, and elaborating an active mental space. There is a recognized tendency for conditional protases (if -clauses) to precede their main clauses, while the reverse order is more common (though not required) for clauses marking cause. (These “unmarked” orders are exemplified by the contrast between If it rains, they’ll cancel the game and They cancelled the game because it rained.) Ford (1993) documents this, and notes that the two clause orders make sense because functionally, a causal clause is an added “explanation” of the main clause, while a conditional clause actually changes the assertion status of the main clause. As Dancygier and Sweetser (2000, 2005) have pointed out, Mental Spaces Theory allows us to capture this difference simply: most causal clauses are elaborating the previously established mental spaces wherein the main clause content holds, while conditional clauses are setting up mental spaces as contexts for the content of the main clause. One possible concern which a reader of Mental Spaces Theory could raise would be how to limit the proliferation of posited mental spaces: the real issue here is how to motivate the spaces brought up in an analysis. And one important sub-issue is how certain spaces seem to be implicitly present, ready for reference without overt marking. Sweetser and colleagues have made strong claims about the inherent accessibility of certain mental spaces, even when not overtly set up or evoked (Sweetser 1990, 1996; Dancygier 1998; Dancygier and Sweetser 2000, 2005): in particular, any communicative use of language necessarily involves the presumption that the speaker has mental states, and that she is expressing some content of her mental states, in some speech act setting, using some set of linguistic forms. This being the case, any communicative speech act rests on the presumed presence of Content, Epistemic, Speech Act, and Metalinguistic spaces. This configuration is a conceptual network of Mental Spaces that represent the basic communicative situation in which a causal connective is uttered. For short, we call this grouping of spaces a Basic Communicative Spaces Network, and we assume that, unlike most other mental spaces, these are evoked “for free” – along with a presumed Base Space of the Speaker’s reality. And since the Basic Communicative Spaces Network comes “for free”, these spaces (but not generally others) are automatically accessible as potential

26

Ted Sanders, Jos´e Sanders and Eve Sweetser

“domains of interpretation” for conditionals, modals, and other linguistic forms – though language-specific semantics will determine whether a given form is ambiguous between these spaces. A Basic Communicative Spaces Network, then, is in essence a specification in mental space terms of the minimum basic structures involved in a speechinteraction ground – not just a speaker and a hearer interacting in an immediate context, but including some Base Space assumed by the speaker as reality, plus the content of the speaker’s epistemic states and the content of the communication. The relation to Langacker’s ground will be obvious here (Langacker 1987, 1991a, 1991b; Coulson and Oakley 2005), as is the relation to terms such as common ground (Clark 1992, 1996) which is in turn related to Goodwin and Harness Goodwin’s (1992) construction of context. We shy from regular use of the term ground, however, for two reasons. First, we believe that the internal multi-space structure of the Basic Communicative Spaces Network is important, as compared to a less differentiated concept of the speech setting which is evoked by Langacker’s ground. Further, we feel it is important to avoid confusion with the broader concept of common ground. The common ground in any discourse setting does necessarily involve the kind of space network just proposed – but it may also involve added, far more complex structures. For example, if we’ve been discussing Jane Austen, then the spaces built up concerning Jane Austen and her work may be part of our common ground; or if we’ve been planning the physical setup of a room for a workshop by laying out objects on a desk in another room, then the common ground could include a shared interpretation of the objects as an envisioned furniture arrangement in another room. A stranger entering the discourse at this point would not share the full common ground of the ongoing participants, but could not avoid constructing a Basic Communicative Spaces Network as the first stratum of shared structure. By acknowledging separate mental spaces for content, speech interaction, speaker’s epistemic processes, and metalinguistic form choice, Dancygier and Sweetser (2005) have been able to show that there are causal and conditional markers which are general across different kinds of spaces (like English because) and ones which are more specific about which kinds of spaces they can mark (English causal since cannot mark content-space relations). These results are of interest because they make it clear that cognitively, humans can conceptualize and label both extremely general causal relations and quite specific ones – one kind of specificity being a restriction as to the level of construal of the causation. Dutch shows another important contrast in its causal markers of result, not (as we shall see) entirely orthogonal to the content-epistemic-speech act contrast set, but independently based. Specifically, Dutch makes a crucial distinction between volitional causal relations, where there is a Subject of Con-

Causality, cognition and communication

27

sciousness volitionally causing an event or situation, and non-volitional ones. Daardoor ‘As a result’ is restricted to non-volitional causation situations such as a sunny day causing a rise in temperature. Daarom ‘that’s why’ is a marker of volitional causal relations. And dus ‘so’ marks causal relations wherein the speaker is directly involved as subject of consciousness (as an initial approximation, non-content-domain causation). Dutch causal conjunction choice thus crucially involves distinguishing between (i) presence and absence of a SoC as a causer and (ii) involvement of the Speaker’s own epistemic and speech-act spaces, and of the Speaker as SoC. We will show how the Basic Communicative Spaces Network allows us to formalize this distinction between a SoC-less nonvolitional relation in which the Speaker is “just reporting”, and the volitional, epistemic and speech act cases in which a SoC is responsible for constructing the causal relation. There is inevitably a special prominence to the Basic Communicative Spaces Network, as opposed to other spaces which are built up as part of a network – in a broad sense, it’s the deictic center of the mental spaces network. What makes this complex is that any SoC has her own Base Space; and any SoC who communicates builds such a sub-network, and constitutes a potential deictic nexus in the mental spaces network. Authors, narrators, speakers – each has a communicative space network. For most Dutch examples from chat-room and newspaper texts we will not need to build as many layers of space networks as would be needed for some narrative works of fiction. However, we include one or two such complex cases from journalistic text. Since the speaker is of course a volitional SoC, there are (as we shall see) complex similarities and differences between the uses of dus and daarom. The concept of blending in Mental Spaces Theory (Fauconnier and Turner 1996, 1998, 2002; Turner and Fauconnier 1995) helps us explain this. Third-person “non-speaker” viewpoints are often blended with speaker viewpoint, setting up a blended space which is formally third person in reference but has other formal characteristics which are very non-standard for a third-person description. The phenomena referred to as Free Indirect Style by narratologists (J. Sanders and Redeker, 1996; Fludernik 1993; Banfield 1982) are one category of such blends. For example, a sentence like She closed the door in Mommy’s face could refer to the speaker’s or narrator’s mother – but it could also refer to the thirdperson agent’s mother, whom the narrator would more normally refer to as her mother rather than Mommy. The viewpoint of the narrator is represented by the third-person she and probably by the past tense, while the agent’s viewpoint is represented in the choice of the word Mommy. Blended spaces of this kind can therefore be understood as involving two inputs, the content (described) space and the speaker/narrator’s here-and-now – which raises the question of how vo-

28

Ted Sanders, Jos´e Sanders and Eve Sweetser

litional causation is to be marked, with dus or daarom. If subjectivity is defined as closeness to the communicative “here and now” (cf. Traugott 1989, 1995), then one way to bring a third-person content space “closer” to the Speaker’s epistemic space is to blend the two spaces.

2.2.

Further details of a Basic Communicative Spaces Network

We will use these ideas on the Basic Communicative Spaces Network to analyze examples of causal connectives from Dutch language use. Figure 1 represents the Basic Communicative Spaces Network. It is this diagram that will be used in all analyses. At the absolute top level, it specifies the literal example in a scattered box, followed by an identification of the segments P and Q that are causally related in the fragment. Below that, the actual Basic Communicative Spaces Network (from now on BCSN) is displayed. It consists of a 2 × 2 grid. Horizontally, it distinguishes between the top level, which is the linguistic level of the explicity realized language, and the bottom level representing the conceptual level of knowledge representation. At the linguistic level, the cause-clause P and the consequenceclause Q are represented: There is coffee and tea (P), SO what do you want? (Q). At the conceptual level, we find the Knowledge Base containing propositions p and q, which correspond to P and Q in the linguistic realization. It is this Knowledge Base that licences the P → Q relationships that are uttered (cf. Sanders et al.’s [1992] Basic operation): Can the fact that there is coffee and tea be a reason to ask what somebody wants? Can falling rain (P) indeed lead to (→) the streets getting wet (Q)? Does the observation that the lights are out (P) count as a valid reason for the conclusion that the neighbors are not at home (Q)? In short, the Knowledge Base contains the adult language user’s representation of encyclopedic knowledge, pragmatic knowledge and human reasoning, as well as the lexicon of the language that is used to express the causal relations (cf. the declarative knowledge in Levelt 1989). This Knowledge Base includes the Base Space (Fauconnier 1985, 1997; Fauconnier and Sweetser 1996), the speaker’s general conceptualization of the world around her. Given a discourse containing a causal relation that is marked linguistically by a causal connective, four interpretations are readily available: It can be interpreted in speech act, epistemic, content or meta-linguistic spaces. Recall that the metalinguistic space is only absent from further discussion in this chapter for non-principled reasons; in fact, we assume it to be part of the network (Dancygier and Sweetser 2005). The three remaining spaces have a fixed position in the diagram, which vertically distinguishes between cases with an explicitly

Causality, cognition and communication

29

Figure 1. Basic Communicative Spaces Network

realized SoC (Jan, she, they) – content space – and implicit SoC’s – epistemic and speech act spaces. The reason why SoC’s are implicit in the latter case is that the Speaker is present in the Deictic Center of Communication.

30

Ted Sanders, Jos´e Sanders and Eve Sweetser

Following a basic insight in pragmatics (B¨uhler [1934] 1990) we start our analysis from the Deictic Center of Communication, where Speaker and Addressee are actually present and communicate with each other. The literal utterance under analysis is always represented in the speech act space, because this is what S has literally said to A. Therefore, P and Q are also represented in the speech act space. As for interpretation, P’, Q’ are the corresponding representations in other spaces. When the relation is to be interpreted in other domains than the speech act one only, the diagram shows this, starting from the speech act space. An SoC is either represented by S = SoC (when the speaker is the SoC) or by a variable x = SoC, which is then in turn specified: x = Jan. The lines in the diagrams denote identity correspondencies. We will use dotted lines for the relations P, Q and their counterparts, and separated lines for the relations between Speaker or X, Y and their counterparts. Space building is indicated by plain lines with an arrow.

3. Analysis and representation of stylized and attested examples The attested examples used in this section were taken from corpus studies presented or published elsewhere: one on newspaper and business texts (Pander Maat and Sanders 1995, 2000), one on chat-texts (Spooren and Sanders 2005) and one on journalistic texts (J. Sanders 2007); also, a number of individual examples were taken from Dutch quality newspapers, an internet exchange, and a children’s book.

3.1.

BCSN at work: four prototypical usage profiles

We start our analyses from the Deictic Center of Communication. Example (1) shows a communicative situation with both Speaker and Addressee present here and now. In this context, S says something to A. This utterance contains two propositions, realized in clauses, and these two clauses are related causally, as indicated by the causal connective dus. (1)

Er is koffie en thee. Dus wat wil je? P (There is coffee and tea.) DUS Q (what do you want?)

Figure 2 represents the mental space configuration of this unattested but unremarkable example. The utterance refers to the here and now in the deictic

Causality, cognition and communication

31

Figure 2. speech act implicit DUS

center of communication, with S and A present. Linguistically, S remains implicit; A is referred to by je ‘you’ in the second part (Q). The causal connection is construed within the speech act space; therefore, the connection between cause (P’) and consequence (Q’), signalled by DUS, is represented in the same space as the utterance parts (P) and (Q).

32

Ted Sanders, Jos´e Sanders and Eve Sweetser

A similar attested example from the VU-chat-corpus is provided in (8). The context is a chat-session between middle school students (source: Spooren and Sanders 2005). (8)

Je krijgt er geen egt cijfer voor, dus late we over iets leuks prate. P (We are not actually being graded on this) DUS Q (lets talk about something fun.)

As in (1), S remains implicit in the first part (P), but S and A are referred to by we in the second part (Q). Again, both causal connection and speech act utterance are construed in one and the same space. This speech act use is one of the ways in which dus is used. A second case is the epistemic use. (2)

Het licht bij de buren is uit. Dus ze zijn niet thuis. P (The neighbors’ lights are out.) DUS Q (they are not at home.)

Here, the speaker observes something and we see her mental processes of inference at work: She concludes here and now that knowing that P (the lights being out) implies Q (the neighbors are gone). Therefore, as Figure 3 shows, the causal relation between P and Q, signalled by DUS, is represented not in the speech act domain, but in the epistemic domain, denoting S’s internal mental processes. S and A are present here and now; SoC is implicit, which is the reason why the representation remains on the implicit side of the diagram; the whole structure is construed without an explicit Subject of Consciousness present. An attested corpus example is provided by (9), which was taken from a corpus of newspaper texts; this particular text was a letter to the editor, in other words, an example of the persuasive genre (Source: Pander Maat and Sanders 1995, 2000). (9)

Drugs verwoesten mensenlevens, dus moeten drugs strafrechtelijk bestreden worden. P (Drugs destroy human lives) DUS Q (drugs must be fought by criminal law).

As in (2), on the basis of knowledge of an ongoing state of affairs (P), the speaker here and now draws a conclusion (Q), signalled by DUS, indicating an epistemic relation, construed in the epistemic space. The prototypical context for the connective daarom is that of volitional action, such as (7). (7)

Het was een warme dag. Daarom ging Jan zwemmen. P (It was a hot day.) DAAROM Q (Jan went swimming.)

Causality, cognition and communication

Figure 3. epistemic implicit DUS

33

34

Ted Sanders, Jos´e Sanders and Eve Sweetser

Figure 4 shows that an explicited Subject (x) in the speech act space, Jan, undertakes a volitional action (Q) for a particular reason (P): In order to prevent getting too hot, or at least to have a nice day, he goes swimming. By using DAAROM, the Speaker expresses that (x) is the Subject of Consciousness responsible for this causal connection. Therefore, the causal connection is not construed in the epistemic domain connected to the Speaker, but rather construed by the SoC in a volitional domain that is connected to Subject (x) in the speech act; this volitional domain is represented on the content-side of the diagram. Note that this analysis does not change in the case of a first person subject with DAAROM. (7a)

Het was een warme dag. Daarom ging ik zwemmen. P (It was a hot day.) DAAROM Q (I went swimming.)

Although it is the Speaker’s perspective that is represented rather than a narrative third person’s, the causal relation is construed in the volitional domain, much in the same way as in Figure 3 of example (7). Even if the “I” is being objectified, this does not automatically mean that the causal relation is construed in the epistemic or speech act space. In other words, the analysis does not change because of the grammatical person. It is the volitional causal relation that distinguishes volitional relations (7) and (7a) from epistemic and speech act relations, regardless of the grammatical person: crucial is the explicitness of the SoC. An example from an expository newspaper text is given in (10) (Source: NRC-Handelsblad 6-6-06). (10)

Elise van de Putte, kinderarts in het Wilhelmina Kinderziekenhuis in Utrecht, wilde weten hoe dat zat. Daarom deed ze onderzoek naar de relatie tussen chronische vermoeidheid bij pubers en vergelijkbare symptomen bij hun ouders. P (Elise van de Putte, pediatrician in the Wilhelmina Children’s Hospital in Utrecht, wanted to know what was going on.) DAAROM Q (she studied the relation between chronic fatigue in puberty and similar symptoms in parents.)

As in (7), the Speaker (author) makes explicit that subject (x) Elise van de Putte is the SoC responsible for the causal connection between reason (P) given in the first part, and action (Q) undertaken in the second part. This prototypical configuration of DAAROM can be used “rhetorically” (Stukker 2005) by the author/Speaker: It can be used to “objectify” causal relations that are in fact connected to the Speaker, as in example (11) which was taken from a business

Causality, cognition and communication

35

Figure 4. content volitional DAAROM

report (Source: Pander Maat and Sanders 1995; context: De Bijenkorf annual report). (11)

Vaste klanten besteden per jaar twee maal zoveel in de Bijenkorfwinkels als andere klanten. Daarom heeft de Bijenkorf aan de Vaste Klantkaart een aantal voordelen verbonden P (In a year, regular customers spend twice as much as other customers do.) DAAROM Q (The Bijenkorf has added a number of advantages to the Regular Customer Card)

36

Ted Sanders, Jos´e Sanders and Eve Sweetser

The Mental Space representation of (15) is similar to that of examples (7) and (14), with one exception: SoC is not a discourse subject (x) – an actor presented in the discourse – but the Speaker (S’). Note that De “Bijenkorf”, a major Dutch department store, writes this text in a annual business report about the company itself. Therefore, the causal relation is construed in the volitional domain connected to the Speaker. In fact, this example can be interpreted as the S acting as if she is an actor in the text, but in fact it is the speaker who is responsible for this causal relation. The use of DAAROM in such cases can be viewed as a discourse strategy, since awareness of the Speaker’s responsibility for the causality can only be inferred by the reader from the context and is presumably made at an implicit or unconscious level.

3.2.

Dus and daarom in speech act and epistemic relations

Apart from these prototypical usage profiles – dus in epistemic and daarom in volitional relations – we also know from earlier work (section 2), that dus and daarom can often be substituted. Is that true for these cases too? In fact, earlier work hardly addressed the speech act use. Closer study of example (1) shows how forward causality in the speech act domain can only be expressed with dus. Even with a long pause and a “:” reading, daarom is odd, see (1b). (1b)

Er is koffie en thee. # Daarom wat wil je drinken? P (There is coffee and tea.) #DAAROM Q (what do you want to drink?)

However, daarom can be used if we make the speech act explicit, as in (1c). (1c)

Er is koffie en thee. Daarom vraag ik je wat je wilt drinken. P (There is coffee and tea.) DAAROM Q (I ask you what you want to drink.)

Figure 5 represents the difference in causal configuration between (1) and (1c): In the case of (1c), there is a linguistically explicit SoC, i.e. the speaker, who is foregrounded as “I” and as such performs an explicit speech act: To ask somebody what he wants to drink. This configuration is very similar to the volitional causal connection in (7) and it is represented likewise: As a causal relation in the volitional domain, in which the SoC – in this case, not x, but explicit S’ – intentionally performs an act (Q’) in order to achieve a goal (P’). Example (12) is a real-life instantiation of this explicit speech act relation signalled by DAAROM (Source http://mail.python.org/pipermail/python-nl/2005May.txt, found May 11, 2007).

Causality, cognition and communication

37

Figure 5. speech act explicit DAAROM

(12)

Komende woensdag zal de eerst volgende Python meeting zijn. We willen graag weten wie er allemaal zullen komen. Daarom vraag ik jullie het a.u.b. te laten weten als je van plan bent te komen. P (This Wednesday we will have the next Python meeting. We would like to know who will attend.) DAAROM Q (I ask you to please make known if you are planning to attend)

As was the case in Figures 4 and 5, the causal connection is construed in the volitional domain of an explicit subject (S).

38

Ted Sanders, Jos´e Sanders and Eve Sweetser

The next question is of course whether daarom can be used to express epistemic relations, too? It is clear that daarom does not fit in epistemic contexts such as (2), even if we use the syntactically right word order, as is demonstrated in (2b). (2b)

Het licht bij de buren is uit. # Daarom zijn ze niet thuis. P (The neighbors’ lights are out.) #DAAROM Q (they are not at home.)

The use of daarom would express that their lights being out is the reason for their not being at home. Since the causal relation is not established between two events in the content domain, but between something that S has observed and her conclusion based on that observation, daarom cannot express the causality. As in the speech act-case, this act of concluding here and now that Q is the case can be made explicit. In that case, it is possible to use daarom. (2c)

Het licht bij de buren is uit. Daarom concludeer ik dat ze niet thuis zijn. P (The neighbors’ lights are out.) DAAROM Q (I conclude they are not at home.)

Figure 6 represents the construal of a volitional causal connection similar to that in (7): The Speaker is made explicit as a Subject (S), and is, as SoC, responsible for undertaking action Q because of P, which is represented in the volitional domain on the content-side of the diagram. It is important to note (this is also stressed by Dancygier and Sweetser 2005) that once a first-person conclusion or speech-act is made explicit, it is treated as a content space, with no special status such as the speaker’s implicit speech-act or epistemic space.

3.3.

How BCSN explains for the relation between Domains, Subjectivity and Perspective

A second matter is the question whether the difference between the speech act and epistemic causal spaces on the one hand, and the content spaces on the other hand, coincides with a difference in perspective. Note that the content relations discussed above show an explicit SoC in the third person, whereas in speech act and epistemic cases, the Speaker is either present and linguistically implicit, or present in first person (I ). In our view, it is not the difference in perspective that determines the relation interpretations. Rather, our claim is that the explicit presence of an SoC opens the way for an epistemic space connected to the SoC in addition to the Speaker’s

Causality, cognition and communication

39

Figure 6. epistemic explicit DAAROM

epistemic space that is always, if often implicitly, present. And in such cases, conceptual blending of the speaker’s epistemic space and the SoC’s epistemic space can take place. Let us systematically explore how the perspective of the Speaker vs. SoC relates to the interpretations in terms of the spaces we distinguish. How do epistemic relations behave in this respect? Epistemic dus-relations like (2) can

40

Ted Sanders, Jos´e Sanders and Eve Sweetser

very well be expressed in narrative contexts, where the perspective is taken not from the Speaker/Implicit SoC/1st person here and now, but from a third personSoC or first person-SoC in the past. The essential characteristic configuration of the causal connection remains, as is shown in example (2d), which is identical to the first person example (11). (2d)

Jan zag dat het licht bij de buren uit was. Dus ze waren niet thuis. P (Jan saw the lights at the neighbors’ house were out.) DUS Q (they were not at home.)

Figure 7 represents the causal construal of this sequence. As is typical for each (fragment of) narrative, subjects such as Jan (x) and objects in the speech act domain are immediately projected in the narrative situation. The narrative situation is construed in a narrative content domain in which the narrative character – Jan (x) is represented. From the narrative content domain, an epistemic space connected to Jan (SoC) is elaborated in which Jan is here and now concluding something (Q) on the basis of some observation (P). This relation could not possibly be expressed with daarom: it is an epistemic relation, in which SoC is responsible for the causal relation, at which he arrives at the conclusion in his own epistemic space. However, the configuration is more complicated than that. The fragment does not say: “so they are not at home, he thought”. In that case, the representation of the causal relation would only concern Jan’s epistemic space (cf. J. Sanders and Redeker 1996). In example (2d), however, the Speaker (author) identifies with the SOC and sees through Jan’s eyes; it is a case of free indirect discourse (thought). The DUS fits in with the SoC-perspective: The reader gets insight into Jan’s space, sees Jan’s internal mental processes. Mental Spaces Theory provides us with an excellent tool to represent this insight: The BCSN-representation shows there is a blend of the epistemic space of the Speaker/narrator (implicit SoC) with the space of the narrative subject (x), who is SoC. In other words, the distance between S and SoC is not only small – it is absent because their epistemic spaces are blended. This becomes even more clear, when the scene is placed in the narrative here and now (2e). Here, the suggestion of free indirect speech by Jan becomes even stronger because of the blending with the deictic here and now of the Speaker/narrator. (2e)

Jan ziet dat het licht bij de buren uit is. Dus ze zijn niet thuis. P (Jan sees that the neighbors’ lights are out.) DUS Q (they are not at home.) # Daarom

Causality, cognition and communication

Figure 7. epistemic 3rd person DUS

41

42

Ted Sanders, Jos´e Sanders and Eve Sweetser

An attested example from a children’s book is given in (13) (Source: C. Slee [1999]. Houten: Van Holkema en Warendorf.) (13)

Tigo fietst weg. Als hij de straat inkomt, ziet hij zijn moeders auto staan. Ze zijn dus al weer thuis. P (Tigo cycles away. When he enters the street, he sees his mother’s car.) DUS Q (they are already at home).

In (13) the conclusion is a clear case of free indirect thought by narrative character Tigo; he is the Subject of Consciousness who concludes Q on the basis of observation P. Note that direct quotation marks are absent, and that S and A are implicit in Q; therefore, no clear boundary can be drawn between the epistemic domain of the narrative character and of the Speaker/narrator; the two are blended. Blending of spaces is possible even without the immediate presence of an observing narrative subject, as is shown by example (2f). (2f)

Het licht bij de buren was uit. Dus ze waren niet thuis. P (The lights at the neighbors’ house were out.) DUS Q (they were not at home.)

In the BCSN-representation (Figure 8) it becomes clear that the only difference between (2d) and (2f) is the presence of a concrete narrative subject (x) making observation (P) and causally connecting conclusion (Q); in (2f), it remains unclear who (x) is. The narrative context will provide plausible candidates for (x). In conclusion, when the distance between S and SOC is small, the causal relation is construed more subjectively, which makes it more difficult, if not impossible, to use DAAROM. Hence: DUS is suited to express implicit SoC cases, which fits in with corpus and experimental results (see section 2). Similar parallels between DUS and DAAROM come to light when we have another look at volitional causal DAAROM-cases with third person (SoC), as the one discussed in (7). Again, such relations are prototypically expressed with a DAAROM, but here DUS is not impossible, see (7b). (7b)

Het was een warme dag, dus Jan ging zwemmen. P (It was a hot day), DUS Q (Jan went swimming).

However, there is a clear semantic difference between (7) and (7b). The DUS in (7b) seems to give the reader insight into the immediate internal mental processes of Jan (SoC) while he is deciding to go swimming: It looks like a monologue interieure. The reader gets involved in a here-and-now moment of decision making, represented in the past. Again, Mental Spaces Theory can be

Causality, cognition and communication

Figure 8. epistemic 3rd person Free Indirect Style DUS

43

44

Ted Sanders, Jos´e Sanders and Eve Sweetser

used to analyze this insight.The BCSN-configuration in Figure 9 shows that DUS in (7b) enables an epistemic interpretation (after all, DUS is the prototypical marker of such relations), in which the S(SoC) concludes that Jan must have thought it was hot and that this must have been the reason to go swimming. A blending of the mental space of the two SoC’s takes place: We are interpreting the epistemic domain of the SoC = X (Jan) as the epistemic domain of the Speaker. Note that this blended reading is even easier to get in the case of 1st person in the past as shown by (7c). (7c)

Het was een warme dag. Dus ging ik zwemmen. P (It was a hot day). DUS Q (I went swimming).

Interestingly, this blended reading was not found in the sequence connected by DAAROM, as was discussed in case example (7a) above. In other words, the blended reading of example (7c) disappears when epistemic DUS is changed in volitional DAAROM. Hence, in the DUS-case (7c), the Speaker is objectified because she is mentioned explicitly, but this objectification does not imply volitionality as in (7a). By contrast, the configuration is similar to that in Figure 8, representing example (7b): From the objectified Speaker an epistemic domain is elaborated which is blended with the epistemic space of the Speaker here and now. In other words, DUS highlights the decision making, enabling the participation of the decision, whereas DAAROM stresses the report of volitionality of SoC’s subsequent action. An attested example of the latter was taken from a chat-corpus (Source: Spooren and Sanders 2005). (14)

(Speaker 1:) “dit gaat f*king langzam, ik had dat al een minuut ingetypt” (Utterances of other participants. . . ) (Speaker 2:) “ja daarom dee ik het twee keer” context: (Speaker 1:) “this goes ** slow, I typed that already a minute ago” (. . . utterances of other participants. . . ) (Speaker 2:) “P (yeah – acknowledging contextual utterance by Addressee) DAAROM Q (I did it twice)”

Note that this analysis sheds a new light on earlier corpus results that show significant differences in the usage contexts of dus and daarom; the prototypical configuration of dus is with first person SoC, and for daarom 3rd person SoC (Pander Maat and Sanders 2000). Daarom prototypically expresses the volitional relation, irrespective of perspective: When there is an SoC undertaking this action, it is a volitional causal relation and it is expressed with daarom. Thus,

Causality, cognition and communication

Figure 9. epistemic 3rd person volitional context DUS

45

46

Ted Sanders, Jos´e Sanders and Eve Sweetser

the difference expressed with daarom versus dus is not so much the third versus first person, but rather the causality that is reported from outside (daarom) versus the inside decision making (dus).This difference can be illuminated with narrative examples from newspapers. Example (15), taken from a narrative newstext fragment, uses daarom to appoint responsibility for the action by the narrative character; by contrast, example (16), taken from a review article, uses dus to express shared responsibility between Speaker and SoC (Source: J. Sanders 2007; context: narrative fragment in an informative article on infanticide). (15)

Sonja van der Z. had haar baby een mooie laatste rustplaats willen bieden, een plekje met zingende vogeltjes, maar dat was niet gelukt en daarom had ze het lijkje in haar rugzak achter in de auto laten liggen. P (Sonja van der Z. had wanted to give her baby a nice resting place, a spot with singing birds, but that hadn’t worked out and) DAAROM Q (she had left the little corpse in her backpack in the trunk of the car.)

As Figure 10 represents, there is blending in this fragment, but it does not concern the causal relation. The first part of the sequence contains content and evaluative lexical choices –“a beautiful resting place, a spot with singing birds” – that are so specific that they ask for attribution to the narrative character Sonja van der Z., and not to the Speaker (journalist). However, no direct quotation marks are used to appoint responsibility to the character alone. Therefore, we cannot but interpret the construal of this part of the sequence as a shared responsibility, blended between character and Speaker. By contrast, the causal relation is solely to be attributed to the narrative character (SoC) and is construed in the volitional domain connected to the SoC. Note what effect the use of DUS would have had: Either via epistemic reasoning by the Speaker, making the character’s decision the S’s decision; or, via blending the epistemic spaces of Speaker (SoC) and character (SoC), involving the Speaker (and the reader) in the momentous decision of the character. Because of the inappropriate nature of this decision, this would be an unwanted effect, which is avoided by using DAAROM, which has a “distancing” effect, compare (7). In principle, DUS would have been possible only in a longer stretch of free indirect thought; a literary device that a journalistic text generally does not use. Let us compare this case with the next attested example, also from a narrative fragment of a newspaper text, this time with dus (Source: Pander Maat and Sanders 1995; context: publishing history of famous poet).

Causality, cognition and communication

Figure 10. content volitional corpus DAAROM

47

48

(16)

Ted Sanders, Jos´e Sanders and Eve Sweetser

Ze had eens een paar gedichten naar Maatstaf gestuurd, en daarover was hij laaiend enthousiast geweest – vond haar po¨ezie meteen af. Dus hij schreef haar meteen of ze nog meer had, en dat werd toen die bundel. P (She had sent some poems to Maatstaf, and he had loved them – found her poetry had an immediate perfection) DUS Q (he wrote her if she had any more, and this became her first collection of poems)

In the case of (16), the Speaker/journalist (SoC) will not mind to participate in the mental act by the character “he” (SoC), since it is an appropriate decision; hence DUS rather than volitional, distance creating DAAROM. The Mental Space Representation is similar to the one of example (7b) in Figure 8: P and Q are projected in a narrative content space, from which an epistemic domain is elaborated, which is connected to the narrative character (SoC) he.This epistemic domain is blended with the Speaker’s epistemic domain; in the blend, the causal relation as signalled by DUS is construed. Another interesting case is provided by example (17), again taken from a newspaper; this time, the genre is not narrative but persuasive, i.e. a letter to the editor (Source: J. Sanders 2007. Context: the prosecution previously has been accused of simplifying the truth in a bizarre case). (17)

(Het is nadrukkelijk niet zo dat als een verhaal maar bizar genoeg is, het ‘dus’ niet wordt geloofd door politie en OM. It is not true that P (if a story is sufficiently bizarre) “DUS” Q (the Police and the Prosecution will not believe it.)

Figure 11 represents the complicated construal of the causal relation in (17). The speech act space merely negates the causal connection; from the embedded negated space, an epistemic space of the quoted subject is elaborated. In principle, the causal relation signalled by DUS would have been construed in the blending of epistemic spaces between Speaker and SoC (x); The use of DUS draws attention to the here and now of the conclusion (p>q), as was shown in Figure 8. However, the direct quotation marks indicate that the conclusion itself (and the makers of this conclusion) are being refuted: The speaker does not want to go in the blend, at least not as far as it concerns the actual causal connection. The direct quotation marks appoint exclusive responsibility for the causal connection to the quoted SoC (x). Using DAAROM would have had a distancing effect, drawing more attention to the SoC’s decision and less to the refusal of the Speaker to join in the concluding act.

Causality, cognition and communication

Figure 11. epistemic reading direct quote corpus DUS

49

50

3.4.

Ted Sanders, Jos´e Sanders and Eve Sweetser

Non-volitional content: daardoor

Dutch offers interesting opportunities for comparison in the specific expression of causality which involves only content relations. In example (3), a causally related sequence of events is simply reported by the Speaker: Hence, there is no SoC involved. The result is a typical non-volitional content configuration, for which Dutch speakers uniquely use the connective daardoor. (3)

De zon scheen. Daardoor steeg de temperatuur. P (The sun was shining.) DAARDOOR Q (the temperature rose).

Figure 12 represents the BCSN for example (3). The causal relation is construed in the content domain, but there is no SoC involved. For this reason we represent it in the upper left of our diagram, with a “fence” around it: there is no SoC involved. Thus, the Basic Communicative Spaces Network allows us to represent a typical feature of the Dutch lexicon of causal connectives: The further specification of the content space in non-volitional versus volitional spaces. This can be viewed as a distinction that is “forced” by the Dutch data. An attested example from a newstext corpus (expository genre) is provided in (18) (Source: Pander Maat and Sanders 1995). (18)

Een kandidaat dient eerst de voorverkiezingen te winnen voordat hij officieel wordt gekandideerd. Een presidenti¨ele campagne duurt daardoor al gauw anderhalf jaar P (A candidate has to win the pre-elections before he is officially nominated) `DAARDOOR Q (a presidential campaign easily takes one and a half years)

Similarly, the causal relation as signalled by DAARDOOR is construed in the content domain without the volitional or epistemic consideration of a Subject of Consciousness.

4. Conclusions and discussion In this chapter, we have described similarities and differences of causal relations in discourse, especially those expressed by Dutch forward causal connectives. We have formulated an integrative, new proposal: The Basic Communicative Spaces Network structure, which, in our view, is a helpful step in developing a Mental Space analysis of Ground. This proposal allows the commonalities and differences of relations and connectives to be described in a coherent framework.

Causality, cognition and communication

51

Figure 12. content non volitional DAARDOOR

We have combined insights from previous accounts (Domain theory, Subjectivity and Mental Spaces Theory) to account for the linguistic categorization in Dutch causal connectives. Domain theory and the study of English connectives gave us the original four-way classification of content, epistemic, speech act

52

Ted Sanders, Jos´e Sanders and Eve Sweetser

and meta-linguistic use. Subjectivity accounts and the empirical study of Dutch pointed up the crucial factor of the Subject-of-Consciousness, who is present as an actor or concluder in many causal relations. Finally, MST provided us with two innovative insights: (i) SoC’s in the discourse can, by their presence, generate their own Mental Spaces, and (ii) these spaces can be blended. The main conclusions can be summarized as follows. The absence of a SoC accounts for the differences between non-volitional content relations and the other relations. Speech act causation invokes the current Speaker as a SoC. The Speaker can also be SoC in volitional relations, but she is explicit in the volitional cases and implicit in the speech act relations. Epistemic and speech act spaces share presence of the Speaker as a SoC, but in the epistemic space the Speaker-SoC is only a participant in reasoning processes; by contrast, in the speech-act space she is an interactive agent in a communicative exchange, which takes place in the shared setting with the addressee and links both to a Deictic Communicative Center. Finally, other actors can also be SoC’s, especially in the volitional content and in the epistemic domains. A crucial insight, which helps us really understand the system and use of dus and daarom is that (i) not only does the Speaker have her own Mental Space, but potentially so do all the SoC’s in the discourse, and (ii) the Speaker’s and the SoC’s spaces may be blended, constituting cases of Free Indirect Speech: “seeing through another’s eyes.” Several interesting observations arise from our analyses at this point. For instance, there seems to be a difference between the accessibility of the Speaker’s epistemic space and that of a narrative character’s SoC. The epistemic space of the Speaker seems to really “come for free” – after all, she is always there – whereas that of a Third Person character requires elicitation (see the discussion of examples [2d] and [2e]). Furthermore, the blending of a Third Person SoC’s and the Speaker’s spaces seems to be even more complicated, given the complexity of the BCSN-configuration. We expect these relative complexities to affect on-line discourse processing. Hence, we believe we have advanced the understanding of narrative voices and genres (following in the steps of Sanders and Redeker 1996; Vandelanotte and Dancygier forthcoming). The BCSN-approach described in this chapter seems useful to make systematic cross-linguistic comparisons, as has been done for conditionals (see Dancygier and Sweetser 2000, 2005); they point out that contrasts such as content vs. non-content domain uses recur as formally grammaticalized categories in multiple unrelated languages.) When even closely related languages like Dutch and English show such interesting similarities and differences, one wonders about other, less-studied and less-related languages. What, then, are our predictions for this cross-linguistic work? We expect the principles that causal relations share to be universal (Sanders 2005). That is, we expect the underlying grid of the

Causality, cognition and communication

53

Basic Communicative Spaces Network with four interpretations that are readily available, the distinction between implicit and explicit SoC and the Mental Spaces attached to SoC’s, to hold for every language studied. The exact way in which the lexicon of connectives “cuts up” the causality will vary. One empirical question is whether Dutch is the only language distinguishing between volitional and non-volitional content relations at the level of connectives. Also, we expect the new insights of the SoC – Speaker blending of Mental Spaces to generate many new results. For one thing, it may account for ambiguities and complexities in language use, as well as for many rhetorical effects on readers, as we have already hinted at in this chapter. Empirical studies should definitely include non-planned, non-edited types of discourse, such as spontaneous conversation (Spooren, Sanders, Huiskes and Degand to appear). Another intriguing issue concerns the cognitive interpretation of the connectives as markers of relations: What is the role of the distinctions and Mental Spaces in the Basic Communicative Spaces Network ? We consider our proposal compatible with psycholinguistic approaches in which the connective is seen as a processing instruction: It informs hearers and readers as to how the relation should be interpreted. In the diagrams, connectives and interpreted segments select the Space where the relation is interpreted. Clearly, the interpretative choice of the domain in which the relation is interpreted, remains a question of interpretation of the relation as a whole: It should be compatible to the content of the segments. An interesting processing hypothesis might be that domain-specific connectives, that is, connectives that clearly select only one of the possible spaces for interpretation, such as Dutch daardoor, should be very informative processing instructors (Sanders 2005). Similarly, the blending analysis is flexible enough to allow for the many actual cognitive interpretations. For instance, we have argued that volitional causal relations in First Person (“So I went swimming”) have the same configuration as those in a corresponding Third Person example (“So Jan went swimming”). In other words, it is the nature of the volitional causal relation rather than Perspective or Grammatical Person which determines the configuration. This proposal can lead to interesting predictions to be tested in processing studies. In fact, it may lead to the right explanations for some existing neurocognitive findings. While discussing neuron activation in purposeful, goal-related (volitional!) hand actions, such as grasping a tool, Feldman (2006: 68) explains how the same neurons are activated during both the execution of these actions – the actual grasping – and the observation of similar actions performed by another individual. These are the kinds of insights into the workings of the causal mind that we hope to achieve in a study of linguistic categories present in people’s everyday language use.

54

Ted Sanders, Jos´e Sanders and Eve Sweetser

References B¨uhler, K. 1990 [1934] Theory of Language: the Representational Function of Language. D.F. Goodwin (trans.). Amsterdam/Philadelphia: John Benjamins Press. Banfield, A. 1982 Unspeakable Sentences. Narration and Representation in the Language of Fiction. Boston: Routledge & Kegan House. Clark, H. H. 1992 Arenas of Language Use. Chicago: University of Chicago Press. Clark, H. H. 1996 Using Language. Cambridge: Cambridge University Press. Cozijn, R. 2000 Integration and inference in understanding causal sentences. Dissertation, Tilburg University. Coulson, S. & T. Oakley 2005 Blending and coded meaning: Literal and figurative meaning in cognitive semantics. Journal of Pragmatics 37(10): 1510–1536. Dancygier, B. 1998 Conditionals and Prediction. Cambridge: Cambridge University Press. Dancygier, B. & E. Sweetser 1997 Then in conditionals. Cognitive Linguistics 8 (2): 1–28. Dancygier B. & E. Sweetser 2000 Constructions with if, since, and because: Causality, epistemic stance, and clause order. In: E. Couper-Kuhlen & B. Kortmann (eds.), Cause, Concession, Contrast: Cognitive and Discourse Perspectives, 111– 142. Berlin: Mouton de Gruyter. Dancygier, B. & Sweetser, E. 2005 Mental Spaces in Grammar: Conditional Constructions. Cambridge: Cambridge University Press. Degand, L. 2001 Form and function of causation: A theoretical and empirical investigation of causal constructions in Dutch. Leuven: Peeters. Degand, L. & Pander Maat, H. 2003 A contrastive study of Dutch and French causal connectives on the Speaker Involvement Scale. In: A. Verhagen & J. van de Weijer (eds.), Usage Based Approaches to Dutch, 175–199. Utrecht: LOT. Fauconnier, G. 1985 Mental Spaces. Cambridge, M.A.: MIT Press. (Reprinted 1994, Cambridge: Cambridge University Press).

Causality, cognition and communication Fauconnier, G. 1997

55

Cognitive Mappings for Language and Thought. Cambridge: Cambridge University Press. Fauconnier, G. & M. Turner 1996 Blending as a central process of grammar. In: Adele Goldberg (ed.), Conceptual Structure, Discourse and Language, 113–130. Stanford, CA: CSLI Publications. Fauconnier, G. & M. Turner 1998 Conceptual integration networks. Cognitive Science 22(2) : 133–188. Fauconnier, G. & M. Turner 2002 The Way We Think: Conceptual Blending and the Mind’s Hidden Complexities. New York: Basic Books. Feldman, J. A. 2006 From Molecule to Metaphor. A Neural Theory of Language. Cambridge: MIT press. Fludernik, M. 1993 The Fictions of Language and the Language of Fiction. The Linguistic Representation of Speech and Consciousness. London: Routledge. Ford, C. E. 1993 Grammar in Interaction:Adverbial Clauses in American English Conversations. Cambridge: Cambridge University Press. Ford, C. E. & S. A. Thompson 1986 Conditionals in discourse: a text based study from English. In: Traugott, E.C. et al. (eds.), On Conditionals, 353–372. Cambridge: Cambridge University Press. G¨unthner, S. 1993 “. . . weil – man kann es ja wissenschaftlich untersuchen” – Diskurspragmatische Aspekte der Wortstellung in WEIL-S¨atzen. Linguistische Berichte 143: 37–59. Goodwin, C. & M. Harness Goodwin 1992 Assessments and the Construction of Context (with Marjorie Harness Goodwin). In: A. Duranti & C. Goodwin (eds.), Rethinking Context: Language as an Interactive Phenomenon, 147–190. Cambridge: Cambridge University Press. Halliday, M. A. K. & Hasan, R. 1976 Cohesion in English. London/New York: Longman. Kamalski, J. 2007 Coherence marking, comprehension and persuasion. Dissertation, Utrecht University. Keller, R. 1995 The epistemic weil. In: D. Stein. & S. Wright (eds.), Subjectivity and Subjectivisation: Linguistic perspectives, 16–30. Cambridge: Cambridge University Press.

56

Ted Sanders, Jos´e Sanders and Eve Sweetser

Knott, A. & R. Dale 1994 Using linguistic phenomena to motivate a set of coherence relations. Discourse Processes 18: 35–62. Knott, A. & T. Sanders 1998 The classification of coherence relations and their linguistic markers: An exploration of two languages. Journal of Pragmatics 30: 135–175. Knott, A., T. Sanders & J. Oberlander 2001 Levels of representation in discourse relations. Cognitive Linguistics 12: 197–209. Lakoff, G. 1987 Women, Fire and Dangerous Things. What Categories Reveal about the Mind. Chicago etc.: University of Chicago Press. Lakoff, G. & M. Johnson 1999 Philosophy in the Flesh. The Embodied Mind and its Challenge to Western Thought. New York: Basic Books. Langacker, R. 1987 Foundations of Cognitive Grammar I:Theoretical Prerequisites. Stanford CA: Stanford University Press. Langacker, R. W. 1990 Subjectification. Cognitive Linguistics 1: 5–38. Langacker, R. 1991a Foundations of Cognitive Grammar II: Descriptive Application. Stanford CA: Stanford University Press. Langacker, R. 1991b Concept, Image and Symbol: The Cognitive Basis of Grammar. Berlin: Mouton de Gruyter. Levelt, W. 1989 Speaking. From Intention to Articulation. Cambridge, MA: MIT press. Millis, K. K., & M. A. Just 1994 The influence of connectives on sentence comprehension. Journal of Memory and Language 33: 128–47. Mulder, G. 2008 Understanding causal coherence relations. Dissertation, Utrecht University. Noordman, L., & W. Vonk 1997 The different functions of a conjunction in constructing a representation of the discourse. In: M. Fayol and J. Costermans (eds.), Processing Interclausal Relationships in Production and Comprehension of Text, 75–93. Hillsdale, NJ: Lawrence Erlbaum. Pander Maat, H. & L. Degand 2001 Scaling causal relations and connectives in terms of speaker involvement. Cognitive Linguistics 12: 211–245.

Causality, cognition and communication

57

Pander Maat, H. & T. Sanders 1995 Nederlandse causale connectieven en het onderscheid tussen inhoudelijke en epistemische relaties [Dutch causal connectives and the distinction between content and epistemic relations]. Leuvense Bijdragen 3: 349–374. Pander Maat, H. & T. Sanders 2000 Domains of use or subjectivity: The distribution of three Dutch causal connectives explained. In: E. Couper-Kuhlen & B. Kortmann (eds.), Cause, Condition, Concession, and Contrast: Cognitive and Discourse Perspectives, 57–82. Berlin etc.: Mouton de Gruyter. Pander Maat, H. & T. Sanders 2001 Subjectivity in causal connectives: An empirical study of language in use. Cognitive Linguistics 12: 247–273. Pit, M. 2003 How to express yourself with a causal connective? Subjectivity and causal connectives in Dutch, German and French. Dissertation, Utrecht University. Amsterdam: Rodopi. Sanders, J. 2007 Subjectivity between narrator and character – Form and function of speaker commitment in narrative and journalistic text genres. Presentation at the Conference The notion of commitment in linguistics, 11–13 januari 2007, Antwerpen (Be) (Universiteit van Antwerpen: Linguistic Society of Belgium). Sanders, J. & G. Redeker 1996 Perspective and the representation of speech and thought in narrative discourse. In: G. Fauconnier & E. Sweetser (eds.), Spaces, Worlds and Grammars, 290–317. Chicago: University of Chicago Press. Sanders, J. & W. Spooren 1997 Perspective, subjectivity, and modality from a cognitive linguistic point of view. In: W. Liebert, G. Redeker & L. Waugh (eds.), Discourse and Perspective in Cognitive Linguistics, 85–112. Amsterdam, etc.: John Benjamins. Sanders, T. 1997 Semantic and pragmatic sources of coherence: on the categorization of coherence relations in context. Discourse Processes 24: 119–147. Sanders, T. 2005 Coherence, Causality and Cognitive Complexity in Discourse. In: M. Aurnague, M. Bras, A. Le Draoulec & L. Vieu (eds.), Proceedings/Actes SEM-05, First International Symposium on the exploration and modelling of meaning, 105–114. Toulouse: Universit´e de Toulouse le Mirail.

58

Ted Sanders, Jos´e Sanders and Eve Sweetser

Sanders, T. & W. Spooren 2001 Text representation as an interface between language and its users. In: T. Sanders, J. Schilperoord & W. Spooren (eds.), Text Representation: Linguistic and Psycholinguistic Aspects. Amsterdam etc.: John Benjamins. Sanders, T. & W. Spooren 2007 Discourse and text structure. In: D. Geeraerts & H. Cuyckens (eds.), Handbook of Cognitive Linguistics. Oxford: Oxford University Press. Sanders, T. J. M., W. P. M. Spooren & L. G. M. Noordman 1992 Toward a taxonomy of coherence relations. Discourse Processes 15: 1–35. Spooren, W., T. Sanders, M. Huiskes, & L. Degand to appear Subjectivity and Causality: A Corpus Study of Spoken Language. In: J. Newman & S. Rice (eds.), Conceptual Structure in Discourse and Language. Spooren, W. & T. Sanders 2005 Chatting about causality. Presentation at the International Workshop on Causality in Language and Cognition. Utrecht institute of Linguistics OTS, Utrecht University, December 1–2, 2005. Stein, D. & S. Wright (eds.) 1995 Subjectivity and Subjectivisation. Linguistic Perspectives. Cambridge: Cambridge University Press. Stukker, N. 2005 Causality marking across levels of language structure. A cognitive semantic analysis of causal verbs and causal connectives in Dutch. Dissertation, Utrecht University. Stukker, N., T. Sanders & A. Verhagen 2008 Causality in verbs and in discourse connectives. Converging evidence of cross-level parallels in Dutch linguistic categorization. Journal of Pragmatics: 1296–1322. Sweetser, E. 1990 From Etymology to Pragmatics. Cambridge: Cambridge University Press. Sweetser, E. 1996 Mental spaces and the grammar of conditional constructions. In: G. Fauconnier & E. Sweetser (eds.), Spaces, Worlds, and Grammars, 318– 333. Chicago: Chicago University Press. Sweetser, E. & G. Fauconier 1996 Cognitive links and domains: Basic aspects of Mental Space Theory. In: G. Fauconnier & E. Sweetser (eds.), Spaces, Worlds, and Grammars, 1–28. Chicago: Chicago University Press.

Causality, cognition and communication Traugott, E. C. 1989 Traugott, E. C. 1995

59

On the rise of epistemic meanings in English: an example of subjectification in semantic change. Language 57, 33–65.

Subjectification in grammaticalization. In: D. Stein & S. Wright (eds.), 31–54. Cambridge: Cambridge University Press. Traugott, E., A. ter Meulen, J. Snitzer Reilly & C. Ferguson (eds.) 1986 On Conditionals. Cambridge: Cambridge University Press. Turner, M. & G. Fauconnier 1995 Conceptual integration and formal expression. Journal of Metaphor and Symbolic Activity 10 (3): 183–204. Verhagen, A. 1995 Subjectification, syntax, and communication. In: D. Stein & S. Wright (eds.), 103–128. Cambridge: Cambridge University Press. Verhagen, A. 2005 Constructions of Intersubjectivity. Discourse, Syntax, and Cognition. Oxford: Oxford University Press. Vandelanotte, L. & B. Dancygier submitted Judging distances: Mental spaces, distance, and viewpoint in literary discourse. In: G. Brˆone and J. Vandaele (eds.), Cognitive Poetics. (Applications of Cognitive Linguistics.) To appear with Berlin: Mouton de Gruyter.

Causal Connectives in Dutch Biblical Translations A cognitive linguistic approach Jos´e Sanders 1. Introduction1 The cognitive linguistic analysis of narrative discourse offers ample possibilities to enlighten diachronic developments in language and in translation styles. In this chapter, such analysis is applied to the study of biblical narrative translations. Biblical narratives are ancient texts from cultures very different from ours today. Translating them necessarily means interpreting and, to a certain extent, adjusting the text to the way we understand text today. In particular, biblical translations can vary dramatically in their choices with respect to discourse coherence. This chapter explores whether the comparison of old and new biblical translations can lead to new insights on the way in which discourse coherence is expressed by biblical narrators; more specifically, how the use of causal connectives in old and new translations expresses subjectification of the biblical narrators and characters. A corpus analysis is conducted on translations of five chapters from the Hebrew Bible. The translations are compared with respect to the use of connecting coherence markers, in particular the causal connectives. The Basic Communicative Spaces Network from the previous chapter is used to analyse various examples in which the use of connective markers was altered dramatically. Narratives from the Hebrew Bible are ancient texts from a primarily oral culture. This has had its influence on many of the linguistic features of written texts rooted in this culture; for instance, the length of clauses is very limited, as is the complexity of the sentence structure; repetition and parallelism play an important role (Frye 1982: 209). Elaborating this observation, Northrop Frye states that in every literature, poetry develops before prose, and that it is only the old prose that sounds primitive, while old poetry is as subtle and mature as any other (209). The predominant rhythm of the Bible is one that can extend in the direction of either verse or prose with a minimum of change (211). 1

This study was supported by additional travel funding by the International funds of the VU University of Amsterdam, Department of Arts. I would like to thank Eve Sweetser, Bonnie Howe, Marie ThereseDesCamp, Hendrik Jan Bosman, Ted Sanders and two anonymous reviewers for their valuable comments on earlier drafts.

62

Jos´e Sanders

Translating these narratives necessarily means interpreting them and, to a certain extent, adjusting the text to the way readers in the goal-language and -culture understand text. This has been a challenge in earlier ages with developing literacy as well as in modern Western cultures, which are dominated by the frequent use of complex and various printed (and electronic) texts. For instance, remarkable to the modern readers’ eye is the homogeneity of discourse coherence markers in older Bible translations such as the 17th century Dutch Statenvertaling (SV, 1637) and the English King James Version (KJV, 1611). The Hebrew connecting particles, that can have various meanings depending on their context, are uniformly translated in the old translations: ‘en/and’ for wa, (expressing succession in time, Arnold and Choi 2003: 83); ‘want/because’for ki (originally a demonstrative with emphatic connotation, Arnold and Choi 2003: 149); and ‘zie/behold’ for (we)hinn`e (a particle of interest indicating a shift in perspective, Arnold and Choi 2003: 157). By contrast, current translations such as the Dutch Nieuwe Bijbelvertaling (NBV, 2002) and the English Contemporary English Version (CEV, 1999) of these particles show more variation, as is exemplified in example 1 from Genesis 27: 11. (1) old

SV Toen zeide Jakob tot Rebekka, zijn moeder: Zie, mijn broeder Ezau is een harig man, en ik ben een glad man. (. . . ) ‘See, my brother Esau is a hairy man, and I am a smooth man’ KJV And Jacob said to Rebekah his mother, Behold, Esau my brother is a hairy man and I am a smooth man: (. . . )

new

NBV Jakob wierp tegen: ‘Maar Esau is toch helemaal behaard, terwijl ik juist een gladde huid heb! (. . . )’ ‘But Esau is completely haired, whereas/while I have a smooth skin!’ CEV“My brother Esau is a hairy man,” Jacob reminded her. “And I am not. (. . . )”

hebr.

%na;w] r[ic; vyai yjia; wc;xe zhe ./Mai hq;b]wi-la, b %qx]y' dm,a %Yw' ql;j; vyai yki ‘And w Jacob said to Rebekah his mother, Behold zhe, Esau my brother is a hairy man, and w I am a smooth man.’

Referential choice, connective markers and the quotation verb in this example all illustrate how readability of the text is adapted to modern standards. In the first place, the referential coherence is established differently in the two translations: Rebekka, zijn moeder ‘Rebekah, his mother’ has zero reference in the NBV. Since it can be inferred from the preceding context and the verb wierp tegen

Causal Connectives in Dutch Biblical Translations

63

‘protested’ that Rebekah is the addressee, reference to her with a full name and role is presumably considered redundant at this point in the text.2 Analysing example 1 further, we see that the referential choice is not the only difference. In the old SV, the utterance from Jacob’s perspective is indicated by the neutral verbal process zeggen ‘say’, whereas the NBV indicates that Jacob’s utterance is ‘protested’ (wierp tegen), i.e., was said in contrast with what was said by Rebekah. Manner is rarely indicated in speech verbs in the source but it is in modern Dutch and English. Thus, introducing the intensifying speech verbs can be seen as an adaptation to what is generally accepted narrative style in the goal language. Such adaptation is not generally accepted; for instance Alter (1996: xxix), argues that “a translation that respects the literary precision of the biblical story must strive to reproduce its nice discrimination of terms, and cannot be free to translate a word here one way and there another, for the sake of variety or for the sake of the context.” In the third place, and of most importance here, differences in connective markers ask for attention. In the old SV translation, the two clauses are connected by the neutral additive en ‘and’, whereas the new translation NBV expresses the negative, contrastive relation between the two clauses by terwijl ‘while’.3 Supposedly, the aim of a comprehensive and readable text has led to the explicitness of causal and contrastive relations in the NBV. In addition, the contrastive character of Jacob’s quote is enhanced by the connective Maar ‘But’, while SV uses the neutral perspective indicator Zie ‘Behold’. As a result, the NBV expresses much more interpretation than does the SV. The claim in this chapter is that these changes in causal connectives are no coincidence, but are part of the more actively interpretative way in which references, verbs of saying and coherence markings are chosen in modern translations. These developments involve increasing levels of Speaker subjectivity in the semantics of the relevant linguistic expressions, since the Speaker signals 2

3

Ariel (1988) states that when a person in the text is highly accessible from short termmemory because of its immediate context – recently mentioned and grammatical subject of the clause – its linguistic expression by which this person is referred to will be a pronoun or even absent. Such a person is “active” in the working memory, easily retrievable and uniquely identifiable; thus, an explicit reference to this person by a full noun, proper name or both would be redundant. The English while, comparable to the Dutch terwijl, underwent a process of grammaticalized subjectification: once either time indicating (“at that time”) or temporal ordering (“at the moment that”), in later days it developed to a connective indicating the simultaneity of two events, and finally transformed to a marker of contrast (Traugott 1989). Example 1 from Genesis 27: 11 illustrates the latter meaning of terwijl in NBV.

64

Jos´e Sanders

more precisely how she construes the discourse relations.4 In a cognitive linguistic framework this can be described in terms of foregrounding. All of these changes instantiate that the conceptualizing subject is more foregrounded; by consequence, the text is construed more subjectively. Subjective is a qualification of that (part of a) text in which the text’s conceptualizing subject is implicitly involved in the conceptualization of the on-stage content (Langacker 1990). Such subjective conceptualization on stage can be established by a variety of linguistic means that express the temporal, spatial, attitudinal/epistemical and social (discourse) position of the Speaker (following Traugott and Dasher 2005: 23). The question of subjectivity is so interesting in the context of biblical narrative and biblical translation, since it is now widely acknowledged that over time, linguistic forms show a development of subjectification and even intersubjectification (Langacker 1990; Sweetser 1990; Verhagen 2005; Traugott and Dasher 2005). Sweetser (1990) showed how connectives have historically become used to express not only content relations (the street is wet WANT/because it rains) but also subjective contributions in the speaker’s domain, for instance a reasoning or epistemic relation (it is raining, WANT/because the street is wet) or even speech acts (bring an umbrella, WANT/because it is raining). Verhagen (2005) describes subjectification as the development towards a higher degree of subjectivity that is linguistically and textually expressed on the speaker’s (writer’s) part. For instance, in Dutch, connectives such as want ‘because, for’, dus ‘thus, so’, nu ‘now, well’ and daarom ‘that’s why’ are, compared to their historically earlier uses, more subjective.5 In narrative discourse, subjectification entails the foregrounding of both the narrator – the Speaker – and the characters – the embedded Speakers. By consequence, in biblical narrative it entails not only the biblical narrator(s) but also the many biblical characters. Can traces of subjectification be found in translations of biblical narrative? Note that biblical translation is not only a matter of translating Hebrew or Greek to English, Dutch or any other language, but also of transferring a very old text to a text formed today; that is, to a text designed for readers used to current discourse conventions (Dayras 1993). Through the centuries, the translation of biblical narrative, however loyal to its source, has 4

5

I follow Dancygier and Sweetser’s (2005) use of Sperber and Wilson’s (1986) convention, referring to the generic Speaker (S) as she and to the Hearer or Addressee (H) as he. In a corpus analysis, Evers-Vermeul and Stukker (2003) found subjectification of forms over grammatical functions (e.g. development of daarom ‘that is why’ from referential expression to connective, and of dus ‘so’ from connective to discourse marker).

Causal Connectives in Dutch Biblical Translations

65

always been an extraordinary challenge with respect to the expression of the conceptualizing subject (who speaks, who thinks) since much of subject of consciousness expression is generally implicit, and is pragmatically prompted in orally derived texts. The Dutch Statenvertaling represents a loyal representation of the source text in 17th century Dutch; its loyalty is so big as to call it “hebraisating”, i.e. using Hebraisms as a way of adapting the then contemporary Dutch to translate Hebrew optimally loyal (Sevenster 1936; Knappert 1936). Today, translations such as the Nieuwe Bijbelvertaling and the Contemporary English Version, aim for loyalty to the source language in combination with naturalness in the target language. The question that will be addressed in this chapter is: Given this dual translation perspective, what are the consequences for the conceptualization of (causal) coherence relations by the “speaking subjects” in the biblical narrative, i.e., the Speaker (narrator) and the embedded Speakers (characters)? This will be investigated by comparing the various realisations of linguistic coherence marking. The method that will be used is to compare selections of older translations to more recent ones and to analyze the influence these divergent choices have on the interpretation of biblical narrative. To this end, both text linguistic (Rhetorical Structure Theory, Mann and Thompson 1988) and cognitive linguistic (Mental Spaces Theory, Fauconnier 1997) models of analysis will be applied.

2. Material An analysis was conducted on a corpus of five chapters from the Hebrew Bible. The chapters were selected because of their presentation of inherent subjective viewpoints of some of the characters: in these stories, characters have knowledge and plans that other characters do not know of (Alter 1981). The chapters are: Genesis 27 (Isaac’s blessing of Jacob before Esau) Exodus 2 (Moses and Pharaoh’s daughter) Ruth 3 (Ruth and Boaz) II Samuel 11 (David and Bathsheba) I Kings 3 (Solomon’s judgment) The texts were selected from two Dutch translations: in the first place, the Nieuwe Bijbelvertaling (NBV), which was published in 2004 as the new, interconfessional translation. It will be compared to the 17th century Statenvertaling (SV 1637). For reference, all examples from these Dutch translations will be

66

Jos´e Sanders

followed by the comparable English standard translations: King James Version (KJV 1611) and the Contemporary English Version (CEV 1999).6 KJV and SV are examples of early translations from the source language that are loyal to the source text. CEV and NBV are modern translations which aim at source loyalty in combination with orientation to the goal language and its conventions (NBG/KBS 1998). The NBV explicitly opts for a text semantic point of view: individual words, sentences and text parts receive their meaning according to their mutual relations and to the whole text in which they are connected (Weren 2005). Thus, discourse elements are translated within their immediate and broader context, where “context” includes both literal and metaphorical material, and both text-conventional and culturally conventional.7 Among other things, this means that text is translated within the conventions of its original genre (letter, narrative, poetry) and that ambiguities are made explicit when potentially incomprehensible for modern readers, resulting in a natural and readable text without additional knowledge of the source language or Bible. An example of the latter we find in example 2 from Genesis 27:3 – 4. In this and following examples, the presence (or absence) and semantics of the connective(s) in the original Hebrew text will be made explicit, so that a real comparison can be made between original and different translations. (2) old

SV (Nu dan, neem toch uw gereedschap, uw pijlkoker en uw boog, en ga uit in het veld, en jaag mij een wildbraad; En maak mij smakelijke spijzen, zo als ik die gaarne heb, en breng ze mij, dat ik ete;) opdat mijn ziel u zegene, eer ik sterve. ‘that my soul bless you, before I die’ KJV (Now therefore take, I pray thee, thy weapons, thy quiver and thy bow, and go out to the field, and take me some venison; And make me savoury meat, such as I love, and bring it to me, that I may eat;) that my soul may bless thee, before I die.

new

6 7

NBV (Neem daarom je jachtgerei, je pijlkoker en je boog, ga het veld in en schiet een stuk wild voor me. Maak dat voor me klaar zoals ik het lekker vind en breng me dat te eten;) het zal mij de kracht geven om je te zegenen voordat ik sterf. ‘Prepare it for me as I like it and bring it to me to eat; it will give me the strength to bless you before I die.’

If necessary, the crucial part of the Dutch translation is translated literally as well. Also, the recent translation accords with current conventions with respect to the addressing of narrative characters: u (‘thy’), jij (‘you’) jullie (‘you’ pl.) (Wenzel 2005).

Causal Connectives in Dutch Biblical Translations

67

CEV (So take your bow and arrows, then go out in the fields, and kill a wild animal. Cook some of that tasty food that I love so much and bring it to me.) I want to eat it once more and give you my blessing before I die. hebr. rWbx]B' .hl;k] %aw] yLi ha;ybih;w] yTib]h'a; dv,a}K' syMix'f]m' yli-hcex]w' tWma; sd,f,B] yvip]n' w;k]r,b;T] ‘And w make me tasty foods such as I love, and w bring to me and w I shall eat; in order to dWbx]B' my soul shall bless you before I die.’ In the SV, the connection between the eating and the blessing is linguistically marked by dWbx]B' ‘opdat / that’ (Hebr. ‘in order to’) with conjunctive mood, which indicates that the first part is to be seen as enablement for the desired goal in the latter part. How precisely this connection is to be interpreted remains implicit. In the NBV, first of all the referential coherence is tightened by insertion of “het / it”, referring to the eating of the venison. Secondly, in the NBV “de ziel / the soul” is not present, and the relation between the eating of the meat and the blessing is semantically connected by the additional process of “giving strength (in order) to”: The food will give strength in order to bless, whereas the biblical concept of the soul8 fulfils no intermediate role. Note that in the SV, the consecutive acts “go out in the field – take venison – make meat – bring – eat – soul may bless” may be interpreted to refer to a ceremonial procedure that, once fulfilled, will make way for the blessing – as if it is a symbolic test for this firstborn before he can receive his final blessing. Possibly, the animate flesh of the venison further enforces the force of this ceremony. In the NBV, the consecutive acts are replaced by an explanation of an enablement process that can be paraphrased as: “go shoot a deer, eating of which will give (physical) strength so that I am able to bless you”. Thus, the possibly spiritual ceremony is overtly materialized. Presumably, the translation anticipates that readers will be looking for an acceptable motivation for the characters’ acts and utterances. In this case, such a motivation is made explicit in Isaac’s instruction. The ritualistic explanation is not taken into account. The example makes clear how pervasive interpretative changes in the translation can be. 8

Heb. yvimn] ' nafshi, ‘my soul’, i.e. the centre of self, the spiritual self. Alter (1996) translates nafshi, ‘life-breath’ as I may solemnly bless you, since nafshi here is an intensive synonym for ‘I’. Its use here as the actor of the blessing emphasizes that Isaac wants this blessing with his full will and strength. The coherence marker “so that” tightly connects the eating to the blessing, which indicates that the ceremony includes the hunt, the ritual meal and the blessing altogether.

68

Jos´e Sanders

We will now concentrate on coherence markings, more specifically on causal connectives. It will be made clear how causal connective choices in the newer translations constitute acts of categorization.

3. Coherence marking Generally, in translations of biblical narrative, the content – events, characters and storyline – is intended to be kept more or less equivalent to the source content, but the linguistic expressions, such as coherence markers are subject to variation. Here, conversational implicature plays an important role: readers will make as many inferences during reading as is possible and reasonable (Levinson 1983). The question is to what degree of explicitness these inferences are brought in translation. In as much as comprehension and readability are aims for a translation, discourse coherence is an important condition. Readers may find reading more or less effortful, depending on the way in which coherence at the clausal level is expressed in discourse. When local coherence is implicit, i.e. not overtly marked and made explicit, readers will have to retrieve their situation model: They go back into their long-term memory to find their accessible knowledge about the text’s theme and genre (Kintsch 1994). Local coherence can be made explicit by linguistically marking the coherence relations, for instance by means of connectives. Coherence relations are meaning relations that connect discourse segments (Hobbs 1979; T. Sanders, Spooren and Noordman 1992). Discourse connectedness is often defined in terms of these relations, such as Cause-Consequence, Contrast and Claim-Argument (T. Sanders and Spooren in press).9 Connectives such as and, but and because are common markers of coherence relations. The use of connectives is restricted by semantic and pragmatic rules. In many languages, there are two or more connectives that 9

A correlation has been found between certain types of coherence relations and the text type under consideration. T. Sanders (1997) found that informative texts, such as encyclopedia entries, are dominated by objective (content, semantic) relations, whereas persuasive texts (such as advertisements) show more subjective (epistemic and speech act/pragmatic) relations. This correlation was shown both in quantitative (number of relations through the whole text) and qualitative terms (which relations are important in the discourse structure?), showing that persuasive text has relatively more subjective relations, which typically appear at dominant positions in the structure; they organize the text as a whole. For the latter type of analysis Rhetorical Structure Theory (RST; Mann & Thompson 1988) was used to produce hierarchical tree representations of the texts. A similar methodology will be used here clarify the coherence marking in biblical narratives.

Causal Connectives in Dutch Biblical Translations

69

express more or less the same semantic context, but – as argued above – the use of connectives is also often restricted to the degree of subjectivity the Speaker intends to express (T. Sanders 2005). In the choice of markers in coherence and reference, readers are supported in understanding and memory of the text. A writer gives readers the best support not simply by maximizing local coherence between text parts, but by making the expression of coherence “optimally explicit”. That is, it is ideal for a reader if the text shows an optimal balance between supportive explicitness of what is new and difficult to understand on the one hand, and challenging to integrate the new with what is already in long term memory on the other hand. Thus, no text will be optimally explicit for all readers; the optimum depends on the specific readers’ knowledge (Kintsch 1994; McNamara et al. 1996; Kamalski, Lentz and Sanders 2008). As stated in section 2, this may lead current biblical translations to sacrifice some of their loyalty to the source text in order to clarify the storyline and characters at points where readers cannot be assumed to understand the biblical world (see NBG/KBS 1998); at other points, the translation may omit what is overtly known and thus redundant. These expectations were put to a test in a corpus analysis.

4. Corpus analysis 4.1.

Method

First, the five chapters were segmented into clauses, using the criterion of the finite clause. The SV corpus has 672 clauses, whereas the NBV corpus counts 605 clauses, which is 10% less. Then, for each clause it was determined by which type of coherence marker, if any, it was connected to its preceding clause.

4.2.

Expectations

The expectation was that compared to the Statenvertaling, the Nieuwe Bijbelvertaling will have less coherence marking that is not necessary to interpret coherential relations. Since many instances of the in biblical narrative ubiquitous additive and sequential connective en ‘and’ are redundant for coherence understanding, it is expected they will be less frequent in NBV, which will result in an overall relatively smaller proportion of linguistically marked coherence relations in NBV. At the same time, additive and sequential coherence markers are in many cases underspecifying the underlying coherence relations, which is why it is expected that temporal (toen ‘when/then’), content-noncausal (maar ‘but’, als

70

Jos´e Sanders

‘like’) and causal (omdat ‘because’, want ‘for’, dus ‘thus, so’) coherence markers would be relatively more frequent, which will result in an overall relatively larger proportion of such explicitly linguistically marked coherence relations.

4.3.

Results

It appeared that overall, the narratives from the NBV translation have fewer coherence markers between text parts at the local level than those in the SV, but these fewer markers are semantically richer and textually more complex (X2 [1] = 40, p < .001). This is shown in Table 1. Table 1. Coherence signals in clauses Statenvertaling (N = 672) and Nieuwe Bijbelvertaling (N = 605)

From a diachronic point of view it becomes clear that the semantically richer coherence markers are more robust over time. Table 2 shows a clear difference between the ways in which additive, temporal/sequential and causal/contentnoncausal connectives in SV were translated in the NBV (X2 [2] = 21, p < .001). Taking the type of coherence marker in the Statenvertaling as standard, it appears that more than half of all additive coherence markers have disappeared in the NBV; for temporal/sequential markers, this is the case for almost half of all markers. The causal and content-noncausal markers however disappeared in just one fifth of all cases. Remarkably, in more than a few cases temporal/sequential markers in the SV are translated with a causal or content-noncausal marker in

Causal Connectives in Dutch Biblical Translations

71

Table 2. Coherence signals from a diachronic viewpoint Statenvertaling → Nieuwe Bijbelvertaling (N = 428)

the NBV (10%); the reverse case, in which a causal or content-noncausal marker in SV is translated as temporal/sequential marker in NBV is very rare. These results further support the expectations. The results of the corpus analysis were studied in some more depth with respect to the degree of subjectivity in causal coherence markers. Since causality is underspecified in the Statenvertaling, first all fragments were selected that had a causal coherence marker in both translations. Then it was noted which coherence marker was used, if any. It appeared that the Nieuwe Bijbelvertaling had relatively more volitional and subjective causal coherence markers than the old SV translation. In other words, the NBV shows relatively more subjective coherence markers which express reasoning on the conceptualizing subject’s part, thus showing more epistemic coherence relations and less content relations (X2 [1] = 6, p < .05). Table 3 gives an overview. Table 3. Coherence signals in causal fragments (N = 41)

In the next section, the implication of these results are discussed in greater detail.

72

Jos´e Sanders

5. Discussion 5.1.

Subjectivity and rhetoric in narration

In the following example, taken from Genesis 27: 1, the NBV translation uses semantically richer and more subjective markers to establish discourse relations than the old SV translation, which follows the Hebrew source text. (3) old

SV En het geschiedde, als Izak oud geworden was, en zijn ogen donker geworden waren, en hij niet zien kon, toen riep hij Esau, zijn grootsten zoon, en zeide tot hem (. . . ) ‘as Isaac had become old, and his eyes had become dark, and he could not see, then he called Esau, his eldest son, and said unto him (. . . )’ KJV And it came to pass, that when Isaac was old, and his eyes were dim, so that he could not see, (. . . )

new

NBV Toen Isaak oud geworden was en zijn ogen zo zwak waren geworden dat hij niet meer kon zien, toen riep hij Esau bij zich, zijn oudste zoon (. . . ) ‘When Isaac had become old and his eyes had become so weak that he could no longer see, he called to him Esau, his eldest son (. . . )’ CEV After Isaac had become old and almost blind, (. . . )

hebr. /nB] wc;x]-ta, ar;qiYiw' t %ar]me wyn;yx] (yh,k]Tiw' qj;x]yi wq]z; yKi yhiy]w' yniB] wyl;a] dm,a %Yw' l %dG;h' ‘And w it came to pass, that yKi Isaac was old, and w his eyes were (too) dim to see t %ar]me (‘from to see-of’), and w he called Esau his elder son, and w said unto him (. . . )’ Firstly, it is remarkable that the translation of this fragment is more compact in the current NBV: it uses two clauses less than SV. Secondly, the sequential connectives in SV have made room for temporal subordinate connective toen ‘when’ and by zo (. . . ) dat ‘so that’, respectively. Isaac’s not being able to see is a consequence of the dimness of his eyes; in the SV, this cause/consequence relation remains implicit and is marked with additional en ‘and’. By contrast, the NBV translation explicitly marks this relation by the linguistic marker zo dat ‘so. . . that’, an expression modifying the nature of the weakness of Isaac’s eyes: they had become weak to such an extent that he could no longer see. The causal

Causal Connectives in Dutch Biblical Translations

73

relation is pragmatically present but the emphasis is on the modification of the weakness causing the blindness.10 In addition, NBV highlights the contrast with the past by adding (niet) meer ‘no longer’. Thus, the narrator is more subjectively foregrounded, interpreting the nature of the coherence between the states and events in the story. The tendency that is illustrated here – on the one hand, omitting redundant coherence markers, on the other hand using explicit connectives where coherence relations are complicated – has remarkable consequences for the discourse rhetoric, as can be elicited by a text analysis using Rhetorical Structure Theory (henceforth RST, Mann and Thompson 1988). Figure 1 depicts the RSTstructure of the SV translation of example 3, configuring the six clauses, two of which express a temporal sequence relation by means of the connectives alstoen ‘if-then’. The other four begin with additional/sequential en ‘and’ which is represented in two joint constructions. 1-6 Sequence 1-4

toen riep hij Esau, zijn grootsten zoon, then he called Esau, his eldest son,

Elaboration 2-4

En het geschiedde, and it happened,

en zeide tot hem, and said unto him

Joint als Izak oud geworden was, when Isaac had become old,

en zijn ogen donker geworden waren, and his eyes had become dark,

en hij niet zien kon, and he could not see,

Figure 1. rst structure of example 3, Genesis 27:1, Statenvertaling

Figure 2 shows an RST-structure of the same sequence in NBV translation. Firstly, the configuration is more compact: two coordinate clauses beginning with en ‘and’ in the SV have been here translated as subordinate clauses. Secondly, and more importantly, the discourse structure has deepened. The third “en” is replaced by “zo. . . dat” which marks a non-volitional result-relation which is represented at a deeper level than was the case in the SV translation. The ultimate result is a double subordination in a three-layered structure. In the rhetorical structure, the last clause – in which the essential act of the sequence takes place, i.e. Isaacs call to Esau – has the highest position, thus being stressed and positioned to receive highest attention. The older translation has a more “flat”, sequentially built structure. In this, it follows the source text with roots in an ancient and mainly oral culture, in which complex sentences 10 An anonymous reviewer pointed this out to me.

74

Jos´e Sanders

with multiple embedding would be unlikely (see Alter 1996: xvii). Presumably, rhetoric fore- and backgrounding would instead have been expressed by other linguistic means. Reading aloud, aspects of prosody such as speed, stress and intonation could have been functional in this respect. Circumstance

1-3 Joint Toen Isaak oud geworden was When Isaac had become old

2-3

Nonvolitional-result

en zijn ogen zo zwak waren geworden and his eyes had become so weak

toen riep hij Esau bij zich, zijn oudste zoon, then he called Esau, his eldest son (...)

dat hij niet meer kon zien, that he could no longer see,

Figure 2. rst structure of example 3, Genesis 27:1, Nieuwe Bijbelvertaling.

Although Rhetorical Structure Theory clarifies the type and rhetorical value of coherence relations, it cannot account for the subjectification that has taken place in the new translation. Mental Space Theory does offer a framework to analyse the narrator’s altered position. Figure 3 represents the MST configuration, in which the narrative content is projected by the narrator on the content side of the Basic Communicative Spaces Network (for a detailed description of the model, see Sanders, Sanders and Sweetser this volume: Chapter 2). In the narrative content space, a temporal subspace is elaborated by the use of toen ‘when’. In Figure 4, the Basic Communicative Spaces Network is used to analyse the NBV translation. Here, again a temporal elaborative subspace is effectuated by toen ‘when’. But within the subspace, a nonvolitional causal relation is expressed. This causal relation is to be construed in the domain of causal content relations, represented in the left upper corner of the diagram (compare the position of daardoor ‘as a result’; for a detailed account of this representation, again see Sanders, Sanders and Sweetser this volume: Chapter 2). By expressing this causal relation, the narrator is more closely connected to the narrative world than in the SV translation where the narrative content was merely projected; in the NBV, it is causally interpreted and signalled. Thus, the narrator becomes subjectified.

Causal Connectives in Dutch Biblical Translations And it happened, as Isaac had become old, and his eyes had become dark, and he could not see

content; explicit SoC

deictic centre of communication implicit SoC

p’ (and it happened) temporal elaborative subspace

speech act q’ (when Isaac was old) and r’ (his eyes were dark) and s’ (he could not see)

speaker / narrator says to addressee / reader:

then t’ (he called Esau, his eldest son) and u’ (he said unto him)

p, q , r, s, t, u

narrative content

knowledge base of additive and sequential ordering p and q and r and s and t and u

Figure 3. Mental space representation of Genesis 27:1, Statenvertaling When Isaac had become old and his eyes had become so weak that he could no longer see, he called Esau, his eldest son

q’’ (his eyes were weak ZO DAT r’’ (he could no longer see)

content, no SoC

deictic centre of communication; implicit SoC

speech act temporal elaborative subspace

p’ (when Isaac had become old) q’ (his eyes were weak r’ (he could no longer see)

speaker / narrator says to addressee / reader: p, q , r, s

narrative content

then s’ (he called to him Esau, his eldest son)

knowledge base of additive and sequential ordering p and q and r and s AND of content causal relations

Figure 4. Mental space representation of Genesis 27:1, Nieuwe Bijbelvertaling

75

76

Jos´e Sanders

The conceptualizing subject in a narrative text can be either the narrator or one of the narrative characters. In example 3, it was shown how the NBVnarrator is subjectified by his interpretation of the coherence relations; it is the character who is subjectified in the NBV in example 4 taken from Genesis 27: 22–23. (4) old

SV Toen kwam Jakob bij, tot zijn vader Izak, die hem betastte; en hij zeide: De stem is Jakobs stem, maar de handen zijn Ezau’s handen. Doch hij kende hem niet, omdat zijn handen harig waren, gelijk zijns broeders Ezau’s handen; en hij zegende hem. ‘Then Jacob came by, unto his father Isaac, who touched him; and he said: (. . . ) But he did not know him, because (. . . ); and he blessed him’ KJV And Jacob went near unto Isaac his father; and he felt him, and said, The voice is Jabob’s voice, but the hands are the hands of Esau. And he discerned him not, because his hands were hairy, as his brother Esau’s: so he blessed him.

new

NBV Jakob kwam dichter bij zijn vader staan en deze betastte hem. Het is Jakobs stem, dacht hij, maar het zijn Esaus handen. Omdat Jakobs handen even behaard waren als die van zijn broer Esau, herkende Isaak hem niet en dus zegende hij hem. ‘Jacob came to stand closer to his father, and he (the latter) touched him. It is Jacob’s voice, he thought, but they are Esau’s hands. Because (. . . ) Isaac did not recognize him and thus he blessed him’ CEV Jacob went closer. His father touched him and said, “You sound like Jacob, but your hands feel hairy like Esau’s.” And so Isaac blessed Jacob, thinking he was Esau.

hebr. WhVemuy]w' wybia; qj;x]yi la, b %qx]y' vG'Y iw' wyd;y; Wyh;-yKi /ryKihi a %lw] wc;xe ydey] syid'Y;h'w] b %qx]y'l/q l %Qh' wm,a %Yw' t %rxic] wyjia; wc;xe ydeyKi Whker]b;y]w' ‘And w Jacob went near unto Isaac his father; and w he felt him, and w said: The voice is the voice of Jacob, and w the hands are the hands of Esau. And w he discerned him not, because yKi (‘that’) his hands were hairy, as his brother Esau’s hands; and w he blessed him.’

Causal Connectives in Dutch Biblical Translations

77

In the last clause of this fragment, the NBV translates not only “en / and” but also “dus / thus”, so. By this marker, the NBV explicitly interprets that the character’s act is the result of what precedes, and additionally, that it follows from reasoning by the character; it is Isaac’s conclusion. Again, the discourse’s rhetoric is enhanced by the grammatical subordination of the grounding circumstances. In the SV-structure, the not recognizing by Isaac and the blessing are two sequentially related events at the same discourse level. In the NBV structure, the blessing – which is the essential act in this fragment – is the volitional consequence of the not recognizing, placed high in the discourse hierarchy. It appears that the NBV translation achieves a more compact and hierarchic discourse structure by marking discourse coherence in a more explicitly causal way, expressing subjectivity by foregrounding the Speaker’s conclusion. The subjectification can be limited to the biblical narrator, as was the case in example 3, but can also spread to the biblical characters, as we saw in example 4. This will be elaborated in the next section.

5.2.

Subjectivity and perspective in characters

In example 4, it was clear that by the use of a subjective causal connective, the narrator can be given entrance to the character’s consciousness. Another, more ambiguous example is taken from Genesis 27: 13–14. (5) old

SV En zijn moeder zeide tot hem: Uw vloek zij op mij, mijn zoon! hoor alleen naar mijn stem, en ga, haal ze mij. Toen ging hij, en hij haalde ze, en bracht ze zijn moeder. ‘And his mother said unto him: (. . . ) Then he went . . . ’ KJV And his mother said unto him, Upon me be thy curse, my son: only obey my voice, and go fetch me them. And he went, and fetched, and brought them to his mother.

new

NBV Maar zijn moeder zei: ‘Die vloek moet mij dan maar treffen, mijn zoon. Doe nu wat ik zeg en ga die bokjes voor me halen’ Dus ging hij ze halen en bracht ze naar zijn moeder. ‘But his mother said: (. . . ) Thus he went to get them and brought them to his mother’

CEV

Rebekah insisted, “Let his curse fall on me! Just do what I say and bring me the meat.” So Jacob brought the meat to his mother, (. . . )

78

Jos´e Sanders

hebr. yli jq' ^]lew] yli %qB] xm'v] ^]a' yniB] ^;t]l;l]qi yl'x; /Mai /l rm,a %Tw' wybia; bhea; rv,a}K' syMix'f]m' /Mai cx'T'w' /Mail] abeY;w' jQ'Yiw' d]l,Yew' ‘And w his mother said unto him: on me your curse, my son; only hear to my voice, and w go fetch me them. And w he went, and w fetched and w brought them to his mother and w his mother made tasty foods, as his father loved.’ In this fragment, NBV translates SV’s additional en ‘and’as contrastive maar ‘but’, which adds an interpretation by the narrator. SV’s temporal connective toen ‘when’, NBV translates as subjective causal connective dus ‘so’. Interestingly, this adds interpretation not only of the narrator’s conceptual structure, but presumably also of the character’s. Whose conclusion – that Jacob went as a reaction on Rebekah’s command – is represented here: The narrator’s, or Jacob’s? The connective dus ‘thus’ is ambiguous in this respect (see Sanders, Sanders and Sweetser this volume: Chapter 2, for a more elaborate discussion on this subject). Van Wolde and Sanders (1994) argue that characters’ perspective in biblical Hebrew can be expressed by various particles such as ki and (we)hinn`e, which indicate grades of focalisation: i.e. the degree to which the speaking subject allows readers to see through the eyes of a person in the text. An illustration can be found in example 2 above. In modern European literary languages such as Dutch, character perspective is not primarily indicated by particles but by a number of more or less narrative techniques, some of which have their origins in modern literary techniques – for example, the representation of embedded perspective by indirect and free indirect speech and stream of consciousness. Other techniques include the representation of observation and evaluation by verbs and adverbs of cognition and sensation, and representation of attitude by verbs and adverbs of volitionality and modality (for an overview, see Sanders 1996). The entrance to thoughts, feelings and goals of characters expressed by these linguistic means, enhance readers’ chances to identify with characters. Recent discourse comprehension studies show that the comprehension and appreciation of texts is enhanced by such linguistic facilitators of character identification (Land et al. 2002; De Graaf et al. 2007). This may explain why current translations such as NBV give direct access to a character’s inner consciousness, thus more explicitly construing the characters as conceptualizing subjects. Example 4 from Genesis 27:23 clarifies this strategy. In the SV translation, Isaac’s quote is represented as spoken; aloud or in himself, this remains implicit. The NBV unambiguously represents the perspective of Isaac’s inner thoughts. Most remarkable is the inversion of order of the embedded thoughts (“It is Jacob’s voice”) and the embedding clause (“. . . ”, he thought), causing a modern narrative discourse

Causal Connectives in Dutch Biblical Translations

79

pattern in comparison to the almost dramatic structure of the older translation: like the Hebrew source, the SV systematically names the speaking characters first, followed by his or her quote, as if they were the spoken lines in a play. The perspective choice not only has consequences for possible referential identification, it may also influence the subsequent discourse structure. In the SV structure, the narrator merely reports what Isaac said. Thus, a contrast exists between what Isaac says (The voice is Jacob’s voice, but the hands are Esau’s hands) and what follows. From his words, it might be inferred that Isaac recognizes Jacob or at least doubts Esau’s identity. However, this expectation conflicts with the outcome: he did not recognize him. These contrastive events are signaled with the connective maar ‘but’, a marker that refers the following to an embedded, counterfactual mental space (Dancygier and Sweetser 2005) which in the following clause will represent Isaac’s inner perspective (Sanders and Redeker 1996); in this space, expectations which had been raised in the basic space are denied. Figure 5 gives the Basic Communicative Spaces Network of example 4. Figure 5, as do the following figures, zooms in on the narrative content-side of the diagram, since the speaker’s speech act and epistemic spaces, although implicitly present, are less relevant here. For the whole configuration reference is made to figures 3 and 4 and to Sanders, Sanders and Sweetser (this volume). The causal relation, signalled by the connective omdat ‘because’ indicates a nonvolitional causal relation. In Dutch, the prototypical causal signal for objective, nonvolitional causality is doordat ‘because of the fact that’; though omdat ‘because’ is also possible, which both SV and NBV use here. character Isaac's space X' (it is Jacob’s voice but Esau’s hands)

counterfactual space

1 Isaac said X

Inference: P (Isaac recognizes Jacob? )

3 but not P

not P' (he did recognize him)

5 Q (his hands were hairy, like his brother Esau’s hands); 6 and he blessed him

not P' OMDAT Q'

nonvolitional space

Figure 5. Mental space analysis of Genesis 27:23, Statenvertaling

80

Jos´e Sanders

In the NBV-version, the narrator has entrance to Isaac’s inner consciousness: he quotes what Isaac thinks (It is Jacob’s voice, but the hands are Esau’s). From this inner perspective, no recognition by Isaac could be expected: the not recognizing is not a subsequent event but simultaneous with the thinking. Thinking and not recognizing are one and the same thing, and a contrastive signal would be misplaced here. Rather, Isaac’s line of thinking asks for an explanation, which the narrator provides in the next clauses. Thus, the relation between the not recognizing and the blessing is not merely a sequence of events, but a consequence, signaled not only by en ‘and’ but also by the narrator’s conclusion: dus ‘thus’. This is not Isaac’s conclusion – he does not know he does not recognize, and so cannot be making statements about the causes and effects of his lack of recognition – it is a conclusion in the narrator’s epistemic space. This analysis is made explicit in the mental space representation in Figure 6. narrative content character Isaac’s inner mental space

1 Isaac thought X 3 Q (his hands were as hairy as his brother Esau’s)

2 X' (Jacob's voice but Esau's hands)

P' OMDAT Q'

4 P (he did not recognize him) 5 and Z (he blessed him)

nonvolitional space

P'' DUS Z'

narrator's epistemic space

Figure 6. Mental space analysis of Genesis 27:23, Nieuwe Bijbelvertaling

In the latter example, the subjective causal connective dus ‘thus, so’ is used to signal the narrator’s conclusion. In a similar way, the character’s conclusion can be represented, as was exemplified above in 5 from Genesis 27: 13–14. In this fragment, the connective “toen / then” in SV is merely indicating a temporal relation. In the case of SV, no conclusion is drawn; by contrast, dus in NBV indicates a conclusion. The question is, whose conclusion? The sentenceinitial position of the connective indicates that what follows can be interpreted as represented (inner) discourse by the character in free indirect style, in which the character’s thoughts sound together with the narrator’s in a dual voice. The

Causal Connectives in Dutch Biblical Translations

81

narrative content character’s space

1 But his mother said X 3 Then he went, and got them, and brought them to his mother.

2 X' ('That curse must be on me, my son. Only hear my voice and go, get them for me.’)

Figure 7. Mental space analysis of Genesis 27:13–14, Statenvertaling

initial position of the connective is comparable to the turn-initial position of connectives in conversation where they function as ties between the participants’ viewpoints (Sotirova 2004). In this case of dus, the subjective causal connective is to be interpreted as indicating a character’s conclusion. In the mental space representation, the elaboration of reasoning in the character’s (Jacob’s) epistemic space is blended with the narrator’s epistemic reasoning. Ultimately, “dus” is constructed in the blended space. The blend entails that the narrator – and with the narrator, the reader – participates in Jacob’s conclusion. narrative content

character Rebekah’s space

2 X' ('Your curse be on me my son! Do as I say and go, get those goats before me.'

1 And his mother said to him: X 3 Q (he went to get them, and brought them to his mother )

X'' dus Q' X'' dus Q'

narrator’s epistemic space

characterJacob's epistemic space

X''’ DUS Q'’ blend

Figure 8. Mental space analysis of Genesis 27:13–14, Nieuwe Bijbelvertaling

82

Jos´e Sanders

An example of temporal ordering which entails reasoning on the part of a character is provided in (6) by the new NBV translation in Genesis 27:27. (6) old

SV En hij kwam bij, en hij kuste hem; toen rook hij de reuk zijner klederen, en zegende hem; en hij zeide: (. . . ) ‘then he smelled the smell of his clothes, and blessed him; and he said: (. . . )’ KJV And he came near, and kissed him: and he smelled the smell of his raiment, and blessed him, and said (. . . )

new

NBV Hij kwam dicht bij hem staan en kuste hem. Toen Isaak zijn kleren rook, sprak hij deze zegen over hem uit: (. . . ) ‘When Isaac smelled his clothes, he spoke this blessing over Jacob: (. . . )’ CEV While Jacob was kissing him, Isaac caught the smell of his clothes and said (. . . )

hebr. rm,a %Yw' Whker]b;y]w' wyd;g;B] j'yre ta, jr'Y;w' /lqV'Yiw' vG'Yiw' ‘And w he came near, and w kissed him. And w he smelled the smell of his raiment, and w blessed him, and w said:’ The reference with the full lexical NP as well as the paratactic Toen ‘then’ in SV indicates a merely temporal ordering, which is represented in the narrative space. Only the semantics of “smelled” are construed in the inner space of

narrative content

Isaac’s inner perspective mental space

1 and he came by 2 and kissed him 3 then he smelled the smell of his clothes

(smelled)

4 and he blessed him

Figure 9. Mental space analysis of Genesis 27: 27, Statenvertaling

Causal Connectives in Dutch Biblical Translations

83

the character Isaac. The smelling is presented as a mere event, with neither intentionality nor causal connection to the subsequent events. In the NBV translation, the smelling is represented in a subordinate clause. Compared to the SV translation, the hypotactic toen ‘when’ in NBV indicates a stronger relation; both intention and causal connection are suggested (as in only when he smelled, he blessed). This entails reasoning on the character Isaac’s part: the smell of the clothes, which is the smell of the field, is motivation for the blessing, or even legitimates it. This inference is elaborated in the character Isaac’s epistemic space. narrative content

1 and he came to stand close to him 2 and kissed him

character's space

4 P' (Isaac smelled his clothes)

3 when P 5 then Q (he gave him this blessing)

inference:

character's epistemic space

Q' because P''

Figure 10. Mental space analysis of Genesis 27: 27, Nieuwe Bijbelvertaling

Not only characters’ causal relations, but also God’s causal force can be described in more or less subjective ways, as expressed by the choice of causal connective. A remarkable example is presented in (7) by Genesis 27: 20.

(7) old

SV Toen zeide Izak tot zijn zoon: Hoe is dit, dat gij het zo haast gevonden hebt, mijn zoon? En hij zeide: Omdat de HEERE uw God dat heeft doen ontmoeten voor mijn aangezicht ‘Because (subjective in expression of character’s reasoning) the LORD your God has made me meet that before my face’ KJV And Isaac said unto his son, How is it that thou hast found it so quickly, my son? And he said, Because the LORD thy God brought it to me.

84

Jos´e Sanders

new

NBV ‘Hoe heb je zo snel iets kunnen vinden, mijn zoon! ’zei Isaak. En hij antwoordde: ‘Doordat de HEER, uw God, alles zo gunstig voor me liet verlopen.’ ‘because (objective in expression of causal course of events) the LORD, your God, let everything go so positively for me’ cev Isaac asked,“My son, how did you find an animal so quickly?” “The LORD your God was kind to me”, Jacob answered.

hebr. yKi rm,a %Yw' yniB] a %xm]li T;r]h'mi hZ,hm' /nB]la, qj;x]yi wm,a %Yw' yn;p;li ^;yh, %la… hw;hy] hr;q]hi ‘And w Isaac said unto his son: ‘How is it that you found it so quickly, my son? And w he said: Because yKi (‘that’) JHWH your God caused (it) to happen to my face.’ This fragment represents the only case in which a more objective causal connective is used in the newer translation. In the NBV, the objective causal connective doordat ‘because of the fact that’ expresses greater distance between the speaking character Jacob and the cause than the subjective causal omdat ‘because’in the SV, which puts more stress on the deliberate causation by God.11 narrative content character Isaac’s space

1. And Isaac said unto his son, X 3. And he said, Y

2 X’ (How is it

that thou hast found it so quickly, my son? )

X''

character Jacob’s space

4. Y' (OMDAT the LORD thy God brought it to me.)

OMDAT God’s volitional space

Y’'

Figure 11. Mental space analysis of Genesis 27:20, Statenvertaling

11 Note that this is the only place in this biblical chapter where the current NBV translation uses a more objective connective than the 17th century SV.

Causal Connectives in Dutch Biblical Translations

85

narrative content character Isaac’s space

2 X’ (How could

1. X, said Isaac

you find something so quickly, my son!)

3. And he answered, Y

character Jacob’s space

4. Y' (DOORDAT the LORD, your God, let everything go so positively for me) X'' DOORDAT Y’'

nonvolitional space (impersonal)

Figure 12. Mental space analysis of Genesis 27:20, Nieuwe Bijbelvertaling

Interdependent with the choice of the causal connective is the choice of the volitional verb. Doen ‘make’ in SV is a causative verb which in earlier centuries was used as laten ‘let’ is used in modern Dutch (Dutch: je laat me schrikken ‘you had me scared’, hij liet haar drie rondjes rennen ‘he let her run three rounds’), but which is now non-typical in referring to causation involving personal subjectivity. At present, it is merely connected to non-personal, inanimate causality (Dutch: de storm deed het beeld omvallen ‘the storm made the statue topple’) (for an overview, see Stukker 2005; Verhagen 2005). This semantic change could pose a problem for a current translation, for a connection between doen ‘make’ with a personal, animate God who would directly intervene in the course of events: is such direct causation by God probable and understandable for readers today? Verhagen and Kemmer (1997), in a discussion of the role of a conceptual model of authority in causality marking with verbs in cultural and religious contexts, give examples which indicate that “God can be conceived of as directly causing anything, including a change in a person’s mind (i.e. the conceived nature of the phenomenon of conversion)” (Verhagen and Kemmer 1997: 74). However, in NBV God’s direct causation in the course of events apparently was not desirable. In NBV, God no longer deed ‘made’ Jakob meet the venison, as in SV, He did not even liet ‘let’ it meet him; in stead, He liet alles gunstig verlopen ‘let everything go positively’. This combination of laten ‘let’ and verlopen ‘go’ ascribes a more distant, not actively interfering but permissive causal role to God, who no longer intervenes in the natural cause of events, and

86

Jos´e Sanders

leaves more room to act to the presently speaking character Jacob – that is, as far as Jacob relates here; from the context it can be inferred that this is a lie, since his caught of the venison is his own doing altogether.

6. Conclusion An analysis involving a Basic Communicative Spaces Network model offers new insights into the subjectification of both narrator and characters in translations of biblical narrative. The analyses presented in this chapter clarified the subjectifcation in the linguistic markers of coherence, specifically causal connectives, in the comparison of two Dutch translations: 17th century Statenvertaling and the recent Nieuwe Bijbelvertaling. The analyses of causal connectives showed that the old translation preserves a great deal of the sequential, drama-like structure of the source text. By the choice of more complex and precise coherence markers, the NBV narrator and characters make stronger interpretative choices. By doing so, the NBV narrator and characters come closer to the reader. In other words, explicitness in the translation enhances the subjective foregrounding of both narrator and the central characters, which creates a more up to date, fiction-like story; at the same time, the new translation is less transparent in that it leaves less room for interpretation by the reader. Thus, it may be concluded that in comparison to its source-loyal 17th -century, predecessor the Statenvertaling, the current Nieuwe Bijbelvertaling is more explicit in its choice of causal coherence markers. In doing so, the Nieuwe Bijbelvertaling aims to increase the reader’s comprehension and imagining of the biblical characters, but not without paying a price: it leaves the reader less room for interpretation and gives less clarity of the deeper meaning of the narrative in the contextual narrative line. Cognitive linguistic analysis of narrative discourse offers ample possibilities to enlighten diachronic developments in discourse and translation styles. The analyses given in this chapter provide some insight how semantic-pragmatic systems behind the use of causal connectives have developed over time.

Causal Connectives in Dutch Biblical Translations

87

References Alter, R. 1981 The Art of Biblical Narrative. Harper Collins. Alter, R. 1996 Genesis. Translation and Commentary. New York: Norton. Ariel, M. 1988 Referring and accessibility. Journal of Linguistics 24: 65–87. Arnold, B. and J. Choi 2003 A Guide to Biblical Hebrew Syntax. Cambridge: Cambridge University Press. Dancygier, B. and E. Sweetser 2005 Mental Spaces in Grammar: Conditional Constructions. Cambridge: Cambridge University Press. Dayras, S. 1993 The Knox version, or the trial of a translator: Translation or transgression? In: D. Jasper (ed.), Translating Religious Texts – Translation, Transgression and Interpretation, 44–59. London: The Macmillan Press. Evers-Vermeul, J. 2005 The Development of Dutch Connectives. Change and Acquisition as Windows on Form-function Relations. Dissertation, Utrecht University. Evers-Vermeul, J. and N. Stukker 2003 Subjectificatie in de ontwikkeling van causale connectieven? De diachronie van daarom, dus, want en omdat [Subjectification in the development of causal connectives? The diachrony of daarom ‘that is why’, dus ‘so’, want ‘for’ andomdat ‘because’]. Gramma TTT 9 (2/3): 111–140. Fauconnier, G. 1997 Mappings in Thought and Language. Cambridge: Cambridge University Press. Frye, N. 1982 The Great Code. The Bible and Literature. San Diego et al.: Harcourt Brace Jovanovich. Graaf, A. de, J. Sanders, H. Hoeken, and H. Beentjes 2007 De rol van identificatie in narratieve overtuiging [The role of identification in narrative persuasion]. Tijdschrift voor Taalbeheersing 29 (3): 237–250. Hobbs, J. 1979 Coherence and coreference. Cognitive Science 3: 67–90.

88

Jos´e Sanders

Kamalski, J., T. Sanders, and L. Lentz 2008 Coherence marking, prior knowledge and comprehension of informative and persuasive texts: Sorting things out. Discourse Processes 45: 323–345. Kintsch, W. 1994 Text comprehension, memory, and learning. American Psychologist 49 (4): 294–303. Knappert, L. 1936 Enkele Opmerkingen Bij De Geschiedenis Der Statenvertaling [Some remarks on the the history of the Statenvertaling]. Nederlands Archief voor Kerkgeschiedenis 29: 197–208. Land, J., T. Sanders, L. Lentz, and H. van den Bergh 2002 Coherentie en identificatie in studieboeken. Een empirisch onderzoek naar tekstbegrip en tekstwaardering op het vmbo [Coherence and identification in schoolbooks. An empirical study of discourse comprehension and appreciation in middle school]. Tijdschrift voor Taalbeheersing 24 (4): 281–302. Langacker, R. 1990 Subjectification. Cognitive Linguistics 1: 5–38. Levinson, S. 1983 Pragmatics. Cambridge: Cambridge University Press. Mann, W. and S. Thompson 1988 Rhetorical Structure Theory: toward a functional theory of text organization. Text 8 (3): 243–281. McNamara, D., E. Kintsch, N. Butler Songer and W. Kintsch 1996 Are good texts always better? Interactions of text coherence, background knowledge, and levels of understanding in learning from text. Cognition and Instruction 14 (1): 1–43. Nederlands Bijbelgenootschap/Katholieke Bijbelstichting 1998 Werk in uitvoering: de Nieuwe Bijbelvertaling toegelicht [Dutch Bible Communion/Catholic Bible Corporation: Work in progress. Explanation of the New Bible Translation]. Haarlem/’s-Hertogenbosch: Nederlands Bijbelgenootschap. Sanders, J. 1996 Perspective and attribution: The cognitive representation of biblical narrative. Poetics 24: 57–80. Sanders, J. and G. Redeker 1996 Speech and thought in narrative discourse. In: G. Fauconnier and E. Sweetser (eds.), Spaces, Worlds and Grammar, 290–317. Chicago: University of Chicago Press. Sanders, T. 1997 Semantic and pragmatic sources of coherence: On the categorization of coherence relations in context. Discourse Processes 24: 119–47.

Causal Connectives in Dutch Biblical Translations Sanders, T. 2005

89

Coherence, Causality and Cognitive Complexity in Discourse. In: M. Aurnague, M. Bras, A. Le Draoulec and L. Vieu (eds.), Proceedings/Actes SEM-05, First International Symposium on the exploration and modelling of meaning, 105–114. Toulouse: Universit´e de Toulouse le Mirail. Sanders, T., W. Spooren, and L. Noordman 1992 Toward a taxonomy of coherence relations. Discourse Processes 15: 1–35. Sanders, T. and W. Spooren 2007 Discourse and Text Structure. In: H. Cuyckens and D. Geeraerts (eds.), Handbook of Cognitive Linguistics, 37–60. Oxford: Oxford University Press. Sevenster, G. 1936 De Statenvertaling En Hare Kantteekeningen [The Statenvertaling and her annotations]. Nederlands Archief voor Kerkgeschiedenis 29: 263– 306. Sotirova, V. 2004 Connectives in free indirect style: Continuity or Shift? Language and Literature 13 (3): 216–234. Sperber, D. and D. Wilson 1986 Relevance: Communication and cognition. Oxford: Blackwell. Stukker, N. 2005 Causality Marking across Levels of Language Structure. A Cognitive Semantic Analysis of Causal Verbs and Causal Connectives in Dutch. Dissertation, Utrecht University. Sweetser, E. 1990 From Etymology to Pragmatics: Metaphorical and Cultural Aspects of Semantic Structure. Cambridge: Cambridge University Press. Traugott, E. 1989 On the rise of epistemic meanings in English: an example of of subjectification in semantic change. Language 65: 1–55. Traugott, E. and R. Dasher 2005 Regularity in Semantic Change. Cambridge: Cambridge University Press. Verhagen, A. 2005 Constructions of intersubjectivity. Discourse, Syntax and Cognition. Oxford: Oxford University Press. Verhagen, A. and S. Kemmer 1997 Interaction and causation. Causative constructions in modern standard Dutch. Journal of Pragmatics 27: 61–82.

90

Jos´e Sanders

Wenzel, V. 2005

Wie tutoyeert God? Aanspreekconventies in de moderne bijbelvertalingen [Who is on first-name terms with God? Addressing conventions in modern Bible translations]. Tijdschrift voor Taalbeheersing 27 (4): 299–314.

Weren, W. 2005

Lucas in de nieuwe Bijbelvertaling [Lucas in the New Bible Translation]. Tijdschrift voor Theologie 45 (1): 75–87. Wolde, E. van and J. Sanders 1994 Kijken met de ogen van anderen. Perspectief in bijbelteksten [Seeing through others’ eyes. Perspective in biblical texts]. Tijdschrift voor Theologie 34: 221–245.

Sources Statenvertaling (Jongbloed-editie), c 2004 Nederlands Bijbelgenootschap, De Nieuwe Bijbelvertaling Copyright  King James Version, c 1999 American Bible Society; all digitally available Contemporary English version,  at www.biblija.net Interlineair Bible Hebrew – English: Westminster Leningrad Codex transliterated; Conc Copyright 2007 Menno cordant Hebrew English Sublinear( for WLC, version 2.0,  Haaijman); digitally available at http://www.scripture4all.org/OnlineInterlinear/Hebrew Index.htm

Causes and consequences: Evidence from Polish, English, and Dutch Barbara Dancygier The pervasiveness of causality in grammatical systems of many languages and the resulting variety of causal conjunctions are often not matched by proportional salience of all the various conjunctions involved. As Sanders et al. and Stukker et al. argue (this volume), the observed variety is due in equal measure to categorial distinctions involving domains, subjective construals, and volitional meanings and to prototype effects, such that the categories are expressed more or less saliently. This paper will discuss some data from Polish, to propose an introductory categorization of causal conjunctions, but then focus on two specific connectives, to and bo, to show the role of discourse-related phenomena in the preference for some conjunctions over others. One possible conclusion which emerges from the overview below is that instead of marking all the fine differences among causal construals consistently through various layers of discourse, language may put more weight on selected all-purpose conjunctions. Interestingly, while individual conjunctions may reveal specific understandings of causality across cognitive domains, levels of subjectivity, and discourse types (see Knott, Sanders and Oberlander 2001; Sanders and Spooren 2001; Verhagen 2005 for in-depth reviews of this awe-inspiring variety), the ones occurring in a broad variety of contexts have to largely ignore such differences. The result is, as the cases of Polish to and bo suggest, that the overarching expressions of causality function as argumentative conjunctions rather than as descriptively adequate markers of causal relations. That is, these conjunctions mark the speaker’s intention to present two situations as linked causally in the speaker’s mind, rather than as indeed linked causally within any specific domain, including the content domain. In what follows, I will rely on Mental Spaces Theory, as introduced by Fauconnier (1985, 1997). The theory has been successfully applied in the analysis of a variety of adverbial constructions (see Dancygier and Sweetser 2000, 2005 for a mental spaces analysis of causals and conditionals; and Verhagen 2005 for a range of constructions relying on intersubjective construals). In the present volume, Sanders et al. propose a more elaborate general mental space configuration, termed the Basic Communicative Space Network, which appears to be particularly useful in representing the natural emergence of subjective construals. I will rely on the basic tenets of this proposal and on Verhagen’s analysis

92

Barbara Dancygier

of intersubjectivity in constructions to show how data from Polish confirm the priority of subjective causality over some concept of objective cause-effect relations. In particular, I will show to and bo to be primarily intersubjective in use. I will start with a brief overview of Polish causal conjunctions, to further focus on bo and to, which apply in a wide range of contexts. To will be shown to present its clause as a consequence (for example, in if -conditionals, paratactic constructions, and some mono-clausal constructions), while bo will be argued to profile its clause as a reason for communicating the main clause. Both types of constructions show very few formal restrictions, and rely primarily on the sequence of clauses as their main constructional parameter. The scope of their usage (from causal constructions, through concessive, to conditional, especially in the case of to) also suggests that the strategy of relying on an intersubjective construal over and above “content”, or real-world one also connects various adverbial conjunctions into a more coherent network of constructional options. The Basic Communicative Space Network proposed in Sanders et al. (this volume) is useful in representing the fact that construals (causal or otherwise), while communicated by a contextually recognizable speaker, may rely on subjectivities other than that of the speaker herself. That is, the role of Subject of Consciousness (from now on SoC), as the locus of a mind where the construal originates, may be evoked independently of the subjectivity of the speaker, and various SoCs may participate in the construal in their own ways. This idea has also recently been brought up in Verhagen’s (2005) proposal of a mental space configuration called a construal configuration, which consists of the object of conceptualization (roughly equivalent to the real-world situations talked about) and the subject of conceptualization, or ground, which represents various contextual factors, but primarily the subjectivities of the speaker, the hearer, and other discourse participants. While both representations capture the role of SoCs in the emergence of constructional meanings, the format of the Basic Communicative Space Network (from now on BCSN) also naturally incorporates the concepts introduced in Sweetser (1990), whereby constructions may be interpreted in domains such as the content domain, the epistemic domain, and the speech act domain, with the subsequent addition of the metalinguistic domain (Dancygier 1992, 1998; Dancygier and Sweetser 2005). Verhagen’s construal configuration, for comparison, is a broad enough framework to allow all of these varieties to be naturally incorporated, but is also more specific in capturing the cases where the speaker is not only relying on the construal attributed to another SoC, but also engages that construal in her own discourse for argumentative or inferential purposes. Indeed, the two frameworks occasionally cover some of the same usage types, such as the cases of indirect discourse. While Sanders et al. describe the case

Causes and consequences: Evidence from Polish, English, and Dutch

93

of Free Indirect Discourse as a blend of two SoCs (the actual speaker and the “reported” speaker), Verhagen talks about instances of Indirect Discourse (such as He said he wanted to come) as instructions for the hearer to construe the situation from the reported speaker’s point of view. Verhagen’s framework is thus more explicit in evoking the hearer’s intended cognitive benefit and the speaker’s argumentative intersubjective goals, independently of the actual construals in the mind’s of the participants. While focusing on somewhat different phenomena, both frameworks attempt to address the potential tension between the speaker’s SoC and those of other participants (whether present or evoked), and both capture the ways in which constructions rely on subjectivity as described in Langacker (1990). However, they seem to be focusing on different levels of constructional meaning and thus are naturally combined in analysis. I will return to the ways in which their concerns can potentially be brought into shared focus in my final comments. The view of causality as a matter of subjective construal rather than objective real-world relationships has been gaining ground in recent research (cf. Pander Maat and Degand 2001; Pander Maat and Sanders 2001; Stukker et al. this volume). In this approach, subjectivity is presented as naturally understood as a matter of degree, or a measure of distance between the current speaker and the SoC of the participant primarily responsible for the construal represented in the construction. Indeed, as I argue below, conjunctions relying on subjective construals are also the most common and most flexible choices in Polish. However, they are possibly more accurately described as intersubjective, in Verhagen’s terms, since they also play argumentative and inferential roles, using the distance across different subjectivities in constructionally determined ways. As Verhagen points out, intersubjective construals are more than a matter of degree of convergence between the speaker and other SoCs, because they profile the viewpoint of another participant as included in the reasoning. As a result, the concept of intersubjectivity as discussed in Verhagen (2005) focuses, among others, on how specific formal choices can be viewed as contributing to the argumentative function of the construction (cf. the discussion of negation, complementation constructions, and concessive connectives, Verhagen 2005). The cases discussed byVerhagen suggest that the behavior of bo and to, which are central to the argument below, are also cases where the argumentative function is central to the construction’s use, but the issues of volitional, epistemic, and speech act meanings, as represented in BCSN, also play a role in the choice of the conjunction. I will start the discussion of intersubjectivity with a brief overview of causal connectives in Polish, in terms of the mental space configurations they set up.

94

Barbara Dancygier

1. Causal conjunctions in Polish The range of causal connectives in Polish is quite broad, but there are two conjunctions which are most common in colloquial usage – to and bo. While bo is typically considered a backward causal connective (though, as I will show, it also appears in constructions where causality is less central to the construal), to is typically talked about as an apodosis marker, rather than a conjunction. To is thus not, strictly speaking, a causal connective, and cannot mark content causality in a natural way, but it participates in causal construals at the argumentative level, also in the cases where causality is but a part of a more specific adverbial construal (conditional, concessive, etc.). Bo and to thus have to be considered in the context of other, more specific, causal connectives in Polish. I will start with the expressions of forward causality, and then continue with a brief review of conjunctions similar to because. In general, it is worth pointing out that causality in Polish is most typically expressed via epistemic or intersubjective construals, with significant attention to subjectivity and little focus on objective causality in the real world. The only expression of objective, content causality (with non-volitional meaning and no SoC suggested) is skutkiem czego ‘as a result (of which), daardoor’, as in (1)

Nie domkn˛ełam zamra˙zalnika, skutkiem czego lody si˛e roztopiły. ‘I didn’t properly close the freezer. As a result, the ice cream melted.’

While the expression is almost literally pointing to the first clause as the cause and the second one as the result, skutkiem czego is not necessarily the preferred option in these cases, and more subjective connectives, such as wi˛ec, are common in colloquial discourse, even though it is clear that allowing the temperature in the freezer to drop caused the ice cream to melt. Other expressions of forward causality focus primarily on the reasoning process, with a set of options signaling inferential chains similarly to so or therefore: wi˛ec (the closest broad equivalent of so/dus), czyli z˙e (strictly inferential), zatem (also inferential, used only in formal discourse), to znaczy z˙e ‘it means that’. In this set, wi˛ec is the most open to volitional meaning, while the remaining expressions are primarily used to mark epistemic reasonings. For example, wi˛ec, but not any of the remaining epistemic connectives, can be used in (2): (2)

Ma by´c gor˛aco, wi˛ec p´ojd˛e jutro popływa´c. ‘It is supposed to be hot, wi˛ec [so/dus] I’ll go for a swim tomorrow.’

Causes and consequences: Evidence from Polish, English, and Dutch

95

The speaker’s decision to go swimming cannot be interpreted in strictly epistemic terms, so wi˛ec is the only acceptable form, but any of the epistemic connectives is appropriate as long as the reasoning remains non-volitional, as in: (3)

Widziałam s´wiatło w oknie, wi˛ec/zatem/czyli z˙e/to znaczy z˙e wr´ocili ju˙z z urlopu. ‘I saw lights in the window, so they must have already returned from vacation.’

In these cases, the speaker can rely on overheard or reported premises. If the first clause of (3) were in the third person, not in the first person, the choice of connective would not be affected, but the conclusion would still be attributed to the speaker (Janek widział s´wiatło w oknie, wi˛ec/zatem/czyli z˙e/to znaczy z˙e wr´ocili ju˙z z urlopu / John saw lights in the window, so [I conclude that] they have already returned from vacation). It appears that the inclusion of another SoC’s reasoning or volitional state is a less salient aspect of the construal in Polish, so that wi˛ec (but not zatem, czyli z˙e, or to znaczy) is the standard option also when the speaker’s subjective state is not profiled in the construction and the third person is used throughout: (4)

Janek usłyszał, z˙e ma by´c gor˛aco, wi˛ec postanowił z˙e p´ojdzie jutro popływa´c. ‘Janek heard that it is supposed to be hot, so he decided to go for a swim tomorrow.’

For comparison, the connective to would not be likely in any of these examples. As I will argue below, it participates in similar reasonings, but at the intersubjective level, rather than as an expression of epistemic or volitional meanings. The speaker’s SoC is defocused (and the causal chain emphasized) in the constructions using the conjunction dlatego (??daarom, ??that’s why, ??which is why). Interestingly, the conjunction is a common choice when epistemic or volitional meanings are backgrounded (when present at all), because the causal link as such is foregrounded. What is more, dlatego appears in two variants (i) dlatego ‘(and) that’s why’, to mark forward causality, and dlatego z˙e ‘because’ to mark backward causality: (5)

a.

Janek jest zm˛eczony, (i) dlatego / (wi˛ec) nie przyjdzie. ‘John is tired, (and) that’s why / so he won’t come.’

96

Barbara Dancygier

b. c. d.

Jestem zm˛eczona, (i) dlatego / (wi˛ec) nie przyjd˛e. ‘I’m tired, (and) that’s why / so I won’t come.’ Janek nie przyjdzie, dlatego z˙e (bo) jest zm˛eczony. ‘John won’t come, because he is tired.’ Nie przyjd˛e, dlatego z˙e (bo) jestem zm˛eczona. ‘I won’t come, because I’m tired.’

Examples c. and d. are interesting for the argument presented here in that bo is allowed in these cases. However, unlike dlatego, it backgrounds the causal link on the object domain, and foregrounds the intersubjective function of the clause. The overview above suggests that the expression of causality through connectives in Polish is broadly structured by epistemic and volitional reasoning (as expressed with wi˛ec on the one hand and the inferential connectives on the other), in contrast to more specific expressions signaling a causal connection, but not limited to it. While wi˛ec highlights the speaker’s reasoning, dlatego highlights causality, but neither one is specifically concerned with the issue of subjective distance between SoC and the speaker. The main body of this paper will discuss Polish connectives whose primary role is to manage argumentation and inferencing across different subjectivities. For example, forward causality is sometimes marked by an apodosis/mainclause marker to, which is not specific to any causal construal. It can be used a sentence like in Było gor˛aco, to poszłam popływa´c ‘It was hot, so I went swimming’, which is very similar to (2). Consequently, wi˛ec could be used instead of to here, but because it refers to past time its first clause is naturally construed as a justification of the decision to go swimming, and may confirm either the fact that swimming took place (independently available in the context) or specify the reason why it happened. In this sense, to seems to be similar both to so and dus, while highlighting the intersubjective accessibility of the events and/or relations between them. As I will show below, the range of its applications is much broader than causal constructions proper, but here is one instance where intersubjective argumentation, in the sense proposed in Verhagen (2005), is associated with a use of a conjunction similar to other causal connectives. An even more salient example of a causal connective which is centrally related to intersubjectivity is bo, a colloquial variant of dlatego z˙e or poniewa˙z ‘because’, as well as an apparent cousin of the Dutch want (see Verhagen 2005 for a discussion of the differences between want and omdat; further exploration of this question is beyond the scope of the present paper). Below, I will take a closer look at bo and to, in an attempt to put them in the context of other related conjunctions, especially the causal ones.

Causes and consequences: Evidence from Polish, English, and Dutch

97

2. Causality and intersubjectivity The argumentative potential of causal constructions has been clearly described in Verhagen’s recent work (cf. Verhagen 2005). Verhagen points out, among other things, that language communication is pervasively argumentative and inferential in nature, so that speakers are more engaged in mutual management of emergent construals then in representing objective ones. Our communicative abilities depend on a construal configuration which distinguishes the domain of the object of conceptualization (roughly equivalent to Sweetser’s content domain [cf. Sweetser 1990]) from the domain of the subject of conceptualization, or Ground (which includes participants, the communicative event, shared context, etc.). Verhagen then shows how a number of constructions (including causal ones) rely on the negotiation in the subjective domain, rather than on attribution of specific objective or subjective construals to conceptualizers. Intersubjectivity is thus achieved constructionally, on the basis of various uses of the overall construal configuration. The approach offers an elegant explanation of the issue of causal chains being involved in constructions which are primarily operating in, let’s say, epistemic domains. For example, it has been noted repeatedly (Sweetser 1990; Dancygier 1998; Dancygier and Sweetser 2005) that epistemic conditionals use, and often reverse, causal chains. For example, we can argue that an epistemic conditional such as If the streets are wet, then it must have rained during the night relies on the causal correlation set up in If it rained during the night, the streets are wet. While both sentences are in some sense argumentative and both rely on the same causal link, they are different in their use of the subject and object of conceptualization, with the epistemic sentence arguing for a connection in the subject domain, on the basis of a correlation in the object domain. Moreover, as Verhagen shows, some conjunctions, especially those involving contrast (such as although or but) involve causal background assumptions alongside epistemic judgments in all of their uses. Thus the same causal chain may be applied differently to the same situation (depending on the intersubjective goals). Dutch causal constructions with dus and want appear to be specifically relevant to the discussion of bo and to connectives below. Verhagen discusses them as similar, in that they both set up a connection between the cause P and the effect Q, but introduce the clauses in a different order (p dus q or q want p), and that they both portray the speaker and another conceptualizer (the hearer) as holding somewhat different attitudes to the q clauses. The speaker is thus using the construction to strengthen the hearer’s attitude to the consequent clause and the primary difference is the direction of argumentation. Because both constructions conventionally evoke different subjective stances towards the same

98

Barbara Dancygier

thought, they can naturally be represented as profiling different SoCs in the BCSN, as well as using the epistemic, speech act, or metalinguistic components of the network. Only some of the uses of dus and want overlap with conjunctions bo and to in Polish, but it seems plausible to refer to bo and to as resembling the functioning of dus and want in introducing intersubjectively argumentative constructions, which choose a different cause/effect sequence, but both assume the profiling of a different SoC, whose construal should be affected by the construction. Both of the Polish conjunctions also rely on causal links in different components of the BCSN. My primary interest in their functions, though, derives from the observation that in colloquial discourse bo and to are overwhelmingly more common than more specific connectives such as dlatego. While to apparently shares its area of usage with wi˛ec (jointly, they seem to cover the scope of dus, at least roughly), bo is undoubtedly the most common of backward causal connectives in Polish. I argue that they serve as examples of how a varied system of causally-based connectives may gradually converge towards options which, while less specific, are open to a greater variety of uses, and which organize causal and other adverbial links at the intersubjective level. One way to interpret the observations outlined below is that a construction whose primary meaning is argumentative may gradually subsume other, more content-based meanings into its scope.

3. Polish to versus English then To has most generally been described as an apodosis marker, especially in conditionals. In some grammars, to is not given an independent status as a conjunction, and is mostly mentioned as the marker of the main clause of a conditional. It is also not obligatory in these constructions, so it has often been dismissed as meaningless, similarly to English then in the same role (cf. Bielec 1998; Fisiak et al. 1978; Grochowski et al. 1984; Jodłowski 1976). The conjunction then itself has been subject to much misunderstanding. Following the logical tradition, it is often mentioned alongside if (if . . . then is a formulation of the connective marking logical implication), but semanticists have repeatedly claimed it to be meaningless. More recent work, however, noted a number of restrictions on the use of then (cf. Iatridou 1994), while the discussion in Dancygier and Sweetser (1997, 2005) argues for then as a meaningful apodosis marker which deictically points to the preceding protasis space. The meaning of then relies equally on its deictic role and on its relationship to the temporal conjunction when, thus supporting those meanings of conditional constructions

Causes and consequences: Evidence from Polish, English, and Dutch

99

which rely on sequentiality (regardless of whether the sequentiality is understood to obtain in the content, epistemic, or speech act domain). There are some similarities between English then and Polish to. Polish to is also described as correlated with if (je˙zeli . . . to) and both conjunctions are common (though not required) as markers of the apodosis. Both appear in conditional constructions with the if p then q/je˙zeli p to q clause order, while the reversed order constructions *then q if p/to q je˙zeli p are unacceptable, which seems to arise out of the anaphoric function of to and the deictic function of then. However, one of the primary differences between the conjunctions is that the etymology of to reaches back to deictic and anaphoric markers, and not to temporal adverbs (see Tabakowska 1997 for a fuller discussion). Its independence from the temporal sequential sense makes to a much more flexible marker, as it can successfully work as an anaphoric main clause marker for most adverbial conjunctions: (6)

Je˙zeli si˛e postara, (to) zda ten egzamin. ‘If he tries hard, (to) he will pass the exam’ – conditional

(7)

Chocia˙z si˛e starał, (to) jednak nie zdał. ‘Although he tried hard, (to) he still didn’t pass’ – concessive

(8)

Poniewa˙z si˛e starał, (to) zdał. ‘Because he tried hard, (to) he passed’ – causal

Unlike then, to can thus mark main clauses of adverbial conjunctions of different kinds. While its presence in causal constructions like (8) is perhaps less surprising, since the q clause can be construed as a “consequence” or “result” both in conditionals and in causals, the concessive construction in (7) is of a different kind. Concessives have been repeatedly described as relying on expected causal sequences (trying hard should result in success), but the constructions represent these sequences as not valid in the particular situation described. It is thus not possible to construe the situation described in the second clause of (7) as resulting from the one profiled in the p clause. The anaphoric function of to, however, allows the speaker to present p as communicable in this context. It is not marking q as a consequence of p, but it is marking the saying of q as a consequence of first saying p – which may also include a conditional, concessive, or strictly causal relation. It is thus argumentative and intersubjective in the sense described in Verhagen (2005), without at the same acquiring an unambiguous epistemic or speech act interpretation. The difference between (8) and other causal constructions introduced in section 1. is of particular importance here. The sentence could naturally be rendered with wi˛ec as the only conjunction (Starał si˛e, wi˛ec zdał/ He worked hard, so he

100

Barbara Dancygier

passed), so it is natural to see the connection between p and q as not holding in the “object of conceptualization”, or content domain, but in the subjective domain. At the same time, the construction in (8) marks causality by the use of poniewa˙z ‘because’ in the subordinate clause, so to in the main clause cannot be seen as introducing it. It does, however, support the causal reasoning by profiling the result in q as not at all surprising – one can expect studying hard to bring results. It thus supports Verhagen’s view of such constructions as clarifying the status of one of the clauses for the communicative benefit of another SoC (discourse participant) whose commitment to q needs reinforcement. The focus on argumentation and coherent flow of discourse, rather than on causality in any specific cognitive domain is what explains the use of to in constructions which may be interpreted adverbially (conditionally or causally) without using overt adverbial constructions. The most interesting area of usage, especially in comparison with English, is the cluster of constructions which are interpreted conditionally, even though they use coordinate conjunctions like and or no conjunctions at all: (9)

You mess up one more time and you’re/ you’ll be fired.

(10)

You mess up one more time, you’re fired.

(11)

Mess up one more time, and you’ll be fired.

Examples like (9), (10) and (11) have been discussed in Dancygier and Sweetser (2005) as allowing for a conditional interpretation on the basis of compositionally relevant criteria such as the use of verb forms (present plus willverb, imperative plus will-verb, or present-present) or the order of clauses which iconically matches the sequence of events (see Haiman 1985, 1986 for further discussion of the role of iconicity in such constructions). In this view, the present tense used in reference to the future and an indication of a sequence of events are grammatical features which prompt a predictive interpretation of the sentence, similar to what is found in futurate predictive conditionals. These compositional parameters are salient in English, but in Polish they do not suffice (see Dancygier and Trnavac 2007 for more discussion). Conjunctionless constructions like (10) are not possible in Polish at all, while coordinate constructions are rare. The restriction seems to emerge, at least partly, from the fact that Polish does not have an appropriate verb form contrast, since the standard expression of futurity is the present perfective form of the verb – clearly distinguishing between the “base” function of the English present and the predictive form like will is thus not natural in Polish. Also, among two coordinate conjunctions in Polish – i and a – the first one appears most commonly in the additive sense, while the

Causes and consequences: Evidence from Polish, English, and Dutch

101

second one, which implies contrast, especially in the cases where the clauses have different subjects, can sometimes be used in this way: (12)

a.

b.

Podpadniesz jeszcze raz, a stracisz prac˛e. (roughly an equivalent of [9]) ‘You mess up one more time, a you’re/ you’ll be fired’ Podpadnij jeszcze raz, a stracisz prac˛e. (roughly an equivalent of [11]) ‘Mess up one more time, a you’ll be fired.’

However, sentences like (12b), with a present perfective verb in an imperative form in p, appear naturally and commonly with to introducing q, as in Podpadnij jeszcze raz, to stracisz prac˛e. To is also used in constructions which do not have to be interpreted as predictions, in which an imperative p clause is followed by a to-clause, and the imperative is genuinely urging the hearer to do something. These constructions are also not open to the epistemic/volitional meaning which would be naturally represented as wi˛ec or dus: (13)

Otw´orz okno, to zobacz˛e czy pada. ‘Open the window to I’ll see if it is raining.’

It appears that when a to-clause follows an imperative with some directive force (which is clearly not the case in [12b]) it is not interpreted conditionally, and does not require expressions of minimal unit semantics such as one more time. On the contrary, as a genuine imperative it is unlikely to appear with such expressions. In such sentences, to is thus maintaining its function of presenting q as a consequence of p, or perhaps arguing that the result of following the imperative will enable the speaker to check the weather. In spite of its flexibility, to is clearly a clausal conjunction – so that it cannot be used in NP, NP constructions such as Another day, another dollar or More work, more pay. This observation, alongside examples considered earlier, might suggest that the link between to and predictiveness is in fact quite loose. In examples like (9) – (11) and the NP, NP constructions, causal meanings are primarily evoked as aspects of the predictive interpretation. Causality, alongside sequentiality and unassertability, is an important part of predictiveness – if speakers predict the q space from the p space, it is because there is a causal link between these spaces, even if it does not apply to the content domain. On closer inspection, to cannot be seen as participating in the construction of predictive meanings as such (which commonly arise in the content, or “object of conceptualization” domain), because its role is to organize discourse in the “subject of conceptualization” – to guide the hearer in the recognition of the

102

Barbara Dancygier

speaker’s course of reasoning. If predictiveness is a part of the interpretation, it is not signaled by to alone, but constructed based on all the parameters of the construction together. The only cases where to is used with expressions other than full clauses occur when some minimal token of (elliptical) clausal structure is retained. For example, to can appear in tautological statements such as Spadnie, to spadnie (It falls to it falls/If it falls, it falls) or as an equivalent ofYou get a cold and that’s it, signaling that the illness would prevent the addressee from doing what has been planned, as in Zazi˛ebisz si˛e, to koniec (You get-cold to end). Also, stand-alone negative forms anaphorically referring to some state of affairs which has just been negated are quite possible, as in Nie, to nie (If you don’t, you don’t/ If he won’t, he won’t, etc.). Finally, unlike English then, to marks the apodosis of speech act constructions, with or without epistemic stance shift: (14)

Je˙zeliby´s chciał, to przynios˛e jeszcze kawy. ‘If you’d like me to, to I’ll get more coffee.’

In his discussion of conditionals, Fillmore (1986, 1990) observes the correlation between the choice of verb forms and the speaker’s epistemic stance. He notes that past tense forms in conditionals consistently mark negative epistemic stance, and that the stance is expressed in both clauses (the observation was further developed in Sweetser 1996 and interpreted in terms of mental space embedding). In Polish, however, epistemic stance shift in one of the clauses is possible. In example (14), the protasis p is marked with epistemic distance by the use of the marker -by – a standard morphological expression of the conditional and/or subjunctive mood in Polish, which also adds to the politeness of this offer. The apodosis is not marked for negative stance, but to links the clauses sufficiently tightly to make the sentence acceptable. Also, while it is natural to interpret to as suggesting that the act of getting more coffee will result from the hearer’s wants, it seems to simultaneously present the speaker as willing to act for the hearer’s benefit as a result of acknowledging his wishes. To is an acceptable link between two components of a speech act construction, also when it does not take the form of a conditional. The example used in Sanders et al. to suggest causal underpinnings of speech act constructions would thus be translated into Polish with to: Mam kaw˛e i herbat˛e – to czego si˛e napijesz? (There is coffee and tea; so/dus what do you want?). As in the other cases, the information about available beverages is not itself an offer (unless in response to a request for something to drink), but is construed that way when it is presented as justifying the question which follows. It is also important to note that to can

Causes and consequences: Evidence from Polish, English, and Dutch

103

here introduce a component in the form of a question – in itself a signal of a construction involving more than one SoC. Finally, the constructional flexibility and versatility of to can be seen in its mono-clausal uses. While to cannot appear at the beginning of bi-clausal constructions, it can still introduce the main clause or its elliptical equivalent when the reason for communicating the clause has emerged in the preceding discourse. Compare (15) with (16) and (17): (15)

Je˙zeli ci˛e boli głowa, to zosta´n w domu. ‘If you have a headache, then (to) stay home.’ ‘Als je hoofdpijn hebt, dan blijf je thuis.’

(16)

To zosta´n w domu. ‘Then (To) stay at home.’ ‘Dan blijf je thuid!’

(17)

To nie! ‘Then (To) no.’ ‘Dan niet / dus niet.’

While (15) is just an ordinary conditional, it also confirms some degree of overlap between to, then, and also Dutch dan. Example (16) is a truncated version of (15), where to picks up on the communicated problem with the headache, similarly to Then stay home or Stay home, then. Example (17) is typical as an offended response when the other participant has just refused or denied something the speaker suggested (similar to Then don’t!) but cannot be understood to be reinforcing the hearer’s resolve if the hearer had previously gone on record about refusing to do x, which the English expression can (it is thus a truncated version of the tautological Nie, to nie. – If you won’t, you won’t). Unlike in some of the constructions mentioned above, this use of to parallels similar constructions in English. While the observation does not add much to what has already been said about to, it might suggest that mono-clausal constructions in English can be seen as specifically using then in the intersubjective domain, where the hearer needs to be informed not only of the speaker’s conclusions, but also of their direct link with preceding discourse. Finally, the possibility of choice between dan and dus in the Dutch version of (17) suggests that the overlap between to and dus, typically restricted to speech act constructions, also appears in such constructions, where the intersubjective meaning is the most salient. To is thus a causal main clause marker in a very broad sense. In toconstructions, the speaker is presenting q as emerging from a causal reasoning from p-space, while neither p nor q have to be assumed to be in the “content

104

Barbara Dancygier

domain” (so that the causality is also not necessarily “content”). More specifically, the p-space in this construal emerges as communicable in the context, and prompts (causes?) the communication of q as the speaker’s proposed understanding of p’s communicative impact. The relationship between p and q in the object of conceptualization (if any) is secondary; constructions with to are thus primarily intersubjective. The above overview of the uses of to does not clearly distinguish among its uses in sentences using other adverbial conjunctions as well, and in those where to is the only link. And yet, to is quite clearly understood in all cases discussed above to present its clause as an elaboration or a consequence of previously communicated discourse. To conclude this section, I will compare the contexts where to is possible with those where other conjunctions, especially wi˛ec, can play a similar role. The uses of to which rely most specifically on the surrounding discourse are those where there is a preceding p-clause marked with an adverbial conjunction of cause, condition, or concession. Examples (6), (7), and (8) above are typical representatives of this area of usage. Characteristically, while p-clauses are marked with their adverbial conjunctions, to is not obligatory, but it also cannot be substituted by any other conjunction implying consequence. In a sense, the presence of a clause unambiguously specifying cause or condition clearly designates the role of to as a marker of consequence picking up on the mental space set up in p, so that the construction as a whole and its accompanying mental space network do not require further elaboration. The anaphoric function of to sufficiently describes these uses, since an anaphoric reference to a causal or conditional space naturally connects to the elaboration of that space in terms of consequences or conclusions. In these cases, to is indeed similar to then, though not explicitly in its temporal function. The constructions where to is the only conjunction present are less specific with regard to the nature of the causal link expressed, and thus more restricted. There are several criteria, established in Sanders et al. (this volume), which help clarify the differences among constructional patterns. First of all, to alone does not appear in strictly non-volitional, epistemic contexts (though it does accompany the conditional conjunction je˙zeli/if as a marker of a conclusion in an epistemic conditional). In such contexts, wi˛ec ‘so’ is the only available choice, as in: (18)

Sło´nce s´wieciło, wi˛ec/*to było gor˛aco. ‘The sun was shining, so it was hot.’

(19)

Było ciemno w oknach, wi˛ec /*to ich nie było. ‘Their windows were dark, so they were out.’

Causes and consequences: Evidence from Polish, English, and Dutch

105

The only situation where to might appear in a construction like (18) is when the hearer has expressed surprise at the accepted fact that the temperature was high, and the speaker offers p, followed by the to-q clause, as an explanation or justification of the otherwise accepted knowledge that it was indeed hot. However, such a construction does not profile an inferential chain, and wi˛ec is a less likely conjunction then. Still, a paraphrase with je˙zeli-to/if-then is perfectly acceptable in rendering the reasoning in (18), because je˙zeli takes the burden of framing the construction as an epistemic conditional. Furthermore, to is in fact not used in predictive sentences which do not present the speaker as performing a speech act (such as warning or offering advice) or otherwise intersubjectively negotiating the hearer’s involvement. For example, (20) is difficult to read as a conditional prediction of possible sunshine causing increased temperature (for this reading, the speaker would have to use je˙zeli-to/if-then). It can only be understood as taking tomorrow’s sunshine as a given and announcing it as therefore tantamount to high temperatures. In such a case, the clauses would have to be pronounced with non-comma intonation. Still, wi˛ec would be a preferred option here. (20)

?Sło´nce b˛edzie s´wieciło, to b˛edzie gor˛aco. Sło´nce b˛edzie s´wieciło, wi˛ec b˛edzie gor˛aco. ‘The sun will be shining to it will be hot.’

For comparison, when the speaker’s volition or speech act performance are involved, wi˛ec is usually not possible, while to is the preferred choice. An example like (21) is a case in point. In (21), in the context of planning a shared walk, the speaker is announcing her decision to wait somewhere along the way rather than inconvenience others; such a volitional and intersubjective meaning of the construction requires to, and makes wi˛ec unacceptable: (21)

Zm˛ecz˛e si˛e, to /*wi˛ec poczekam. ‘I’ll get tired, to/*wi˛ec I’ll wait.’

The intersubjective and speech act import of a construction relying on to alone is still clearer if we consider examples referring to the past, and not the future (where predictive reasoning is typically available). In situations where a past factual situation is already on the conversational table, to can spell out the consequences of that situation, especially as they are relevant to the hearer, while wi˛ec is much less likely here, and unacceptable with the imperative:

106

(22)

Barbara Dancygier

Zgubiłe´s moje r˛ekawiczki, to /*wi˛ec kup mi nowe (to/?wi˛ec musisz mi kupi´c nowe. ‘You lost my gloves, to / *wi˛ec buy me a new pair (to/?wi˛ec you must buy me a new pair). ’

The uses discussed above consistently support the understanding of to (especially when it is the only conjunction present) as presenting consequences of situations communicated in preceding discourse in a way that primarily benefits the hearer’s appreciation of the speaker’s thoughts, needs, and expectations. In the contexts where the speaker uses to in a construction not specified as conditional, causal, or concessive, the hearer is “instructed” to construe the to-clause as representing the consequence of some relevant subjective understanding of the preceding clause. In this way, to forces the hearer to interpret its p-clause (in whatever spoken or implied form) as somehow prompting the communication of q. Thus to not only presents q as a result, but (re)construes the hearer’s understanding of p as a cause, albeit not on an objective level. Such a construal organizes on-going discourse in a specific way, dynamically negotiating the interpetations across different SoCs. Overall, the description of to proposed above supports the role of subjective construal in understanding causality in discourse. Quite clearly, the most common marker of consequence clauses in Polish focuses primarily on intersubjective meanings, while consistently downplaying content causality. What is more, it constructionally involves the hearer in negotiating the most appropriate interpretation.

4. Polish bo versus English because The scope and frequency of the use of to seems to be matched quite well by another conjunction – bo. Bo is a causal conjunction roughly equivalent to because, or its colloquial version ‘cause – it introduces subordinate clauses of causal constructions. Bo is the most common causal conjunction in Polish. It emerged as a shortened version of bowiem, which could be roughly glossed as ‘because-I-know’, which correlates with the fact that its scope is broader that content causality alone. Similarly to to, bo introduces the second clause of the construction, but the construction profiles backward causality (apparently similarly to Dutch want). Also, bo-clauses are typically not sentence-initial (the restriction seems stronger than the one limiting the sentence-initial use of because), but, similarly to to, but unlike want, bo can introduce mono-clausal elliptical constructions (as in [26]). The salience of the q, bo p pattern of boconstructions is not matched by other causal conjunctions in Polish, such as

Causes and consequences: Evidence from Polish, English, and Dutch

107

poniewa˙z (content because) or dlatego ‘that’s why’, which are more flexible as regards clause order options. Examples of bo-constructions range over a broad area of usage: (23)

Zdał, bo si˛e bardzo starał. ‘He passed bo he tried hard.’

(24)

Musiał si˛e bardzo stara´c, bo przecie˙z zdał. ‘He must have tried hard, bo he did pass after all.’

(25)

123 dzieli si˛e przez 3, bo 1+2+3 dzieli si˛e przez 3. ‘123 is divisible by 3, bo 1+2+3 is divisible by 3.’

(26)

Bo nie. ‘Bo I didn’t/I don’t /I won’t.’

In all these examples, which range across all cognitive domains, bo-clauses are presented as causes of or reasons for the communication and/or factual status of the main clauses. Example (23) may be seen as a straightforward causal construction, with a presupposed q and an asserted p, (24) and (25) are epistemic (while relying on content causality), while (26), similarly to (17), is an uncooperative response to a question like Why didn’t you finish your homework?. Example (24) is interesting in that it represents an abductive reasoning (i.e., a reasoning from the result to the cause), but it requires overt expressions of an inferential sequence in both clauses (musiał ‘must have’and przecie˙z ‘after all’). Using bo epistemically without these expressions is very awkward (consider?? Starał si˛e, bo zdał /He tried hard, bo he passed) and tends to be reconstrued into a rather unnatural scenario where trying hard is caused by passing. At the same time, (25), a classic example of an inferential reasoning or rule does not require additional expressions of epistemic forms to be present, possibly because both p and q are considered factual, and all that bo does then is show the speaker’s reasoning. Furthermore, an example like (23) can easily be rephrased as an argumentative construction if the speaker adds an epistemic expression to the q clause referring to the future, as in Na pewno zda, bo si˛e bardzo starał /He will pass for sure, bo he tried hard. It seems justified, then, to treat bo as an argumentative conjunction, presenting causes for the speaker’s communicative acts, while possibly also evoking causal links in the “object of conceptualization”. Alternately, one could argue for inherent subjectivity of causal statements shining through all the uses of bo. It seems, however, that the entrenched directionality of bo-constructions is better explained in terms of the conjunction’s argumentative role. Also, such an interpretation reveals interesting parallels between bo and to.

108

Barbara Dancygier

5. Comparing bo and to As the examples above show, bo and to are used in a similar range of constructions and they are both functioning in the “subject of conceptualization”. Both argue for a causal link between two spaces, but the direction of argumentation is different: to uses p to foreground q (result), while bo uses q to foreground p (cause/reason). In spite of obvious differences between to and dus, the pairing of bo and to resembles the contrast identified in Verhagen (2005), distinguishing Dutch conjunctions dus and want. The contexts where bo is used indeed seem to resemble at least some of the central uses of want, although, naturally, only a thorough study could reveal more facts. For example, bo (and not dlatego z˙e or poniewa˙z ) would be used in an example which supports Verhagen’s central argument about the use of want, as opposed to omdat: (27)

Jammer als ‘t niet duidelijk is, want ‘tis laat en ik heb geen zin om ‘t uit te leggen. ‘Too bad if it isn’t clear, because it’s late and I don’t feel like explaining it.’ ‘Je´sli to nie jest jasne, to trudno, bo jest ju˙z p´oz´ no i nie mam ochoty ci tłumaczy´c.’

The use of the causal conjunction in (27) suggests argumentation in the speech act domain, and in fact many uses of bo are best described in such terms, including the following classic speech act causality example used in Sweetser (1990): (28)

What are you doing tonight, because there’s a good movie on? ‘Co robisz dzi´s wieczorem, bo jest dobry film?’

As I argued above, conjunction to also naturally appears in speech act argumentation, so that bo and to are indeed similar in this respect. It is also seems true of both bo and to that while “content level” causality or “causality in the object of conceptualization” may be a part of what is communicated, it is independent of the argumentation in “the subject of conceptualization” which the conjunctions engage in. Consequently, some examples of bo-constructions may rely on a presupposed q to explain its source via p (as in [23]), while in other cases the construction asserts q, justifying the assertion with p (as in [25]). To-constructions also vary in a similar way (p is presented as presupposed or not, depending on the context and the choice of p-space builder). It can further be argued that the relative independence of clauses in both constructions in terms of epistemic stance marking (see Fillmore 1986, 1990) is

Causes and consequences: Evidence from Polish, English, and Dutch

109

at least to a degree the result of their argumentative nature. The cases of mental space embedding in conditionals (discussed in Sweetser 1996) are distinguished, among others, on the basis of the required inheritance of stance marking from the higher space to the lower one. More specifically, if the if -clause mental space, which is higher in the network, uses past tense forms to mark negative epistemic stance (rather than past time), its main clause will also inherit the same past tense marking, so that the embedded space does not conflict with the embedding space. In the case of conditionals, the projection of stance into embedded spaces explains why if -clauses marked for negative stance are typically followed by apodoses which inherit the same degree of epistemic distance. For instance, If you had some time is more likely to be followed by we could go for a walk tomorrow than by we will go for a walk tomorrow. In Polish, with to marking the apodosis, such a combination is acceptable, as in Je˙zeliby´s miał czas, to p´ojdziemy jutro na spacer (If you had some time, to we will go for a walk tomorrow). The protasis is then tentative and polite, as in English, but the tentativeness does not have to (though of course it can) be marked in the apodosis. Such sentences in Polish can roughly be glossed as ‘I am not assuming that you will have time tomorrow, but if you did, here is what I propose’. In a sense, it is clear that the hearer’s time is the issue to be negotiated, while assuming that the walk is welcome is justified. It is also possible to say Je˙zeli masz jutro czas, to mo˙ze by´smy poszli na spacer (If you have some time, to we could go for a walk tomorrow) – with negative stance only in the apodosis. The suggestion then is that it is quite likely for the hearer to be free the next day, but his wish to go for a walk requires negotiation. As regards bo, clauses with differing stances are also acceptable, as in Poszłabym na spacer, bo jestem zm˛eczona (I would go for a walk, bo I’m tired). The conditional/distance marker -by in the first clause suggests a wish, then justified in the bo-clause with the present state of affairs. It is harder to come across examples where stances are reversed, but it seems to be the result of the fact that the bo-clause is causal, and thus is more likely to refer to factual situations rather than wishes or suggestions. The only situation where the main clause can be marked with positive or neutral stance (no -by in Polish), while the bo-clause is marked with negative stance occurs when the main clause uses overt negation, as in Nie zostawi˛e dziecka samego, boby si˛e przestraszyło (I won’t leave the baby alone, bo he would get frightened). But this construction, similarly to English, seems to imply that the bo-clause is in fact an apodosis of a truncated conditional, so that it can be glossed as ‘I won’t leave the baby alone, bo he would get frightened if I did’. It seems, then, that the choices of stance shifts and combinations are dictated by the argumentative role of both conjunctions.

110

Barbara Dancygier

While bo clearly is a cause or reason conjunction, the status of to is less transparent, in that it accompanies different conjunctions, but also links clauses on its own. However, since the constituents it works on are clausal, even if elliptical, it should probably be referred to as a conjunction. Moreover, its ability to support the main clause in adverbial constructions of many kinds further confirms its argumentative status, as it co-occurs with causal meanings of more specific kinds (conditional, concessive, content causal, etc.). Interestingly, however, it cannot co-occur with bo. The primary reason seems to be the direction of argumentation (from cause to effect or from effect to cause). For the same reason both conjunctions have the same clause-order restrictions, namely, they cannot be sentence-initial. At the same time, the very fact that they share the restriction and do not appear in other clause-order patterns distinguishes them from many more specific causal or conditional constructions where both clause order patterns are possible, with added subtlety emerging from comma or non-comma intonation. It appears, then, that bo and to are used in the intersubjective domain to mark causality in argumentation, and in some sense cover the two main clause order patterns, since one construction is, loosely speaking, a mirror image of the other.

6. Alternatives and imperatives The argumentative aspect of to could be seen particularly clearly in the role it plays in constructions which construct conditional meaning without conditional conjunctions. In English, the constructions in question are quite varied – starting from ordinary coordinate constructions with and, through conjunctionless ones, to non-clausal NP, NP combinations. In Polish, as I noted above, the range of such constructions is limited, and relies primarily on to, which seems to confirm its role as an argumentative organizer of causal meanings of various kinds. There is, however, another group of coordinate constructions with conditional meaning – the constructions with imperatives and or as a conjunction. In fact, the conditional, and therefore also causal, meaning of such constructions in English is so salient that they were typically described, albeit in different ways, as conditionals. Typical examples of such or-constructions (or Dutch of -constructions) function as threats or deterrents (see Fillenbaum 1986): (29)

Stay where you are or I’ll shoot.

As Dancygier and Sweetser (2005) have argued, sentences like this prompt a mental space set-up involving two conditionals: in one, the expected compliance

Causes and consequences: Evidence from Polish, English, and Dutch

111

results in no harm to the hearer, while in the other, the (discouraged) noncompliance results in the hearer being hurt.The construction, as it stands, profiles the protasis space from one construction, and the apodosis space from the other, but relies on the hearer’s understanding that the choice is between two causal sequences. In spite of its broad range of applications, to cannot apply here in Polish, as it would presumably suggest that issuing an order to stop is the reason for threatening the hearer with shooting him. However, the reverse relation, whereby threatening to shoot gives the hearer a reason to follow the order to stop is roughly what the sentence does. Thus the standard equivalent of (29) is a construction with bo, as in (30): (30)

St´oj, bo strzelam! ‘Stand-Imperative bo I shoot.’

A threat like (29) can be expanded to highlight its conditional meaning, into something like Stay where you are or I’ll shoot if you don’t, and similar rephrasing is possible in Polish: (31)

St´oj, bo jak nie, to strzelam! ‘Stand-Imperative, bo [if not, to I shoot]!’

In this way, bo and to can appear in the construction, in their respective roles. In fact, it seems possible to use ‘cause in English in a similar way, as in (32): (32)

Stay where you are, ‘cause I’ll shoot if you don’t.

What this seems to suggest is that the mental space set-up in Polish and in the ‘cause constructions like (32) is different from the alternative set-up of (29). Instead of two alternative conditional configurations, (31) and (32) use the imperative to set up a future space of expected compliance (the hearer stops moving), and then justify the interactive strength of an unhedged imperative with a conditional profiling of the potential non-compliance and its consequences. The explicitly expressed conditional is thus not an equal alternative of another conditional, but is instead embedded in the causal follow-up to the imperative already on the table. Interestingly, this construal, even though not explicitly profiling alternativity, still relies on the concept of alternativity described in Dancygier and Sweetser (2005), whereby profiling a future potential situation comes with an unavoidable setting up of its unrealized alternative. Predictive protases of conditionals and imperatives are both instances of this kind of alternativity in the content domain, since the whole point of engaging in conditional prediction or issuing orders is that the implied unrealized scenarios be downplayed or back-

112

Barbara Dancygier

grounded without being rejected. The choice between or (or of in Dutch) on the one hand and ‘cause or bo (but apparently not want) on the other thus plays on the same idea of alternativity, but the latter constructions also choose a different strategy at the argumentative level – instead of building the alternativity into the argument via or, they rely on presenting the link between the order and the ensuing threat as causal. Interestingly, bo-constructions in Polish can be used as threats even if the actual desired state is not explicitly identified through an order. For example, an exhausted parent watching a child repeating an undesirable action (such as making a noise or touching a forbidden object) can use a mono-clausal construction with bo, such as Bo po˙załujesz! / Bo you will regret it. Again, there is an imperative in the context (possibly repeated many times in preceding discourse), as well as an unsaid conditional part to this utterance (if you don’t stop), but it is still interesting that the bo-clause can prompt for the whole construction in the right context. However, not all constructions with imperatives and bo-clauses are understood as equivalents of deterrents with or. The bo-clause does not have to be understood as a threat, even if it describes some unwelcome situation and performs a speech act. In such a context it still functions in the same way as in other contexts – to clarify the cause of what is communicated in the main clause. In (33), the imperative is justified with the speaker’s need for fresh air, and furthermore, the bo-clause then describes a present state of affairs. (33)

Otw´orz okno, bo duszno. ‘Open-Imperative the window bo (it is) stuffy.’

If the bo-clause of (33) were further changed into a prediction of the speaker collapsing for lack of fresh air (as in Otw´orz okno, bo si˛e udusz˛e /Open the window, bo I’ll suffocate), the urgency of the imperative would increase, but it would not necessarily be understood as a threat. It seems, then, that bo supports a wide range of speech acts, from complaints to threats, while in English or is the preferred option in the case of threats. To sum up, in all the examples with imperatives and bo the addressee is urged to follow the request or order (issued or signaled, verbally or not), with the bo-clause describing a current or future situation giving him a reason to do what the speaker suggests. Bo-clauses thus consistently support the communicative goals of their qclauses (in content, epistemic, or speech act domain), by specifying the reasons the speaker invokes in the argumentative event represented by the construction as a whole. Finally, the foregrounding of p (the threat or a state of affairs to be

Causes and consequences: Evidence from Polish, English, and Dutch

113

recognized), while arguing for the causal connection between a request/order and the accompanying complaint/threat is what gives the whole construction its argumentative power.

7. Argumentation and blending Specific p/q relations in the various bo and to constructions in Polish emerge from the broader format of such constructions – their being argumentative in the subject of conceptualization, rather than descriptive in the object of conceptualization. The two constructions both signal causal links in the discourse, but in a different order: to marks the construed result, while bo marks the construed cause. The broad discourse scope of both conjunctions suggests that argumentative causal links may play a significant role in discourse organization and interact in complex ways with other types of causality. Among others, they may shed light on the ways in which causal meanings participate in supporting the emergence of other meanings. At the same time, being rooted in discourse, intersubjective connectives discussed here seem to call for an expansion of the BCSN model in the direction of the representation proposed in Verhagen (2005). While BCSN offers an elegant way of distinguishing degrees of SoC involvement and of representing subjectivity blends in many causal constructions, it does not offer a straightforward way of incorporating intersubjective meanings. BCSN starts with the deictic anchoring of the speech event and distinguishes it from its content, which may evoke more than one SoC, but it does not present the ground as an independent locus of meaning construction. The speaker and the hearer are naturally part of the Ground as a source of deictic anchoring, but not as subjectivities which can be related in specific ways. However, what the intersubjective set-up is intended to capture is the fact that there may be differences within the Ground which are specifically addressed through the constructional means chosen (see Sweetser 2008). Specifically, intersubjective connectives like bo and to rely on the difference in the degree of the hearer’s and the speaker’s alignment with one of the clauses or the understanding of the situation described. Negotiation across these viewpoints, and the speaker’s argumentative attempt to bring the hearer to share her construal of the situation are the goals of intersubjective constructions. The examples of BCSN discussed in Sanders et al (this volume) focus on different phenomena – primarily on those where a SoC other than the speaker is evoked alongside the speaker’s implicit involvement in the reasoning, but there is no tension between the speaker and the other SoC, so that their aligned viewpoints in these cases are correctly presented as blends.

114

Barbara Dancygier

Among others, Sanders et al represent some examples of the uses of dus – the conjunction described by Verhagen as intersubjective. For example, the epistemic construction P (the lights are out) DUS Q (the neighbors are not at home) is represented with one SoC, that of the speaker, while the epistemic third person construction P (Jan saw their lights were out) DUS Q (they were not at home) is presented as blending the explicit SoC (Jan) with the implicit SoC of the speaker. Sanders et al. are thus offering a very subtle explanation of how an epistemic argument with the use of dus can at least partially rely on evoked subjectivities other than the speaker. For comparison, Verhagen’s treatment of dus focuses on a different aspect of the construal. Rather than consider the ways in which the speaker in an argumentative construction may be speaking for other subjectivities accessible in the ground, he assumes that all of its uses are intersubjective in the sense that the hearer is assumed to hold a different or uncertain view of the result situation in the proposed causal chain, and the speaker’s use of the dus-construction is intended to bring the speaker’s and the hearer’s views to be aligned. As a result, some important questions arise about the role of intersubjective or blended arguments and the possible interaction among them. Let us recall that when Verhagen talks about reporting constructions (such as Jan saw that the lights were out), he treats them as instructions for the hearer to consider the lights being out from the point of view of Jan, and not the speaker, while the speaker is realizing her own argumentative goals as well. That seems to be the first “intersubjective” part of the argument. But when a dus-conclusion is drawn, then, if I understand Verhagen correctly, it would be intended to strengthen or prompt the hearer’s understanding of the impact of the earlier statement. However, if that statement is itself intersubjective (deferring the evidential modality to Jan), the conclusion (they were not at home) would be drawn by Jan and reported by the speaker, or drawn by the speaker, who assumes that Jan would have drawn the same conclusion, or simply drawn by the speaker. But whichever way the SoC blend works, the conclusion is still argumentatively benefiting the hearer (not Jan and not the speaker). Presumably, the SoC blends postulated in Sanders et al. seem to create a blended subjectivity to present the argument profiled by dus, while Verhagen would possibly consider some of them as intersubjectively deferred. However, the reasoning spelled out for the benefit of the hearer, the very intersubjective core of the dus-construction, is from the speaker’s perspective alone. We could perhaps distinguish, then, the ways in which specific links in the argument may arise as blends or intersubjective constructions, from the ways in which the construction as a whole may do its intersubjective job. The lowest level of any mental space set-up should represent the Ground as the locus of

Causes and consequences: Evidence from Polish, English, and Dutch

115

information needed to understand the speaker’s and the hearer’ purposes in the exchange – this would include, among others, the possibility of different factual knowledge and different attitudes, as well as shared knowledge and shared attitudes. Furthermore, any communicative or argumentative event can potentially rely on other such events, whether present in the Ground or explicitly brought up by the interlocutors. We should thus be able to represent the various levels of constructional embeddings, and then distinguish the cases where they lead to blends from the cases where they lead to an argument which relies on multiple subjectivities within and outside of the Ground. Crucially to the interests of this volume, this may involve various causal links, ranging across different cognitive domains, and brought into the argument through evocation of various subjectivities. It is beyond the scope of the present paper to conduct a thorough study of bo and to on all these levels of constructional complexity. Nevertheless, the range of uses of bo and to gives support to some general claims made elsewhere in this volume. First, subjective construals, rather than objectively existing causal chains, seem to lie at the core of all the uses of bo and to. The question of blended SoCs in epistemic constructions is also not salient in my examples, because to is not acceptable in inferential reasoning sentences (see the discussion of examples [2] – [4] above, where wi˛ec [the closest equivalent of dus] is used instead of to). The only cases where a to-clause relies on another SoC are truncated sentences in (16) and (17), where the reason for the speaker using to is a statement available in immediately preceding discourse. However, it would still be difficult to argue that the “concluding” subjectivity is blended – staying home is still the speaker’s suggestion alone. Bo is also not common in such uses – even the speech act example (28), while evoking the hearer’s expected cooperation in q, uses the bo-clause to express the speaker’s reasons for asking. To conclude, while other causal and conditional conjunctions have more specific and more naturally objectifiable uses, as well as more varied types of blended SoC uses, the two conjunctions discussed in more detail in this paper are primarily used in the intersubjective domain, with a very broad range of uses. Interestingly, their intersubjective and argumentative status accounts for rather unexpected restrictions – consider the fact that to is practically barred from marking inferential conclusions when no conjunction marks its p. In this case, contrary to, for instance, conditionals, where the main dividing line lies between the predictive and non-predictive ones, the conjunction is not appropriate when the reasoning is not volitional, so that no subjective construal is foregrounded. At the same time, the use of bo with imperatives shows how the openness to intersubjective reading can be exploited in different mental space set-ups, with different intersubjective or speech act consequences. Finally, we should note that

116

Barbara Dancygier

the two conjunctions, in spite of their broad application, are possibly the most common expressions of causality in Polish, even though other, more specific, and more objectively framed conjunctions are available. Finally, it seems that subjective and intersubjective causal construals are more communicatively useful than the more specific and more explicitly objective ones, precisely because of their interpersonal function. Depending on the specific choices a language makes, such causal construals may further depend on lower levels of subjective blends and intersubjective arguments. But the intersubjective connectives as such operate on the top level of this constructional pyramid, and thus in the end the overall meaning of the construction will crucially be located there – at the level of intersubjective argumentation.

References Bielec, D. 1998 Dancygier, B. 1992 Dancygier, B. 1998

Polish: an essential grammar. London/New York: Routledge. Two metatextual operators: negation and conditionality in English and Polish. BLS 18:61–75.

Conditionals and prediction. Cambridge: Cambridge University Press. Dancygier, B. & E. Sweetser 1997 Then in conditionals. Cognitive Linguistics 8(2): 1–28. Dancygier, B. & E. Sweetser 2000 Constructions with if, since, and because: Causality, epistemic stance, and clause order. In: E. Couper-Kuhlen & B. Kortmann (eds.) Cause, concession, contrast: Cognitive and discourse perspectives, 111–142. Berlin: Mouton de Gruyter. Dancygier, B. & E. Sweetser 2005 Mental Spaces in Grammar: Conditional Constructtions. Cambridge: Cambridge University Press. Dancygier, B & R. Trnavac 2007 Conjunctions, verb forms, and epistemic stance in Polish and Serbian predictive conditionals. In: Divjak, D. and A. Kocha´nska (eds.), Cognitive Paths into the Slavic Domain, 179–216. Berlin: Mouton de Gruyter. Fauconnier, G 1985 Mental Spaces. Cambridge MA: MIT Press.

Causes and consequences: Evidence from Polish, English, and Dutch Fauconnier, G. 1997 Fillenbaum, S. 1986

Fillmore, C. 1986 Fillmore, C. 1990a

117

Mappings in Thought and Language. Cambridge: Cambridge University Press. The use of conditionals in inducements and deterrents. In: E. Traugott, C. Ferguson, J. S. Reilly, A. Meulen (eds.), On Conditionals, 179–195. Cambridge: Cambridge University Press. Varieties of conditional sentences. ESCOL 3 (Eastern States Conference on Linguistics): 163–182.

Epistemic stance and grammatical form in English conditional sentences. CLS 26: 137–162. Fisiak, J., M. Lipi´nska-Grzegorek & T. Zabrocki 1978 An introductory English-Polish contrastive grammar. Warszawa: Pa´nstwowe Wydawnictwo Naukowe. Grochowski, M., S. Karolak & Z. Topoli´nska 1984 Gramatyka wsp´olczesnego j˛ezyka polskiego: Składnia [A grammar of contemporary Polish – Syntax]. Warszawa: Pa´nstwowe Wydawnictwo Naukowe. Haiman, J. (ed.) 1985 Iconicity in Syntax. Amsterdam: John Benjamins. Haiman, J. 1986 Constraints on the form and meaning of the protasis. In: E. Traugott, C. Ferguson, J. S. Reilly, A. Meulen (eds.), On Conditionals, 215–227. Cambridge: Cambridge University Press. Iatridou, S. 1994 On the contribution of conditional “then”. Natural Language Semantics 2: 171–199. Jodłowski, S. 1976 Podstawy polskiej składni [Foundations of Polish syntax] Warszawa: Pa´nstwowe Wydawnictwo Naukowe. Knott, A., T. Sanders & J. Oberlander 2001 Levels of representation in discourse relations. Cognitive Linguistics 12: 197–209. Langacker, R. W. 1990 Subjectification. Cognitive Linguistics 1: 5–38. Pander Maat, H. & L. Degand 2001 Scaling causal relations and connectives in terms of speaker involvement. Cognitive Linguistics 12: 211–245. Pander Maat, H. & T. Sanders 2001 Subjectivity in causal connectives: An empirical study of language in use. Cognitive Linguistics 12: 247–273.

118

Barbara Dancygier

Sanders, T., J. Sanders & E. Sweetser this volume Causality, cognition and communication: a mental space analysis of subjectivity in causal connectives. Sanders, J. & W. Spooren 1997 Perspective, subjectivity, and modality from a cognitive linguistic point of view. In: W. Liebert, G. Redeker & L. Waugh (eds.) Discourse and perspective in cognitive linguistics, 85–112. Amsterdam et al.: John Benjamins. Stukker, N., T. Sanders & A. Verhagen this volume Categories of subjectivity in Dutch causal connectives: a usage-based analysis. Sweetser, E. 1990 From etymology to pragmatics. Cambridge: Cambridge University Press. Sweetser, E. 1996 Mental spaces and the grammar of conditional constructions. In: G. Fauconnier & E. Sweetser (eds.), Spaces, Worlds, and Grammars, 318– 333. Chicago: Chicago University Press. Sweetser, E. 2008 Viewpoint and perspective in language and gesture. Paper presented at Emory University Linguistics Colloquium, April 21, 2008. Tabakowska, E. 1997 Conceptualization: conditionals as an instance of figure/ground alignment. In: Angeliki Athanasiadou, and Ren´e Dirven (eds.), On conditionals again, 273–289. Amsterdam: John Benjamins. Verhagen, A. 2005 Constructions of Intersubjectivity. Discourse, Syntax, and Cognition Oxford: Oxford University Press.

Categories of subjectivity in Dutch causal connectives: a usage-based analysis Ninke Stukker, Ted Sanders and Arie Verhagen 1. Introduction1 The meaning and the use of different types of causality markers has often been described with reference to our conceptual understanding of causality (e.g. Talmy 1988; Verhagen and Kemmer 1997; Wolff and Song 2003). We focus on causal connectives – causality markers functioning at the discourse level of linguistic structure, relating discourse segments into a coherent whole (cf. Hobbs 1979; Mann and Thompson 1988; Sanders, Spooren and Noordman 1993). In this paper, we adopt the cognitive semantic view that causality markers specifically have a categorizing function: when selecting one of the options available in a language, the speaker assigns the causal relations expressed to a specific conceptual type of causality. The meaning and use of causal connectives has recently been characterized in terms of the concept of “subjectivity” (e.g. Pander Maat and Sanders 2000, 2001; Pander Maat and Degand 2001; Pit 2003; Degand and Pander Maat 2003; Stukker 2005; Verhagen 2005; see also the contributions to the present volume). Thus, typical examples of the frequently used Dutch connectives daardoor, daarom and dus marking “forward” causality (cause precedes effect in order of presentation) can be characterized as follows: (1)

Het was extreem koud. Daardoor waren de waterleidingen gesprongen. ‘It was extremely cold. Daardoor the water pipes had burst.’

(2)

Het was extreem koud. Daarom zochten we een caf´e op. ‘It was extremely cold. Daarom we entered a caf´e.’

(3)

Het is onbewolkt. Dus het zal wel koud worden vandaag. ‘The sky is clear. Dus it will probably be cold today.’

Daardoor is typically used for marking objective, “non-volitional” causal processes occurring in observable reality (1). Daarom is typically used in contexts of objective, ‘volitional” causality, in which an volitional action performed in the real world (“entering a caf´e” in [2]), is motivated by a situation presented as the cause (“the extreme cold”). Dus is typically used for marking subjective, 1

This chapter goes back to Stukker (2005) – notably chapter 4.

120

Ninke Stukker, Ted Sanders and Arie Verhagen

epistemic causal relations, in which a causal relation is constructed on the illocutionary level, between a conclusion of the speaker, presented as the causal effect (“it will be cold today” in [3]), and an argument functioning as the causal antecedent (“the clear sky”). The idea that each one of the connectives is related to a specific conceptual model of causality will be referred to in this paper as the “categorization hypothesis”. This hypothesis was tested in several corpus studies (e.g. Pander Maat and Sanders 1995; Pander Maat and Degand 2001; Pit 2003; Stukker 2005). These studies revealed that, indeed, the majority of connectives’ natural usage contexts reflect these conceptual categories of causality more or less directly. In a minority of cases, however, the relation of connectives to their assumedly typical causality category appears to be less straightforward. Under specific circumstances, connectives are used in contexts which are taken to belong to other connectives’ causality categories. Well-known examples are the usage of dus in volitional causal relations as in (4), and the usage of daarom in epistemic causal relations as in (5). (4)

Het was extreem koud. Dus zochten we een caf´e op. ‘It was extremely cold. Dus we entered a caf´e.’

(5)

Het is onbewolkt. Daarom zal het wel koud worden vandaag. ‘The sky is clear. Daarom it will probably be cold today.’

Should these findings be interpreted as evidence against the categorization hypothesis? This is indeed what has been proposed by several of the studies mentioned above. In this paper, we follow a different line of reasoning. We re-interpret the findings from previous connective studies within a usage-based framework.The “usage-based approach to language” assumes that variation is an inherent characteristic of language use, and seeks to explain occurring patterns of variation with reference to more general cognitive mechanisms (cf. Langacker 1987; Bybee 1985, 2006, 2007; contributions to Barlow and Kemmer 2000). In line with this framework, we propose that an interplay of conceptual and usage factors can explain why the usage of Dutch causal connectives does not always conform to abstract definitions that seem to be quite straightforward otherwise. Our hypothesis is that the apparent “counter examples” are actually non-typical, or: peripheral, members (referred to in this study as “non-prototypical usages”: NPU) of the very same conceptual category the connectives refer to in their more typical usage contexts (referred to as “prototypical usages”: PU). In this paper, we focus on one specific factor causing variation in connective use, namely: the idea that language users categorize causal relations not on the basis of “objective reality”, but on the basis of their subjective construal of the situation (cf. Lan-

Categories of subjectivity in Dutch causal connectives: a usage-based analysis

121

gacker 1987; Verhagen 2000), and on one specific factor constraining variation, namely: the prototypicality structure of semantic categories discussed above. The approach proposed here has several advantages: not only can it adequately describe the flexibility connectives show in language use, it can also explain their flexibility as a consequence of their categorization function and as a consequence of more general cognitive mechanisms governing natural language use. In section 2, we discuss previous analyses of Dutch causal connectives, focusing on the usage types considered “problematic” for the categorization analysis. In section 3, we discuss theoretical assumptions underlying the usage-based perspective on language use that are relevant to our analysis of causal connectives in terms of prototypicality structure. Section 4 presents results of an analysis of PU and NPU of daardoor, daarom and dus in a corpus of newspaper texts.

2. Patterns in causal connective use The meaning and use of causal connectives have been described with reference to the cognitively basic concepts of subjectivity and volitionality. The importance of the concept subjectivity in determining linguistic phenomena is widely attested (see for example Traugott 1989, 1995; Langacker 1990; Lyons 1995; cf. discussion in Sanders, Sanders and Sweetser this volume). Notwithstanding terminological differences (Pit 2003; De Smet and Verstraete 2006), theorists agree that subjectivity is to a great extent equivalent to “speaker involvement”. We define “speaker involvement” as: referring to the degree to which the present speaker is involved in the construal of the causal relation.2 A causal relation is “subjective” if for its interpretation reference to the speaker is needed. Conversely, a linguistic element counts as “objective” if speaker involvement is absent in the interpretation of the causal relation. The results of the studies mentioned above suggest that prototypical usage contexts of daardoor, daarom and dus can be related to different categories of “objective” and “subjective causality” defined in terms of “speaker’s roles”: the speaker acts as a “concluder”, as an volitional agent, or is absent. Typical corpus examples of each category are (6) – (8)3 : 2

3

This characterization is based on the definition presented by Pander Maat & Degand (2001: 214). The original definition is: “the degree to which the present speaker is implicitly involved. . . ” Since SI can also be signaled explictly (with modal elements, perspective markers, etc.), we did not include the element “implicit”. Examples discussed are (unless stated otherwise) fragments taken from the Dutch daily newspaper Trouw (year of publication: 2001), part of the corpus analyzed by Stukker (2005).

122

Ninke Stukker, Ted Sanders and Arie Verhagen

(6)

De Boeing 747, het duurste vliegtuig dat rondvliegt, daalt steeds sneller in waarde. De afgelopen jaren is het vermogen van vliegmaatschappijen daardoor met vele miljarden dollars verminderd. ‘The Boeing 747, the most expensive plane in the air, is continuing to diminish in value rapidly. Daardoor airlines’ capital has decreased by billions of dollars over the last few years.4 ’

(7)

(In Denmark and in the Netherlands, carcasses of cows older than 30 months are tested for the cattle plague BSE). Andere landen zijn nog niet klaar om elk voor de slacht aangeboden rund te onderzoeken. Zij vernietigen daarom op grote schaal dieren. ‘Other countries are not yet ready for testing any individual bovine destined for consumption. Daarom they destroy animals on a large scale.’

(8)

(Dutch soldiers who served in Bosnia relate the high incidence of leukemia among them to frequent exposure to impoverished uranium). Maar de huidige hypothese wijt de leukemie¨en aan een virus (. . . ). Het is dus denkbaar dat de soldaten die nu leukemie hebben gekregen, slachtoffer zijn van iets anders dan verarmd uranium. ‘But the current hypothesis attributes the leukemias to a virus. (. . . ) Dus it is conceivable that the soldiers who suffer from leukemia now, are victims of something else than impoverished uranium.’

The differences between the causal relations represented above can be described as follows: Dus is typically used for marking epistemic causal relations5 , in which a causal relation is constructed between a conclusion of the speaker presented as the causal effect (“it is conceivable that. . . ” in [8]), and an argument functioning as the causal antecedent (the fact that the current hypothesis attributes leukemia to a virus). In epistemic causal relations, the speaker functions as the source of the causal relation (it is he who relates argument and conclusion). In other words: reference to the speaker is obligatory in order to interpret the causal relation correctly, hence the causal relation is subjective. 4

5

We focus on the causal relations mainly from a conceptual perspective. Therefore, the English glosses of our Dutch text material do not contain literal translations; neither did we attempt to reproduce specific syntactic characteristics of the Dutch causal connectives (see for a discussion e.g. Evers-Vermeul, 2005). The category of subjective causal relations also contains “speech act” relations, in which the causal relation is constructed between a speech act and a proposition functioning as a justification of that speech act (Sweetser 1990); an example would be What are you doing tonight? Because there is a good movie on). Speech act relations hardly occurred in our corpus; therefore, they will not be taken into account in this paper.

Categories of subjectivity in Dutch causal connectives: a usage-based analysis

123

This type of speaker involvement is lacking in causal relations typically marked with daardoor and daarom. These types of relations can be seen as inherently “objective”.6 Fragments (6) and (7) both describe causal relations between states of affairs in the observable world, having their source outside the speaker. Yet, they differ with respect to the concept of volitionality. The causal relation in (7) describes an volitional act which is motivated by the situation described in the first segment. Volitionality is absent in fragment (6) where one physical process induces another one, without intervention of a human being. In Dutch, volitional causality is typically expressed with daarom; non-volitional causality is typically expressed with daardoor. The distinction between volitional and non-volitional causality is closely related to a distinction exhibited by Dutch causal auxiliary verbs, causality markers operating at the clause level of the linguistic structure (Stukker, Sanders and Verhagen 2008). The causal verbs doen and laten distinghuish between physical, “inanimate” causal processes and causal processes in which animate beings are involved. Just like the distinction objective-subjective, the distinction volitional-non-volitional is considered to be a cognitively important one; it reflects the conceptual model of “Naive Dualism” (Verhagen and Kemmer 1997; see also D’Anrade 1987). “Typicality”, or “prototypicality” is defined in this study in terms of usage frequency. The more frequent a usage-type occurs in natural language use, the more prototypical it is taken to be (see section 3). In the remainder of this paper, the connectives’ prototypical usage contexts are referred to as “PU”. As we already noted in section 1, in a minority of cases, usage-contexts of daardoor, daarom and dus do not conform to the patterns described above. Under specific circumstances, connectives are used in contexts which are taken to belong to other connectives’ causality categories. These non-prototypical usage types (which we will call “NPU”) occur with each of the connectives under investigation, in all of the causality categories discussed. By way of illustration, we discuss the use of dus in volitional causal contexts, assumed to be the prototypical contexts of use for daarom. Consider (9) and (10): 6

Approaches disagree on the question whether volitional causality should count as inherently subjective (see for example Pander Maat & Degand 2001), or as inherently objective (Stukker 2005), In this paper we assume that subjectivity is not an intrinsic characteristic of volitional causality; the fact that some “subject of consciousness” necessarily plays a role opens the way to introduce elements of subjectivity in such contexts. The source of subjectivity is then transferred from the speaker to another “subject of consciousness” (“Perspectivization”: J. Sanders & Spooren 1997; specifically with reference to causal relations, see Sanders, Sanders & Sweetser this volume).

124

Ninke Stukker, Ted Sanders and Arie Verhagen

(9)

(Bystanders rush to help out at the Volendam pub fire.)“Ik woon vlakbij, dus ik ben brandwondencr`eme gaan halen.” “‘I live nearby dus I ran to get burn ointment.”’

(10)

(Letter to the editor discussing the consequences of lowering the age limit for child adoption in the Netherlands.) Het is allemaal heel goed te begrijpen dat de realistische adoptieouders, na de wachttijd van vele jaren, hun kindje zo snel mogelijk willen hebben. Dus kiezen ze voor een kindje uit China of een ander “snel” land. ‘It is only natural that realistic adoptive parents, after having waited for many years, want to have their child as soon as possible. Dus they opt for a child from China or from another “quick” country.’

NPU of dus and daarom exhibit systematic patterning. Pander Maat and Sanders (2000) and Pander Maat and Degand (2001) observed that volitional causal relations marked with dus show a higher degree of subjectivity than those marked with daarom. The SOC of dus-marked contexts referentially coincide with the speaker (see [9]) more often than the daarom-marked contexts do (see for example [7]; Pander Maat and Sanders 2000: 73–74; Pander Maat and Degand 2001: 239–240). In addition, volitional contexts marked with dus or daarom differ in terms of perspective configuration (Pander Maat and Sanders 2000: 71–3; for backward causal connectives, see Pit 2003; see also the chapter by Sanders et al. in the present volume). Volitional causal relations marked with dus more often contain “continuous speaker perspective” than those marked with daarom. Both (9) and (10) contain continuous speaker perspective. In the former, the speaker is continuously referred to by “I”. In the latter fragment, the causal relation is reported from the perspective of “they”, who functions as an embedded speaker, from whose perspective the causal relation is reported. Linguistic signals construing this interpretation are the occurrence of perspective markers in the cause-segment (mental state verb “want” and evaluative element “as soon as possible”, both grammatically attributed to the primary SOC “they”7 ). Since this actor perspective is not explicitly “blocked” in the second segment, we assume that the perspective is continued. In fragment (7), which is representative of volitional causal relations marked with daarom, explicit indications of perspectivization are lacking. In the remainder of this paper, we will argue that NPU of a specific connective are adequately analyzed with reference to the causal category the connective is taken to refer to in its PU. As we will argue, a recurring pattern seems to be that 7

The fact that in this specific fragment the perspective of “they” is embedded under the speaker’s own evaluation does not make a difference for the analysis presented here.

Categories of subjectivity in Dutch causal connectives: a usage-based analysis

125

non-typically used connectives seem to import specific elements of their PU into the foreign context, creating ambiguity in terms of causality category. In order to characterize the contrast between volitional causal relations marked with dus rather than with daarom, for example, reference to dus’ more typical context of use (epistemic causality) is essential, no matter how subtle these differences are, or with what linguistic indications “speaker involvement” is construed. The inverse pattern seems to occur in epistemic causal relations marked with daarom. These contexts systematically contain less subjective elements than those marked with dus (Pander Maat and Sanders 2000; Pander Maat and Degand 2001), which can be interpreted as an indication that daarom’s epistemic NPU are in fact conceptually related to its PU of descriptive (content, objective) volitional causality. Previous studies did not analyze NPU within the objective causal domain (of daardoor and daarom), but we will see below that a similar pattern holds in these cases.

3. A usage-based interpretation of patterns in causal connective use Before we start analyzing the usage of daardoor, daarom and dus in terms of PU and NPU (see section 4), a crucial question is: How is the occurrence of NPU to be accounted for at all? We propose to do so with reference to general mechanisms of language use. The relation between linguistic knowledge and language use has been studied within the “usage-based approach” to language (Langacker 1987, 2000; Bybee 1985, 2006; for an overview see Barlow and Kemmer 2000 and contributions to that volume). The usage-based approach assumes that variation is an inherent characteristic of language use, and it seeks to explain occurring patterns of variation and stability with reference to more general cognitive mechanisms. In this section, we discuss a number of usage mechanisms which, we believe, mediate the relation between the mental representation of the conventional meaning and function of daardoor, daarom and dus, and their occurrence in language use. The first issue we want to address is: Under the assumption that the meaning of daardoor, daarom and dus can adequately be described with reference to clearly delineated, well-defined conceptual categories of causality – how can we explain the variation occurring in contexts of natural language use? We specifically focus on “subjective construal” as a mechanism underlying linguistic categorization (3.1) The second issue is: in view of this variation, how can we maintain the idea that daardoor, daarom and dus have a constant meaning? We focus on prototypicality structure as a mechanism allowing for flexibility while

126

Ninke Stukker, Ted Sanders and Arie Verhagen

maintaining the stability required for language to function as a more or less stable system of conventional symbols (3.2). Our claim will be that these mechanisms in combination explain the patterns of variation and stability observed in daardoor, daarom and dus, starting from the categorization hypothesis.

3.1.

Cognitive factors producing variation in language use

How can we explain that the meaning of connectives sometimes is, and sometimes is not completely congruent with overt signals in the linguistic contexts they are used in? The incongruence observed in Section 2 seems to be of a specific type. The NPU discussed are ambiguous with respect to causality type; they contain characteristics congruent with the causality type the connective is hypothesized to belong to, and at the same time they contain characteristics incongruent with it, but congruent with a different causality type. We propose to analyze NPU as cases of “subjective construal” of the causality category. It means that the same causal relation in reality may be categorized differently by different speakers, according to their particular understanding of the situation or their rhetorical purposes. This proposal is in line with the suggestion put forward in several studies of linguistic categorization that an expression’s meaning is not just an objective characterization of the situation described. Equally important for linguistic semantics is how the speaker chooses to “construe” the situation and portray it for expressive purposes (Langacker [1990] 2002: 315, 1987; Verhagen 2007; and references cited there). Subjective construals of categorization in terms of causality were found with causal verbs, marking causal relations at the clause level (Verhagen 2000; see also Verhagen 1997; Verhagen and Kemmer 1997). Thus, we can analyze the NPU of dus in volitional causal contexts, discussed in section 2 and repeated here, as follows. (11)

(Bystanders rush to help out at the Volendam pub fire.)“Ik woon vlakbij, dus ik ben brandwondencr`eme gaan halen.” “‘I live nearby dus I ran to get burn ointment.”’

(12)

(Letter to the editor discussing the consequences of lowering the age limit for child adoption in the Netherlands.) Het is allemaal heel goed te begrijpen dat de realistische adoptieouders, na de wachttijd van vele jaren, hun kindje zo snel mogelijk willen hebben. Dus kiezen ze voor een kindje uit China of een ander “snel” land. ‘It is only natural that realistic adoptive parents, after having waited for many years, want to have their child as soon as possible. Dus they opt for a child from China or from another “quick” country.’

Categories of subjectivity in Dutch causal connectives: a usage-based analysis

127

Volitional causal relations are inherently objective. They refer to processes which occur in observable reality. This is indicated in the contexts of (11) and (12) by the assertive speech acts, of which the truth is not questioned. At the same time, the related segments of (11) and (12) contain linguistic elements construing subjectivity (speaker reference in [11], perspectivizing devices in [12]; see section 2), leaving the context of use as a whole ambiguous for causality type. The overall interpretation of the causal relation seems to be disambiguated by the speaker’s choice for a specific connective. Marking with dus in fragments (11) and (12) indicates that the reader is expected to construe the causal process as a case of subjective causality, despite its inherently objective character. The subjective construal analysis of causal connectives proposed here is in line with the assumption that the mental representation of a linguistic utterance (in a natural context) is only partly based on the overt signals it contains. This assumption is accepted in various branches of linguistic study. The representation a language user builds from a given utterance is, apart from the linguistic signals, a product of the previous discourse, background knowledge and inferencing (Sanders and Spooren 2001: 3; for more elaborate discussions see Fauconnier 1984; Sperber and Wilson 1995; Langacker 1987, 2000; Verhagen 1997, 2000; Radden et al. 2007). This inherent context dependency and underspecification of language implies that individual usage contexts of a linguistic element may vary to the extent that their characteristics conform to the more abstract semantic representation associated with this element in language user’s long term memory (Verhagen 1997).

3.2.

Cognitive factors retaining stability in language representation

The mechanisms discussed in Section 3.1 explain why variation may occur in the way abstract semantic knowledge (in our case: concerning causal connectives) is mirrored in actual language use. But in view of this variation, how can we maintain the idea that daardoor, daarom and dus have a constant meaning? Or put in more general terms: how is it possible that language users are able to interpret conceptual models invoked by linguistic elements in a consistent way? A common assumption in functionally oriented branches of linguistic theory is that an individual’s linguistic system is fundamentally grounded in “usage events”. This assumption has been elaborated notably within what has become known as the “usage-based approach to language” (Langacker 1987, 2000; Bybee 1985, 2006; Goldberg 2005; for an overview see Barlow and Kemmer 2000 and contributions to that volume). According to this approach, the language system does not consist of rules generating grammatical instances of language

128

Ninke Stukker, Ted Sanders and Arie Verhagen

use, but rather of generalizations over individual usage events, which in turn categorize or license other usage events. In other words, according to the usagebased approach to language, “grammar is the cognitive organization of one’s experience with language” (Bybee 2006: 711). Thus it is assumed that grammar ultimately emerges from individual usage events. This process is often compared to the more general psychological process that the occurrence of any (psychological) event leaves some kind of trace that facilitates their reoccurrence. Frequently encountered patterns become “entrenched” in memory; acquire the status of cognitive routine that is retrieved and applied without requiring conscious attention. Another process, which occurs parallel to entrenchment, is that of abstraction: the emergence of a structure through reinforcement of the commonality inherent in multiple experiences that differ in some other way (Langacker 2000: 4). In this process, peculiarities of individual instances are filtered out from the representation the language user has of the entrenched unit. This process is reflected in language acquisition patterns of children (Barlow and Kemmer 2000: xii, see references cited). In the process of abstraction, the case of “schematization” is of particular importance for our purposes. Langacker (2000: 4) defines a “schema” as the commonality that emerges from distinct structures when one abstracts away from their points of difference by portraying them with lesser precision and specificity. The phenomenon of schematization lies at the basis of language users’ability to recognize a variety of specific instances of use as an instantiation of a specific entrenched structure. This is the cognitive ability of categorization: recognize similarities and differences between phenomena. Or from the perspective of our problem: recognize a specific instance of use as a member of an entrenched schema. A number of factors seem to play a role in this process, among which contextual priming (cf. discussion of the context dependency of language in Section 3.1) and the amount of overlap between the “target” and the potential categorizing structure (Langacker 2000: 15–17; for a more elaborate discussion see e.g. MacWhinney 2000). Categories that emerge from linguistic experience exhibit prototype effects. “Prototypicality” refers to the well-known cognitive phenomenon that conceptual categories are not homogenous; some members of a category are better examples than others. Oranges, apples and bananas are better examples of FRUIT than nuts and olives, which are nevertheless recognized as members of the same conceptual category (Rosch 1973)8 . The “best examples” are the category’s pro8

The participants in Rosch’ experiment were US college students. It is highly likely that prototypicality effects vary with regional or cultural factors (e.g. Lipka 1987; cf. discussion in Ungerer & Schmid 1996: 49–52).

Categories of subjectivity in Dutch causal connectives: a usage-based analysis

129

totypical members; other members vary to the degree to which they have features in common with the prototype (cf. Wittgenstein 1953; Rosch 1973; Rosch and Mervis 1975). The usage-based conception of language predicts that the more frequent, or: the more entrenched, contexts of use are the more prototypical members of a category, while the less frequently encountered usage contexts are more “peripheral” members belonging to the same category, related to the prototype by way of conceptual affinity.9 We started this section with the question: How can we explain that language users maintain a more or less stable representation of the meaning of causal connectives, in view of the variation in usage-contexts encountered? Referring to the usage-based theory of language, we may answer this question as follows. On the basis of the linguistic and non-linguistic contexts of use, language users will be able to recognize the PU and NPU of causal connectives as members of one and the same category. We hypothesize that NPU of causal connectives are motivated by the phenomenon of subjective construal based on the conceptual model of subjectivity. Findings reported in previous studies that seem to favor our hypothesis, are: the fact that the connectives’ PU are relatively frequent, and observations on the connectives’ NPU reported in previous studies (and more particularly: our interpretation of them, see Section 2).

4. Corpus analysis The usage-based perspective discussed in the previous section yields specific predictions concerning the patterns manifested within the variation of connectives’ contexts of use. In this section, we report a corpus analysis in which we investigate to what extent patterns of use found with daardoor, daarom and dus can be accounted for by the categorization hypothesis, under the assumption that the “appearance” of the causality categories in language use is mediated by

9

Empirical evidence in favor of this effect of “token frequency” is found for example in the phenomena of phonetic reduction of high frequency words and phrases (cf. Bybee 2006, 2007), the “conserving effect” – the finding that high-frequency sequences become more entrenched in their morphosyntactic structure (e.g. Bybee 1985), and the “autonomy effect” – the fact that morphologically complex forms of high frequency can lose their internal structure as they become autonomous from etymologically related forms (Bybee 1985 – see discussion of these phenomena in Bybee 2006; see for discussion of other types of frequency effects Hasher 1984; Geeraerts, Grondelaers & Bakema 1994; Verhagen 2000; Bybee 1985, 2007; Goldberg 2005; Schmid 2000).

130

Ninke Stukker, Ted Sanders and Arie Verhagen

prototypicality effects. In Sections 4.1 and 4.2 we present our hypotheses and methods of analysis. Section 4.3 presents the results of our corpus analysis.

4.1.

Hypothesis 1: Frequency of use as an indication for prototypicality of PU

Building on the “frequency of use reflects the cognitive entrenchment” assumption (section 3.2), our first – quantitative – hypothesis is: the connectives daardoor, daarom and dus are significantly more frequently used in usage types conforming to the categorization hypothesis (their PU). Hypothesis 1: Daardoor is prototypically used in content non-volitional relations; daarom is prototypically used in content volitional relations; dus is prototypically used in epistemic relations. Notice that this hypothesis reflects the “semasiological” perspective on word meaning. Given that a lexical items couple word forms to semantic contents, the semasiological question is: “Given linguistic item X, what meaning does it express?” This perspective is complemented by the “onomasiologial” perspective on word meaning. The onomasiological question is: “Given concept Y, what linguistic item(s) can it be expressed with?” (Geeraerts 1997: 17). Although our present research question is primarily of a semasiological nature (PU and NPU are inherently semasiological concepts), both of the levels of analysis are actually relevant for understanding how the meaning and use of causal connectives relate to conceptual structure. In the final analysis, we need to know what the connectives “mean” (denote in terms of conceptual reference – the semasiological perspective) and we need to know how (with what element or elements) an articulate conceptual category can be expressed linguistically (the onomasiological perspective). We will return to this issue in the discussion.

4.1.1. Method and operationalization In line with the usage-based assumptions presented in section 3, a usage context’s degree of prototypicality is operationalized in terms of usage-frequency: the more frequent a specific context is, the more prototypical it is taken to be. The categories of causal relations mentioned in the hypotheses will be determined making use of the “basic operation paraphrase test” of Sanders (1997; cf. Knott and Dale 1994). The paraphrase test makes it possible to determine the

Categories of subjectivity in Dutch causal connectives: a usage-based analysis

131

relational meaning with elimination the connective’s contribution to the overall interpretation of the relation.10 The categories are operationalized as “paraphrases” (see Figure 1), explicating the relational meaning they add to the connected segments. In doing so, the type of causal relation that relates the segments can be established in an objectified and reliable way. The paraphrases are adapted from Pander Maat and Sanders, 1995 (cf. Evers-Vermeul and Stukker 2003). The paraphrases represent the connectives’ hypothesized prototypical usage schemas on two aspects which have proven to be crucial for determining causality category (cf. Section 2): SOC type (implicit speaker SOC, explicit [speaker or actor] SOC, no SOC) and event type in the consequent-segment (conclusion, volitional action, nonvolitional situation). “S1” refers to the segment containing the causal antecedent. The paraphrase itself refers to segment “S2”, containing the causal effect. Relation category Paraphrase Content non-volitional De situatie in S1 leidt tot de volgende situatie: S2 The situation in S1 leads to the following situation: S2 Content volitional De situatie in S1 is een motivatie voor de volgende handeling: S210 The situation in S1 is a motivation for the following action: S2 Epistemic De situatie in S1 is een argument voor de volgende conclusie: S2 The situation in S1 is an argument for the following conclusion: S2 Figure 1. Paraphrases of categories of causal relations marked by daarom, daardoor and dus.

The paraphrase test consists of three steps, which we will illustrate analyzing (13), a PU context of daarom, already discussed as fragment (7) and repeated here: (13)

(In Denmark and in the Netherlands, carcasses of cows older than 30 months are tested for the cattle plague BSE). Andere landen zijn nog niet klaar om elk voor de slacht aangeboden rund te onderzoeken. Zij vernietigen daarom op grote schaal dieren.

10 Our analysis focuses on lexical elements (and sometimes contextual elements). Constructional aspects are not taken into account. For hypotheses concerning constructional differences entailing differences in conceptual interpretation, see e.g. EversVermeul (2005). 10 With S2 understood to be actually performed.

132

Ninke Stukker, Ted Sanders and Arie Verhagen

‘Other countries are not yet ready for testing any individual bovine destined for consumption. Daarom they destroy animals on a large scale.’ Step 1: Determine the text segments that are related by the connective. [Andere landen zijn nog niet klaar om elk voor de slacht aangeboden rund te onderzoeken.]S1 [Zij vernietigen daarom op grote schaal dieren.]S2 [Other countries are not yet ready for testing any individual bovine destined for consumption.]S1 Daarom [they destroy animals on a large scale]S2 Step 2: Remove the connective marking the causal relation. [Andere landen zijn nog niet klaar om elk voor de slacht aangeboden rund te onderzoeken.]S1 [Zij vernietigen op grote schaal dieren.]S2 [Other countries are not yet ready for testing any individual bovine destined for consumption.]S1 [They destroy animals on a large scale]S2 Step 3: Insert the paraphrases and determine which one fits the context under consideration best.12 [Andere landen zijn nog niet klaar om elk voor de slacht aangeboden rund te onderzoeken.]S1 [Other countries are not yet ready for testing any individual bovine destined for consumption.]S1 #13 The situation in S1 leads to the following situation: The situation in S1 is a motivation for the following action: # The situation in S1 is an argument for the following conclusion: [Zij vernietigen op grote schaal dieren.]S2 [They destroy animals on a large scale]S2 Since the connective’s contribution to the interpretation of the relation as a whole is eliminated in this procedure (step 2), “PU-hood” or “NPU-hood” of a given context can be established as follows. If the paraphrase category coincides with the connective’s hypothesized prototypical context of use, the fragment at hand is classified as PU. This is the case with (13), and also with (15) below, which is a typical PU of dus (discussed in section 2 as [8]). In fragment (14), on 12 In order to construct a maximally natural text, tense and aspect of the paraphrase may be adapted to the fragment. If necessary for a adequate interpretation, propositions may be reconstructed (write out referential expressions, integrate relevant implicit information that can be deduced from the context). 13 The # symbol indicates that the suggested relation leads to an incoherent interpretation.

Categories of subjectivity in Dutch causal connectives: a usage-based analysis

133

the other hand (discussed earlier as [9]), the paraphrase fitting bst (of volitional causality) does not coincide with the hypothesized PU of the marking connective dus. Hence, (14) is classified as NPU. (14)

[“Ik woon vlakbij, ]S1 “[‘I live nearby,’] S1 # The situation in S1 leads to the following situation: The situation in S1 is a motivation for the following action: # The situation in S1 is an argument for the following conclusion [Ik ben brandwondencr`eme gaan halen.” ]S2 [‘I ran to get burn ointment.’]S2”

(15)

[Maar de huidige hypothese wijt de leukemie¨en aan een virus.]S1 [‘But the current hypothesis attributes the leukemias to a virus.’]S1 # The situation in S1 leads to the following situation: # The situation in S1 is a motivation for the following action The situation in S1 is an argument for the following conclusion [Het is denkbaar dat de soldaten die nu leukemie hebben gekregen, slachtoffer zijn van iets anders dan verarmd uranium.]S2 [‘It is conceivable that the soldiers who suffer from leukemia now, are victims of something else than impoverished uranium.’]S2

4.1.2. Sample and procedure The frequency of use hypothesis is tested against a corpus of newspaper texts, taken from an electronic version of the Dutch national newspaper Trouw, from the year 2001.14 In order to control for possible genre-effects (cf. Sanders 1997), the sample was built from genres belonging to different text types. For a more detailed description of the sample, see Appendix 1. From each of the sub corpora, 50 occurrences per connective were selected.15 Only forms functioning as markers of causal coherence relations were included,16 and only relations holding between segments of minimally one clause were included in the sample. Occurrences were included in the sample in chronological order. The newspa-

14 Available from “Krantenbank” Factlane, (Lexis Nexis Nederland bv), a service that provides electronic access to the archives of a number of Dutch daily newspapers. 15 In order to minimize risks of bias according to individual authors’ styles and topic, maximally two occurrences per article were included in the corpus. 16 For an overview of other usage-types of daarom and dus, see Evers-Vermeul & Stukker, 2003; Evers-Vermeul, 2005.

134

Ninke Stukker, Ted Sanders and Arie Verhagen

pers were analyzed “issue by issue” (day by day).17 The sample was analyzed quantitatively making use of the statistical method of contrast analysis (see Appendix 2).

4.2.

Hypothesis 2: Conceptual affinity of NPU to the prototype as an indication of category membership

Our second hypothesis concerns the connectives’ NPU, which are expected to occur less frequently (see hypothesis 1). The degree of (non-)typicality of usage contexts is operationalized quantitatively in terms of relative frequency of use (see hypothesis 1). Non-typical contexts can thus be defined as “contexts of use, defined in terms of causality category, occurring significantly less frequently than others”. On the basis of the assumptions discussed in section 3, we expect that NPU are adequately analyzed as non-typical members of the same conceptual category the connective’s PU belong to. Our general expectations are that a) connectives’ NPU can be analyzed as subjective construals of categorization, and b) the overall interpretation of NPU shows conceptual affinity to the resp. connectives’ PU. The following hypotheses, specified per connective, are tested: Hypothesis 2: Given a connective a which is prototypically used in category A, and which can non-prototypically be used in category B (C, D), the nonprototypical usage of a in B (C, D) is characterized by elements of category A more often than when category B (C, D) is marked with connective b (c, d).

4.2.1. Method and operationalization Neither the speaker’s subjective interpretation of a situation nor his rhetorical purposes are directly accessible for analysis. Therefore, subjective construals are reconstructed on the basis of the connectives’ (linguistic or non-linguistic) contexts of use. It can only be hypothesized that non-typical usage types show conceptual affinity to the more typical usage types of a given connective; it cannot be known in advance what kind(s) of connection will occur; given the role of individual creativity it seems to be impossible in principle to predict all possible

17 Therefore, the distribution over text genres mirrors composition of the complete newspaper, rather than that proportional selection of all of the genres was strived at.

Categories of subjectivity in Dutch causal connectives: a usage-based analysis

135

usage types of a given connective. Therefore, our analysis consists of identifying general tendencies in the usage of the connectives under investigation.

4.2.2. Sample and procedure The primary causality category of an NPU was determined on the basis of the linguistic context, making use of the paraphrase test described in Section 4.1; this step is performed when testing hypothesis 1. The analysis testing hypothesis 2, then, consists of looking for demonstrable signs from the linguistic or extralinguistic contexts of use – either within the segments connected in a causal relation, or in the broader context – which motivate the categorization of the specific causal relation as the causality type associated with the connective chosen.

4.3.

Results

Our data largely reflect patterns found in previous corpus studies of daardoor, daarom and dus (Pander Maat and Sanders 1995, 2000; Pander Maat and Degand 2001).

4.3.1. Hypothesis 1: Usage frequency of connectives’ PU Our first hypothesis is corroborated by the connectives in our sample (Z = 4.30; p < .001; for a more detailed discussion of the statistical analysis see Appendix 1). Figure 2 presents an overview of the relation between connectives and causality categories in the corpus from a semasiological perspective (cf. section 4.1). When read vertically, it answers the question: “Given connective a (b, c), what category of causality does it typically express?”. Our paraphrase method revealed that the connectives’ PU contexts may be linguistically realized with different degrees of explicitness with respect to the categories they represent. In this section we discuss the most important of the PU patterns we encountered in our corpus.

Daardoor In accordance with our PU hypothesis, we found that daardoor prototypically is used in contexts of content non-volitional causality (96/100 cases). Half of the cases in our sample consist of physical processes, such as fragment (16) (discussed in section 2 as [8]):

136

Ninke Stukker, Ted Sanders and Arie Verhagen

Figure 2. The relation between connectives and causality types in our sample.17

(16)

De Boeing 747, het duurste vliegtuig dat rondvliegt, daalt steeds sneller in waarde. De afgelopen jaren is het vermogen van vliegmaatschappijen daardoor met vele miljarden dollars verminderd. ‘The Boeing 747, the most expensive plane in the air, is continuing to diminish in value rapidly. Daardoor airlines’ capital has decreased by millions of dollars over the last few years.’

The other half of the non-volitional daardoor contexts concerned causal processes with animate beings as a locus of effect, an example is (17). (17)

Mag je een onderzoek van vorige week geloven, dan speelt in Amsterdam een derde van de kinderen tussen zeven en negen nooit buiten. [. . . ] Daardoor leren ze niet goed om te gaan met andere kinderen en dat is weer slecht voor later, als ze voortdurend nieuwe mensen leren kennen. ‘If you are to believe a study from last week, a third of the children in Amsterdam between the ages of seven and nine never play outside. Daardoor they do not learn to interact with other children: that has a negative effect later on when they continue to meet new people.’

17 Two of dus’ contexts of use and three of daarom’s contexts of use in the sample were categorized as speech act causality. Because of the extremely low frequency, these cases were not further taken into account. For this reason, the sums total of daarom and dus in figure 3 do not add up to 100. The speech act cases were discarded from statistical analysis. We did not discuss the category of speech act causality in the present paper for reasons of space. See for a discussion of speech act causality Sanders et al., this volume).

Categories of subjectivity in Dutch causal connectives: a usage-based analysis

137

Despite animacy of the locus of effect, (17) is not a case of volitional causality; the paraphrase “The situation in S1 is a motivation for the following action: S2” does not adequately reflect the purport of the relation. Since (17) reports on an investigation conducted by other persons than the writer of the fragment, categorization as an epistemic causal relation is not very likely either. This is reflected in the fact that the paraphrase for epistemic causality “The situation in S1 is an argument for the following conclusion: S2” does not fit this context either. The paraphrase for non-volitional causality “The situation in S1 leads to the following situation: S2” is the only one fitting this context adequately. This can be explained by the fact that “they” cannot be analyzed as a subject of consciousness. “Learn to interact with other children” is a mental process that can only take place in an animate being, but occurrence of the process is not dependent on the experiencer’s intentions of doing so. Daardoor may also occur in non-volitional relations containing an action predicate in S2, as in fragment (18). (18)

De schaatser Frans de Ronde omschreef de Jaap Edenbaan als een grote kattebak. “Overal lag zand. Daardoor schaatsten velen met bramen op hun ijzers. [. . . ]”. ‘The skater Frans de Ronde defined the Jaap Eden rink as a big kitty litter bin. “There was sand everywhere. Daardoor many skaters skated with scratches on their blades.”’

Interestingly, in spite of the presence of the inherently volitional action in the consequence segment, the paraphrase of volitional causal relations “The situation in S1 was a motivation for the following action: S2” does not fit very well. The adequate interpretation of this relation is not that the fact that “sand was all over the skating rink” led to the volitional action of “skating of many”, but that the situation depicted in S1 led to the unvolitional situation of “many skaters skating with scratches on their blades”. Note that the deviant instantiations of daardoor’s PU non-volitional causality we discuss here, and other connectives’PU we discuss in this section, are not to be classified as NPU. The reconstructions indicate that classification in accordance with the connective’s PU causality type yields the most plausible interpretation.

Daarom As noted in previous studies (cf. Pander Maat and Sanders 2000; Pander Maat and Degand 2001), daarom is the most “generalist” of the three connectives analyzed in this study. It is the only connective that is conventionally used in all of the categories distinguished. Still, in accordance with our PU hypothesis,

138

Ninke Stukker, Ted Sanders and Arie Verhagen

daarom has a statistically significant preference for volitional causal relations (50/100). The majority of these contexts contain an overtly expressed action in S2. An example of this pattern is (19), discussed in section 2 as (9). (19)

(In Denmark and in the Netherlands, carcasses of cows older than 30 months are tested for the cattle plague BSE). Andere landen zijn nog niet klaar om elk voor de slacht aangeboden rund te onderzoeken. Zij vernietigen daarom op grote schaal dieren. ‘Other countries are not yet ready for testing any individual bovine destined for consumption. Daarom they destroy animals on a large scale.’

However, volitionality of the causal relation need not be explicitly specified with an action predicate. The paraphrase test identified volitional causal relations without overtly expressed actor SOC’s. As an example, consider (20), in which the inherent volitionality of the causal relation is “hidden” in an agentless passive construction. (20)

Toen Napoleon ook Holland in bezit kreeg, was de vaart ineens niet meer nodig. De Noordervaart is daarom nooit verder gegraven dan tot Beringe. ‘Once Napoleon controlled Holland as well, the waterway was no longer necessary. Daarom the Great Northern Canal was never dug any further than Beringe.’

Some contexts in our sample even contained state predicates which appeared to be relevant in the interpretation of the relation as “result of an action”. These cases were analyzed as volitional causal relations. An example is (21). (21)

Microsoft-oprichter Bill Gates waagde zichzelf als eerste aan het spel tijdens een computerbeurs in gokparadijs Las Vegas. “Wij wilden iets bouwen dat een doorbraak voor computerfreaks is”, omschrijft de voormalige topman Microsofts ambities. Het zwart gekleurde apparaat heeft daarom vier aansluitingen voor besturingsknuppels meegekregen, waar de gangbare “gameboys” er maar twee hebben. ‘Microsoft’s founder Bill Gates was the first to try the game during a computer conference in gambler’s paradise Las Vegas. “We wanted to build something that would be a breakthrough for computer freaks” explains Microsoft’s ex- top executive. Daarom the black colored machine has been provided with four connections for game pads, while the current “gameboys” have only two.’

Categories of subjectivity in Dutch causal connectives: a usage-based analysis

139

The perfect heeft meegekregen in (21) (literally: ‘has received’) suggests an act of transfer, including an volitional actor (hence the translation ‘has been provided with’). Thus both in (20) and in (21), the paraphrase for volitional causal relations, “The situation in S1 was a motivation for the following action: S2” is the only one that fits these contexts adequately.

Dus Dus is prototypically used in the category of epistemic causality (76/100). This is in accordance with our PU hypothesis. A typical example of this category is (22), discussed in section 2 as (10). (22)

(Dutch soldiers who served in Bosnia relate the high incidence of leukemia among them to frequent exposure to impoverished uranium). Maar de huidige hypothese wijt de leukemie¨en aan een virus (. . . ). Het is dus denkbaar dat de soldaten die nu leukemie hebben gekregen, slachtoffer zijn van iets anders dan verarmd uranium. ‘But the current hypothesis attributes the leukemias to a virus. (. . . ) Dus it is conceivable that the soldiers who suffer from leukemia now, are victims of something else than impoverished uranium.’

Epistemic causal relations are, in a large majority of cases, easily recognized by way of subjective elements (signs of “self-expression” of a SOC: evaluations, modal elements, etc., cf. Langacker 1990; J. Sanders and Spooren 1997; Pit 2003). But sometimes, the intended epistemic interpretation is linguistically “underspecified”. Yet, the paraphrase of epistemic causality “The situation in S1 is an argument for the following conclusion: S2” is the only one that fits these contexts adequately. An example is (23). (23)

(A “pavese” is a weapon shield) Twee eeuwen lang, tussen 1300 and 1500, zijn paveses overal in Europa in gebruik geweest, dus ook in Nederlandse legers. ‘During two centuries, between 1300 and 1500, paveses were used everywhere in Europe, dus in Dutch armies as well.’

Underspecified epistemic causal relations concern almost without exception instances of “noncausal epistemic relations” in which either the real-world causality has a different direction than the epistemic one (so-called “abductive” causality19 ), or real world causality is not relevant at all, as in the case of (23), 19 An example from our corpus is “Dan moet je eerst uitvinden van welk station je vertrekt. Rouaan ligt ten westen van Parijs. Dus het Noordstation, waar je in de

140

Ninke Stukker, Ted Sanders and Arie Verhagen

an instance of assumption based reasoning (cf. Pander Maat and Degand 2001: 221–224). Noncausal epistemic relations can be identified making use of the paraphrase test. A non-standard linguistic construal of an epistemic causal relation is (24). Judging from form characteristics, the relation between S1 and S2 can be interpreted both as a motivation for the (genuine) question “why shouldn’t Braakhekke sing on stage?” (in which case it should have been categorized as speech act causality, cf. Sweetser 1990; see also Sanders et al. this volume), and as giving an argument for the conclusion that is formulated as a rhetorical question “Braakhekke should sing on stage as well”, which leads to categorization as an epistemic causal relation. The latter interpretation seems to be the more adequate one in this context, where the cook justifies his involvement with theatre. This is corroborated by the fact that the paraphrase for epistemic causality “The situation in S1 is an argument for the following conclusion: S2” adequately reflects the purport of the relation in (24). (24)

(Television chef Braakhekke also appears on stage these days) Is Braakhekke wellicht een aandachtsjunk, wil de interviewer weten. “Neeeee”, kaatst de lange uithaal tot over het Leidseplein. Braakhekke “likt alleen graag aan het theater en zingt altijd onder het koken, dus: waarom niet op het toneel?” ‘The interviewer wonders whether Braakhekke is perhaps an attention junkie?, “Noooo”, resounds his answer across the Leidseplein. Braakhekke “only likes to lick the theatre and sings while he is cooking, dus: Why not on stage?”’

Conclusion from the PU analysis On the basis of the “frequency reflects entrenchment assumption”, then, we interpret the results of our PU analysis as an indication that the usage contexts conforming to the categorization hypothesis are more entrenched in the language user’s semantic knowledge than the usage contexts which do not conform to the categorization hypothesis.

buurt Belgische frieten kunt kopen?” Then first you have to find out from which station you’ll be leaving. Rouen lies west of Paris. “Dus” the North station, in which surroundings you can buy Belgien fries?

Categories of subjectivity in Dutch causal connectives: a usage-based analysis

141

4.3.2. Hypothesis 2: Conceptual affinity of connectives’ NPU to their PU Our second hypothesis concerns the non-typical contexts of use (NPU) of daardoor, daarom and dus. We hypothesized (section 4.2) that the connectives’ NPU reflect characteristics of their PU, to such an extent that the NPU can be analyzed as peripheral members from the same semantic category the connectives’ PU belong to. In this section, we describe NPU patterns we found in our sample. Per pattern we discuss one representative example. We describe in what way the NPU is conceptually related to the connective’s PU, and also how the NPU pattern diverges from the same causality category marked with one of the other connectives.

Daardoor Daardoor is typically used to mark non-volitional causal relations (see section 4.3.1). In a small minority of cases in our sample (4/100), it is used to mark volitional causal relations. An example is (25): (25)

De Chinezen krijgen dit jaar veel meer vrije dagen. De regering hoopt dat de bevolking daardoor meer spaargeld gaat uitgeven om de groei van de economie op peil te houden. ‘The Chinese get more vacation days this year. The government hopes that daardoor the populace will spend more of its savings to keep the economy growing.’

The causal relation in (25) is constructed between the first sentence the complement clause of the second sentence. The fragment is taken from an article which discusses a characteristic of the economic climate in China, namely that the Chinese people tend to save all their money instead of spending it and thus stimulate economic activity. The causal relation in (25) is best explicated with the paraphrase for volitional causality, “The situation in S1 is a motivation for the following action: S2”. The paraphrase for non-volitional causality, “The situation in S1 leads to the following situation: S2” doesn’t seem to reflect the purport of (25) adequately. However, in accordance with our second hypothesis, daardoor-marked volitional causal contexts systematically exhibit conceptual affinity to daardoor’s PU on the aspect of the relation’s “locus of effect”. They differ from daarom and dus marked volitional contexts with respect to degree of intentionality of this locus of effect.

142

Ninke Stukker, Ted Sanders and Arie Verhagen

In the case of (25), the locus of effect “the populace”, is ambiguous with respect to SOC-hood. On the one hand, the predicate “spend” in S2 is inherently intentional. On these grounds, we would have to interpret (25) as containing an actor SOC. On the other hand, “spend” is presented as instantiating systematic patterns. The measure of allotting the people more holidays is prompted by an (assumed) social law: “holidays are spent shopping” or “the more time off, the more shopping is done”. A linguistic indication in favor of this non-volitional interpretation is that the causal relation is strongly presented from the perspective of Chinese government. Although a “normal” intentional causal interpretation of the situation reported is certainly conceivable on a conceptual level (Chinese people consider more holidays a reason for going shopping more frequently), the present wording favors an interpretation from the perspective of the government. The first indication of this is the embedding of the cause-segment in a matrix sentence that conveys a mental state of the government. The next indication is located in the second part of the effect-segment, a goal presented internally in the sentence, namely, the action of spending. It is improbable that “to keep the economy growing” specifies the goal that “the populace” has for “spending”, as the construction of the sentence suggests. Most probably, this element should be understood as a motivation for introducing the measure from the perspective of the government. This is possibly why the volitional causal paraphrase “The situation in S1 is a motivation for the following action: S2” fits this context well, while at the same time the intentions of the actual intentional agents are not very relevant in interpreting the situation. Similar elements indicating “restricted intentionality” can be identified in the other volitional contexts from daardoor, but they are absent in any of the daarom- or dus-marked volitional relations (see table 1; cf. the examples of volitional daarom discussed in section 4.3.1 and the volitional dus example in [31] discussed below). Although the effect-denoting segments of the daardoormarked cases contain an action predicate and an animate locus of effect, these are not to be construed as a genuinely intentionally acting “SOC”s (cf. Section 2). It is highly unlikely that the loci of effect in these cases consciously interpreted the situation presented in the cause-segment as a valid reason for performing the action depicted in the effect-segment; their “intentionality” seems to be of a restricted kind.These observations indicate that daardoor-marked cases show a conceptual relatedness to daardoor’s PU, non-volitional causal relations, where human intentionality does not play a relevant role.20 20 Note that fragment (25), just like the three other volitional contexts of daardoor, differ from daardoor-marked fragments with “non-volitional action predicates”, one of the patterns instantiating daardoor’s PU (cf. [18], discussed in section 4.3.1). In the latter case, ambiguity of the intentional aspect in the relation is solved at

Categories of subjectivity in Dutch causal connectives: a usage-based analysis

143

Table 1. Intentionality in content volitional causal relations Restricted intentionality

Full intentionality

Total

Daardoor

4

0

4

Daarom

0

50

50

Dus

0

22

22

Total

4

72

76

Daarom Daarom occurs in NPU contexts more frequently than daardoor and dus do. It is also the only connective that has NPU in two different causality types, namely non-volitional causality (16/100) and epistemic causality (31/100). We start our discussion with the latter type, an example is (26). (26)

(Selecting tall players for volleyball increases chances of international success) “Nederland heet een lang volk te zijn, maar via de clubs vinden we de lange talenten niet. Ze lopen wel op straat rond; vaak gefrustreerd al vroeg met sport gestopt, omdat hun motoriek tijdens de eerste puberjaren achterloopt bij die van kleinere leeftijdsgenootjes. Daarom moeten wij zelf naar de scholen gaan om ze te vinden en om ze te overtuigen dat ze juist door volleybal meer eigenwaarde kunnen krijgen.” “‘The Dutch are supposed to be a tall people, but we can’t find the tall talents through the clubs. They’re walking around on the streets; often already long frustrated and having given up the sport, because their motor skills are not that of their smaller peers when they’re teenagers. Daarom we have to go to schools ourselves to find them and convince them they’d be greatly appreciated in volleyball.”’

Fragment (26) contains an intentional action predicate in S2, but this predicate is embedded under a modal verb. Therefore, the paraphrase for volitional causality, which is daarom’s PU, “The situation in S1 is a motivation for the following action: S2” does not fit this context adequately. The paraphrase for epistemic causality “The situation in S1 is an argument for the following conclusion: S2” by contrast, does. Epistemic causality is distinguished from other types by its inherent high degree of speaker involvement.The source of the causal

the segmental level (see discussion in 4.3.1), while ambiguity of intentionality in volitional relations with daardoor, such as (25), occurs at the relational level.

144

Ninke Stukker, Ted Sanders and Arie Verhagen

relation is by default the speaker21 , relating an argument to a conclusion in ongoing discourse. This characteristic is lacking in daarom’s PU volitional causality. Our hypothesis therefore predicts that epistemic causal relations marked with daarom are relatively less subjective than the dus-marked ones. We expect that the contexts of use of epistemic daarom relatively more often contain characteristics associated with daarom’s PU: causality. In our data, we found two patterns corroborating this expectation. The first indication concerns modality type.The category of epistemic causality arguably contains two different types of conclusions. The first involves “real” epistemic modality: conclusions regarding the probability of the truth of proposition P. An example of this type is S2 of (27), discussed as a PU context of dus in section 4.3.1 (22), and repeated here as (27). (27)

(Dutch soldiers who served in Bosnia relate the high incidence of leukemia among them to frequent exposure to impoverished uranium). Maar de huidige hypothese wijt de leukemie¨en aan een virus (. . . ). Het is dus denkbaar dat de soldaten die nu leukemie hebben gekregen, slachtoffer zijn van iets anders dan verarmd uranium. ‘But the current hypothesis attributes the leukemias to a virus. (. . . ) Dus it is conceivable that the soldiers who suffer from leukemia now, are victims of something else than impoverished uranium.’

The other type involves so-called deontic modality (conclusions about the desirability of some course of action expressed in P (Pander Maat and Sanders 2000: 74; cf. Sweetser 1990). An example of a causal coherence relation expressing deontic modality in S2 is (26) above. Following Sweetser (1990:49) “deontic” and “epistemic” modality are defined as follows: Modality type

Definition

Deontic modality

real-world obligation, permission or ability

Epistemic modality

necessity, probability or possibility

Figure 3. Distinction between deontic and epistemic modality (Sweetser 1990).

We found an asymmetry in distribution of dus and daarom over modality types (Z = 3.97; p < .001) which is in line with our expectation concerning their differences. Both connectives can express both of the modality types, 21 This may also be an embedded speaker, to whom the actual speaker “lends” perspective (cf. J. Sanders 1994; see also discussion of this type of fragments in Sanders et al. this volume).

Categories of subjectivity in Dutch causal connectives: a usage-based analysis

145

but dus appears to prefer “real epistemic” modality, while daarom appears to be distributed over the two types rather evenly. Deontic modality with dus is exceptional. Table 2 presents an overview. Table 2. Modality type in dus and daarom. Epistemic modality

Deontic modality

Total

Daarom

14

12

Dus

70

6

76

Total

84

18

102

2621

Both in epistemic and in deontic modality, the speaker expresses an attitude towards the proposition, which classifies both types as subjective. However, deontic and epistemic modality differ in terms of degree of subjectivity. Deontic modality (or: agent-oriented modality [Bybee et al. 1994], root modality [J. Sanders 1994]) reports the existence of internal and external conditions on an agent with respect to the completion of the action expressed in the main predicate. As a report, the agent-oriented modality is part of the propositional content of the clause (Bybee et al. 1994: 177). Deontic modality refers to a situation in observable reality, and is not uniquely linked to the speaker as a source of judgment of the situation depicted. Epistemic modality, on the contrary, is typically concerned with the “truth” of a proposition in which the speaker is a primary source of responsibility of the epistemic judgment (cf. Traugott 1989; Sweetser 1990; Bybee et al. 1994). The relatively large proportion of deontic contexts – which count as relatively objective contexts, despite its inherent subjectivity – marked with daarom can be interpreted as an indication that daarom’s epistemic causal NPU exhibit conceptual affinity to daarom’s PU of objective, volitional causality. A second pattern distinguishing epistemic daarom from epistemic dus concerns the “information status” of the segment functioning as the conclusion in the epistemic relation. Whereas in dus-marked epistemic relations the conclusion is always performed in the actual discourse situation, in epistemic contexts marked with daarom, the conclusion may concern “given information”. An example of an epistemic relation with a conclusion performed in actual discourse is (27) above. An example of an epistemic relation with a “given” conclusion is (28): 21 Five of the epistemic daarom-fragments in the corpus could not be classified as epistemic or deontic modality. This is why the sum total in this analysis amounts to only 26.

146

(28)

Ninke Stukker, Ted Sanders and Arie Verhagen

“Maar bij ernstige brandwonden is het hele lichaam ziek”, zegt Hermans. “De lever, de nieren, alle organen doen mee. Daarom is de zorg voor deze pati¨enten zo ingewikkeld.” “‘Serious burns make the whole body ill”, says Hermans. “The liver, the kidneys, all organs are affected. Daarom the care for these patients is so complicated.”’

Fragment (28) is taken from a “background” article which discusses the question whether the capacity of care for burn injuries in the Netherlands is to be extended or not. The interviewee, Professor Hermans, argues against this proposition. The cause-part of the relation in (28) must be interpreted as an argument for the conclusion in the effect-part, stating that care for burn victims is complicated; the paraphrase for epistemic causality “The situation in S1 is an argument for the following conclusion: S2” conveys this interpretation best. However, this is not the only aspect communicated in this fragment. Judging from the referential expression zo (roughly translated as ‘this’) preceding the evaluative expression ingewikkeld (‘complicated’), this evaluation is not new. As a matter of fact, Hermans is quoted twice in the article. Fragment (28) represents the second quote. Fragment (29) precedes the fragment cited in (28) and contains the first quote: (29)

“Het gaat niet om het personeel. Voor deze zorg zijn zeer gespecialiseerde verpleegkundigen, internisten, anesthesisten en chirurgen nodig. Als daar meer van komen, hebben ze ieder te weinig werk om routine op te bouwen. En dat kan riskant zijn.” “‘The amount of personnel is not relevant. For this type of patient care, highly specialized nurses, internists, anaesthesists, and surgeons are needed. If their number increases, individually, they won’t be able to practice their skills to a sufficient extent. That may be dangerous.”’

The proposition that “the care for these patients is (. . . ) complicated” is conceptually present immediately after this first quote, as an inference from Hermans’ statement that “for this kind of patient care, highly specialized . . . are needed”.Yet, the inferred proposition is stated explictly in the second quote, and the proposition immediately preceding it functions as a justification for stating this proposition. Thus, the conclusion segment of the causal relation in (28) is ambiguous for information status. The information conveyed in the conclusion part itself is not entirely “new”, but its status as a conclusion, and the justification

Categories of subjectivity in Dutch causal connectives: a usage-based analysis

147

for this conclusion are.23 Table 3 summarizes the distribution of “performativity of the conclusion” over epistemic causal relations marked with dus or daarom24 Table 3. Distribution of “performativity of the conclusion” over epistemic causal relations marked with dus or daarom. Conclusion performed

Conclusion given

Total

Daarom

24

7

31

Dus

76

0

76

Total

100

7

107

23 This ambiguity of information status of the relation as a whole renders “nonperformative conclusion relations” somewhat ambiguous for causality type as well, namely between epistemic and non-volitional causality. Since the paraphrase test explictly addresses the “informational surplus” of a causal relation (cf. Sanders, Spooren & Noordman 1993) we chose to categorize this type as epistemic. Whether categorization of “non-performative conclusion relations” as epistemic causal relations is actually the most adequate analysis remains a matter of debate (see also Sanders & Spooren 2007, who claim that the inference of “new” knowledge is a distinguishing feature of epistemic relations). Categorization of these cases as content non-volitional would not alter the overall picture of the analysis of distribution of daardoor, daarom and dus presented in this paper (Z=4.28; p Background in this case” (Asher and Lascarides 1998: 107). In a similar vein, Levinson (2000: 122), when discussing conjunction, notes that, “when events are conjoined, they tend to be read as temporally successive, and if at all plausible, as causally linked”. He does not present this as an on-line processing preference, but these arguments might give rise to the Causality-by-Default hypothesis: Because readers aim at building the most informative representation, they start out assuming that the relation between two consecutive sentences is a causal relation (given certain characteristics of two discourse segments). Subsequently, causally related information will be processed faster, because the reader will only arrive at an additive relation if no causal relation can be established.

216

Ted Sanders and Wilbert Spooren

There is a second way out of the Paradox of Causal Complexity: although causal relations may conceptually be more complex, discourse processing is often sped up by content- or structure-based schematic expectations on the basis of the text read so far. Such an explanation focuses on the top-down processes involved in the comprehension process: Reading about a problematic situation elicits expectations about text continuation: a Solution to the Problem just stated. In the case of a List context, fewer or weaker expectations will be generated. Problem-Solution structures are widely recognized examples of conventional structures that trigger such expectations, in both text linguistics (Hoey 1983) and text processing (Meyer 1985). Text processing is largely influenced by readers’ schematic knowledge. Knowledge of text structure is considered one type of schematic knowledge (Van Dijk and Kintsch 1983). The structure of a text under consideration triggers expectations regarding the text passages yet to come. Textual information that fits in with such expectations will be processed faster. The crucial difference between the hypotheses is in the role of causality: only under the Causality-by-Default hypothesis, the causal nature of the relation per se influences processing. If causality-by-default is true, the processing difference between causal and additive relations remains, irrespective of the amount of expectations. By contrast, under the Schematic Expectation hypothesis, processing is determined by expectations, irrespective of the causal nature of the source of the expectations. Recently, the two theoretical explanations were compared in an experimental set-up in which the comprehension of causal relations with and without strong schematic expectations were investigated (Mulder and Sanders 2005; Mulder 2008). Mulder (2008) finds evidence in favor of a top-down schematic expectations account: Only when readers have strong expectations on the causal coherence relation that connects a previous segment (like a Problem) to an upcoming one (like a Solution), they process the new information faster. In conclusion, there is substantial evidence from processing studies demonstrating the psychological relevance of causality as a conceptual category at the discourse level. Causal relations result in better memory representations than additive relations. It remains to be seen what the precise explanation is for the relatively fast on-line processing of causal relations. But how about the different types of causality, defined in terms of distinctions between domains or differences in subjectivity? There is not too much experimental work on these distinctions. In text production, Den Ouden (2004) found a difference between objective (semantic or content) and subjective (pragmatic, i.e., epistemic-speech act) relations: Subjective connections were preceded by longer pauses and were produced with a higher mean pitch range, indicating

Causal categories in discourse – Converging evidence from language use

217

that these sentences were more discontinuous than objective connections. Similar findings have been reported for differences in the prosodic patterns English because-clauses that either expressed objective (content) or subjective (epistemic) causal relations (Couper-K¨uhlen 1996). Studying on-line text processing, Traxler, Bybee and Pickering (1997) focused on the difference between objective (10) and subjective (11) causal relations (they talk about causal and diagnostic sentences). (10) Heidi felt very proud and happy, because she won first prize at the art show. (11) Heidi could imagine and create things, because she won first prize at the art show. In (10), the fact in S1 is caused by the fact in S2. In (11), S1 is a claim that is supported by the argument in S2. In an eye-tracking experiment, a processing difference appeared between the two types of relation: The total reading times are longer on “first prize” in the subjective version (11). Traxler, Sanford, Aked and Moxey (1997) replicated this finding in a self-paced reading paradigm: Objective sentences are processed quicker than subjective ones. Noordman and De Blijzer (2000) arrive at similar conclusions: “epistemic relations require more processing time than content relations [. . . ] Understanding a sentence that describes a relation between states of affairs in the world is easier than understanding a sentence that expresses a reasoning about that relation” (Noordman and De Blijzer 2000: 54). Now do these findings indeed show that subjective relations are more complex than objective relations?As in the case of schematic expectations for causality, the question is whether this effect is actually due to the inherent complexity of the relation, or to its fit in the discourse structure. Subjectivity of relations has been shown to correlate with its surrounding context, such as text type. For instance, persuasive texts show more subjective relations than informative texts do (Sanders 1997). Therefore, the ultimate test of the relative complexity of objective versus subjective relations should take place in fitting natural contexts, such as an epistemic relation in an argumentative genre and a content relation in an informative genre6. Another type of evidence for the cognitive relevance of subjective versus objective categories was recently reported by Kamalski, Lentz, Sanders and Zwaan (2008); Kamalski (2007). They investigated the processing and representation of causal connectives and cue phrases in persuasive texts. Fragments including 6

These issues are the topic of Anneloes Canestrelli’s current PhD project at Utrecht University.

218

Ted Sanders and Wilbert Spooren

subjective markers like Therefore in (12) and objective markers like As a result in (13) were compared to their implicit equivalents – that is to identical fragments, but without the causal connective and cue phrase. Text fragments (12) and (13) are taken from larger persuasive texts on genetic manipulation. (12) (. . . ) Therefore, there is only one clear answer to genetic manipulation: stop eating genetically manipulated food. Biological foods are safer for the environment. . . (13) Extra genes are introduced. As a result, the organism changes. The experimental texts were manipulated at several places, so that four versions could be compared: Objective implicit, Objective explicit, Subjective implicit and Subjective explicit. Readers were asked whether they considered the author convincing. It turned out that the objective explicit text versions were considered most convincing, whereas the Subjective explicit were considered less convincing. When asked to judge the author’s intentions, participants appeared to consider subjective explicit versions most “persuasive”. Perhaps even more interestingly, on-line processing results show that subjective and objective markers result in different processing. Reading times indicated that readers slowed down at the end of the sentence in the Subjective explicit condition, but not in the Objective explicit condition. Kamalski et al. suggest that subjective marking in persuasive text leads to a so-called forewarning effect: readers resist the author’s message because this author is very clearly present (“on stage”), making an argument in favor of his own claims. Even though the exact set-up of the experiment leaves some questions unresolved (Mulder 2008), the results illustrate the cognitive relevance of the categories marked by objective versus subjective causal markers. The distinction between objective and subjective causal markers affects the on-line interpretation and the effectiveness of a persuasive text. Apparently, connectives and cue phrases – sometimes considered “only” small, subtle linguistic elements – have a substantial influence on on-line processes of text interpretation and representations. In our view, these results are an important extension of the study of coherence phenomena to several discourse genres. An important issue for further research is of course how these findings relate to the results on the relative complexity of subjective versus objective relations reported above. In this section we have reviewed evidence showing that categories like Causality and Subjectiviy play a role in the processing of natural language: Causal relations are processed differently than Additive relations and Causal Subjective relations are processed differently than Causal Objective relations.

Causal categories in discourse – Converging evidence from language use

219

And this is exactly what one would expect in a cognitive interpretation of CCR. Hence, by abductive reasoning, these experimental results provide an argument for a psychological interpretation of our categorization of coherence relations.

5. Causal categories in language acquisition We expect the causal categories to be relevant for the way in which children acquire connectives and relations. More precisely, we expect the different conceptual categories to show a different pattern in acquisition. We will also consider whether the data predict differences in cognitive complexity: Do they allow for conclusions on the relative order of acquisition? Most children build their first multi-clause discourse before the age of three; instead of uttering one clause at a time, they start producing combined clauses (cf. Clark 2002). At first, the coherence relations between these clauses remain implicit. Later, children learn how to use connectives to make coherence relations explicit. The overview in Spooren and Sanders (2008) shows that both naturalistic and experimental studies of first language acquisition show that additive relations are acquired before causals, and that additive connectives are acquired before causal connectives, which appear around age two (Clark 2002). More precisely, Bloom, Lahey, Hood, Lifter, and Fliess (1980, reprinted in a slightly revised version in Bloom 1991) found that English children follow the same route in acquiring coherence relations: (14) additive < temporal < causal < adversative Studying the transcripts of spontaneous speech data of twelve Dutchspeaking children in the CHILDES database (MacWhinney 2000), EversVermeul (2005) and Evers-Vermeul and Sanders (2009) show that the first causal connective (as in their example 15b) does not appear before an additive connective (as in 15a) has occurred. (15a) Kim is een meisje. En jij bent een jongetje he? ‘Kim is a girl. And you are a boy, aren’t you?’

(Thomas, 2;8.23)7

(15b) Ik heb (een) beetje griep. Want ik ben laat gaan slapen. ‘I’ve got a touch of flu. Because I went to bed late.’

(Abel, 3;3.8)

All Dutch children use the additive en before they come up with the causal connective want. Hence, in terms of first emergence, empirical studies of En7

The notation “2;8.23” stands for age in years;months;days.

220

Ted Sanders and Wilbert Spooren

glish and Dutch corroborate the idea that causal relations are more complex than additive relations. Learning the conceptual relation of causality as well as acquiring the linguistic tools to express it, only seems to be possible on the basis of an understanding of simple additive relations: children have to understand that events can be related, that they can be ordered in time and, subsequently, that the first event, which precedes the second, can be the cause of the second. But how about the different types of causality? There are only few acquisition studies that distinguish between different types of causal relations and connectives. Interestingly, the data from these studies contradict each other. In a corpus of naturalistic data, Kyratzis et al. (1990) found that speech act causal relations are frequent even at a very early age, whereas epistemic causal relations are acquired very late (they hardly occur, even in the oldest age group of their 6;7 – 12;0 year olds). For Dutch 6–11 year olds, Spooren and Sanders (2008) report how younger children use more objective content relations than older children do. Looking at much younger Dutch children, Evers-Vermeul (2005) found that epistemic connectives were clearly acquired last, but she found no difference between the emergence of connectives expressing content versus speech act relations. In sum, there is no clear picture in the area of Subjectivity. Studying the acquisition of French parce que in the childes database, Zufferey (2007) replicates these findings. She finds that children acquire the epistemic use later than content and speech act uses. An interesting additional finding is that epistemic verbs (penser, croire) are also acquired later than speech act verbs (dire, demander). Although Kyratzis et al. (1990) present the most sophisticated picture, their findings are based on data collected in a context that is heavily biased toward noncontent (more specifically, speech act) relations. Several studies have shown that the type of relations children produce depends on the communicative setting. For instance, in a conversation task, children use more subjective (epistemic, speech act) relations, whereas in a descriptive task, objective relations prevail (Spooren and Sanders 2008). This implies that researchers should pay attention to this interaction, and check whether their corpora contain a wide variety of contexts. There is one additional crucial issue for further research. Up till now, we explored cumulative cognitive complexity as a major factor in determining the developmental connective sequences. However, frequency of use in the parental input can be taken as an alternative determinant of the acquisition process. Since the recent introduction of the usage-based approach to child language acquisition, interest in parental input has been renewed (cf. Diessel 2004; Tomasello 2003). For example, Diessel (2004) found that English connectives that frequently appeared in the mother’s data emerged earlier in children’s talk, and connectives that were relatively infrequent in the mother’s data emerged

Causal categories in discourse – Converging evidence from language use

221

relatively late. An analysis of parental input is lacking in the existing connective studies, although Evers-Vermeul (2005; Evers-Vermeul and Sanders 2001) did a first attempt to check whether frequency of use by parents determined the acquisition order. For a small subset of the data, parents’ utterances in the CHILDES corpus were analyzed. Input-based predictions were based purely on these frequencies. The results showed how input-based hypotheses do not necessarily differ from complexity hypotheses. Hence, input- and inherent complexity theories may be dependent: Parents show “audience design” in that they monitor their own language, check whether their children understand what they are saying, and consequently adapt their own language use. There is only one way to decide between theories of cognitive complexity and parental input: to predict input-based acquisition orders and compare it to the actual orders of emergence the children show (Van Veen, Evers-Vermeul, Sanders and Van den Bergh 2009).

6. Variation across media and genres So far we have looked at evidence for the categorization of both relations and connectives. The data we have used come from on- and off-line processing experiments, and from studies of language acquisition. We now return to the investigation of causal connectives. In this section we report on corpus investigations of the categorization of connectives. Unlike most available corpus studies, we do not focus on written text, but rather study different media and text genres. We expect an influence from medium type on the type of relations to be found. The written language that is usually studied often comes from newspapers. These are highly edited and relatively detached texts, which have been produced in a communicative situation characterized by a large distance between authors and readers. Authors can (re)consider and revise their lexical choices and formulations. The focus is usually on content and not on interpersonal issues. It is for instance no coincidence that speech act relations are extremely rare in this type of written discourse. Contrast this with spontaneously produced spoken language (Chafe 1994; Clark 1996). In this case, language use is heavily situated, in that conversational partners share the communicative situation and background. Also, interpersonal issues are often topical. Furthermore, as a rule, speakers do not edit or revise their utterances to a degree comparable to written genres. As a consequence, it can be expected that spontaneous conversations display a higher overall degree of subjectivity. An interesting third communicative situation is that of chat language. Chatters do not communicate as directly as interlocutors in a spontaneous conver-

222

Ted Sanders and Wilbert Spooren

sation do. They also produce their utterances more autonomously than conversational partners do: In most chat systems, utterances are produced completely and then sent to the server to be displayed to the fellow chatter. As for editing, the production circumstances allow for revising one’s contribution, although time pressure caused by the fellow chatters is undoubtedly much higher than for the average journalist writing a newspaper article. In this section we report an analysis of the use of omdat and want in spoken language. We want to find out whether the difference between the two connectives, as found in written language, also occurs in spontaneously produced discourse, be it spoken (spontaneous conversation) or written (chat). We focus our analysis on a number of properties that in the recent past have been described as characteristic for the distributional differences between want and omdat, namely the type of coherence relation and the degree of subjectivity that the connectives express (Pander Maat and Degand 2001; Pander Maat and Sanders 2001; Pit 2003).

6.1.

Omdat and want in spontanous conversations8

6.1.1. Materials From the Corpus of Spoken Dutch, a 10 million words corpus of completely digitalized material, annotated in several ways (Oostdijk 2000), 149 fragments with want and 124 fragments with omdat were selected more or less randomly from the conversations and interviews in the corpus. For the sake of interrater agreement each fragment was analyzed independently by two coders. In the case of disagreement, the difference was discussed.

6.1.2. Analysis We determined the amount of subjectivity in the corpus examples by analyzing a number of properties of the discourse context (i.e. the segments surrounding the connectives) that provide information on the subjectivity of the relation. Our selection was based on suggestions in the literature.9

8 9

This section is based on Spooren, Sanders, Huiskes and Degand (to appear). The ordering in terms of degree of subjectivity expressed in this section is based on Pit (2003).

Causal categories in discourse – Converging evidence from language use

223

Modality of the first segment (S1) Each segment was coded as expressing either a judgment or another modality (fact, general knowledge, an intentional act, individual knowledge, a perception, an experience). A segment expresses a judgment if it gives both a conceptualizer – the person responsible for the causal relation; compare the notion Subject of Consciousness in chapter 2 – and that what is judged. The segment expresses a state and uses a so-called scalar predicate (a predicate that can be modified with degree expressions, such as very much X ; more than X ), which can be paraphrased with “I believe that. . . ”. Fragment (16) gives an example. (16) Speaker A [S1 en ik vind ‘t niet meer leuk op die manier te werken ook dat nog ‘ns een keer.] ‘and I don’t like it any longer to work that way also.’ Speaker B nee ja. ‘no yes.’ Speaker A h`e? ‘right?’ Speaker A omdat [S2 er geen uh geen wisselwerking is.] ‘because there is eh no interaction.’ Judgments were considered more subjective than other modalities.

Relation type The causal relation expressed in each fragment was analyzed in terms of domains (Sweetser 1990): content (in which the speaker describes a causal relation in the world), epistemic (in which the speaker infers a conclusion on the basis of an argument) and speech act relations (in which the speaker motivates a speech act). Furthermore, within the content relations we distinguish between volitional and non-volitional relations (see Stukker, Sanders and Verhagen this volume): Does the relation involve an intentional act or not? Examples are: (17) Non-volitional content Speaker A [S1 waarom is ie dan al o alle warm water al op dan?] ‘why is he then all o all of the hot water already gone?’ Speaker B omdat [S2 ik de kraan aan heb laten staan vanochtend.] ‘because I left the tap running this morning.’

224

Ted Sanders and Wilbert Spooren

(18) Volitional content Speaker A maar [S1 ze kunnen ook rustig uh toch naar een gewone brugklas gaan] omdat [S2 ze dus alleen maar uitval hebben voor nou noem maar wat uh Nederlands of alleen maar uitval hebben voor nou het lezen of of of rekenen of zo terwijl de rest heel goed functioneert.] ‘but they can also easily uh still go to a regular first year because they have only a cancellation of classes for let us say reading or or or arithmetic and the like whereas the rest functions very well.’ (19) Epistemic Speaker A [S1 en en wa ik het nu heb dat is geen noodsituatie] want [S2 ik kan donders goed inschatten kwart over vier half vijf dat er dan geen student meer boven gaat kijken.] ‘and and wha I have it now that is not an emergency because I can estimate very well quarter past four half past four that no student will go and look upstairs then anymore.’ (20) Speech act Speaker A [S1 en uh a als iemand mij belt ja dan ben ik er niet] want [S2 ik ben bezig met dit werk en dat moet vandaag af punt]. ‘and uh i if someone calls me yes then I am not in because I am busy working at this and it has to be finished today period.’ The causal relations can be ordered from least subjective to most subjective, as follows: Non-volitional content < Volitional content < Epistemic / Speech act Type of conceptualizer in the first segment (S1) The conceptualizer is the person responsible for the causal relation that is constructed. In spoken dialogues there can be either no conceptualizer (as in example [17]), or the conceptualizer is a third person (example [18]), a second person (example [21]) or a first person (examples [19] and [20]). (21) Second person conceptualizer Speaker A [S1 waarom dacht je dat ze in een tent zaten dan?] ‘then why did you think they were in a tent?’ Speaker B ja xxx. ‘yes xxx.’ Speaker B omdat [S2 ze altijd in een tent gaan.] ‘because they always go in a tent.’

Causal categories in discourse – Converging evidence from language use

225

These options can be ordered in degree of subjectivity, as follows: No Conceptualizer < Third person < Second person < First person In the analysis presented below we only use the distinction between third person conceptualizers and first person conceptualizers. Cases without a conceptualizer were deemed irrelevant as these express facts. Second person conceptualizers were very rare in the corpus. Linguistic realization of the conceptualizer The final property we will report on is the linguistic realization of the conceptualizer. We have followed Langacker’s (1990) suggestion that an explicit reference to the conceptualizer objectifies the conceptualizer. Consequently, implicit reference to the conceptualizer is considered more subjective than explicit reference: Explicit reference to the conceptualizer < Implicit reference to the conceptualizer

6.1.3. Results The general hypothesis was that want-fragments show a higher degree of subjectivity than omdat-fragments. This hypothesis will be discussed for each of the four properties that we have looked at. Because of the low cell frequencies we have aggregated groups in a theoretically motivated way. Modality Our subhypothesis with respect to modality is that want-fragments occur relatively more often with judgments in S1 than omdat-fragments. This hypothesis is borne out (χ2 [1] = 4.78, p < .05). Table 2 summarizes the data. Table 2. Number of judgments vs. other modalities as a function of connective (want, omdat) in the spoken corpus (for seven omdat-fragments and one want-fragment the data are missing). judgments

other modalities

omdat

49 (41.9 %)

68 (58.1 %)

want

82 (55.4 %)

66 (44.6 %)

226

Ted Sanders and Wilbert Spooren

Type of coherence relation With respect to coherence relation we expected to find epistemic and speech act relations to occur typically with want and non-volitional cause and volitional reason with omdat. In other words, we expected to find more content relations with omdat than with want. This hypothesis is borne out, as can be seen in Table 3 (χ2 [2] = 64. 41, p < .01). Table 3. Number of content relations vs. epistemic/speech act relations as a function of connective (want, omdat) in the spoken corpus.10 Content

Epistemic

Speech act

omdat

106 (90.6 %)

8 (6.8 %)

3 (2.6 %)

want

61 (42.7 %)

56 (39.2 %)

26 (18.2 %)

Type of conceptualizer With respect to type of conceptualizer we expected to find more third person conceptualizers with omdat and more first and second person conceptualizers with want. This hypothesis has to be rejected on the basis of the data (χ2 [1] = .52, p = .47), as can be seen in Table 4. Table 4. Number of facts/3rd person conceptualizer vs. 2nd/1st person conceptualizer as a function of connective (want, omdat) in the spoken corpus. fact/3rd person

2nd /1st person

omdat

25 (21.4 %)

92 (78.6 %)

want

37 (25.2 %)

110 (74.8 %)

Coding of the conceptualizer We expected the omdat-fragments to have more explicit codings of the conceptualizer than the want-fragments. In this analysis we excluded the fragments with a fact, because there is no conceptualizer in these cases. Table 5 gives a summary of the data, which show that the hypothesis is supported (χ2 [1] = 5.07, p < .05). 10 In seven omdat-fragments and six want-fragments, the relation was a so-called textual relationship, in which the second segment gives a paraphrase or a metacomment. These have been discarded.

Causal categories in discourse – Converging evidence from language use

227

Table 5. Number of explict vs. implicit concepualizers as a function of connective (want, omdat) in the spoken corpus (NB. facts were excluded from the analysis). explicit

implicit

omdat

80 (74.1 %)

28 (25.9 %)

want

84 (60.4 %)

55 (39.6 %)

Intermediate conclusion There is substantial evidence that want differs significantly from omdat in our spoken corpus. Compared to omdat, want occurs more often with judgments, with non-content relations and with implicit conceptualizers. The one exception is the identity of the conceptualizer, on which characteristic want and omdat do not show any difference. It has been noted (Pit 2003) that omdat is often used to express explicit acts of speaking and thinking (“I assume this because. . . ”, “I say this because . . . ”).This use of omdat might account for the high frequency of first person pronouns with omdat. If that is correct then the frequency of first person pronouns was not a good operationalization of high degree of subjectivity. We could even argue that these constructions show a low degree of subjectivity: The speaker describes his or her speaking or thinking and thereby creates a distance (see also Langacker 1990). Our analysis shows a relationship between type of coherence relation (content-epistemic-speech act) and amount of subjectivity expressed by a connective. This correlation resonates with the analysis presented in chapter 2 where content relations were shown to be systematically different from non-content relations in terms of the presence and role of a subject of consciousness.

6.2.

Omdat and want in chat

6.2.1. Materials For the study of connectives in chat language, more work needed to be done than for the spoken data. Unfortunately, there is not much chat data available for linguistic study. Chat language is notoriously ephemeral and a systematic collection of chat data is difficult because of privacy requirements. Recently we have started collaborating at the VU with a grammar school in Amsterdam. We had children of grade 1 and 3 (age 13 and 15, respectively) come to our computer rooms and chat with each other about a number of predetermined topics. Some of these were assumed to create involvement (for example, “what is your favorite t.v. show?”), whereas others were considered to generate less involvement (for

228

Ted Sanders and Wilbert Spooren

example, “what do you think about the Classical Music Awards?”). We used a chat facility in a virtual learning environment, in order to be able to archive the chat sessions. Today we have a small corpus, with 120 speakers. It is a corpus that we just started to build, so it is not nearly as big as we want it to become (ca. 90,000 tokens, over 11,000 types to date). We analyzed the fragments with want and omdat in the same manner as in the case of the spoken data. For the analysis we have selected 43 examples with want and 41 examples with omdat. The use of omdat in our chat corpus is relatively scarce. In order to reach a total of 41 occurrences of omdat we supplemented our dataset with 11 occurrences of omdat from Internet Relay Chat conversations in the CONDIV corpus (Grondelaers, Deygers, Van Aken, Van den Heede and Speelman 2000). The latter were the first tokens of omdat in the IRC-CONDIV corpus. The following fragment is an example of the way omdat is used in the chat corpus. The children are talking about the relative slowness of the system used to record the chat conversations (compared to MSN). (22) 1 Pupil A: [S1 waarom doen ze dit niet via msn????/ ] ‘why don’t they do this via msn????/’ 2 Pupil A: daar duurt verzenden neit een half uur ‘there transmission doesn’t take half an hour’ 3 Pupil B: weet ik niet onm ‘I don’t know xxx [uninterpretable]’ 4 Pupil C: en nu? ‘what now?’ 5 Pupil A: cool hoe doe je prive ‘cool how do you do private [messaging]’ 6 Pupil C: weet k veel, omdat [S2 ze dom zijn] ‘I don’t know, because they are stupid’ The omdat sequence is the question by Pupil A in segment 1 and the answer to the question in segment 6 by Pupil C (note that, typically for chat, this question answering is delayed, due to system properties, and therefore interrupted by a number of other, less related contributions; pupil C also contributed utterance 4, which is not related to Pupil A’s question). The modality of the first segment is an action by an unspecified 3rd person (ze ‘they’). The modality of the second segment is a judgment from the speaker. The conceptualizer in the first segment is made explicit (ze), whereas the conceptualizer of the second segment (the speaker) remains implicit. The relation between the segments is volitional content (the second segment gives the reason, albeit a peculiar one) for the action in the first segment.

Causal categories in discourse – Converging evidence from language use

229

The next fragment gives an example of the way want is used in the chat corpus. Due to system requirements the pupils were identified by numbers. A substantial amount of the conversations is about the identification of the pupils. (23) is an example. (23) Pupil A:

wie is vijf [S1 neit belle] want [S2 dan ga ik spac] ‘who is five. It better not be Belle, because in that case I’m gonna space’

The modality of the first segment is a speech act of warning by the speaker, pupil A, who remains implicit as a conceptualizer. The modality of the second segment is individual knowledge (it can be paraphrased as “I know that in that case I’m gonna space”) of the speaker, who is explicitly referred to. The relation between the two segments is a speech act relation: the second segment gives the motivation for uttering the warning in the first segment.

6.2.2. Results Modality Our subhypothesis with respect to modality is that want-fragments occur relatively more often with judgments in S1 than omdat-fragments. Although want occurred more often with judgments than omdat, this did not turn out to be significant (χ2 [1] = 2.10, p = .15). Table 6 summarizes the data.

Table 6. Number of judgments vs. other modalities as a function of connective (want, omdat) in the chat corpus. judgments

other modalities

omdat

11 (26.8 %)

30 (73.2 %)

want

18 (41.9 %)

25 (58.9 %)

Compared to the spoken corpus there are relatively few judgments in chat. It is unclear what the cause is of this difference. It may be the setting. Some of the topics in the chat conversations did not generate much discussion, and consequently the conversations digressed into everyday casual conversation like “are you going to the party tonight?”. It should be noted however that both in spoken discourse and in chat the difference in percentages between omdat and want is comparable: in spoken language 13.5 %, in chat 15.1 %. This suggests that the analysis of more data may reveal a significant difference.

230

Ted Sanders and Wilbert Spooren

Type of coherence relation With respect to coherence relation we expected to find epistemic and speech act relations to occur typically with want and non-volitional cause and volitional reason with omdat. In other words, we expected to find more content relations with omdat than with want. This hypothesis is borne out, as can be seen in Table 7 (χ2 [2] = 26.33, p < .01). Table 7. Number of content relations vs. epistemic and speech act relations as a function of connective (want, omdat) in the chat corpus.

omdat want

Content 35 (89.7 %) 14 (34.1 %)

Epistemic 2 (5.1 %) 8 (19.5%)

Speech act 2 (5.1 %) 19 (46.3 %)

Note. In four cases of omdat the type of relation was missing.

The main difference with the spoken data is that want in chat is used much more often to express speech act relations. This may well be typical for the conversational style used by children in our chat sessions (example [24]). (24) 1 Pupil A: [S1 geen praatjes he kleine man] ‘no big mouth ay little man’ 2 Pupil A: want [S2 anders zet ik je in der prullenbakk] ‘because otherwise I will put you in the wastepaperbasket’ At the level of content versus non-content relation categories, the distinction between omdat and want is comparable in spoken language and chat. Type of conceptualizer With respect to type of conceptualizer we expected to find more third person conceptualizers with omdat and more first and second person conceptualizers with want. This hypothesis has to be rejected on the basis of the data (χ2 [1] = 1.05, p = .31), as can be seen in Table 8. Table 8. Number of facts/3rd person conceptualizer vs. 2nd/1st person conceptualizer as a function of connective (want, omdat) in the chat corpus.

omdat want

fact/3rd person 8 (19.5 %) 4 (9.3 %)

2nd /1st person 33 (80.5 %) 39 (90.7 %)

Causal categories in discourse – Converging evidence from language use

231

The main difference with our spoken data is that want in chat is used even more subjectively. Coding of the conceptualizer We expected the omdat-fragments to have more explicit codings of the conceptualizer than the want-fragments. In this analysis we excluded the fragments with a fact, because there is no conceptualizer in these cases. Table 9 gives a summary of the data, which show that the hypothesis is not supported (χ2 [1] =.02, p =.90). Table 9. Number of explict vs. implicit concepualizers as a function of connective (want, omdat) in the chat corpus (NB. facts were excluded from the analysis). explicit

implicit

omdat

19 (59.4 %)

13 (40.6 %)

want

22 (57.9 %)

16 (42.1 %)

Compared to our spoken data the main difference seems to be that omdat in chat is used much more implicitly. This may again be a reflection of the fact that the chat setting is overall more subjective. Intermediate conclusion To a certain extent we see similar differences between want and omdat in spoken language and in chat. In both media omdat occurs much more often with content relations than want, and both in spoken language and in chat the first segment in a want construction tends to express a judgment, even though the difference between want and omdat fails to reach significance in the chat corpus, probably due to the limited number of analyzed cases. Nevertheless, it seems obvious that the medium influences the use of omdat and want. The overall number of judgments in the first segment of causal constructions is much lower in chat than in spoken language. The amount of want constructions that express speech act relations is much higher in chat than in spoken language. The number of first person conceptualizers is higher in chat than in spoken language. Also the number of omdat constructions in which the conceptualizer of the first segment remains implicit is substantially higher than in spoken language. Most of these differences seem to point in the direction that chat communication is more subjective than spoken language (the only exception is the relatively low number of judgments in chat).

232

Ted Sanders and Wilbert Spooren

To a certain extent this is surprising, given that chat communication is more detached than spoken communication because the immediate situational context (which includes the conversational partner) is less available in chat communication. In terms of the Basic Communicative Space Network presented in chapter 2, a communicative situation of direct speaker-hearer interaction can be expected to be more subjectively grounded than chat communication, because of the direct availability of Speaker and Addressee. So the question is: why is our spoken corpus less subjective? A possible explanation is the type of tasks that the speakers in the corpus had to carry out. For example in the CGN corpus the subcorpus of spontaneous conversations that we have analyzed includes among others interviews with teachers, which have a certain amount of prestructuring, and may consequently be less spontaneous than unstructured communication. It also includes conversations that give the impression of being forced: apparently contributors to the corpus were asked to record conversations on a DAT recorder for a particular amount of time, which leads them sometimes to talk about the amount of time left to talk. This reduces the degree of spontaneity in the conversations. Another possible explanation are the characteristics of the participants in the chat conversations: They were relatively young, compared to the speakers in the CGN corpus, and knew eachother relatively well. All in all our results can be seen as a plea for researching comparable communicative settings.

7. Conclusion People communicate through discourse. Understanding the coherence relations between the utterances in a discourse is a crucial aspect of this communication. When language users (learn to) understand discourse, they connect discourse segments by inferring coherence relations, such as Cause-Consequence and Conclusion. In our cognitive approach to coherence relations (CCR) we have argued that they do so on the basis of a very limited set of cognitive principles (Sanders et al. 1992, 1993 and elsewhere). In this chapter we focused on one type, or rather “family” of coherence relations: causal relations. We have looked at causality as such, but also at subcategories of causality, as defined in terms of domains (content-epistemicspeech act) and subjectivity (objective-subjective). We have shown that causal categories are fundamental to human cognition and natural language at the discourse level. Following the methodology of converging evidence, we have presented empirical studies of language use, with respect to both causal relations and causal connectives. The results lead us to conclude that Causality and

Causal categories in discourse – Converging evidence from language use

233

Subjectivity are two of those cognitive principles that organize our knowledge of coherence relations. Notions like Causality and Subjectivity can help us explain the system and use of causal relations and their linguistic expressions in everyday language use, in language acquisition and in discourse processing. Many recent and ongoing corpus studies of causal connectives have argued that the systematic use of a particular lexical item to express a certain type of causal relationship implies that people distinguish between several types of causality. In fact, several chapters in this book focus on the system behind meaning and use of these causal connectives in several languages (see the chapters by Sanders et al.; J. Sanders; Dancygier and Stukker et al.). These and other recent studies provide a wealth of data, suggesting just how promising the Subjectivity account is. Needless to say that differences in Subjectivity fit neatly into a CCRacount of discourse coherence. Ongoing further work, focusing on systematic cross-linguistic comparison, will provide the ultimate answers.11 In the current chapter, we have considered this organization of the lexicon of causal connectives as one window on conceptual categorization. Most new data were presented in the last section of the paper and concerned connective use in spoken discourse and in chat language. We argued that it is important to investigate whether insights from existing work on written corpora can be generalized to spoken discourse, especially because claims concerning cognitive reality are at stake. After all, our most natural and spontaneous way to communicate is not merely in discourse, but in spoken discourse. Now that spoken corpora have become available in many languages, it is possible to test hypotheses against spoken corpus data. A second, more specific reason to be interested in spoken data is directly related to the interpretation of causal connectives as acts of categorization: How “basic” is this act? Are the distinctions only realized by highly proficient language users in a production context with many editing opportunities? Or are the same differences realized in the totally different production context of spontaneous conversations, characterized by limiting time constraints and few planning and editing options? Our study of Dutch spontaneous spoken and chat discourse clearly answers these questions for the two causal connectives we have studied: There are substantial differences in the way omdat and want are used in both spontaneous conversations and chat communication, which may be due to the fact that our corpus of chat language displays an extremely high degree of subjectivity. To a certain extent this characteristic of chat diminishes differences between the two connectives. Nevertheless, the overall difference between want and omdat (omdat typically signals a content relation, want typically signals a 11 This is the topic of Ninke Stukker’s postdoc project at Utrecht University.

234

Ted Sanders and Wilbert Spooren

non-content relation) is found across the board. This replicates earlier results reported by Pander Maat and Degand (2001) and Pit (2003) on the distribution of these connectives in written language. These results generate a great number of interesting research questions, for instance about the comparability of different communicative settings. One issue is to what extent our results can be compared with data from written language studies. An interesting case in point is our finding that in chat we find on average 85 % use of 1st person conceptualizers and in spoken language 76 %. Pit (2003: 184) reports in an analysis of written language 37 % 1st person conceptualizers. Obviously, the spoken and chat corpora are much more comparable than the written corpus used by Pit. Despite this difference, in all three types of language use there are clear differences between the use of omdat and want. Both Causality and Subjectivity can be identified as important principles behind spontaneous connective use. Furthermore, it seems worthwhile to feed theories of connectives and coherence relations with corpus studies of spontaneous language use in communicative situations that allow for direct interaction; for one thing, we have never before seen so many attested speech act relations as in our chat corpus. Systematic comparison of various communicative situations is imperative. Another important methodological question is how we can move away from the existing dominance towards explicitly marked relations. In this chapter we have underlined the importance of studying implicit relations, that is, coherence relations that are not signalled by connectives and cue phrases (see the chapter by Fauconnier). And even though we have assembled important data from categorization and processing studies, all existing corpus studies focus on explicit relations. In the earlier sections of the chapter, we concluded that Causality and Subjectivity can indeed be identified as basic categories underlying coherence relations, for instance in categorization tasks like sorting cards. In processing studies, causal relations appear to have a special status: Causally related information is represented better in memory than non-causally related information. Furthermore, causally related information is processed differently than non-causally related information. Similarly, processing studies show different processing patterns for Objective versus Subjective markers and relations, indicating the relevance of causal subcategories as well. Studies of language acquisition also show how the different conceptual categories have different patterns in acquisition. We have shown that the relevance of our categories leads to a clear prediction regarding the order of acquisition of coherence relations and connectives: relations and connectives that are cognively complex will be acquired later than less complex relations and connectives. This is the so-called cognitive complexity

Causal categories in discourse – Converging evidence from language use

235

hypothesis.A straightforward causal complexity hypothesis – causal connectives do not appear in first language acquisition before additive connectives – was indeed corroborated for English and Dutch.As for the causal subcategories, we can conclude that epistemic relations are acquired later than content and speech act relations. However, the cognitive complexity hypothesis should be confronted with a different type of explanation for the order of acquisition, namely the role of parental input. Further studies will have to show to what extent parental input provides a more sufficient account than a CCR-account does.12 Another remaining crucial intriguing question concerns cognitive complexity in processing. New processing research will have to show whether the fast processing of causally related information should be explained in terms of inherent complexity or rather in terms of non-linguistic expectations about the content of the subsequent text. Similarly, the processing of subjective versus objective causal relations, embedded in natural context, should provide ultimate conclusions on the relative complexity of the different types of causal relations. All in all, the existing results and the remaining questions guarantee a stimulating research program. Triangulation is at the heart of our enterprise. We use the strategy of converging evidence to zoom in on the cognitive properties of causal connectives and the coherence relations they encode. We believe that the use of evidence from different research strategies (in-depth linguistic analysis, corpus studies of various types of discourse, online and offline measurements of the reading process and the resulting representation) provides a set of windows on causal categorization in discourse, resulting in a panoramic overview on causality in the human mind. Therefore, (cognitive) linguistics can benefit from a research strategy of converging evidence.

Appendix: List of coherence relations Causal, semantic, basic order, positive Cause-consequence (i)D (i)E

Doordat er een lage drukgebied ligt hoven Ierland, wordt het slecht weer. Because there is a low-pressure area over Ireland, the bad weather is coming our way.

12 This is the topic of Rosie van Veen’s PhD project at Utrecht University.

236

Ted Sanders and Wilbert Spooren

Causal, semantic, basic order, negative Contrastive cause-consequence (ii)D

(ii)E

(Een probleem voor de Artificiele Intelligentie is de soms indrukwekkende effici¨entie van het menselijk geheugen.) Hoewel het aantal overeenkomsten tussen gezichten enorm) is, hebben wij er niet de minste moeite mee zeer grote aantallen mensen van elkaar te onderscheiden. (A problem for Artificial Intelligence is the sometimes impressive efficiency of the human memory.) Although the number of similarities between faces is enormous, we do not have the slightest difficulty in distinguishing a very large number of people.

Causal, semantic, nonbasic order, positive Consequence-cause (iii)D (iii)E

Een pianoconcert van Beethoven werd van het programma genomen omdat de solist Anthony di Bonaventura ernstig ziek werd. A piano concerto by Beethoven was removed from the program because the soloist Anthony di Bonaventura fell seriously ill.

Causal, semantic, nonbasic order, negative Contrastive consequence-cause (iv)D

(iv)E

(Met de vrijlating van Hetzel eindigde een van de meest sensationele processen die West-Duitsland sinds jaren heeft beleefd.) Hans Hetzel werd in 1969 tot levenslange dwangarbeid veroordeeld wegens moord, hoewel hij bij boog en bij laag had volgehouden onschuldig te zijn. (The release of Hans Hetzel ended one of the most sensational trials West Gennany has seen in years.) In 1969 Hans Helzel was sentenced to life-long hard labor because of murder, although he had stoutly maintained his innocence.

Causal categories in discourse – Converging evidence from language use

237

Causal, pragmatic, basic order, positive Argument-claim (v)D (v)E

Door nesten of dode vogels kan een schoorsteen verstopt raken. Laat uw schoorsteen dus ieder jaar nakijken en zonodig vegen. Nests or dead birds may clog up chimneys. Therefore, have your chimney checked once a year and swept when necessary.

Instrument-goal (vi)D (vi)E

We geven ook de Portugese benamingen van de voornaamste bezienswaardigheden om het vragen te vergemakkelijken. We will also present the Portuguese names for the most important places of interest to make the questioning easier.

Condition-consequence (vii)D (Film-en fotomateriaal kunt u verkrijgen bij de souvenirwinkels in de wandelsafari en bij het safarirestaurant.) Klaar? Dan gaan we nu op safari. (vii)E (Film and photomaterials can be obtained from the souvenir shops on the walking safari and from the safari restaurant.) Ready? Then we’re now off on safari.

Causal, pragmatic, basic order, negative Contrastive argument-claim (viii)D (De hoeveelheid berichten over ongelukken in de dagbladen zegt maar weinig over de belangrijkste doodsoorzaken.) Al schreven de kranten vorig jaar diverse keren over enkele gasongevallen, als gasgebruiker loop je heel wat minder risico dan als verkeersdeelnemer. (viii)E (The number of reports about accidents in the newspapers does not say much about the most important causes of death). Although the papers wrote about gas accidents several times last year, the risk run by the gas user is much smaller than that of the traffic participant.

238

Ted Sanders and Wilbert Spooren

Causal, pragmatic, non basic order, positive Claim-argument (ix)D

(xi)E

(Veel mensen schijnen alleen de top van het struikje van de broccoli te eten.) Dat is zonde, want de stronk smaakt ook goed. (Many people seem to eat only the flower head of the broccoli.) That is a pity, because the stalk tastes good too.

Goal-instrument (x)D

(x)E

(Als u van plan bent een huis te gaan kopen, en u wilt gebruik maken van een subsidieregeling, dan heeft de Postbank iets voor u.) De koper zal in het algemeen direct zijn woonlasten met de toegezegde subsidie willen verminderen. Daartoe biedt de Postbank aan deze subsidie voor te financieren. (If you intend to buy a house and you want to make use of a subsidy arrangement. the Postbank has something for you.) In general the buyer will want to diminish his costs of living with the promised subsidy. To that end the Postbank offers to finance this subsidy in advance.

Consequence-condition (xi)D

(xi)E

(Het boek wordt besloten met ten wijze raad van een Middeleeuwse vinoloog:) Wijn is een zeer gezonde drank, die de levensduur van de mens niet onaanzienlijk kan verlengen, mits de wijn in geringe hoeveelheden en met niet al te grote regelmaat wordt genoten. (The book ends with a wise piece of advice by a medieval judge of wine:) Wine is a very healthy beverage that can lengthen man’s life not insignificantly, provided that the wine is drunk in small quantities and not too regularly.

Causal categories in discourse – Converging evidence from language use

239

Causal, pragmatic, non basic order, negative Contrastive claim-argument (xii)D U moet er wel rekening mee houden dat er langs de hele Joegoslavische kust haaien voorkomen, al wordt dat bepaald niet van de daken geschreeuwd. (xii)E You will have to take into account that there are sharks along the whole Yugoslavian coast, although this is certainly not shouted from the rooftops.

Additive, semantic, positive List (xiii)D (Uit her onderzoek blijkt dat diepvriezers en koelkasten de afgelopen jaren steeds zuiniger zijn geworden:) Het energieverbruik van een koelkast is 17% minder geworden en een diepvriezer verbruikt 18-20% minder stroom dan tien jaar geleden. (xiii)E (It appears from the investigation that deep freezers and refrigerators have become more economical in recent years.) The energy consumption of a refrigerator has decreased by 17%, and a deep freezer uses 18 to 20% less electricity than 10 years ago.

Additive, semantic, negative Exception (xiv)D Een diersoort kan een zekere bejaging verdragen, maar de Californische condor kan dat niet. (xiv)E A species can stand a certain amount of hunting, but the California condor cannot. Opposition (xv)D (Niet alle Nederlandse bedrijven deden het gisteren even goed op de Amsterdamse effectenbeurs:) Bergoss verbeterde twaalf punten, evenals Van Haltum, Holee en SmitTak. Philips verloor daarentegen tien punten.

240

(xv)E

Ted Sanders and Wilbert Spooren

(Not all Dutch companies did equally well at the Amsterdam stock exchange yesterday.) Bergoss improved by 12 points, as did Van Haltum, Holec and Smit-Tak. By contrast, Philips lost 10 points.

Additive, pragmatic, positive Enumeration (xvi)D Reageerbuisbaby’s doen vragen rijzen over ethische en maatschappelijke aspecten. Bovendien, wat te denken van de juridische problemen die ze oproepen? (xvi)E Test-tube babies raise questions concerning ethical and social aspects. Moreover, what about the legal problems they evoke?

Additive, pragmatic, negative Concession (xvii)D (De Consumentenbond raadt het drinken van bronwater af.) De consumptie van bronwater is in Nederland de laatste jaren sterk gepropageerd, maar bij een onderzoek in Duitsland naar de samenstelling van flessenwater zijn minder gunstige ervaringen opgedaan. (xvii)E (The consumer’s association advises against the drinking of mineral water.) The consumption of mineral water has been advocated strongly over the last few years in the Netherlands, but the results of an investigation in Germany on the composition of bottled water were not so good.

References Asher, N. & A. Lascarides 1998 Bridging. Journal of Semantics 15: 83–113. Bestgen, Y., L. Degand, & W. Spooren 2006 Towards automatic determination of the semantics of connectives in large newspaper corpora. Discourse Processes 41(2): 175–194.

Causal categories in discourse – Converging evidence from language use Bloom, L. 1991

241

Language Development from Two to Three. Cambridge: Cambridge University Press. Bloom, L., M. Lahey, L. Hood, K. Lifter & K. Fiess 1980 Complex sentences: acquisition of syntatctic connectives and the semantic relations they encode. Journal of Children Language 7, 235– 261. Chafe, W. 1994 Discourse, Consciousness and Time. Chicago: Chicago University Press. Clark, E. V. 2003 First Language Acquisition. Cambridge: Cambridge University Press. Clark, H. H. 1996 Using Language. Cambridge: Cambridge University Press. Clark,H. H. , & E. V. Clark 1977 Psychology and Language:An Introduction to Psycholinguistics. New York et al.: Harcourt, Brace, Jovanovich. Couper-K¨uhlen, E. 1996 Intonation and clause-combining in discourse: The case of because. Pragmatics 6: 389–426. den Ouden, J.N. 2004 Prosodic realizations of text structure. Dissertation, Tilburg University. Diessel, H. 2004 The Acquisition of Complex Sentences. Cambridge: Cambridge University Press. Evers-Vermeul, J. 2005 Connections between form and function of Dutch connectives: Change and acquisition as windows on form-function relations. Dissertation UiL OTS, Utrecht University. Evers-Vermeul, J. & T. Sanders 2009 The emergence of Dutch connectives; how cumulative cognitive complexity explains the order of acquisition. Journal of Child Language, 36/4, 829–854. Garnham, A. & J. Oakhill (eds.) 1992 Discourse Representation and Text Processing. A special issue of Language and Cognitive Processes. Hove, UK: Lawrence Erlbaum Associates. Gernsbacher, M. A. & T. Giv´on (eds.) 1995 Coherence in Spontaneous Text. Amsterdam et al.: Benjamins. Graesser, A. C., M. Singer & T. Trabasso 1994 Constructing inferences during narrative comprehension. Psychological Review 101: 371–395.

242

Ted Sanders and Wilbert Spooren

Graesser, A. C., P. Wiemer-Hastings, & K. Wiemer-Hastings 2001 Constructing inferences and relations during text comprehension. In: T. Sanders, J. Schilperoord & W. Spooren (eds.), Text Representation: Linguistic and Psycholinguistic Aspects, 249–271. Amsterdam: Benjamins. Grondelaers, S., K. Deygers, H. Van Aken, V. Van Den Heede, & D. Speelman 2000 Het CONDIV-corpus geschreven Nederlands [The CONDIV corpus of written Dutch]. Nederlandse Taalkunde 5(4): 356–363. Halliday, M. A. K. & R. Hasan 1976 Cohesion in English. London: Longman. Hobbs, J. 1979 Coherence and coreference. Cognitive Science 3, 69–90. Hoey, M. P. 1983 On the Surface of Discourse. London: Allen & Unwin. Hovy, E. H. 1990 Parsimonious and profligate approaches to the question of discourse structure relations. Proceedings of the 5th International Workshop on Natural Language Generation. Kamalski, J. 2007 Coherence marking, comprehension and persuasion. On the processing and representation of discourse. Dissertation, Utrecht University. Kamalski, J., L. Lentz, T. Sanders, & R. Zwaan 2008 The fore-warning effect of coherence markers in persuasion: off-line and on-line evidence. Discourse Processes 45(6): 545–579. Keenan, J. M., S. D. Baillet & P. Brown 1984 The effects of causal cohesion on comprehension and memory. Journal of Verbal Learning and Verbal Behavior 23: 115–126. Knott, A. & R. Dale 1994 Using linguistic phenomena to motivate a set of coherence relations. Discourse Processes 18: 35–62. Kyratzis, A., J. Guo & S. Ervin-Tripp 1990 Pragmatic conventions influencing children’s use of causal constructions in natural discourse. In: K. Hall, J.P. Koenig, M. Meacham, S. Reinman & L.A. Sutton (eds.), Proceedings of the 16th annual meeting of the Berkeley Linguistics Society, 205–214. Berkeley, CA: BLS. Langacker, R. W. 1990 Subjectification. Cognitive Linguistics 1 : 5–38. Levinson, S. C. 2000 Presumptive Meanings. The Theory of Generalized Conversational Implicature. Cambridge, MA: MIT Press. Louwerse, M. M. 2001 An analytic and cognitive parameterization of coherence relations. Cognitive Linguistics 12 (3): 291–315.

Causal categories in discourse – Converging evidence from language use

243

MacWhinney, B. 2000 The CHILDES project, 3rd edition. Mahwah, NJ: Lawrence Erlbaum. Mann, W. C. & S. A. Thompson 1988 Rhetorical structure theory: towards a functional theory of text organization. Text 8, 243–281. Meyer, B. 1985 Prose analysis: Purposes, procedures and problems. In: B.K. Britton & J.B. Black (eds.), Understanding Expository Text, 11–62. Hillsdale, NJ: Erlbaum. Mulder, G. 2008 Understanding Causal Coherence Relations. Dissertation, Utrecht University. Mulder, G. & T. Sanders 2005 Causality by Default or Top Down Expectations? On the processing of causal relations in expository text. Annual Meeting of the Society for Text and Discourse, VU Amsterdam, July. Noordman, L. 2005 Taal en kennis [Language and Knowledge]. Afscheidsrede Universiteit van Tilburg, Tilburg. Noordman, L. & F. De Blijzer 2000 On the processing of causal relations. In: E. Couper-Kuhlen & B. Kortmann (eds.), Cause, Condition, Concession and Contrast: Cognitive and Discourse Perspectives, 35–56. Berlin: Mouton de Gruyter. Noordman, L. & W. Vonk 1997 The different functions of a conjunction in constructing a representation of the discourse. In: M. Fayol & J. Costermans (eds.), Processing Interclausal Relationships in Production and Comprehension of Text, 75–93. Hillsdale, NJ: Erlbaum. Pander Maat, H.L.W. 1998 Classifying negative coherence relations on the basis of linguistic evidence. Journal of Pragmatics 30: 177–204. Oostdijk, N. 2000 The Spoken Dutch Corpus Project. The ELRA Newsletter 5 (2): 4–8. Pander Maat, H. & L. Degand 2001 Scaling causal relations and connectives in terms of speaker involvement. Cognitive Linguistics 12(3): 211–245. Pander Maat, H. & T. Sanders 2000 Domains of use or subjectivity? The distribution of three Dutch causal connectives explained. In: E. Couper-Kuhlen & B. Kortmann (eds.), Cause, Condition, Concession and Contrast: Cognitive and Discourse Perspectives, 57–81. Berlin: Mouton de Gruyter.

244

Ted Sanders and Wilbert Spooren

Pander Maat, H. & T. Sanders 2001 Subjectivity in causal connectives: An empirical study of language in use. Cognitive Linguistics 12(3): 247–273. Pit, M. 2003 How to express yourself with a causal connective: Subjectivity and causal connectives in Dutch, German and French. Dissertation, Utrecht University. Amsterdam: Rodopi. Redeker, G. 1990 Ideational and pragmatic markers of discourse structure. Journal of Pragmatics 14: 305–319. Sanders, T. 1997 Semantic and pragmatic sources of coherence: On the categorization of coherence relations in context. Discourse Processes 24: 119–147. Sanders, T. 2005 Coherence, causality and cognitive complexity in discourse. In: M. Aurnague, M. Bras, A. Le Draoulec & L. Vieu (eds.), Proceedings/Actes SEM-05, First International Symposium on the exploration and modelling of meaning, 105–114. Toulouse: Universit´e de Toulouse le Mirail. Sanders, T. & L. Noordman 2000 The role of coherence relations and their linguistic markers in text processing. Discourse Processes 29: 37–60. Sanders, T. & H. Pander Maat 2006 Cohesion and Coherence: Linguistic approaches. In: Brown. K. et al. (eds.), Encyclopedia of Language and Linguistics, 2nd edition. London: Elsevier. Sanders, T. & W. Spooren 2001 Text representation as an interface between language and its users. In: Sanders, T., Schilperoord, J. & Spooren, W. (eds.), Text Representation: Linguistic and Psycholinguistic Aspects, 1–26. Amsterdam et al.: Benjamins. Sanders, T., W. Spooren, & L. Noordman 1992 Toward a taxonomy of coherence relations. Discourse Processes 15 (1): 1–35. Sanders, T., W. Spooren, & L. Noordman 1993 Coherence relations in a cognitive theory of discourse representation. Cognitive Linguistics 4(2): 93–133. Spooren, W. 1997 The processing of underspecified coherence relations. Discourse Processes 24: 149–168. Spooren, W. & T. Sanders 2008 The acquisition order of coherence relations: on cognitive complexity in discourse. Journal of Pragmatics 40, 2003–2026.

Causal categories in discourse – Converging evidence from language use

245

Spooren, W., T. Sanders, M. Huiskes, & L. Degand in press Subjectivity and causality: A corpus study of spoken language. In: J. Newman & S. Rice (eds), Empirical and Experimental Methods in Cognitive/Functional Research. Stanford et al.: CSLI/University of Chicago Press. Stukker, N. M. 2005 Causality Marking across Levels of Language Structure: A Cognitive Semantic Analysis of Causal Verbs and Causal Connectives in Dutch. Dissertation, Utrecht Institute of Linguistics, Utrecht University. Sweetser, E. 1990 From Etymology to Pragmatics. Cambridge: Cambridge University Press. Tomasello, M. 2003 Constructing a language: A usage-based theory of language acquisition. Harvard: Harvard University Press. Trabasso, T. & P. Van den Broek 1985 Causal thinking and the representation of narrative events. Journal of Memory and Language 24: 612–630. Traxler, M. J., M. D. Bybee, & M. J. Pickering 1997 Influence of connectives on language comprehension: Eye tracking evidence for incremental interpretation. The Quarterly Journal of Experimental Psychology 50A: 481–497. Traxler, M. J., A. Sanford, J. Aked, & L. Moxey 1997 Processing causal and diagnostic statements in discourse. Journal of Experimental Psychology: Learning, Memory, & Cognition 23: 88– 101. Van den Broek, P. W. 1990 Causal inferences in the comprehension of narrative texts. In A. C. Graesser and G. H. Bower (Eds.), Psychology of learning and motivation: Inferences and Text Comprehension (Vol. 25, pp. 175–196). New York, NY: Academic Press. Veen, R. van, J. Evers-Vermeul, T. Sanders, & H. Van den Bergh 2009 Parental input and connective acquisition: a growth curve analysis. First Language 29, 266–288. Verhagen, A. 2005 Constructions of Intersubjectivity: Discourse, Syntax and Cognition. Oxford: Oxford University Press. Van Dijk, T. A. 1979 Pragmatic connectives. Journal of Pragmatics 3: 447–456. Van Dijk, T. A. & W. Kintsch 1983 Strategies of Discourse Comprehension. New York: Academic Press.

246

Ted Sanders and Wilbert Spooren

Van Lier, R., P. van der Helm, & E. Leeuwenberg 1994 Integrating global and local aspects of visual occlusion. Perception 23: 883–903. Wason & Johson-Laird 1972 Psychology of Reasoning: Structure and Content. London: Batford. Wolfe, M., J. P. Magliano, & E. Larsen 2005 Causal and semantic relatedness in discourse understanding and representation. Discourse Processes 39 : 165–187. Zufferey, S. ´ 2007 Pragmatique lexicale et m´etarepr´esentation. Etude th´eorique et empirique de l’utilisation et de l’acquisition des connecteurs pragmatiques. Ph.D. dissertation, Universit´e de Gen`eve, Facult´e des Lettres. Zwaan, R. A. 1996 Processing narrative time shifts. Journal of Experimental Psychology: Learning, Memory, and Cognition 22: 1196–1207.

Index accessibility 25 acquisition of connectives 207, 213, 219–221, 233–235 affectee 174 and 110 animacy, 137, 148, 159, 160, 162 approach cognitive 206–246 usage-based 220 argumentation 113 audience design 221 Basic Communicative Spaces Network 7, 24–59, 61, 74, 79, 86, 113, 232 because 20, 26, 106 Biblical Narrative 61–90 blending 27, 39, 44, 53, 113 blend(ed) 48, 52, 81 bo 8, 91–118 categorization 11, 121–171, 206–246 conceptual 205 linguistic 24, 51 causal verbs 123, 124, 160, 172–204 causality 215, 218 categorizations of 1 levels of 20 causality-by-default 215, 216 causee 174 causer 174 causation human 176 immaterial 176 indirect 176 physical 176, 177 chat 227–231 cognitive approach of coherence relations 206–246 cognitively plausible 22, 23

coherence relations 2, 19, 119, 206, 207 classifications of 2, 119, 147, 209 collocation, significant lexical 182 collocational analysis 182 communication chat 231 deictic center of 29, 30 communicative situation 221 conceptual complexity 213 conceptualizer 97, 224, 225, 226, 230, 231 conditional 26 connective 53 Polish 8, 91–118 construction type 181 content domain / relation 2, 20, 25, 28, 64, 68–71, 74, 79, 97, 208, 216, 217, 224 non-volitional 50, 130, 131, 135, 136–171 volitional 37, 124–171 context, construction of 26 converging evidence 11, 13, 207, 209, 232, 235 conversation, spontaneous 53, 219–222, 232–234 coreferentiality 181 daardoor 6, 19–59, 119–171 daarom 6, 19–59, 119–171 prototypical 44 deictic center of communication 27, 29, 30 denn 20 diachronic development 8, 61–90 discourse, spoken 12, 233 discourse processing 11, 52, 214–219 distancing effect 46 doen 10, 173–204

248

Index

doordat 21, 22 dus 6, 19–59, 119–171 prototypical 44 Dutch Belgian 175, 187 Netherlandic 175, 187 Spoken 10 Entrenchment, cognitive 128 epistemic connective / domain / relation 2, 20, 21, 25, 28, 64, 68, 71, 79, 80, 81, 83, 97, 208, 216, 217, 224 first-person 6, 23, 34, 40, 44, 53 fixation, lexical 182 forewarning effect 218 free indirect speech 24, 52, 78, 80 free indirect style 27 genres 133, 162, 164–221 gesture 12 ground 22, 26, 97 common 26 hypothesis causality-by-default 215, 216 (in)direct causation 173–204 parental input 220 schematic expectation 216 schematic structural expectation 215 idiomatization, collocational 174 if-clauses 25 imperatives 110 inanimateness 180 information status 145–147 instigator, causal 174 instruction, processing 13, 22, 52–53, 206 intersubjectivity 8, 97 Knowledge Base 28 language chat 221, 233 spoken 221–227, 231

language acquisition 11, 219 laten 10, 173–204 lectal variation 175 logistic regression 10, 187–196, 200 mediality 162 medium 221, 231 Mental Spaces 5, 25, 79, 80, 81, 83, 91–110 blended 27 base 25, 28 meta-linguistic 25, 28 SoC’s epistemic 39 speaker’s epistemic 39 speech act 30 Mental Spaces Theory 8, 19–53, 61–90, 91–118 metalinguistic use 20 modality (deontic, epistemic) 144–148, 151, 162, 165 monologue interieure 42 narrative 46, 61–90 newspaper (corpora) 30, 119–171 newstext, narrative 46 objective 216–218, 220 omdat 21, 96, 108, 222–232 onomasiological perspective on word meaning 130, 157, 158 paradox of causal complexity 216 paraphrase test 5 patterns collocation 10 statistical collocation 182 perspective 7, 23, 24, 38, 53, 62, 63, 65, 77, 78, 79, 80, 124, 143, 149 speaker’s 34 perspective continuity 124, 150 perspectivization 122, 123, 127, 142, 149–152 Polish 91–118

Index processing instructions, connectives as 13, 52, 206 processing of causal relations 213–219 propositional 208 protases, conditional 25 prototypical 44 prototypical usage of connectives 119–171 puisque 20 relations coherence 119, 206, 207 content 208, 216, 226, 227, 230, 231, 233 content non-volitional 50, 130, 131, 135, 136–171 content volitional 21, 37, 53, 124–171 epistemic 21, 38–47, 12–122, 130, 131, 137, 139, 140–161, 208, 216, 220, 226, 230 pragmatic 208, 216 propositional 208 semantic 208, 216 speech act 208, 216, 220, 226, 229, 230, 231 subjective 209, 217 response variable 180 Rhetorical Structure Theory 65, 68, 73, 74 Rhetorical use of connectives 34 Scientific Method 173, 174, 204 semasiological perspective on word meaning 130, 157, 158 setting, communicative 220, 229 since 20, 26

249

speaker involvement 4, 121–123, 144, 149, 156, 208 specificity, connectives vary in 26 speech act domain / relations 2, 20, 25, 28, 30, 31, 36, 64, 79, 208, 216, 224 explicit 36 spoken discourse 12, 233 subject, intransitive 174 subject of consciousness 21, 26, 27, 30, 52, 113, 123, 137, 154, 157, 223, 227 subjectification 9, 61, 63, 64, 74, 77, 86 subjective 119–171, 217–218, 220, 224, 232 subjective construal 10 subjectivity 4, 22, 63, 64, 69, 71, 77, 85, 207, 216, 218, 220, 222, 225, 227, 233 substitution 5 texts business 30 chat- 30 journalistic 30 then 98 third-person 6, 23, 27, 40, 52, 53, 224–230 thought, free indirect 46, 78, 80 to 8, 81–119 underspecification of connective meaning 206 usage-based 9, 119–171, 172–204 verbs, causal 123, 124, 160, 172–204 want 21–22, 96, 98, 108 222–232