University of Southern California Dissertations and Theses

Discourse functional units: A re-examination of discourse markers with particular reference to Spanish

(USC Thesis Other)

Discourse functional units: A re-examination of discourse markers with particular reference to Spanish

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content DISCOURSE FUNCTIONAL UNITS: A RE-EXAMINATION OF DISCOURSE MARKERS WITH PARTICULAR REFERENCE TO SPANISH by Magdalena Romera A Dissertation Presented to the FACULTY OF THE GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (LINGUISTICS) May 2001 Copyright 2001 Magdalena Romera Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. UMI Number: 3027770 Copyright 2001 by Romera, Magdalena All rights reserved. ___ ® UMI UMI Microform 3027770 Copyright 2002 by Bell & Howell Information and Learning Company. All rights reserved. This microform edition is protected against unauthorized copying under Title 17, United States Code. Bell & Howell Information and Learning Company 300 North Zeeb Road P.O. Box 1346 Ann Arbor, Ml 48106-1346 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. UNIVERSITY OF SOUTHERN CALIFORNIA THE GRADUATE SCHOOL UNIVERSITY PARK LOS ANGELES. CALIFORNIA 90007 This dissertation, written by ^ Magdalena,.Romera.... under the direction of fcrJ .. Dissertation Committee, and approved by ail its members, has been presented to and accepted by The Graduate School, in partial fulfillm ent of re quirements for the degree of DOCTOR OF PHILOSOPHY Dean o f Graduate Studies Date ................ DISSERTATION COMMITTEE ..... Chairperson — _ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 Magdalena Romera Dr. Elaine Andersen ABSTRACT DISCOURSE FUNCTIONAL UNITS: A RE-EXAMINATION OF DISCOURSE MARKERS WITH PARTICULAR REFERENCE TO SPANISH. The issue of the definition of Discourse Markers (DMs hereafter) is still a very controversial one. In this dissertation, this question is newly addressed in the light of a functional perspective. The proposal defended here is to consider these expressions as part of a general discourse framework according to which different categories can be established on the basis of independent principles. A definition of discourse categories is suggested that takes into account functional features primarily rather than structural ones. The particular discourse model adopted is based on the notion of coherence relations, understood as the mechanisms a listener/reader makes use of in order to obtain a coherent interpretation of a piece of discourse (Mann and Thompson, 1988; Sanders et al., 1992; Hobbs, 1990; 1995). Within this framework, a new category of elements, Discourse Functional Units (DFUs hereafter), is proposed. Many of the traditionally considered DMs can be also included in the set of DFUs. However, membership in this category is determined exclusively on the basis of functional criteria. DFUs are identified as those forms whose function is to carry information on the particular coherence relation that holds between two or more propositional units. Therefore, pragmatic operators such as modal particles Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 would not be included in the DFU category. On the other hand, adopting this functional definition also implies that DFUs need not be restricted to pragmatic contexts; that is, a linguistic expression can be a DFU as long as it performs the same function in every context. In accordance with these premises, the main properties of DFUs and the constraints in the interaction between DFUs and particular coherence relations are discussed. The theoretical model adopted is tested against natural data from natural conversations in Peninsular Spanish. Conclusions on several DFUs in Spanish are presented as part of this general new category. The theoretical implications of the findings of this study and their relevance for a general theory of DFUs are also discussed. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. II ACNOWLEDGEMENTS I would like to express my gratitude to all the people that have helped me in the arduous job of completing this dissertation. First of all, I would like to thank my dissertation committee members. I am indebted forever to my chair Dr. Elaine Andersen for her advice and her support. She has showed me in countless occasions the value of good work. To Dr. Carmen Silva-Corvalan my warmest and deepest thanks for her perennial encouragement. Through our discussions I have learned a lot about linguistics as well as Spanish itself, and how to do serious and scientific work. Also, I thank Dr. Edward Finegan for his comments, always accurate. Thank you as well to the people at the Information Sciences Institute, Ed Hovy, Daniel Marcu, Kevin Knight, Bonnie Stalls for giving me the opportunity to work with them in several projects. I have much benefited from my professional experience with you. I leave there very good friends. I also want to thank other professors in the Linguistics Department at USC. I am indebted to Dr. Jean-Roger Vergnaud for his constant support and his friendship, especially during my final years in graduate school. Thanks also to Dr. M. Luisa Zubizarreta for her help when I first arrived at USC. I am grateful as well to the staff in the department: Linda, Jessica, and Wenona, and most of all to Melanie for her Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Ill invaluable help. Special gratitude to Laura Reiter, who has helped me so much and who has become a friend to all of us over these years. Love ya!. I cannot forget the instructors and professors at the Department of Spanish and Portuguese, my second home while at USC. Gayle Vienna has always tried to ease the job of teaching while getting a Ph.D. She has also become a good friend. My gratitude goes also to Bertha Arce and Martha Galvan for always solving innumerable bureaucratic glitches. I thank J. Miguel for helping me obtain the data for this dissertation. I am also indebted to Pedro Romera and all the instructors at Colegio Publico Manuel de Falla, for allowing me access to their students, from whom I obtained data for a pilot study previous to this dissertation. I would like to acknowledge the Del Amo Foundation and the Program for the Cultural Cooperation between the Spanish Ministry of Education and Culture and American universities for the doctoral dissertation grants they awarded me. Also, I would like to thank my friends in Los Angeles for all that necessary emotional support, as important as all the intellectual knowledge I have gained all these years. Liliana Sanchez, Jose Camacho, Esti Amorrortu, Diana Luxenberg, Ana Tere Martinez, Marta Jevenois, Pilar Agusti, Margarita Ravera, Maite Zubiaurre, Marisa Checa, Lucia Rodriguez, Rogelio Adobbati, Philip Khoen, Jose Luis Ambite, Helen Franks, Daniel Garcia, Jim and Wanda Blake, Betty Krsnarik. Many thanks Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. IV also to my friends at the California Institute of Technology. Working there was a great experience because of you. Barbara Estrada, Doreen Domb, Christiane Orcel. Thanks to Alan Hajek for his personal and professional support. Special thanks go to my dear Carlos Jordan, colleague, best friend and much more who always supported me every step of the way, and to my family in Spain: my mom and my dad, unfortunately he could not get to see this dissertation finished, and to my sister and my brother for always being there for me. And last but not least, Gorka. This has been one of the hardest, most exciting, disappointing at times, arduous and fruitful journeys of my life, and I am happy that you were always there to let me share it with you. Thank you for your generosity, for spending countless hours discussing my linguistic ideas with me, for being always willing to read my work, for your careful comments as a reader and as a linguist. They have made me improved constantly all these years. I appreciate all these things in you very much. Mi querido Gorka, siempre gracias, por ser asi y por estar a mi lado, sin ti nunca hubiera conseguido llegar hasta aqui. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. V TABLE OF CONTENTS ACKNOWLEDGMENTS II CHAPTER 1. INTRODUCTION. 1.1. Introduction. Spanish Discourse Markers and Discourse Functional Units. 1 1.2. The proposal. 4 1.3. Thesis Organization. 7 CHAPTER 2. PROBLEMS WITH THE THEORETICAL NOTION OF DISCOURSE MARKERS. 2.1. Introduction. The nature of the problem. 11 2.2. Problems in defining DMs by a set of features. 14 2.2.1. Defining DMs by features. 14 2.2.1.1. Structural features. 15 2.2.1.2. Prosodic Criteria. 22 2.2.1.3. Distributional features. 26 2.2.1.4. Summary. 32 2.3. Functional approach to the notion of DMs. 33 2.3.1 The meaning of DMs. 33 2.4. Conclusions. 40 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. VI CHAPTER 3. DISCOURSE MODELS. DISCOURSE UNITS. 3.1. Introduction 42 3.2. Models of discourse structure 43 3.2.1. Halliday and Hasan 44 3.2.2. A wide notion of coherence. Schiffrin’s model 47 3.2.3. Redeker’s model 50 3.2.4. The Rhetorical and Coherence Relation Theories 52 3.3. A Discourse model based on coherence relations 55 3.3.1. Coherence relations 55 3.3.2. Types of coherence relations 59 3.3.3. Discourse units 70 3.3.3.1. Sentences 72 3.3.3.2. Speech acts 74 3.3.3.3. Intonation units or idea units 77 3.4. Conclusions 87 CHAPTER 4. DISCOURSE FUNCTIONAL UNITS. 4.1. Introduction 89 4.2. Definition of Discourse Functional Units 90 4.2.1. The connectivity function 91 4.2.2. DFUs as linguistic expressions that make explicit a coherence relation 97 4.2.3. The expression of coherence relations and the meaning of DFU s. 100 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. VII 4.3. Discourse Functional Units and the types of coherence relations they represent 4.3.1. Types of DFUs or types of relations? 4.4. Summary CHAPTER 5. EMPIRICAL STUDY OF DFUs ON MODERN SPANISH. 5.1. Introduction 5.2. Data Source 5.2.1. The corpus 5.3. Method of analysis 5.3.1. The coding process 5.4. Results and discussion 5.4.1. Forms and frequency 5.4.2. Forms and relations. The semantic, pragmatic and textual sources of coherence 5.4.2.1.Forms most frequent in semantic relations: si, y, porque, osea. 5.4.2.2.Forms most frequent in pragmatic relations: pero, es que, entonces, pues. 5.4.2.3 Forms most frequent in textual relations: bueno. 5.4.3. Summary 5.4.4. Extractability 5.5. Conclusions Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 110 114 127 128 129 129 134 134 147 148 159 160 190 241 254 255 272 VIII CHAPTER 6. CONCLUSIONS. 6.1. Summary 276 6.2. Conclusions from the empirical study and implications for the theory of DFUs. 284 REFERENCES 298 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. IX LIST OF FIGURES Figure 3.1. Taxonomy of relations as it appears in Hovy et al. (1997) 67 Figure 5.1. Forms and frequency of DFUs 149 Figure 5.2. Adversative 151 Figure 5.3. Contrast 151 Figure 5.4. Concession 151 Figure 5.5. Contrast to Expectation 152 Figure 5.6. Clarification 152 Figure 5.7. Objection 152 Figure 5.8. Request 152 Figure 5.9. Percentage of representation of si in sem., pragm. and textual domains. 161 Figure 5.10. Percentage of representation ofy in sem., pragm. and textual domains. 167 Figure 5.11. Percentage of representation of porque in sem., pragm. and textual domains. 174 Figure 5.12. Percentage of representation of osea in sem., pragm. and textual domains. 181 Figure 5.13. Percentage of representation of pero in sem., pragm. and textual domains. 191 Figure 5.14. Percentage of representation of entonces in sem., pragm. and textual domains. 206 Figure 5.15. Percentage of representation of pues in sem., pragm. and textual domains. 216 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. X Figure 5.16. Percentage of representation of es que in sem., pragm. and textual domains. 229 Figure 5.17. Percentage of representation of bueno in sem., pragm. and textual domains. 242 Figure 5.18. Percentage of extractable DFUs according to source of coherence 262 Figure 6.1. Discourse constituency 278 Figure 6.2. Textual component of language 289 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 CHAPTER 1. Introduction. Spanish Discourse Markers and Discourse Functional Units. 1.1. Introduction. Goals and Contributions. This study started out as an attempt to contribute to a better understanding of the linguistic expressions which have been classified in previous discourse analyses as Discourse Markers (DMs hereafter) and to develop principled criteria for the characterization of these forms as a linguistic category. As I delved deeper into the issues and the empirical data, it became clear that I needed to reassess and redefine the notion of DMs and to propose the coinage of another category, Discourse Functional Units (DFUs hereafter), for which membership is defined by a clearer set of principles. The study of DMs has evolved significantly in the last decade. From their treatment by traditional grammars as ‘expletives’, ‘pause-filler’ forms that do not contribute to the syntax of language to any extent, there has been a more recent move towards a much deeper consideration of their role in structuring discourse. Many researchers have tried to develop a characterization of DMs, and some of them have Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 made an attempt at classifying these forms as a linguistic category of grammar. The criteria and the methods used for their study vary greatly, however, depending on the theoretical perspective of the researcher, and so far the definition of what constitutes a DM is still an open issue. This thesis addresses the question of what a DM is and proposes a definition of these forms as basic units of a model of discourse based on the notion of coherence relations. Within this framework, I try to offer an independent definition of the category of DMs in terms of their functionality (hence the new term that I propose to refer to these elements: discourse functional units) and a characterization of these units on the basis of an empirical analysis of natural conversation in Spanish. From a general linguistic point of view, this study of discourse functional units contributes to a deeper knowledge of linguistic forms that may encode semantic, pragmatic and textual meanings in discourse. And in particular with respect to the issue of DMs, this dissertation contributes to the categorization of some ‘fuzzy’ forms operating at the discourse level for which there is no unitary agreement either on the terminology used to refer to them, or on the features they are supposed to have in common. The study of these forms in Spanish should constitute a serious contribution to Spanish Linguistics as well. Unlike other languages for which the study of discourse phenomena in general have received more comprehensive treatment, the literature on Spanish DMs has dealt primarily with pragmatic and textual functions of isolated forms such as pues ‘so’ or o sea ‘that is’(Portoles, 1989; Martin Zorraquino, 1992, among Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 others; Brizuela, 1995) or on linguistic variables that also reflect the social characteristics of the communication (Brizuela and Andersen, 1999). Very few attempts have been made in the last decade to provide general accounts of the features of these linguistic expressions in Spanish. Among the studies that have proposed a general classification of Spanish DMs, the most exhaustive ones are the one by Fuentes Rodriguez (1987), and more recently the one by Pons (1998). Under a framework of textual coherence (Halliday and Hasan, 1976), Fuentes Rodriguez provides a very detailed study of what she calls connective/textual links {enlaces conectivos). Her classification includes a wide set of forms which can indicate logical or intradiscourse relations between utterances. Although the value of her work is unquestionable, two main problems can be identified. First, it is not clear what criteria she relies on to select certain forms over others. For example, when dealing with the logical relation of addition, she excludes from her study the formy ‘and’, the primary form used to express this relation in discourse (cf. Romera, 1998). Second, although she mentions that some of the forms studied are more frequently used than others, she does not provide quantitative information that could help establish how these forms are used by speakers or whether the functions of these forms bear different status (i.e., whether there are primary and secondary functions). Pons (1998) on the other hand offers a very exhaustive analysis of many forms that in the Spanish literature have been considered DMs, both from a qualitative and quantitative point of view. He combined most all definitions of DMs ever proposed and Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. produces a characterization of DMs on the basis of the common features these definitions suggest. However, as in Fuentes Rodriguez’s case, the criteria used to select the forms analyzed remain unclear, and that is a serious theoretical and methodological handicap. The study of DMs in Spanish is also empirically motivated. As many researchers in different languages have pointed out (Scott, 1982; Jisa, 1987; Andersen, 1996, among others), the existence of a classification of the different types of DMs in an adult language model offers an interpretive basis for studies dealing with the role of pragmatic expressions in language development, either in first or in second language acquisition. Especially in the case of second language acquisition, a classification could prove invaluable for teachers, as absence or misuse of these forms is a clear indicator of non native status and can negatively impact communication. 1.2. The proposal. The issue of the definition of DMs is a controversial one. Finding unitary agreement on what constitutes a DM is has seemed thus far to be an almost impossible task so far. In the current literature, there are many different perspectives from which DMs are studied, and as a result there are considerable differences in the inventories of DMs for each approach, as well as serious terminological diversity and confusion. The main theoretical approaches to the definition of DMs as a category range from characterizing them by a set of features they are supposed to share to analyzing them as Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 5 part of a general discourse framework. Within the first approach, several groups of structural and/or functional features have been proposed in the literature, but none of these features alone nor any combination of them are able to provide full characterizations of all DMs. The main problem lies in the fact that only a series of a priori selected groups of expressions have been the basis for the analysis of these features, and such groups or lists of forms do not constitute exhaustive lists of the set of DMs that can be found in the languages in question. This creates a serious theoretical problem: the definition of DMs is not independent from the particular set of forms chosen to be studied, and therefore it is unlikely that it can be perfectly applicable to a different set of forms. It would clearly be better to develop a different approach which would allow a non-a priori account of the concept of DMs. The perspective proposed in this dissertation is to consider these expressions as part of a general discourse framework in which different categories can be established on the basis of independent principles. I suggest a definition of discourse categories based on functional features rather than structural ones. It is commonly agreed that the forms usually called DMs lack propositional meaning and bear only a procedural meaning. I argue that this is basically a functional distinction that allows us to establish a main differentiation between types of entities in discourse. On the one hand, there are discourse segments that bear propositional content, defined primarily in terms of the semantic information which is transmitted and secondarily on phonological features (intonation contour, pauses, etc.). I refer to this type of unit as a Discourse Propositional Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 6 Unit (DPU). On the other hand, there are units that bear primarily a procedural meaning, which basically transmit information on how to establish a coherent relation between DPUs. I call this type of unit a Discourse Functional Unit (DFU). Obviously, many of the forms traditionally considered DMs could now be considered DFUs as well. However, membership in this category is determined on the basis of functional criteria exclusively, that is, only those forms whose function is to offer information on the relation between DPUs are considered DFUs. Therefore, pragmatic operators such as model particles for example, which modify the illocutionary meaning of a DPU, would not be included in the DFU category. On the other hand, adopting this functional definition also implies that DFUs need not be restricted only to pragmatic contexts; that is, a form can be considered a DFU as long as it performs the same function in every context. (Vid. Chapter 3 for semantic, pragmatic and textual characterization of DFUs). As mentioned above, the notion of coherence is crucial in the definition of the functions of DFUs. The particular model I adopt here adopts the tenets of Coherence Relation Theories (Grosz and Sidner, 1986; Mann and Thompson, 1988; Sanders et al., 1992; Hobbs, 1990; 1995), which claim that coherence is constructed through a series of relations between discourse segments. A coherent discourse is then structurally organized into a series of units interconnected by these relations. Coherence relations are the fundamental criteria around which the discourse model is built; they are the ones that make possible the coherent interpretation of a piece of discourse. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 7 Within this framework, I explore the notion of DFUs as elements that make coherence relations explicit. I also discuss the constraints in the interaction DFU- coherence relations. After presentation of this theoretical framework, the model adopted and the notion of DFUs in particular are tested against natural data from Spanish conversations. The corpus studied is drawn from the samples compiled in the CREA (Corpus de Referencia del Espanol Actual ‘Corpus of Reference of Modem Spanish). A total of approximately 16 hours of recording are analyzed. The results of this empirical analysis provide many insights into the nature of Spanish DFUs, as well as contributing to a more general understanding of this discourse category. 1.3. Thesis Organization. Following the introductory discussion of the general issues, the rest of this thesis is organized as follows: In Chapter 2. Problems with the theoretical notion o f DMs, I critically review the different approaches to the definition of DMs in general and in Spanish in particular. I show how a definition created through a series of postulated formal features has resulted in a ‘catch-all’ category in which there is no a consistent principled way to determine membership. It seems that the problem arises in large part through analysis based on a set of intuitively selected forms without much independent justification of the set, then trying to draw general conclusions based on that arbitrary set. Instead I argue for a different methodological approach, which defines DMs as a part of a Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 8 discourse model, on the basis of functional features only. A distinction between units in discourse is proposed according to this view. Chapter 3. Discourse Structure. Discourse Units presents a review of the models that give an account of DMs in these functional terms, based on the theoretical that discourse shows a coherent structure. First, I review the precursor model ofHalliday and Hasan (1976), then its inheritor models: Schiffrin (1986), Redeker (1986; 1990; 1991) and the Coherence Relation Theories (Grosz and Sidner, 1986; Mann and Thompson, 1988; Sanders et al, 1992; Hobbs, 1990; 1995). The most important point I would like to emphasize about these theories is that, unlike the previously mentioned studies, they are able to provide an independent definition of discourse units. With the exception of Schiffrin’s model, which is constructed basically for the purpose of analyzing DMs’ functions, in the rest of these theories their discourse framework is created independently from the use of DMs. It is precisely this fact which allows us to offer an independent definition of these forms. DMs do not need to be defined in relation to the features any particular set of forms exhibit, but rather as a category the members of which perform a function within a model of discourse, and that is the expression of coherence relations. This approach to the definition of discourse elements avoids the circularity problem mentioned above. In chapter 3, the notion of discourse that I adopt in this dissertation is also presented. Discourse is seen as a structurally organized compound of units which are related to each other through a series of coherence relations. Since the concept of units Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 9 and relations are the fundamental basis of the definition, the most important issue is how we determine what a unit is, and what are the relations that we claim create coherence among those units. I review the most common definitions of discourse units used in the field. First, those based on syntactic criteria (sentences, clauses); then, the ones based on pragmatic information (speech acts); and finally, those based on phonological/semantic criteria (intonation/idea units) which I found to be the most reliable for the purpose of my study. Additionally, based on the taxonomies provided in the literature, I suggest a tripartite categorization of coherence relations: semantic, pragmatic and textual. This distinction is made on the basis of the information in each case that contributes to the relation: semantic, pragmatic or textual information. I also consider the relations to be paratactic when both the segments related have a central role in the relation, and hypotactic when one of the segments performs a central role and the other one a subsidiary role. Chapter 4. Discourse Functional Units presents the definition of this new category of DFU within the discourse model. DFUs are one of the two basic units in discourse, along with the elementary discourse units (DPUs) that get connected through the coherence relations, and are defined as functional categories in discourse. They function as explicit markers of the coherence relation that operates over the propositional units. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 10 The interaction between a particular DFU and the relation or relations it expresses is not an arbitrary one. For a DFU to come to represent a relation, its primitive meaning must be compatible with the meaning of that relation. In addition, a DFU can also expand its use into expressing other relations, but they must be also compatible in meaning with the ones that the DFU already represent. Furthermore, the new relations should be as specific or more specific than the most frequent relations the DFU represents. Finally, DFUs are more or less susceptible of extractability in accordance with the type of relation they represent. A DFU is less easily extractable when expressing semantic or pragmatic relations than when representing textual ones. This is because each type of relation contributes differently to the coherence of discourse: semantic and pragmatic relations contribute directly to coherence whereas textual relations only do so indirectly, and therefore the former are more relevant for a coherent interpretation than the latter. Chapter 5. Empirical Study presents the empirical analysis of the Spanish data according to the premises of the proposed model. The chapter begins with an explicit description of the coding procedures of the corpus, based on oral discourse samples compiled in the CREA (Corpus de Referencia del Espanol Actual). It then presents a quantitative analysis of several DFUs, along with qualitative discussion on the results obtained. Chapter 6. General Conclusions, contains the general conclusions in the light of the empirical results obtained. I also discuss the theoretical implications of the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. findings for the study of DMs. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 11 CHAPTER 2 Problems with the theoretical notion of Discourse Markers. 2 .1. Introduction. The nature of the problem. It is widely accepted by now that in the study of Discourse Markers (DMs) the main issue that still remains controversial is the definition of a DM, that is, understanding what a DM is, what characteristics the members of such a class may share. In a recent book on DMs within Relevance Theory, Jucker and Ziv (1998) devote a substantial amount of their introduction to discussion of this issue. In the same fashion, from a grammatical-pragmatic theoretical perspective, Fraser (1999) also deals with the same problem. The issue is complicated and everyone concludes that there is no unitary agreement in their definition, either on the terminology to be used to refer to these forms or on the features they are supposed to have in common. As Jucker and Ziv (1998) put it, ‘the range of features proposed in the literature as being characteristic of discourse markers is not always shared by the various putative members of the class. It appears that discourse marker is a fuzzy concept’ (Op. Cit.:2). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. . . . . . . 12 The great variety of terms used to refer to these forms includes discourse markers (Schiffrin, 1987; Hovy, 1994), discourse operators (Redeker, 1986, 1991), discourse particles (Schourup, 1985; Abraham, 1991), cue phrases or connective phrases (Hovy, 1997; Knott et al., 1997), pragmatic markers (Fraser, 1988, 1995), pragmatic particles (Ostman, 1981), punctors (Vincent and Sankoff, 1992; Vincent, 1993) or discourse connectives (Blakemore, 1987, 1988) among others1 . Under these tags, a varied group of grammatical categories can be found. Among others, conjunctions (and, or, but, so), adverbs (before, however, then, now, if), prepositional phrases (on the other hand, after all, for example), exclamations (ah!, oh!), and utterances (I'm not sure, y 'know, if you like, before Iforget). The approaches to the definition of DMs are also varied. The properties considered to characterize the set of forms called DMs are almost as varied as the grammatical forms mentioned above; they usually include a mixture of formal and functional characteristics2 : Formal or identificational features take DMs as linguistic units characterized by a set of common, mainly structural (distributional, categorial), properties. Functional features are established according to the role DMs play at the different levels in the analysis of language. Basically, in most studies both types of criteria tend to be combined in order to delimit the set of DMs analyzed. (Vincent and Sankoff, 1992; Vincent, 1993; Brinton, 1990; Sankoff et al. 1997, among others). 1 For a more detailed set of references see Pons (1997). 2 Brinton (1996) includes a good summary of these characteristics drawn from general and individual studies. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 13 As becomes apparent once formal features are looked at in detail, serious problems arise in feature-based definitions of DMs. None of the criteria suggested to define this class can be applied in a consistent, principled way to all the items that have been proposed as DMs. This is why the proposal presented in this dissertation is the consideration of functional features alone for a principled definition of DMs. Structural features may be appropriate for practical purposes; that is, they provide operational criteria that allow researchers to focus on a particular set of forms, but the features proposed in the literature seem to offer contradictory results in the identification of forms as DMs. That is, according to some features a particular form may be classified as a DM, whereas by another feature the same form may not be considered a DM. As I demonstrate in this chapter, this problem derives from a more serious methodological issue. All the structural features proposed to identify DMs have been usually suggested a posteriori, that is, they all derive from the analysis of forms that have been established a priori by the researcher as DMs. This inductive methodology creates a very serious theoretical problem of circularity. A general definition of DMs is given on the basis of a set of features extracted from the observation of the behavior of a group of forms established as DMs a priori. Therefore, the definition is not independent from the particular set of forms chosen to be studied; and consequently it is unlikely that it will be applicable altogether to any other set of forms. What we need is criteria to identify those forms independent from the properties of the particular forms themselves. To achieve this goal the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 14 methodology proposed here is of an onomasiologic nature: start with a well defined discourse model in which DMs can be characterized as a part, on the basis of independent principled criteria. Since structural features cannot be systematically applied, the criteria that permit an independent definition of units in discourse must be essentially of a functional nature. The reexamination of the category of DMs from a functional perspective implies a reconsideration of the notion of DMs and the proposal of a class of discourse units, Discourse Functional Units, for which are defined by a clear set of functionally principled criteria. 2.2. Problems in defining DMs by a set of features. 2.2.1. Defining DMs by features In this section I review the criteria most commonly used in the literature to define the category of DMs. These include structural features (sentential exclusion, meaning), prosodic features (being preceded and followed by pause, special intonational patterns) and distributional features (restricted to oral discourse, sentence initial positions, and co-occurrence of several DMs in the same utterance). After reviewing each of these characteristics, a clear conclusion emerges: formal features do not constitute clear criteria to support an independent definition of DMs or a characterization of their status as a discourse category. The problem, as mentioned, stems from the methodology followed: all the properties used to define the category of DMs come from analysis of partial sets of expressions which vary greatly depending on the researcher's perspective. The most obvious implication for Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 15 the field is the need for a different theoretical approach to the study of DMs that provides a general definitional criteria of DM’s structural and functional properties. 2.2.1.1. Structural features Sentential exclusion One of the most common features mentioned when describing DMs is that they occur outside the structure of the sentence, i.e., they are not integrated in the structure of the sentence. This argument has been put forward by many researchers in past years, especially those that understand the organization of discourse in terms of hierarchical units above the sentential level. This view has been especially fruitful in the European tradition and among those non-European researchers that approach the study of DMs from a grammatical perspective. Within the Spanish tradition, for example, Martin Zorraquino (1998) characterizes DMs as entities with external value to the predication because they are not selected by the subject or the combination NP-VP. se trata de entidades invariables, con rango extemo a la funcion predicativa oracional -no son seleccionadas ni por el verbo ni por la combinacion de la relation suj eto-predicado- y que reflejan un contenido secundario respecto del contenido de toda la oration o de alguno de sus miembros3 Martin Zorraquino, 1998:25 Additionally, a second characteristic also mentioned by Martin Zorraquino that argues in favor of their external value to the sentence structure is the fact that they cannot be substituted by pronouns or deictics as other sentential constituents. 3 They are invariant entities, with external value to the sentential predicative function -they are neither selected by the verb nor by the combination subject-verbal phrase, and they reflect a secondary content with respect to the content of the whole sentence or any o f its members. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 16 Fraser (1994; 1998), Malamud-Makowski (1997) and Schiffrin (1986) also advocate for a definition of DMs that includes this “out of sentence structure” feature, although they do not offer any syntactic arguments in favor of this explanation. But despite considering the very same feature, all these researchers end up delimiting different sets of DMs. For example, for Fraser (1994) and Malamud- Makowski (1997), sentential adverbs like frankly, obviously and stupidly are not included within the selected group of DMs. Schiffrin does not even talk about sentential adverbs. Nevertheless, Martin Zorraquino, along with other European researchers, bases a great part of her characterization of DMs on forms like logicamente ‘logically’, sinceramente ‘sincerely’, naturalmente ‘naturally’ when these function as sentential adverbs. If we turn to the grammatical studies on the syntactic status of some of the forms that have been defined also as DMs in the literature, such as conjunctions (or, but, so, like, or therefore, however, etc.), the opinions among grammarians are divided as to where these forms should go in the syntactic structure of sentences. Warner (1985), in an attempt to provide a syntactic characterization of forms such as otherwise, like, but, so, unless, suggests that they appeal to ‘the discourse fragment as a level of suprasentential syntactic organization’. But at the same time he states that ‘they [discourse connectives] occur as constituents of sentences, to be sure’ (Warner, 1985:73). With respect to coordination conjunctions (and, but), syntacticians agree on treating them as deficient heads, that is, they have the same features as other heads but they do not project a Conjunction Phrase maximal projection. Rather, they project a category of the same type as its conjuncts (Camacho, 1997). However, most scholars that have worked on coordination agree that conjunctions have a structure Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 17 that falls within the X-bar template. That is, conjunctions are heads that have specifiers and complements. Finally, there are few syntactic studies on consequential adverbs and conjunctions like thus, then or so, but some studies on their Spanish counterparts suggest that there is a plausible sentential treatment for asi ‘thus’, entonces ‘then’, en consecuencia ‘as a consequence’, pues ‘so’. These analyses propose that these forms head adverbial adjunct clauses, adjoined to the matrix clause, and that they obtain a peripheral predication, forcing the tense from both clauses to be compatible (Ticio, 1997) So, there seems to be enough evidence from syntactic studies to conclude that within current syntactic theories grammarians find plausible sentential analyses for these forms. This indicates that unless we are able to provide strong syntactic arguments in favor of the exclusion of these forms from the regular sentential tree, we should use the ‘out of the sentence feature’ cautiously to identify these forms as DMs. It could be argued that what researchers typically consider DMs are precisely those types of conjunctions that fall out of the regular conjunction/adverb behavior. But this is only partially true. For example, if we take the case of and, in proposals like Fraser’s (1990) or in Rhetorical Structure Theories (Mam and Thompson, 1988; Hobbs; 1990, 1995; Sanders et al., 1993), we see that there is not much difference between the sentential use of a conjunction and what they call DMs. Even in Schiffrin’s analysis of and and but, it is not always clear that these forms are not functioning as conjunctions. Consider examples (1) and (2). (1) is extracted from Soria and Ferrari (1998), who follow RST models of discourse, and (2) from Schiffrin (1987). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 18 (1) he received, then, important commissions for the Venetian churches and dedicated himself to the decoration of palaces and villas of the high Venetian aristocracy. (Soria & Ferrari) (2) Jack: I keep thinking of the French Catholic. It’s entirely different than the Spaniard Catholic. And the Spaniard Catholic is entirely different from the Italian Catholic. and the Italian Catholic is entirely different from the Hungarian Catholic. (Schiffrin: 134) In both (1) and (2) the uses of and resemble very much that of a conjunction between simple sentences. As we can see, in (2) both occurrences of and function as sentential conjunction: ‘The French Catholic is entirely different from the Spaniard Catholic and the Spaniard Catholic is entirely different from the Italian and the Italian is entirely different from the Hungarian’. These cases of and do not seem different from the common sentential use of this conjunction. What I propose is that in functional terms conjunctions are DMs as long as they perform the same functions regardless of the context in which they appear: sentential or suprasentential. Source expressions o f DMs. The second characteristic that has been proposed as a defining feature for DMs is that they derive from a word-source which is a lexical expression4 that can be part of the syntactic structure of the clause (Vincent, 1993; Vincent and Sankoff, 1997; 4 The term lexical is used here to refer to words that bear autonomy within the language system. This contrast with term grammatical which means a non-autonomous form that only bears grammatical features (function words, clitics, inflections, etc.) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 19 Brinton, 1990; Traugott; Ferrara, 1997; Romera, 1998; Fraser, 1998). These source forms pertain to many different grammatical categories, and DMs differ from them in that DMs have lost or changed many of the original morpho-syntactic and semantic characteristics of these word-sources, along with their original functions. In the last years, an increasing number of researchers have began paying attention to the study of a process by which lexical forms come to evolve into grammatical ones. This is called grammaticalization. By this process, first studied in the beginning of the century by Meillet (1912), lexical forms such as verbs or nouns are pressed into new uses to serve grammatical functions, that is, to express grammatical features. One example is the case of the English verb manage, which from being a verb with full sense like in Amy manages the sales office o f a large corporation, changed into a grammatical use as an auxiliary verb, as in Amy manages to get a salary increase every year (Ftopper, 1996:217). Some researchers have described the common principles that underlie this process and define the direction of the change (Lehmann, 1985; Hopper, 1991; Hopper and Traugott, 1993; Heine, Claudi and Hiinnemeyer, 1991). The following is a summary of the main characteristics of grammaticalization, extracted from Hopper (1996:230). Layering. Old and new emergent layers in a functional domain may coexist and interact. For instance, the several ways of expressing future tense in English. Divergence. The original form and the newly grammaticalized one may undergo new different changes. The source form may continue undergoing the same regular changes as other lexical items. Specialization. Each form in a functional category tends to specialize for a specific function. For example, pas in French initially coexisted with ne, but now it is taking over as ‘the expression’ of negation. Persistence. Some parts of the meaning of the original form can be traced back Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 20 within the new grammaticalized form. De-categorization. Grammaticalization usually involves a change of word category. The direction of the change is always from noun or verb to another category such as adjective or adverb. It never occurs in the opposite direction. In addition to these principles, other researchers have also suggested a loss or lessening of part of the phonetic body of the lexical item as part of the process. This is the process known as phonological attrition (Bybee and Pagliuca, 1985; Lehmann, 1985; Heine, Claudi and Hunnemeyer, 1991). It is hence not surprising that some researchers have suggested the formation of some DMs as a case of grammaticalization. DMs seem to adjust to the main patterns mentioned above. DMs like but, well, however, anyhow, besides, right, like (Finell, 1992; Traugott, 1982; Brinton, 1996) come originally from lexical forms of time, space, manner, etc. pushed into expressions of ‘more subjective meanings’ (Brinton, 1996:61). The tendency in the process of grammaticalization is to evolve from semantic to pragmatic meanings. Traugott (1989; 1990) identifies this tendency as an evolution from propositional to textual to interactional meanings. This theory is extremely appealing and it certainly provides an excellent insight into the obscure area of DMs, offering explanation for how certain forms that in their lexical use express a propositional meaning, evolve and come to serve textual and interpersonal functions. The grammaticalization process certainly seems to account for the evolution of some lexical forms into DMs, the question now is whether it can help us establish a criterion to identify them. That is, we could suggest that a grammatical form can be called a DM if it has a traceable lexical source from which it distinguishes in meaning (evolved form semantic to pragmatic meaning), although part of the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 21 original meaning is still noticeable in the DM. The new form presents some degree of phonological attrition, and belongs to a different grammatical category from its source form. This hypothesis presents some problems. In the first place, not all forms that have been studied as DMs are consistent with all the principles mentioned above. For example, in some cases it is almost impossible to trace the source form of the DM. In some of these cases the lexical item may have already disappeared from the language. An example pointed out to me by B. Comrie (p.c.) would be the Russian form ved you know’. This is a form of a verb ‘to know’ that is no longer productive in contemporary Russian. In other cases, what happens is that the lexical form was never actually part of the lexical repertoire of the language. C. Silva-Corvalan (p.c.) pointed out the clear example of some South American languages in contact with Spanish that borrow grammatical markers directly from Spanish. Likewise, the Basque dialects in contact with Peninsular Spanish borrow forms like bueno ‘well’ pues ‘so’ in their DM function. For all these forms there is no traceable source in the host language. Other characteristics or tendencies are not general to all DMs, either. De categorization is not so evident in the case of conjunctions like and, but, or. Conjunctions are mainly function words that perform the same function in different contexts, sentential or suprasentential. Phonological attrition does not always happen. It affects basically to words more frequent in oral discourse and that additionally have a reduced phonological body already, such as but or OK (Romera, 1998). In the second place, not all forms that evolved into grammatical elements are DMs. Traugott (1989) mentions modal auxiliaries, temporal markers, speech-act Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 22 verbs, evidential markers, all as examples of the rise of pragmatic meanings in English. Thus, this rather obvious conclusion explains that in fact it is not the process of grammaticalization what identifies DMs, since DMs are only one possible type of grammatical markers. DMs are only one more example of the result of a semantic tendency in the language. So, it is possible that most or all DMs could have evolved diachronically from lexical items either present or absent from the source language, but this characteristic is not sufficient by itself to identify their status as a category. 2.2.I.2. Prosodic criteria. The prosodic characteristics of DMs constitute the second set of criteria for their identification. Several researchers in discourse analysis have noticed that DMs bear certain prosodic features that make them different from other segments in discourse. Schiffrin (1987:328) mentions that they show a range of prosodic contours, such as tonic stress followed by a pause and phonological reduction. Other researchers have claimed that DMs form a separate intonation unit (Quirk et al., 1985). However, not much work that undertakes a serious prosodic analysis has been done in the field. To my knowledge, the only work that uses a computerized pitch- extraction program in the study of the prosodic characteristics of DMs is that of Ferrara (1997). She uses prosodic information to characterize anyway as a DM. Her analyses are based on the study of the pitch contours (Fo) of this form in samples of male and female oral discourse. By examining the different intonational patterns of anyway she distinguishes three types of anyway, one is a DM and two are adverbial homophonous forms. The analysis clearly showed three different prosodic contours and clearly singled out the DM from the other two. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 23 In my opinion, only a study of the Fo contours where DMs appear will help us determine the prosodic properties of DMs, that is, only by these means will we be able to determine whether DMs are prosodic units per se and whether there is lessening of phonetic body, as some researchers suggest (Cf. Schiffrin, 1987). The concept of prosodic unit coincides mainly with what has been called intonation unit. The main difference lies on the fact that intonation units are identified in terms of auditory perception, whereas prosodic units, also called declination units because of the general decline in pitch level, are identified using acoustic measurements. Additionally, some of the criteria used in the identification of intonation units are of little relevance acoustically. An intonation unit as defined by Chafe (1987,1992) or Dubois et al. (1993) is a part of discourse that (a) is preceded and followed by a noticeable pause (0.3 second or greater); (b) presents an overall decline in pitch level; and (c) presents a falling pitch contour in the end. In order to adjust intonation units to prosodic units, the first of these three properties must be modified, because intonation units are not always followed by a pause. It seems that when acoustic measurements are made on natural speech, in many cases pauses do not play an important role in marking boundaries between prosodic units. Schuetze et al., (1991) showed that out of a total of 455 acoustically identified prosodic units nearly two thirds of them showed no boundary pause. According to these authors, it seems that the most influential prosodic feature in the identification of prosodic units is pitch reset, that is, a change in pitch direction relative to the following units. A pilot study of the Spanish DMs (bueno ‘well’, pero ‘but’, pues ‘so’, y ‘and’) was set up by Romera and Elordieta (in progress) to test whether these forms met all the necessary requirements to be identified as prosodic units: overall decline in pitch Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 24 level and final tone with falling contour accompanied by pitch reset. Additionally, using Beckman and Ayers’ (1994) notation system for prosodic labeling, some more detailed prosodic information were added. Thus, an additional feature called break index, which marks the subjective strength of association between one word and the next, scaled from 0 for the ‘strongest perceived conjoining’ to 4 for the ‘most disjoint’ was incorporated. The break index between a DM and the next word should rate between 3 and 4 to be identified as a prosodic unit. Only in cases in which the DM has a very reduced phonological body (such as but, and, so) the break index can be lower, because of a tendency shown by function words with little phonetic weight to cliticize. The prosodic characteristics set out to study were only confirmed for some cases of DMs with a relatively big phonetic body, such as bueno and pues. Only the examples of bueno and pues analyzed presented in the majority of cases a significant decline in pitch (from 170 Hz to around 100 Hz), a final falling contour and a break index of 4. The cases of pero and y analyzed failed to show these characteristics in every instance. Overall, they presented an imperceptible break index; at the same time, it was difficult to find a clear falling final tone in the occurrence of these forms. Using a more traditional methodological approach, Silva-Corvalan (1998) in her study of ahora ‘now’ in Chilean Spanish reports similar conclusions. That is, not in all cases the DM presents characteristics of what she defines as ‘final tonal group’. Ahora conectivo no aparece en todos los casos seguido de una pausa breve o de un patron entonacional de final de grupo tonico que lo separe del enunciado que encabeza.5 5 Ahora as a connective does not appear in all cases followed by a brief pause or with a characteristic final lowering intonation pattern that separates it from the utterance it precedes. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 25 So it seems that judging from some of the few available studies on intonation of DMs, there is no conclusive evidence that prosodic characteristics are a relevant factor or at least a defining factor for the status of DMs. Different forms in discourse that have been studied as DMs, such as the Spanish bueno or pues, seem to show characteristics that can determine that they are prosodic units in some cases, but not in all the examples examined, and those characteristics are not found in other DMs with less phonetic material, such as y. With respect to the other feature often mentioned in the prosodic definition of DMs, their phonological reduction, there is also no empirical evidence that this process occurs in all instances. Although for many researchers, especially many of those working under the grammaticalization framework, this seems to be almost a clear indication that DMs are part of the grammatical marker group which evolves from original lexical items, it is not clear that DMs have completed the process of phonetic loss yet. On the contrary, phonological reduction is recognized by other researchers as a feasture that distinguishes DMs from other grammaticalization processes. When talking of pragmatic particles, Brinton (1998) explicitly states that “there is only one respect in which the development of modal particles differs from other instances of grammaticalization: namely, there is no loss of phonetic substance” (:64). In many cases, as Romaine and Lange (1991) point out, the forms analyzed are very recent and “the discussion lacks historical depth” (:251). Due to this recency in historical evolution, DMs show variation with respect to phonological attrition. A DM may be clearly reduced only in some contexts, (especially in informal registers), as for instance cauze (because) or y ’ know in English, but not in others, e.g. more formal registers. Other factors, mainly frequency of use, will probably determine whether some DMs get permanently reduced in the future. It is Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 26 probable that due to these factors, not all DMs end up losing part of their phonetic body, as is observed with other phonological processes. In conclusion, for now there is no conclusive evidence to support the idea that prosodic features alone can determine the status of DMs. In the first place, there are not enough studies that empirically prove that all forms called DMs share the same prosodic characteristics. Some DMs can be considered as prosodic units in many instances, since they present pitch reset, that is, a change in pitch direction relative to the following units, final falling contour and break index of 3 or 4. But these characteristics are not extendible to all the DMs, especially to those that have a more reduced phonetic body, such asy ‘and’. Additionally, there seems to be evidence that phonological reduction shows a great range of variation depending on many circumstances. The same DM varies in this respect when appearing in different communicative contexts. 2.2.1.3. Distributional features This section contains a review of several characteristics of DMs that are commonly mentioned but that do not have an intrinsic structural character; rather, they are defined by observation of the distribution of DMs. The three most common are: DMs are features of oral discourse; they are often restricted to sentence initial position; and they may co-occur in one single sentence. Features o f oral discourse The first distributional feature of DMs is their predominantly oral character, that is, their frequency in oral discourse is much higher than in written discourse (Ostman, 1982). This assumption is implicitly accepted by discourse analysts, but Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 27 there are few studies that have tested it comparing both modalities. Among the few, a recent experiment that tries to establish this relationship is the one by Soria and Ferrari (1999), who set out to study the frequency and types of relations that DMs express in oral versus written discourse. In their experiment, 19 subjects were asked to retell a story that had been previously presented to them in the form of mute cartoons. They had to retell the story first in the form of an oral narrative and then in writing. Results showed that, in fact, the same subjects used more DMs in the oral version: 1 DM every 1.6 clauses in the spoken part versus 1 DM every 2.5 clauses in the written part. So, initially, we could conclude that the use of DMs is more frequent in oral than in written contexts. But apart from this initial distinction, a second question was investigated: whether different types of DMs were used in each context, that is, whether the relations expressed by DMs were different in oral and written discourse. The findings showed that the relations established by DMs were basically the same in the oral and in the written narratives; the percentages of each established relation were virtually identical in both contexts. So, initially DMs seem to be indeed more frequent in oral discourse than in written discourse, however they perform the same functions, that is, they express the same type of relations in both modalities. A further question to investigate which Soria and Ferrari do not address in their study is the difference in frequency of each particular form in both registers. Although the relations expressed by DMs may be the same, it is important to know whether certain DMs are more frequent or maybe exclusive of oral contexts whereas others are more frequent or restricted to the written register. The assumption that DMs are a predominant feature of oral discourse may be a result of the selection of a particular set of forms. For example, if we decide that we will study well, now, so, Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 28 but, oh, because, or, I mean, and y'know, then , as Schiffrin (1987) did, chances are that we conclude that some of these forms are almost exclusive of oral discourse. Even without doing a statistical analysis we are able to see that oh ox y'know are very unlikely to appear in written registers, unless these portrait informal dialogs. But if we extend our set and include forms such as however, as a consequence, as a result, on the other hand, while, as Soria and Ferrari (1999) did, then we could expect similar results to the ones they obtained. So, it seems that again we are facing a circular problem here: the conclusions drawn and used to define the status of DMs are based on a particular set of forms that varies from more prone to appear in oral discourse to more prone to appear in written discourse, depending on the researcher's approach. Therefore, as with prosodic features there is not enough evidence that the assertion that DMs are predominantly features of oral discourse can be valid as a general criterion for their identification. Restricted to sentence initial positions. Another feature often mentioned as a possible common characteristic of DMs is their occurrence in sentence or discourse unit initial positions. For Keller (1979) this is one the defining features of DMs, but as other researchers have pointed out (Brinton, 1996), this is one of the least general features, as the exceptions to this prototypical appearance are countless. In particular and general studies of DMs, it has been proved that the same DMs may appear in initial, medial and final positions in a discourse unit. For example, it seems that the placement of a DM such as like is particularly flexible. Among other places, it may appear preceding a whole interrogative proposition, or in the middle of a unit before a numeral expression (Andersen, 1998; Miller and Weiner, 1995). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 29 Additional examples of the exceptional character of this feature are provided by those studies that claim the existence of a type of DMs which only appear in final positions. In their study of Brazilian Portuguese DMs, De Oliveira and Tavares (1992) mention a particular group that only occurs at the end of utterances, such as ne? ‘right?’, sabe? ‘y'know?’ and entendeu? ‘do you understand?’. All three are categorized as markers that express request for feedback, and that distributionally they appear only at the end of the utterance they accompany (:236). Vincent (1993) and Vincent and Sankoff (1992) describe a particular type of DMs which they call punctors. These appear at the end of the unit they belong to and show complete prosodic assimilation to the preceding phrase. Among them the French la, osti, hein, spa, etc. Finally, there are some authors that suggest precisely the opposite feature as one of the defining characteristics of DMs. Martin Zorraquino (1994; 1998) suggests that one of the syntactic properties of DMs is their distributional flexibility, as they may appear in initial, medial or final sentence positions. She attributes this mobility to the DMs’ essentially external relation to the sentence structure. Nevertheless, although she proposes this flexibility in distribution mainly as an essential feature of attitudinal sentence adverbs such as evidentemente ‘evidently’ or claro ‘of course’, she later recognizes certain restrictions such as grammatical category or discourse coherence that apply over this distributional freedom. Thus, it seems clear that there is enough evidence in the literature on DMs to disregard the claim that they appear in sentence or discourse unit initial position and that their distribution is an essential feature for the identification of their status in the language. There is little doubt about the possibility that at least some DMs can appear in other positions, such as medial or unit final. For example, there is some Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 30 literature that describes a particular group of DMs based precisely on their appearance in final position in their discourse units. Finally, there are contradictory arguments in the literature with respect to the nature of the distribution of DMs. Some researchers have suggested that it is precisely their distributional freedom the feature is one of the main characteristics of DMs. Again, as in the previous section, it is necessary to pose the question of whether all this confusion is a product of the particular set of forms analyzed in every case. Every group of researchers focuses on a certain group of DMs, and from that limited group they extract a series of characteristics that are true for that particular group. The problem comes when these characteristics are used as general features for forms other than the ones originally studied. In many cases, these features only apply to the DMs they were drawn from. DMs may co-occur in the same sentence. The final characteristic that I will review here that has been proposed as a defining feature of DMs is the co-occurrence of several DMs in the same unit. These combinations have been called DM clusters or stacks of DMs (Andersen, 1990; Andersen 1996). Although almost all researchers would agree that DMs may appear combined with other DMs (Fraser, 1990; Brinton, 1996 among others), to this date there is not much investigation on the conditions, order of combination or frequency of co-occurrence of particular forms. As Fraser (1990) points out, an important question still unanswered is how these forms combine in general. For example, he points out that there are clear preferences in the ordering of stacks of DMs 'there is clearly some preference among alternative orders of occurrence, e.g. Well, anyway, let's... rather than Anyway, well, let's' (Fraser, 1990:395).An additional question with respect to the ordering of stacks of DMs is whether these choices are uniquely Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 31 controlled by speaker preferences or whether in some cases these alternative choices involve the expression of different contents. One could easily imagine certain contexts in which the least preferred alternative suggested by Fraser (1990) is still plausible. The admission that it is possible to use least-preferred combinations of DMs in different contexts would constitute, along with the fact that DMs are not interchangeable, an additional argument to accept that DMs in fact bear some meaning. This point would lead us to an important conclusion, which is that DMs might not be different from other lexical categories in the language. In some respects they may still behave as lexical items. That is, lexical categories such as nouns, verbs, adjectives or adverbs may have more than one instance occur in a single sentence. The main requirement for this to happen is semantic compatibility, that is, the unit in which they appear needs to have a global coherent interpretable meaning in some possible context. The semantic compatibility of lexical items is determined by their specific semantic features, as well as by the context they appear in. Two lexical items from the same or different category may not co-occur in one sentence unless they are coherent. For example, the adjectives ‘heavy’ and Tight’ express incompatible meanings in every possible context, and therefore they may not co-occur in the same constituent, as in the example in (4): (4) The desk is heavy and light Obviously, further studies on the co-occurrence of DMs and their distinction in this respect from other lexical items are needed to prove the veracity of these claims. But given these initial premises, it seems that DMs do not necessarily behave Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 32 differently from other lexical items in the language. Therefore, it seems safe to argue that the fact that there are co-occurrence restrictions among DMs is not a defining feature for the category of DMs. It does not seem to be an exclusive characteristic of DMs, and consequently it is not relevant to prove the nature of their status. 2.2.I.4. Summary. In this chapter I have shown that the assumption that features alone or in combination can serve as criteria to identify a DM category presents problems. For each of the features there is enough evidence in the literature that proves that they fail to be shared by all expressions called DMs. We saw that structural features widely accepted in the field, like sentential exclusion, or the origin of DMs, both present many exceptions among DMs. In the same fashion, there is no conclusive evidence from the studies available on DM intonation to sustain that prosodic features are a defining characteristic of the status of DMs. Finally, distribution features (oral discourse elements, sentence initial position restrictions and co occurrence within the same utterance) offer again a partial description of the behavior of these expressions. Our discussion has also shown that the problem of the definition of the properties of DMs lies on the way the data analysis has been approached. The conclusions drawn by the analysis of DMs must be necessarily partial, since the group of expressions studied are very reduced and arbitrarily chosen by each researcher. Therefore it is almost impossible that characteristics extracted Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 33 from a limited and ‘a priori’ (with no clear definitional criteria) selected set of expressions are general enough to be applied to a hypothetical entire category. The need for a different methodological approach is evident. The one endorsed in this dissertation is based on the idea that DMs must be studied as discourse units, independently of the particular linguistic expressions realized as DMs, and as part of a general discourse framework, following independently defined cirteria. This theoretical perspective becomes apparent in the next chapter, where I present several models of discourse based on the notion of coherence. 2.3. Functional approach to the notion of DMs. 2.3.1. The meaning o f DMs. The meaning of DMs has often been included as a structural attribute for the characterization of DMs. Independently from the particular view researchers adopt for the study of DMs, there is general agreement that DMs have no propositional meaning and that the meaning they bear is basically procedural (Brinton, 1996; Schiffrin, etc.). In my opinion, this observation is crucial in order to distinguish a possible class of units in discourse. In particular, the consideration that DMs bear only procedural meaning allow us to consider the meaning of DMs as a functional feature rather than a structural one. This feature alone can help us distinguish a class of units in discourse that I will call Discourse Functional Units which include the majority of forms usually treated as DMs. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 34 , but procedural one. For some authors, such as Schourup (1985) and Schiffrin (1987), some DMs are supposed to encode a core of invariant meanings derived from their contextualized interpretation in discourse. Others simply do not have any semantic meaning. For example, the marker so always involves a notion of necessary result from the content of a previous unit, but well and oh are semantically empty. This minimalist perspective is based on theories of the meaning of conjunctions (and, but, or), which are semantically defined in truth-conditional terms (Gazdar, 1979; Kempson, 1975). Not everyone seems to agree on this view, however. The major criticism to comes from the fact that the theoretical foundations of this assumption are not clearly defined and in some cases it is far from obvious that a DM has no meaning, as it is proved by the simple fact that DMs are not interchangeable (cf. Redeker, 1991:1164). A different perspective is the one taken by the maximalist approach, which claims that DMs have several identifiable meanings. According to this approach, DMs are polysemous forms that have a particular meaning in every context. In this view, there is no limit to the number of possible meanings of a word and the task of the linguist is to isolate and identify every one of them (Cf. Hansen, 1998). Finally, a third alternative that shares many characteristics with the minimalist approach is the ‘polysemy’ approach (Hansen, 1998; Koning, 1991; Kroon, 1995). In this perspective, words may have different senses but instead of suggesting that every sense constitutes a different meaning as the maximalist approach would, these authors propose an alternative account closer to the minimalist view, which is that all the meanings of a word are related throughout a core representation: ‘these various senses are related in an often non-predictable, but nevertheless motivated way, either Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 35 in a chain-like fashion through family resemblances, or as extensions from a prototype’ (Cf. Hansen, 1998: 241). It seems clear that for all researchers, minimalist, maximalist or polysemic, the investigation focuses on the types of non-propositional meanings DMs have. That is, DMs do not have propositional meaning but they have another type of meaning: essentially metalinguistic. Maschler (1997:192) for example proposes that in order for a form to be identified as a DM it must have a metalingual interpretation, that is, it ‘must refer metalingually to the realm of the text and not to the extralingual world’. Fraser (1996:186) suggests that DMs contribute to the procedural meaning of the discourse they are inserted in, that is, they help the speaker interpret the messages attached to them. Hansen (1998) also incorporates this characteristic onto her definition of DMs: T define markers as linguistic items which fulfill a non- propositional, metadiscursive (primarily connective) function’ (:236). In my previous work I called this meaning communicative meaning (Romera 1998), due to the fact that it can only be interpreted in communicative events. DMs have metalinguistic content, since they are only interpretable in the realms of the act of communication; outside of it they have no referent. Speakers use DMs to react to communicative situations or to express relations within one’s own speech. DMs may express contrast, conclusion, elaboration, explanation or a purely interactional meaning/function to maintain the communication. It is precisely this part of the characterization of the meaning of DMs which in my opinion is the only relevant feature concerning the status of DMs. I claim that the metalinguistic or procedural meaning of DMs is the only ‘structural’ feature that successfully offers a principled basis for the characterization of the category of DMs. The relevance of this feature comes from the fact that it can be understood not only Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 36 as a structural feature but also as a functional one; that is, it is difficult to separate the functional side of the procedural content of linguistic items from the structural side. In linguistic terms the distinction between propositional vs. procedural meaning involves a differentiation with respect to the reference the item invokes. The expression of propositional content implies a reference to objects, individuals and states of affairs in the real world or in the world represented in the discourse, whereas the term procedural refers to the expression of metalinguistic knowledge, that is, the reference to some information regarding the interpretation of ‘the text or the interaction between its participants’ (Cf. Maschler, 1991:192). Therefore, rather than providing content-based information, DMs offer information on how to interpret a piece of discourse, that is, they offer functional information. So, when talking about the procedural content of a form we are not only talking about a structural feature alone, but of a functional one as well. This double nature of this feature is reflected in its treatment in the literature. Traugott (1982; 1989) offers evidence of how functional properties of lexical items become part of their permanent meaning. For example, what starts by being inferential or functional implicatures such as preference or concession ends up being a permanent interpretation of certain linguistic expressions. This is illustrated by the development of why from a question to a complementizer to the hearer-engaging why of such expressions as: I f you have any trouble reaching her, why, just feel free to call me. (Cf. Traugott, 1982: 254) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 37 Additional examples are seen in some of the so-called ‘functions’ of DMs. Brinton (1998), for example, mentions that among the commonly agreed functions of DMs are those of ‘subjectively, to express a response or a reaction to the preceding discourse or attitude towards the following discourse’; and also ‘interpersonally, to effect (sic.) cooperation, sharing, or intimacy between speaker and hearer, including confirming shared assumptions, checking or expressing understanding, requesting confirmation, expressing deference, or saving face (politeness)’ (:38-39). These functions are not very different from the hearer-engaging function/meaning of why described by Traugott, and they could perfectly be characterized as the procedural meaning of these forms. The application of this functional distinction in the characterization of discourse elements seems to be an appropriate venue to establish independent bases to identify DMs. It allows us to differentiate categories in discourse applying criteria with independence of the particular forms analyzed. The proposal defended here is to consider types of categories in discourse based on the type of meaning they represent. On the one hand, elements that express propositional meaning would constitute LEXICAL categories. On the other hand, elements that bear procedural meaning, rather than propositional would be FUNCTIONAL categories. The distinction between LEXICAL and FUNCTIONAL categories is not new in linguistics; most to the contrary, it is shared in other areas of analysis. For example, Fukui (1986) and Fukui and Speas (1986) assume a general distinction within the classification of the entries in the lexicon: categories which bear certain semantic features and categories that do not (Op. Cit. 129). The first type is a general class of items called lexical categories; all the lexical entries in them can be defined in terms of +/- Verb, +/- Noun. The second are functional categories, since they also Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 38 appear as independent lexical entries, but have different status from the lexical ones. According to these authors functional categories are ‘closed-class items’ and lack the sort of semantic value associated with lexical categories. (Op.Cit.: 133). Determiners and complementizers are among the elements analyzed as functional categories under this framework. The similarity between the classificatory principle of lexical entries and the explanation of the meaning of DMs versus other elements in discourse is rather obvious, the elements in one category lack propositional content vs. the elements in the other which have it. Some authors have even suggested a common explanation of functional units at different levels. Within Relevance Theory, Escandell (1997) proposes the analysis of determiners as elements that bear certain instructions to the listener on how to interpret the constituent they accompany. This analysis is based on the identical premises that define DMs in this framework. Elements that transmit instructions on how to interpret an utterance. The theoretical choice of considering elements in discourse strictly in functional terms leads to the re-examination of the category of DMs. I propose the coinage of another term: discourse functional unit. The new category of discourse functional elements constitutes a broader category than the composed of forms traditionally considered DMs. On the one hand, Functional Elements need not be limited to certain contexts, as long as they exhibit the same functional role they must be considered as pertaining to the same category. For instance, in the example below, the connective pero ‘but’ is a functional unit in both cases, since it performs the same function: expressing the relation between two units. However, the first case of pero ‘but’ in line 3 would not be considered as a DM according to traditional views. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 39 (5) <H2> 1 y si no... 2 £que te... que te puede pasar si te pillan en esas cosas? <H1> 3 ipff! pues un multazo pero... de ordago. <H2> 4 <simultaneo> Y entonces, ^por que lo haces? <H1> 5 yaellos... </simultaneo> Y aellos... 6 pues porque no habia mas manera de demostxar quien... <H2> 7 pero, en realidad, ^quien tuvo la culpa? <H1> 8 el. (5’) <H2> 1 and if not... 2 what can happen to you if they catch you in those things? <H1> 3 ipjf! well a big fine but... a huge one. <H2> 4 <simultaneous> and then why do you do it? <H1> 5 and to them... and to them... <lsimultaneous> and to them... 6 well because there was no other way to prove who... <H2> 7 but actually who was at fault? <H1> 8 him. (CREA. F19L55) On the other hand, depending on the notion we adopt for the definition of discourse functions, we find that some forms considered DMs in certain frameworks are not functional categories. In this dissertation, as we will see next, I consider functions in discourse to be tightened to the notion of relations to establish coherence. Thus, Modal Particles (Helbig, 1988; Rudolph, 1989) for example, are excluded from the category of functional elements, since they are part of the Elocutionary meaning of an utterance, but not used in the expression of coherence relations. Given the problems discussed above with the notion of discourse marker and working exclusively under functional principles, it seems more appropriate to abandon the concept of discourse marker (or maybe at least the term) and focus instead on the concept of discourse functional unit. In the next chapter, I will Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 40 concentrate on the discussion of a model of discourse capable of offering an account of functions in discourse. 2.4. Conclusions. As one can see there are still enough theoretical and practical problems in the study of DMs. At the beginning of this chapter, we saw that due to the huge variety of expressions analyzed as DMs, the great terminological diversity and the many different perspectives from which DMs are studied, finding unitary agreement with respect to what we understand by a DM is almost impossible. Several groups of structural features have been proposed in the literature. In this chapter I reviewed the most frequently used by researchers in the field of discourse analysis: structural features (sentential exclusion), prosodic features (preceded and followed by pause, especial intonation pattern) and distributional features (restricted to oral discourse, sentence intial positions, and co-occurrence with other DMs in the same utterance). We saw that each one of these features leads only to partial characterizations of DMs. After an exhaustive analysis of every feature, there is enough evidence in the literature to reject the assumption that they may serve as general criteria to identify DMs. It was shown that the main problem lies on the fact that only a series of ‘a priori’ selected groups of expressions have been the bases for the analysis of these features. This creates a serious theoretical problem: the definition of DMs is not independent from the particular set of forms chosen to be studied, therefore it is unlike that it can be applicable to a different set of forms. A second approach to the definition of DMs is the analysis of these expressions from a functional point of view, as part of a general discourse framework. This Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 41 approach avoids entirely the circularity problem since it provides independent criteria for the study of these forms. In this respect, I have proposed to focus on the a category of DISCOURSE FUNCTIONAL UNITS instead of the fuzzier group of DMs. This category would be composed of those forms whose main function in discourse can be analyzed in terms of the relations they make explicit. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 42 CHAPTER 3 Discourse models. Discourse units. 3.1. Introduction. After reviewing in the previous chapter the problems with the theoretical notion of discourse marker, it is clear that only a functional approach is capable of offering a ‘non a prioristic’ account of this forms as a discourse category. The definition of the notion of discourse functional elements must be made as a part of a discourse framework in which we are capable of independently identifying the functions these units perform. This chapter is devoted to the presentation of the discourse model followed in this dissertation. I share the idea that discourse is composed by a series of units which relate to each other in meaningful ways, that is through a series of relations. Section 2 focuses on the notion of discourse structure and discusses different models based on the notion of coherence. Section 3 presents the general framework followed in this dissertation under the premises of discourse relations. Based on the distinction between categories suggested before, I distinguish two types of basic elementary units in discourse: A) Discourse Propositional Units which bear propositional and pragmatic content and which relate to one another through a series of relations. The Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 43 combination of these units represents the structure of discourse and creates the patterns by which a hearer is able to interpret a piece of coherent discourse. And B) Discourse Functional Units, which in agreement with a theory that understands discourse structure in terms of coherence relations, are taken to be those units which have only a functional value within the structure of discourse, and that is to make explicit the specific relation through which the hearer obtains a coherent interpretation of two discourse units. Chapter 4 deals in its totality with the concept of discourse functional units. 3.2. Models of discourse structure. The majority of the models that offer an account of these forms from a functional perspective are found under the premises of discourse coherence1 . In all of them, discourse is accounted for through a series of interrelated modules or components for which different functions or relations become more relevant. The notion of coherence itself is not unproblematic, and as we will see, it is precisely the different entailments of this notion that determine wider or narrower definitions of 1 In addition to the models based on notions of coherence, several approaches in the literature have also offered general definitions of DMs within a global discourse framework. Within the studies on the English language, two are the most relevant and productive models: the grammatical-pragmatic perspective that defines DMs as one class of pragmatic category (Fraser, 1990; 1996; 1997); the accounts within the Relevance Theory (Blakemore, 1979; 1992) which treat DMs as a type of expression that conveys only procedural meaning in discourse, that is, DMs express only instructions on the interpretation of the conceptual representation of utterances. With respect to the study of discourse in other languages, the most common approaches include the grammatical-pragmatic perspective (Martin Zorraquino, 1994; Martin Zorraquino et al. 1998; Silva Corvalan, 1999) and the study of discourse from the Argumentation Theory perspective (Anscombre and Ducrot, 1983), which considers that language is used to construct argumentative strategies and that connectors articulate two members o f an argumentative unit. This theory is particularly fruitful in the study of connectors in French (Anscombre and Ducrot, 1977; Ducrot et al. 1980; among others) and in Spanish (Portoles, 1998; among others). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 44 what functional units are. Although coherence is a notion that intuitively needs no explanation (coherence means “making sense of a piece of spoken or written discourse”), unfortunately it cannot be that easily defined in formal terms. One may ask why it is important to formally know what coherence is. The reason probably lies in the fact that depending on what is understood by this notion the range of linguistic forms that are considered to contribute to the “making sense of the text” may be wider or narrower. More precisely, the different entailments of this notion determine wider or narrower conceptions of what DFUs are. 3.2.1. Halliday and Hasan. In the pioneer coherence model of Halliday and Hasan (1976) (H.&H. here on), one of the four mechanisms analyzed to establish cohesive relations in discourse is what they call ‘conjunction’. Conjunction creates cohesion through “a specification of the way in which what is to follow is systematically connected to what has gone before.” (H.&H.:227). The mechanism of conjunction may be expressed through a range of different structural ways. More specifically, semantic relations may be expressed through the use of 'conjunctives ’ , 'conjunctive adjuncts' or 'discourse adjuncts'. These include adverbs and compound adverbs {furthermore, nevertheless, anyway, instead, besides, therefore, thereupon), conjunctions {but, so, then, next), prepositional phrases {on the contrary, as a result, in addition) and prepositional expressions {as a result o f that, instead o f that, because o f that) (Op.Cit.:231). These conjuncts express several relations in discourse, characterize by H.&H. as additive, adversative, causal and temporal. Interestingly enough, the final Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. section of H&H’s chapter on conjunction is dedicated to other 'conjunctive items'. A series of different forms that will later be called DMs are included here. The common criterion to group these forms together is their continuative function in discourse, i.e., to continue with the discourse in similar or different fashion. This mechanism may be expressed through a range of different structural ways. More specifically semantic relations may be expressed through the use of 'conjunctives ’ , 'conjunctive adjuncts' or 'discourse adjuncts'. These include adverbs and compound adverbs, ‘furthermore, nevertheless, anyway, instead, besides, therefore, thereupon, conjunctions ‘but, so, then, next’, prepositional phrases ‘on the contrary, as a result, in addition’ and prepositional expressions ‘as a result of that, instead of that, because of that’(Op.Cit.:231). These conjuncts express several relations in discourse, characterize by H.&H. as additive, adversative, causal and temporal. Interestingly enough the final section of their chapter on conjunction is dedicated to other ‘conjunctive items'. A series of different forms that later will be called DMs are included here. The common criterion to group these forms together is their continuative function in discourse, i.e. to continue with the discourse in similar or different fashion. The forms analysed are, now, o f course, well, anyway, surely, after all. According to H.& H, each of these also expresses a particular type of semantic relation. Now opens a ‘new stage in the communication’ (:268). O f course, means ‘you should have known that already’ or ‘I accept the fact’ (:269). Well, means ‘ I acknowledge the question, and will give a considered answer’(:269). The meaning of anyway varies between dismissive and resumptive with respect to the preceding sentence (:270). Surely on the other hand ‘invites the hearer to assent to the proposition being enunciated’ (:270) and after all means ‘after everything relevant has been considered, what remains is...’ (:270). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 46 It is absolutely clear that in H.&H.’s classification of conjunctive forms there is a clear distinction between a more formalized type of conjunctions such as additive, causal, adversative or temporal relations, and other ‘looser’ types of relations, such as accepting facts, acknowledging questions, inviting the hearer to assent, or dismissing a previous part of discourse. In the first case, attention is put in the identification and explanation of a particular semantic relation, while in the second case, there is no systematic semantic relation to be explained. Rather the relation, if any, comes from the study of particular forms. Additionally, the very basic postulate proposed for the characterization of conjunction as a cohesive force in discourse does not seem to hold in the case of continuative forms. An important point that H.& H. make about the completeness of the set of structures representing each relation is that the real forms do not seem to matter that much because the cohesive relation is not achieved through the conjunction expression but on the contrary, It is the underlying semantic relation [...] that actually has the cohesive power. This explains how it is that we are often prepared to recognize the presence of a relation of this kind even when it is not expressed overtly at all. We are prepared to supply for ourselves, and thus to assume that there is cohesion even though it has not been explicitly demonstrated. (:229) This point may hold for relations like addition, adversativity, causality, etc. but if forms like well, o f course, surely were absent from a piece of discourse, would the listener be able to supply such a relation on her own? Take example (1) from H.&H. (:270) used to illustrate the use of ‘surely’ (1) They’ll think you’re serious. Nobody could be so stupid as to think that, surely. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 47 Now observe (5') in which the form has been taken out from the sentence. (! ’) They’ll think you’re serious. Nobody could be so stupid as to think that. It is clear that once surely is not there anymore, the underlying semantic relation, if any, disappears with it. The same can be said about well, anyway, after all or now. The point is that from the very first beginning in the study of conjunctive forms, different forms have been grouped together as performing the same function, when in fact they might not have much in common other than the fact that are not easily classifiable. 3.2.2. A wide notion o f coherence. Schiffrin’s model. An inheritor of H. & H.’s model is Schiffrin’s (1987) model of discourse. This researcher’s notion of discourse coherence is the widest and the most inclusive of all coherence models. Schiffrin offers an integrative approach in which coherence entails all the factors that intervene in a successful (mainly oral) communication. In fact, coherence in discourse is equated to success in communication and for it to happen not only linguistic factors but also other non-verbal contextual factors must be taken into consideration. Following Gumperz’s (1982, 1984) integrative view of interpretation of communicative meaning, Schiffrin advocates for a model of discourse coherence in which verbal (prosodic, phonological, morphological, syntactic, rhetorical) and not verbal (kinesic, proxemic) aspects of communication must be interpreted to achieve successful communication. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 48 This wide notion of coherence leads Schiffrin to propose a model of discourse in which several different sources of information are included. She devises a discourse framework composed of five planes. Three of these planes include linguistic information, in the sense that they offer information related to the message and how this is developed. These are Ideational, Action and Exchange structures. As she explains, her model focuses mainly on local coherence, that is. ‘coherence that is constructed through relations between adjacent units in discourse’ (Op. Cit.:24). In formation in the Ideational structure refers to the relations between semantic units. Information in the Action structure refers to relations between speech acts. And information in the Exchange structure refers to the relations between turns or adjacency pairs. Besides these three planes, two other structures incorporate additional sources of information. Participation framework provides pragmatic information, in which the term pragmatic “concerns the relation of language to its users” (:28) and Information state provides information about the management of knowledge and metaknowledge. In the Participation framework relations between the speaker and the hearer and between the speaker and the utterance are at stake. Information state concerns the assumptions of speaker/hearer’s knowledge and the importance of those assumptions for the production or the interpretation of current discourse. Within this model Schiffrin then defines and introduces the term discourse marker as indicator of the location of utterances on each of these particular planes of talk. (:24 and 326). DMs may function at one or at several of these planes at the same time. The introduction of these different sources of coherence in Schiffrin’s model is what makes it possible to consider as DMs expressions whose main function is played at the participation framework and the information state planes. It is the case Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 49 of oh! or y ’ know for example. These expressions, as we will see next are not considered DMs by other researchers. The nature of all these sources of information may present a problem for the formalization of the model. As some researchers have noticed (Redeker 1991), there exist noticeable differences with respect to the status of each plane. On the one hand, ideational, action and exchange planes are organized around units: semantic units, turns and speech acts. But on the other hand, no specific unit defines relations in either the participation framework or the information state. Given that local coherence is defined by Schiffrin as the relations of adjacent units, it is not clear how these five planes contribute equally to it. It is clear that ideational, action and exchange structures are able to provide information about units in discourse (turns, speech acts, etc.), but participation framework and information state operate on a different level. The types of relations established in these modules are of a different nature, they do not develop between linguistic units but between speaker and hearer, speaker and utterance or between assumptions about common knowledge. These relations seem to have a scope wider than any two particular units, operating over the whole process of interaction. This is why it seems that expressions which function primarily at the participation framework and the information state are somehow different from forms whose main function is realized at the other three planes. In accordance to this idea other researchers have suggested a tighter notion of coherence that restricts the discourse model and also the set of expressions that act as DMs. This is the case of Redeker’s model (1986; 1991) which I present next. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 50 3.2.3. Redeker’s model. Redeker (1990; 1991) presents a model of discourse based on a more restricted notion of coherence. As for Schiffrin, for Redeker coherence is realized as a set of relations between units in discourse. But unlike Schiffrin, Redeker establishes only three sources of information. A semantic one, basically equal to the Schiffrin’s Ideational structure, a pragmatic or rhetorical one, similar to Schiffrin’s Action structure, and then what she calls sequential source of coherence, which roughly corresponds to Schiffrin’s exchange structure. Redeker’s model of discourse is therefore composed only of three components: semantic or ideational, pragmatic or rhetorical and sequential. In her view, the components which incorporate speaker’s attitudes and the cognitions in the exchange of information contribute to coherence ‘only indirectly’, since they only motivate the speaker’s choices at the pragmatic planes’(1990:l 162). Of the three components, the first two types of relations (semantic and pragmatic) receive preferential treatment and clearer definitions, while the sequential component seems to be defined as a default or a subtype category. On the one hand, Ideational relations are given when the context ‘entails the speaker’s commitment to the existence of [a] relation in the world the discourse describes’ (Rederker, 1991:1168). Rhetorically related units are those in which the relation is given ‘between the illocutionary intentions the convey’ (1991:1168). Sequential relations, on the other hand, are defined more loosely and seem to implicate a lesser degree of coherence. Sequential relations are given ‘when two adjacent discourse units do not Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 51 have any obvious ideational or rhetorical relation -while still being understood as belonging to the same discourse’ (Redeker, 1990:369). As one can see, sequential relations seem to have a secondary place in the model: whenever no ideational or rhetorical relations are identifiable, then sequential relations are given. In Redeker (1991) the concept of sequential relations gets a more developed explanation. This time the sequential component turns into more of a sub-type category: ‘sequential transitions are paratactic or hypotactic relations between ideationally and rhetorically only loosely related adjacent discourse segments.’ (1991:1168). This new definition does not clarify much the notion of sequential relations itself, but seems to imply that these relations are just a categorization of ideational and rhetorical relations. These latter relations can be of two types: either strong or loose; if they are loose then they are sequential. Obviously, it is not so clear whether this is the right way we should be interpreting sequential relations. Therefore, as in the case of Schiffrin’s Participation framework and Information State, the role of the sequential component in Redeker’s model of discourse needs further explanation. In what way sequential relations are different or maybe similar to the other two types of relations remains unanswered. Within this model, discourse operators (as Redeker calls them) are defined according to each of the components. ‘A discourse operator is a word that is uttered with the primary function of bringing to the listener’s attention a particular kind of linkage of the upcoming utterance with the immediate discourse context.’ (1991:1168). The types of linkages that a DM may establish are within the scope of each of the planes: semantic, rhetorical and sequential. This more restricted vision of Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 52 discourse structure with only three sources of information instead of five allows Redeker to consider DMs as somewhat less open range of expressions. For example, she excludes what she calls ‘clausal indicators of discourse structure’ (let me tell you a story, as I said before..) (1991:1169), deictic expressions if their use is other than anaphoric (the case of now, here, today). And DMs as y ’ know and I mean are considered as such only if they are not syntactically and intonationally integrated into the utterance( 1990:1166). As we can see, these two coherence-based models identify DMs or discourse operators according to the function these expressions played at the different components. But one important distinction between the two models concerns the independent definition of the components. As Redeker (1991) points out, the discourse structures must be independently defined from the functions that these units play. This implies that coherence relations are realized with or without a DM being present. This is an important question that places DMs on a different perspective. DMs only realize explicitly unit relations that exist regardless of their use. This is the perspective that Redeker’s model assumes different from Schiffrin’s, and the main concept around which Rhetorical and Coherence Relation Theories are developed. 3.2.4. The Rhetorical and Coherence Relation Theories. Under the generic name of Rhetorical and Coherence Relation Theories I have grouped a series of discourse models which work under the premises that discourse can be divided into compositional units which hold together by a series of Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 53 relations. This point is also shared by the other three models seen so far: relations between discourse units make coherence possible. What makes discourse relations theories different is that the emphasis is placed in the definition of such relations. (Grosz and Sidner, 1986; Mann and Thompson, 1988; Sanders et al., 1992; Hobbs, 1990;1995;). Whereas in Schifffin's model the components of the discourse structure are given primarily to account for the use of DMs, in rhetorical models discourse functional units are studied only as mechanisms that make possible the explicitation of the coherence relations. These coherence relations stand between discourse units regardless of the DM's presence. This point is also shared by Redeker's model, but whereas for Rekeder the components of the discourse structure (Semantic, Pragmatic and Sequential) still play an important role to explain how these forms are used, for some rhetorical models the definition of each specific relation is the only important point for the theory. In some cases these studies provide long lists of relations often divided into groups that classify them according to a specific criterion. For example, in Hobbs (1995) some relations express 'a coherent set of states or events in the world', such as temporal or causal relations. Others serve some discourse plan the participant has in mind: these are evaluation relations. Others provide a necessary linkage it the text, for example background or explanation relations. And finally others 'manipulate the listener's inference process to lead him to the correct interpretation' (Op.Cit.:41), for example elaboration and contrast. In other cases like Sanders et al. (1992), the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 54 relations are organized into a system of four cognitive-based primitives: basic operations -additive and causal; source of coherence -semantic or pragmatic; order of the segments -basic and non-basic; and polarity -positive or negative. Other studies are much more conservative in the definition of relations, providing only a general binary division to classify them. These theories are what Hovy et al. (1993) call the parsimonious theories (Grosz and Sidner, 1986). The division is made only between hypotactic and paratactic relations, in which units relate in terms of dominance. In hypotactic relations a satisfaction-precedence relation is given and one unit is at a lower level than the other. In paratactic relations both units stay at the same level. A further distinction between these theories and the ones interested in defining every relation lies on the fact that dominance-based divisions place only intentional constraints in the distribution of the relations, but do not specify semantic meanings for them. As we can see, in most of these studies there is no specific mention of discourse components as was the case in Schiffrin's or Redeker's models. The only possible components are the groups constituted by the principles underlying the different types of relations. Each of these principles can be considered a specific source of information that contributes to coherence. Within this framework, some authors see connectives (as they are usually called by these researchers) as explicit realizations of a relation. That is, connectives play their role only in the interpretation of a relation: they make explicit a relation from a set of possible inferrable relations Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 55 between two units. For some authors the only necessary constraint connectives must meet is that their meaning should be compatible with the meaning of the relation given between the two units (Sanders et al., 1993:94). The important advantage of all these theories is that the framework for the study of these units is created independently from their use. The types of relations established in discourse are not motivated by the presence of a connective or any other linkage word. Relations between units and not the use of a discourse connective is what constitutes the bases for coherence. Once the framework is delimited by itself, connectives turn out to be functional units that make explicit a particular rhetorical or coherence relation. This is the most important premise in which the discourse framework used in this dissertation is based. 3.3. A Discourse model based on coherence relations. 3.3.1. Coherence relations. As we just saw, segments in discourse are interconnected in such a way that the content of a part is not only relevant because of the information it provides per se, but also because it may be essential for the interpretation of previous and subsequent material. In some cases, a discourse segment may provide the background information to interpret the next segment. In other cases, several units together may form a structure of a list of narrated events, and yet in other cases one or more units may provide additional information about some element of the previous discourse. Schiffrin provides a clear illustration of a case in which several Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 56 units constituted the background information needed for the interpretation of the upcoming unit. This was Henry’s turn of speech, composed of units (d)-(g), repeated below in (2): As she explains, the information in (i) and (ii) ‘provides background conditions that enable the EVENT in (iii). After introducing a new referent (i), [the speaker] presents an event [...] that makes possible an outcome: going to a big specialist (ii) necessarily preceded the wrong diagnosis (iii).’ (Op. Cit: 87). In example (3), Ioffer another case in which units (ii) and (iii) constitute the elaboration on the content the speaker has uttered in (i). (2) PRIOR TEXT BACKGROUND (new referent) sometimes it works (i) there’s this guy Louie Gelman, BACKGROUND (ii) he went to a big specialist, (enabling event) ENABLED EVENT (iii) the guy... analyzed it wrong (Schiffrin, 1997:87) (3) STATEMENT (i) el sentido de la responsabilidad en e l ... el trabajo ELABORATION (ii) tratar de mejorarlo (iii) y no conformarme con cualquier cosa. (3’) STATEMENT (i) the sense o f responsibility in your in your job ELABORATION (ii) trying to improve it (iii) and not just being happy with anything. (CREA. Conv. 20) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 57 Units (ii) and (iii) elaborate on the content of unit (i): what ‘sense of responsibility’ means for the speaker. It means always demanding more from yourself (not being satisfied just with any results) and therefore always trying to improve what you do (ii). Thus, (2) and (3) are examples that show how the structure of discourse is organized around certain relations that make the units hold together and make speakers/hearers interpret what others want to communicate. The definition of discourse in terms of relations is the basic tenet of the studies that belonged to the generic group that I called rhetorical/coherence relation theories. However, within this framework the nature and definition of discourse relations still remains a somehow controversial issue. On the one hand, for some authors, the main objective of a discourse model is to successfully account for the description of the structure of a text. Thus, relations constitute an analytical tool to explain how these are constructed (Cf. Mann and Thompson, 1986; 1987; 1988; Grosz and Sidner, 1986). On the other hand, other researchers suggest that the purpose of analyzing discourse should be something else than the pure description of it as a linguistic object; that is, a plausible discourse theory should be able to explain how discourse is understood and produced by actual users. For example, within this perspective Hobbs (1990), understands relations as the types of inferences a reader needs to draw in order to obtain a coherent interpretation of a text; Dahlgren (1996) defines discourse relations as a part of a theory of discourse interpretation (naive theory) which understands that ‘the hearer build(s) a cognitive representation such Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 58 that the relations among individuals, events, states and other abstract types in the representation correspond with her/his understanding of the way the actual world individuals and events relate’ (:112); and Sanders, Spooren and Noordman (1992), go a step further and try to demonstrate by experimental investigation the psychological nature of relations as cognitive entities. But despite this basic difference with respect to the nature (descriptive tool vs. psychological entity) of relations, more or less both approaches coincide in adopting a general operational definition of coherence relations. Relations are basically binary semantic linkages between a current utterance and the preceding discourse. For Sanders, Spooren and Noordman (1992), a coherence relation is ‘an aspect of meaning of two or more discourse segments that cannot be described in terms of the meaning of the segments in isolation. [...] it is because of this coherence relation that the meaning of two discourse segments is more than the sum of the parts’ (:2). Put in more formal terms a coherence relation is a binary predicate whose arguments are discourse segments. In this dissertation the definition of coherence relation adopted coincides with this operational definition. A relation is a binary predicate which has two discourse segments as arguments, and introduces some semantic aspect between them. The particular modification in the meaning of the combination of the two segments is what determines the types of coherence relations. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 59 3.3.2. Types o f coherence relations. Many studies in this framework have been devoted to the creation of inventories of relations that make possible coherence in discourse. However, as with issue of the nature of the relations, their number and taxonomy is still debated. On the one hand, some studies have proposed long taxonomies of relations (Hobbs, 1990; Mann and Thompson, 1988; Sanders et al., 1992). On the other hand, other researchers have argued that establishing a finite number of relations is an impossible task in discourse and have proposed two general types of structural relations: dominance and satisfaction-precedence, based on the goals and plans the speaker/writer has in mind when building a text (Grosz and Sidner, 1986). But in spite of the different approaches, all these theories agree on the fact that there exists at least one fundamental distinction between the types of relations that can be established in discourse This is the semantic versus pragmatic categorization, as many researchers have pointed out (Van Dijk, 1979; Martin, 1983; Sweetser, 1990; Redeker, 1990, 1991; Sanders et al., 1992). This distinction has also been referred to as ideational vs. interactional categorization (Schiffrin, 1987) or informational vs. intentional categorization (Hobbs, 1996), but as a general definition of this concept, however, we can say that semantic, ideational or informational relations are established between the ideational meanings or propositional contents of the segments related, and that pragmatic, interactional or intentional relations connect the intentions or illocutionary meanings expressed by those segments. Sweetser (1990) offers a clear explanation of both notions. When talking about Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 60 conjoined segments in discourse, she distinguishes three types of possible interpretations: ideational, epistemic and speech-act readings. Consider the following examples extracted from her work (4a-c): (4a) John came back because he loved her. (4b) John loved her, because he came back. (4c) What are you doing tonight, because there’s a good movie on. (Sweetser, 1990: 77) In (4a) the real world causality relates the two propositions. That is, the relation established between them is that of consequence/cause between the two propositional contents. According to Sweetser this sentence can be paraphrased as: (4a=) John’s love caused him to come back. In (4b) the relation is not the real cause of John’s return. Rather, it is the speaker’s knowledge and the conclusion this knowledge led him to that are conjoined. So (4b) could be paraphrased as: (4b’) The knowledge that he came back enables me to conclude that he loves her. In (4c) on the other hand, what is joined is two speech acts. The because clause gives the reason of the main clause speech act. The paraphrasis would be: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 61 (4c’) I am asking you whether you are free tonight because I want to invite you to watch a good movie which is on tonight. These three readings are what Sweetser calls content, epistemic and conversational readings respectively. These examples clearly represent the different ways in which two segments can be related. In (4a) it is clear that the connection between the two clauses is established through the propositional content of each of them. In (4b) and (4c) on the other hand, the connections or relations between the clauses are of a different nature. In (4b) the knowledge of some information enables the speaker to get to certain conclusions; in (4c), the relation is established between the illocutionary acts or the intentions of the speaker. These distinctions are fundamental in most discourse studies; however, the majority of proposals make only one categorization or distinction between semantic and pragmatic readings of DMs. Sweetser’s content readings are considered to be the semantic readings, and her epistemic and speech-act readings are grouped together as the pragmatic readings (Martin, 1983; Redeker, 1990; Sanders et al. 1992; Sanders, 1997): A pragmatic relation refers to the illocutionary meaning of an utterance whereas a semantic relation refers to the locutionary meaning (Sanders et al., 92:8) According to this distinction, researchers in the field have proposed concrete taxonomies of semantic and pragmatic relations. For example, among the most commonly agreed semantic or ideational relations are temporal sequence, cause, result, consequence, circumstance, purpose, condition, comparison, etc. Examples of Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 62 pragmatic or interactional relations are opinion, interpretation, evaluation, evidence, justification, etc. (See Hovy, 1997 for an extensive set of references). As we can see, in the case of relations such as cause or temporal sequence the relation is established through the content of the units. For example, in (5) it is the interpretation of the events expressed by 1 and 2 that make the units coherent. (5) First you open the document, then you make the changes. In cases of relations such as justification, however, it is necessary to count on the speaker’s intentions to obtain the correct interpretation. There is a necessary goal or plan the speaker is following in order for her to introduce justification of a certain part of the discourse. Justification in itself implies the involvement of an speech act on the part of the speaker. Additionally, there is still a third distinction that needs to be made regarding types of coherence relations. Sometimes neither the semantic meaning nor the intentions or beliefs expressed in the segments are sufficient to find out what makes a fragment interpretable. For example, in (16) the units introduced after 6 do not seem to be primarily related with the previous units (3-5), neither by their propositional content nor by their illocutionary meaning: (6) 1 A: te gusta lo que estas haciendo ahora 2 en la universidad? 3B: s i = Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 63 4 si 5 no no esta mal 6 bueno 7 yo llevaba ya= casi 8 casi cinco anos sin dar clases 9 yo= habia estado dando clases 10 un ano ahi en- 11 en un colegio privao [...] 12 y ahora pues 13 otra vez volver (6’) 1 A: Do you like what you are doing now 2 at the university? 3B: yes= 4 yes 5 it=s not it=s not bad 6 well 7 I had been almost= 8 almost five years without teaching 9 1= had been teaching 10 for a year in= 11 in a private school 12 and now so 13 again to go back What speaker B does from units 6 to 11 is to introduce additional optional information that can make his answer that ‘it is not so bad’ more understandable, but this information is not relevant for the understanding of his main argument. He has introduced a digression. This type of relation is what has been called sequential or textual (see also Hovy, 1997). Sequential relations have been considered from different perspectives. Redeker (1990) suggests reserving the term ‘sequential’ for those metalinguistic or purely textual relations in discourse, that is, those relations that lead the transitions to Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 64 a new topic or a new issue in discourse and those ‘leading into or out of a commentary, correction, paraphrase, aside, digression, or interruption segment.’ (Redeker, 1990:369). So basically, sequential relations are either pure textual relations, such as topic shift, or metalinguistic relations which introduce parts in discourse that are only tangentially related to the main argument. Their inclusion in the discourse may not be essential, and in most cases their exclusion does not affect the interpretation of the main argument. Other researchers like Schiffrin (1987) maintain that sequential relations also facilitate transitions between speakers’ exchanges, that is, turns or adjacency pairs (question-answer) in interactional events. For example, in (17) the relation established through pues ‘so= facilitates the transition of the answer in the question- answer pair. (7) A: ^Con quien vas a ir? B: pues(...) con mis padres (7’) A: who are you going with ? B: so (...) with my parents (Personal Corpus) In (7) pues does not establish a relation between semantic or pragmatic meanings. It is basically used to take the turn and answer a question. It seems that these two interpretations of sequential relations are important and since both contribute to the creation of coherence they should be considered Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 65 when studying the construction of discourse. Thus, there is enough evidence showing that discourse relations can be divided into three main categories: semantic, pragmatic and sequential. Each of them establishes coherence through the connection of the different meanings a discourse unit may convey. In this dissertation I use this tripartite categorization of discourse relations: semantic, pragmatic and sequential, as defined above. For the determination of the particular relations that are included in each of these groups I follow basically the taxonomy offered in Hovy et al. (1997) with some variations necessary to account for oral discourse analysis. In this taxonomy there is a first division between ideational, interactional and textual relations on the basis of the meanings we just saw. Then, within each of these primary divisions, more specific relations are defined. The main categories included in the Semantic or Ideational group are: Elaboration, Circumstance, Sequence, Cause/Result, General-Condition. In the Pragmatic group we find Interpretation, Enablement, Antithesis, and Exhortation. Finally, in the Textual group of relations we find Logical-Relation, Joint and Sequence. Each of these categories at the same time also includes several more specific relations. Under the generic label of Circumstance for example we find location, time, means, manner, instrument and parallel-event. I have included modifications in two groups: the pragmatic one and the textual one. In the pragmatic group it was necessary to incorporate a set of ‘interactional’ relations to account for the conversational moves the participants make use of. For example, a common case is represented by a unit which relates to a previous one as a request for clarification; Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 66 that is, this unit represents the need of one of the participants of obtaining clarification on some information just given by the other participant(s). The group of ‘interactional’ relations I used was adapted from Tsui (1994). I also introduced modifications in the set of textual relations. More specifically in the set of Logical- Relations, Joint and Sequence I have included several others which Hovy et al. do not account for. Following Redeker’s (1990) and Schiffrin’s (1987) definition of textual relations I introduced the following relations: commentary, correction, paraphrase, aside, digression, interruption and turn transition. Figure 1 shows the entire list of relations considered. The relations not originally in Hovy’s taxonomy are marked with an asterisk (*). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 67 SEMANTIC PRAGMATIC TEXTUAL Elaboration Interpretation “ “ Commentary Elab-Generality Evaluation *Aside General-Specific Conclusion “ "Digression Abstract-Instrance Enablement “ "Interruption Identification Background Pre-Sequence/ Restatement Antithesis “ "Expressive sequence: Correction Exhortation (ritualistic act for civi Summary Support Joint/Continuation Circumstance Solutionhood Topic Shift Location. Answer Time Evidence Means Proof Manner Justification Instrument Motivation/Explanation Parallel-Event Concession/ *Contrary to Sequence expectation Cause/ Consequence “ "Interactional Purpose Initiatiations: General-Condition informatives: Condition reports Exception assessments Comparative expressives Equative confirmation Otherwise commitment Comparison repetition Analogy clarification “ "Disjunction directives: “ "Adversative command * Contrast advice *List (Addition) Responses: request for confirmation request for information request for clarification request for repetition request (offer, proposal, invitation) agreement/endorsement disagreement/objection doubt Fig.3. 1. Taxonomy of relations as it appears in Hovy et al. (1997). The asterisk (*) marks the new relations introduced by us. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 68 Additionally, for each the three groups also include a distinction between relations that are established at the same level of analysis versus those relations for which one of the members is considered to be more relevant than the other. This is because in addition to recognizing the three main types of coherence, it has also been proposed in the literature that it is necessary to consider the different role a unit bears with respect to the other units to which it is related (Grimes, 1975, Halliday and Hasan, 1976; Martin, 1992). That is, some units relate to other units at the same structural level, whereas others relate at different levels of analysis. I refer here to the distinction between paratactic vs. hypotactic relations. Paratactic relations are those established between units of equal importance, and hypotactic relations are those established between a unit that plays a central role and one that is subsidiary to the role played by the other unit. For example, in (8) both units are equally important; that is, the absence of one of them does not prevent the interpretation of the other, as they both add equally important information to the discourse. In (9), however, unit 3 is structurally dependent on unit 2; the absence of 2 leads to the ‘no coherent interpretation’ of 3. (Mann and Thompson, 1988): (8) 1 [en la terraza se fuma] 1 2[se esta muchisimo mejor en la terraza] 2 (S’) 1 [we can smoke in the balcony] 1 2[it is much more pleasant in the balcony] 2 (CREA, Conv. 19) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 69 (9) 1[ 2 [no, no lo he pensado todavia]2 3[porque tambien estoy con la idea de preparar temas para la oposicion]3 ]1 (9’) 1[ 2[ no, I have not thought about it yet]2 3[because I also have the idea o f preparing exams]3 ]1 Personal Corpus The distinction between paratactic versus hypotactic relations is relevant primarily in terms of practical analysis. It provides us with more criteria to determine the order in which relations are produced. For example, it is often the case that the most relevant information is not produced in the first place but later in the discourse, as in cases in which the first information provided constitutes the background information that allows us to interpret the upcoming units. So initially it may seem that these background units may not have a place to be attached to, that is, they may seem not to be related to any unit in the previous context. Obviously, in these cases the relevance of the unit is given in relation to the upcoming material. The paratactic vs. hypotactic distinction is then the operational mechanism that helps us determine the direction of the relation. To summarize, we have seen that there are fundamental differences in the way relations are established in discourse. The most important distinction corresponds to semantic vs. pragmatic types of relations. This distinction is made on the basis of the different nature of the meanings related. Two segments are connected through a semantic interpretation if it is the relation between their propositional contents what makes a Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 70 coherent interpretation possible. Two segments are said to have a pragmatic reading if the relation is established between the attitudes or beliefs of the speaker towards the reality of the events expressed by them. Finally, there is also evidence of the existence of a third type of relation, which is the sequential or textual one. Sequential relations can be considered to introduce argumentation mechanisms in discourse that express transitions to a different topic or a commentary (Redeker, 1990) and facilitate the transitions between speakers or turns and/or transitions to a different transaction (Schiffrin, 1987). In this dissertation, I use the semantic/pragmatic/textual categorization, and for each of these categories I also define several subcategories which indicate more specific types of relations (cf. Fig. 1). Additionally, I adopt the paratactic versus hypotactic distinction which helps us to determine the direction in which units are related. Once the basic tenets of the model are settled it is necessary to explain how the concrete realization of the discourse structure is produced. The view defended here is that discourse relations are the key mechanism through which a coherence interpretation is achieved. As we saw previously in the discussion, discourse relations are realized between discourse segments; obviously at this point it turns necessary to make explicit what we take to be discourse segments, or more properly what we identify as discourse units. 3.3.3. Discourse units. Based on the functional perspective discussed in the previous chapter, I suggest a basic distinction between units in discourse: a first type of unit which bears propositional content and a second type which only carries a functional value. Units of the first type are Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 71 the ones found as arguments of the coherence relations. Units of the second type are what I called discourse functional units and their role is to represent explicitly the relations established between propositional units. In the literature, we can find different proposals regarding mainly the first type of units, commonly referred only as discourse unit. The second type of unit, discourse functional units (DFUs) has also been dealt with extensively in the literature, mainly under the consideration of DMs. In the remaining of this chapter, I discuss several definitions of discourse propositional units that have been proposed in the literature and suggest a definition of them in terms of primarily semantic criteria and secondarily on phonological criteria. The discussion about the second type of unit (DFUs) is dealt with in the next chapter. The issue of what can be defined as ‘the unit’ in discourse is an extremely complicated one. Again, the problem lies on the difficulty of agreeing on a set of criteria. Depending on the theoretical perspective taken, the criteria vary considerably. Sentences, clauses, idea units, utterances, speech acts, and intonation units are among the most frequently proposed units. Sentences and clauses are the units of analysis for studies that take into consideration essentially syntactic properties of discourse (Polanyi, 1988; Redeker, 1986; 1990; Fraser, 1990). Idea units, utterances and speech acts are the units of study for approaches based on the analysis of semantic properties and interactional relations (Hymes, 1972; Schiffrin, 1986; Schegloff, 1987; Redeker, 1990). Finally, intonation units or tone units are the units of study for approaches that rely basically on the prosodic features of oral speech (Chafe, 1992). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 72 3.3.3.1. Sentences. Besides the multiple criteria and different approaches one can take, defining discourse units is complicated because for each of these definitions there are always some gray areas in which exceptions are to be found. For example, as I mentioned before, taking sentences as units faces an immediate problem of what to do with cases like (10). (10) Pero buen- ah por lo menos (..) la idea de las (..) es decir bueno te presentas no tienes nada que perder. (10’) But well eh at least (..) the idea o f the (..) that is well you get there you have nothing to lose. Personal Corpus It is obvious that applying sentential units to this piece of oral discourse would not take us very far. Only two possible full-fledged sentences can be identified here: te presentas ‘you get there’, and no tienes nada que perder ‘you have nothing to lose’. The researcher is then left alone to decide what to do with the rest. It is absolutely necessary to introduce additional criteria to accommodate truncated sentences or even isolated words in hesitating discourse. Besides not being able to account for some parts of discourse, relying only on syntactic criteria to identify units would not allow us to incorporate interactional information into our analysis. As I mentioned previously, this information most of the time has the power to overcome syntactic constructions. For example, consider the case of the hypothetical encounter in (11), in which the interaction is based on a formulaic dialogue perfectly clear and understandable from an interactional point of view, but not analyzable into sentences. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 73 (11) Attendant: good morning Client: good morning cheese burger and a coke, please Att.: for here or to go? Client: for here Att.: five fifty, please Client: here Att.: here is your change, sir Client: thank you Att.: thank you sir have a good day The only sentences used during the whole transaction is ‘here is your change, sir’ and ‘have a good day’, but for the rest of the interaction not a single complete sentence has been used. So for the analyst that wants to keep sentences as her basic units of analysis, the only possible solution is to propose that under the surface structure each of these utterances are sentences. As many researchers have pointed out (Schiffrin, 1986; and references therein), the problem with suggesting the existence of deep sentential structures for these constructions is that surface structure and deep structure do not always have a one-to-one correspondence. In the case of ‘ for here or to go?’, one can think of several possible ‘deep structures’ for it: 'do you want it fo r here or to go?’ 'should I make it fo r here or to go?’ 'is it for here or to go?’. The problem then is to decide what structure to choose among the possible ones and to define the criteria to base our decision on. From an interactional perspective it is clear that each of these sentences evokes differences as to who performs the action or what is the illocutionary force of the sentence, and these differences should be considered if I are interested in tracking the speaker’s intentions when constructing discourse. So it seems important to add other criteria besides the syntactic ones. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 74 33.3.2. Speech acts. The case of sentential units is probably one of the clearest in which exceptions are easy to see, especially when dealing with oral discourse. But it is not an isolated case. Those definitions that take into consideration semantic, pragmatic and phonological features have also their own problems. The units proposed under these criteria are speech acts and intonation units or idea units. Speech acts could apparently be a less problematic unit of analysis. They have been widely used by a number of linguists and sociolinguists and although speech act theory (Austin, 1962; Searle, 1969) is not a discourse theory per se, it is accepted and utilized by many discourse analysts (Hymes, 1974; Labov and Fanshel, 1977). Speech acts are clearly based on pragmatic and interactional criteria, which syntactic definitions lack. Speech acts are the linguistic expressions of speakers’ intentions and beliefs, that is, speakers convey through language not only pure neutral statements, but they add to them information about their beliefs, intentions, and commitment to the messages expressed. Searle (1979) proposes a taxonomy of five different types of speech acts according to different speakers’s intentions and beliefs. He classifies speech acts into representatives, when the speaker is asserting a statement; directives, when the speaker is requesting something from the hearer; commissives, in which the speaker commits herself to the action expressed by the utterance, for example when promising something; expressives, when the speaker is expressing her feelings, for example, thanking the hearer; and declarations, in which the speaker realizes a particular action through the act of talking, for example in the case in which the speaker is accepting a job or a marriage. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 75 Although this taxonomy seems exhaustive enough as to what illocutionary acts one performs when using language, there are also a number of unsolved issues in adopting speech acts as units of analysis, as Schifffin (1994) shows. The most important problem is the identification of a speech act itself. On the one hand, the correspondence between a discourse segment and a speech act is not unique, that is, for the same portion of linguistic material more than one speech act can be identified. On the other hand, in some cases the analyst may face the issue of what to do with utterances for which no possible speech act can be assigned (Andersen, 1991; Schifffin, 1994;). Schiffrin (1994:62-76) exemplifies extensively these two cases with several examples. She shows how the utterance y ’ want a piece o f candy? can be interpreted as a question, a request for information and as an offer, following Searle’s principles of speech act identification. The decision of what of these three speech acts to assign in this case is especially important if I think that depending on it the segments to follow may be interpreted differently. Intimately related to one speaker’s utterance is the response of the hearer, especially in interactional contexts. The continuation to a question is different from the continuation to a request or an offer; they must represent different types of speech acts. In this case, Schifffin suggests that answering no to y ’ want a piece o f candy? can be easily identified as an answer, but it is more difficult to identify it with an illocutionary act that follows a request or an offer: if no is the only utterance after a request or an offer, the analyst must explain why that utterance alone somehow does not represent structural completion. That is, after the negation of a request or the rejection of an offer the usual part to follow is an explanation of the reasons for the negation/rej ection. So, the problem is that if I can not interpret no as a speech act that follows a request then what speech act Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 76 should I assign?. If no speech act can be assigned to it, the question is what to do when the discourse segment identified does not correspond to any of the proposed speech acts. Schifffin mentions additional problematic examples, which she calls expansions, as in (12), to support her point. (12) (a) Do you think there’s much prejudice between like other groups- (b) Other ethnic groups? (c) Other nationalities? (Schifffin, 1994:79) In this example, I should consider how to analyze utterances (b) and (c). If I have to assign a speech act for these units, it seems plausible to call them expansions of (a). Some researchers have identified them as a specific type of expansion, a self-repair, given that (b) and (c) can be considered to replace (a) (Schegloff et al. 1973). However, since obviously (b) and (c) are totally dependent in terms of propositional content on (a), analyzing (b) and (c) as units and assigning a speech act to them would force us to create rules to identify expansions that are totally dependent on the prior speech act. As Schifffin explains, one can add an expansion like (b) or (c) to any type of speech act, so the risk is that if the analyst identifies the first speech act incorrectly, then she will necessarily make a wrong identification of the speech acts that come afterwards. Again as in the previous example, if (b) and (c) are not treated as speech acts, there is little a speech act theory can tell us about how to segment units that do not represent speech acts. So, it seems that although speech acts are excellent units to analyze pragmatic information in terms of speaker’s beliefs and intentions in the structure of discourse, they Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 77 may fail in the identification of utterances that do not clearly fit into the speech act taxonomy proposed in the theory. The problem is relevant since, as I just saw, the identification of a speech act clearly seems to rely on the correct identification of the previous speech acts. This dependency may create a ‘chain reaction’ for which the wrong identification of an utterance leads to subsequent sequential mistakes. 3.3.3.3.Intonation units or idea units. Finally, other researchers have suggested one more type of discourse unit, taking semantic and phonological criteria into account: idea unit or intonation unit. The concept of idea unit or intonation unit implies the fusion of criteria based on the information transmitted by a segment and on suprasegmental features. The combination of these two types of criteria seems to be the most appropriate way to define discourse units, both from a theoretical and empirical perspective, since they are general enough and can avoid the problems seen with syntactic and pragmatic criteria. However, since the ideal correspondence between the two of them does not always occur, semantic criteria take precedence when the aim of our analysis is to uncover structural patterns. Units defined only in phonological terms do not offer enough information as to what the structure of a piece of discourse is. Chafe (1980; 1992) suggests considering suprasegmental features to identify units in speech. In particular, he defines the notion of unit as a part of discourse ‘supported by a convergence of (a) the pauses preceding and following it, (b) a pattern of acceleration- deceleration, (c) the overall decline in pitch level, (d) the falling pitch contour at the end, and (e) the creaky voice at the end’ (1992:60). All these phonological features correlate Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 78 with the expression of an ‘idea unit’ or an ‘information unit’. For Chafe the organization of discourse into intonation units supports the theory that discourse represents the way the human mind works in communication. He suggests that the segmentation of speech in intonation units ideally represents the verbalization of the idea of an event or a state of affairs at a time, and therefore ‘we can conclude that each such idea is active, or occupies a focus of consciousness, for only a brief time, each being replaced by another idea at roughly one- to two-second intervals.’ (Op. Cit.: 66). So actually, under this theory, speech is a representation of the cognitive processes speakers undergo when communicating orally. Units represent the time a speaker takes to verbalize an idea. Pauses and other suprasegmental features are the indicators of when an idea has been expressed and gives way to a new one. The combination of semantic information with phonological to define units is very appealing initially; as analysts, the more criteria we can turn to in interpreting discourse, the more resources we have to decide when to break a unit. However, as I mentioned at the beginning of this section, it is difficult to always find correspondence between semantically and phonologically determined units. The concept of ‘idea unit’ implies the assertion of an event or a state, for which there are participants, also called referents (Chafe, 1992:67). But a unit determined by phonological criteria may not verbalize an event idea or a state idea, but rather just one single referent. That is, an intonation unit can be as simple as the introduction of only one of these pieces of information. See for instance the following example offered by Chafe to illustrate the braking down of units; it is obvious that the units are being told apart exclusively according to prosodic features. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 79 (13) 1(A) ... (0.9)There’s a gir 2(A) .. .Did you ever hear that? 3(B) ...(O.l)No, 4(A) .. ,(0.3)Some elephants and these 5(A) ...(O.l)they 6(A) ,..(0.7)there 7(A) these gals were in a Volkswagen, 8(A) ,..(0.4)anduh, 9(A) .. .(0.3)they uh kept honkin’ the horn (Chafe, 1992: 67-68) Only units 2, 7 and 9 take the form of a complete clause and represent an idea unit in terms of introduction of an event or state. Units 4, 5 and 6 only introduce single discourse referents. In addition, intonation units may not provide us with enough information if our objective is to uncover the structure of discourse. It seems it is necessary to introduce additional conditions such as semantic criteria to decide when to break a unit. For example, in (14) below it is obvious that in line 3 the speaker is merely pausing because he needs time to retrieve some information from his memory: he is about to give a temporal reference and he needs to remember the exact time. The question is whether to consider line 4 as a new unit: according to phonological features it is, because it is separated by pauses from the previous and following pieces of discourse, but informationally line 4 seems to belong to the previous unit. (14) 1(A). 2 jQuepena! ^no? 3(C) A mi me nombraron presidente y administrador hace (...) 4 6 anos. 5 Y... 6 al mes ya lo habia deja<(d)>o. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 80 7 jEra horrible! 8 j Horrible! 1(A). yeah? @@ 2 what a shame, right? 3(C) I was named president and manager about (...) 4 6 years ago 5 a n d ... 6 in a month I quit already 7 it was awful! 8 awful! (CREA, C25) Additional support for the idea that sometimes intonation units may not offer enough information to uncover structural patterns in discourse is provided by the fact that speakers organize the information to be conveyed not only according to cognitive needs but also to informative needs. The expression of a certain piece of information is interconnected and motivated by the information given before and the information the speaker still plans to convey. There is evidence that speakers organize units of information into larger and interconnected structures, and although in many cases it is possible to find that information units are also separate units in terms of intonation and syntactic structure, in other cases there is no correspondence between them. Schiffrin (1997) makes this point very clear when discussing different criteria to determine discourse units. She focuses on an example of the introduction of a new referent. The new referent does not provide relevant information per se; rather, it is a part of a whole series of intonation units which, once connected, provide support to the argument made in the previous part of the talk. Let us see her example, reproduced here in (15): Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 81 (15) Debby: (a) Yeh. Well some people before they go to the doctor, they’ll go to a friend, or a neighbor. (b) Is there anybody that [uh Zelda (c) [Well:: well I guess- Henry (d) [Sometimes it works. (e) Because there’s this guy Louie Gelman, (f) he went to a big specialist, (g) and the guy. ..analyzed it wrong. (Schiffrin, 1997:75) Schiffrin suggests that when analyzing a unit like (e), it is important to focus not only on the intonational, syntactic and semantic features: (e) conforms to all the requirements to be considered an intonation unit: it shows a break after it, syntactically it takes the form of a clause, and semantically it represents the introduction of a new referent. That is an idea unit according to Chafe’s criteria. However, looking only at these features would lead us to overlook ‘other cues suggesting that Henry’s Because there’ s this guy Louie Gelman, [is] also a part of a larger unit of analysis.’ (Schiffrin, 1997:84). More specifically, she proposes that units (e), (f) and (g) intonationally may be analyzed separately, but informationally they are all part of a larger unit which provides support to the assertion Sometimes it works, which is the answer to Debby’s question is there anybody that uh... In support of her argument, Schiffrin mentions other cues not commonly taken into account when dividing discourse into intonation units. For example, the pending final tone of (e) and (f) indicate that there is yet more information to come; on the contrary, the final lowering intonation of (g) indicates final information. Also, (e), (f) and (g) are connected through the planned conjunction structure (X, Y and Z), which indicates the expression of several connected terms. And finally, both actions in (f) and Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 82 (g) support the answer given by Henry, that is, ‘sometimes it works to seek medical advice from people who are not doctors’. Equally, in example (16) from my personal Spanish corpus, it can be seen that the division in intonation units does not tell us much about whether or how the information contained in each unit contributes to the understanding of the previous talk. (16) 1 (A). i a que vienen? 2(B). pues a... 3 cada uno a quejarse de lo suyo. 4(C). somos cuatro vecinos y mal avenidos. 5 todos viejos, 6 solterones, 7 aburridos, 8 y... y... 9 no se que pasa, 10 pero nos llevamos de mal... (16’) 1(A). what do they come here for? 2(B). well to... 3 everyone to complain o f their own thing 4(C). we are four neighbors and getting along bad 5 all old 6 bachelors 7 boring 8 and... and... 9 I don’ t know what’ s wrong, 10 but we get along so bad (CREA C20) Lines 5, 6 and 7 have been divided into units according to phonological criteria; after each of them the speaker paused and each of them introduced a new referent, i.e., each of them represents a category which the neighbors in the building are part of: old, bachelors, and boring. But none of these units alone contributes to support the claim Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 83 made in 4: we are few and we don’t get along well. The speaker wants to convey the information that the conjunction of these characteristics is the cause of the neighbors not getting along. So, although we are probably right in following phonological criteria and dividing this information into three separate units, the three of them together give reasons as to why the neighbors are not getting along, not each of the units separately. So, depending on the specific focus of the analysis, sometimes phonological features (such as pauses before and after a discourse segment or the interpretation of a final tone) may not give us all the information we require to see the larger picture. That is, in some cases I would take into account larger units than the intonation unit as the basic unit of analysis. This is not to say that intonation units are not of any use in discourse segmentation. In fact, in many cases basic units of analysis may coincide with intonation units, but they do not necessarily coincide. The question of what conditions should prevail when breaking a unit are indeed narrowly tied to the objectives of our analysis. But if the premise we depart from is that discourse has a structure and our analysis aims to uncover the structural patterns that the speakers follow when communicating, then a unit that is able to offer more information on what its relationships with the previous and the upcoming units are could be of more use to us. This is why for the purpose of this dissertation semantic and not phonological features have been chosen as the primary criteria for discourse segmentation. I adopt here the basic definition of discourse units proposed by Polanyi (1995). This author focuses on information conditions as the primary criteria to identify discourse units and proposes a definition of units formally characterized by the introduction of semantic requirements the unit must fulfill. A unit is defined as the expression of an event Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 84 or a state of affairs e, by a certain number of participants, on a period of time t and in space s. We define dcu [discourse constituent unit] as a contextually indexed representation of information conveyed by a semiotic gesture, asserting a single state of affairs or partial state of affairs in a discourse world. Each dcu expresses an event or in general a state of affairs in some spatio- temporal location, involving some set of (defined or as of yet undefined) participants (Davidson 1967). The event will be either positive or negative, generic or specific. (Op. Cit.:5-6). Since a definition exclusively in terms of semantic conditions may be difficult to apply to natural interaction, in which an ideal complete information unit is not the norm, additional criteria should be introduced to allow the analysis of semantically truncated segments or units that do not represent complete syntactic utterances. So, an operational definition of how discourse is segmented must allow more flexibility, introducing phonological criteria as well. Thus, in practice, the semantic conditions stated above are the fundamental criteria that the analyst should apply first, but in cases in which the utterance is not represented by a complete syntactic structure prosodic information such as breaks and pauses should be the guiding criteria. The definition of units primarily in semantic terms has certain advantages. The most important one is that it allows us not to get committed to a particular syntactic structure. An event or state of affairs is taken to represent an instantiation of a predicated relation even if this is not a reported one (Cf. Hengeveld, 1990). For instance, Dahlgren (1996) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 85 suggests that in the case in which an observer sees a car crashing she may say something like: ‘Oh, that car is falling off of a cliff, we have a verbal utterance in which the actors and the goals are made explicit. But the observer may also say only ‘Oh! an accident’, in which case we must infer the existence of some participants and some goals which are not explicit. (: 114). So in this latter case we still can find a predication, in the sense that this expression represents concrete particulars individuated by their causes and consequences: When a person has seen a car fall over a cliff, he/she may interpret it as a whole with two causally related subparts, as in (i) (i) (el) The car came to the edge. (e2) the car fell over the edge. (Dahlgren, 1996:114) This is why in many cases the structure of the unit taken in semantic terms as the representation of an event or a state of affairs (SoA) may coincide with a sentence, a clause, or an intonation unit, but in general the length and structure of a unit need not correspond to any of these; it might be equal to or maybe smaller than any of them. As I mentioned before, I consider these to be the basic units in discourse that bear propositional content, as opposed to other units which only carry procedural or functional value. That is, Discourse Propositional Units represent an instantiation of actual events and states of affairs. I also mentioned that DPUs are the arguments of coherence relations, that is, adopting Dahlgren’s view, I understand that discourse coheres precisely because the hearer can interpret the relations between the representations of events and SoAs as a correspondence of how events and SoAs relate in the real world. In this Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 86 respect, in a formal representation of discourse, the segments that instantiate events or SoAs are put in relation through a series of semantic connections, so these segments, that I called DPUs must appear to the left and right to the relation. discourse propositional unit <=RELATION=> discourse propositional unit This proposal of discourse segments as arguments of coherence relations is not only supported from a theoretical point of view, but it also seems to receive support from empirical analyses. The boundaries of discourse segments are generally said to be placed at the points where a new relation is established. That is, a change in relation at any point in discourse is a signal that indicates the break of a new segment. The correlation between relation change and the segmentation of new units seems to be a very strong one. In a study set up by Dahlgren (1993), she examined the most relevant factors that made a reader place boundaries at specific points in discourse. She reports that ‘the most consistent correlate of a new segment was change in coherence relation.’ (Op. Cit.: 127). In 92%1 of the cases the breaking down of a new segment was motivated by a change in the coherence relation holding between the new piece of discourse and the previous one. The second type of unit which bears primarily functional content is the discourse functional unit (DFUs). In the following chapter I offer a proposal in which the particular role of these categories is discussed in detail. 1 92% corresponds to the breaking o f what Dahlgren calls sister segment, which refers to a segment that relates to the previous one at the same level of analysis (in coordination). She also analyzes another type of segment, a subsegment, which is subordinated to the previous one. With respect to this second type of segment, change in coherence relation caused segmentation in 88% o f the cases. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 87 3.4. Conclusions. I began this chapter on the premises that the best way to give a ‘non a prioristic’ account of what DMs are and how they work is to have a general theoretical discourse model that can offer a functional explanation of the organization of discourse and at the same time define and explain the notion of DMs as a part of that organization. The theoretical framework presented here departs from the theoretical premise that discourse has a structure which can be accounted for in terms of linguistic patterns. A detailed definition of discourse structure from a functional/structural perspective like the one adopted here is that discourse is composed of a series of elementary units which are interconnected through a series of relations. These relations of a semantic, pragmatic or textual nature are the ones that enable the hearer to obtain a coherent interpretation of a piece of discourse. Relations are the ones that make possible a coherent interpretation in discourse; therefore, their identification provides us with enough information to determine how the different pieces of information that a speaker wants to convey get related, and how the hearer is able to interpret them coherently. To fully account for the actual structure of discourse, the most important issue is then how we determine what constitutes the minimal piece that get interpreted through coherence relations. These minimal discourse segments are what I called discourse propositional units (DPUs) because they consititute the representation of an event or state of affair. The criteria I proposed here to identify these units are primarily semantic and secondarily phonologic. The definition of DPUs in information terms allows us not to get committed to a specific syntactic structure, avoiding in this way the problems associated with the notion of units as sentences or clauses. Also adopting phonological criteria as secondary Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 88 conditions provides us with enough information to make proper coherence interpretations in the cases of incomplete units (such as truncated sentences or short turns of speech, composed of one or two words). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 89 CHAPTER 4. Discourse Functional Units. 4.1. introduction. After the presentation in the previous chapter of the theoretical discourse model followed in this dissertation, the most important issue to deal with and the focus of this dissertation is the study of the second type of unit proposed: DFUs. As I mentioned before, the functionality of these forms has been extensively studied under the notion of DFU. In this chapter I review the most influential proposals on the classification of DFUs’ functions according to different criteria. I propose that the new category of DFUs perform only a major function, which is the representation of coherence relations. In contrast with other proposals that claim the existence of types of DFUs (Polanyi, 1986; Fraser, 1999) I suggest there is not unique correspondence between a particular DFU and a certain type of representation; that is, the same linguistic form may be able to act as a functional unit in the representation of coherence relations in every discourse component: semantic, pragmatic and textual. A form such as pero ‘but’ in Spanish is capable of representing semantic, pragmatic and textual relations. However, in each specific domain a DFU presents particular characteristics in accordance with the type of Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 90 relation it represents. For example, a DFU representing a semantic relation may not be extractable and the members related by the DFU may not allow any intervening material between them. In contrast, the same DFU when representing a textual relation may be subject to extractability and to the insertion of additional material between the members it relates. I claim these characteristics are not intrinsic properties of DFUs, since the same form representing different types of relations may or may not be subject to them. 4.2. Definition of Discourse Functional Units. As we saw in chapter 2, I propose to identify categories in discourse on the bases of their functionality. A functional definition alone does not face the a prioristic problems structural definitions do. First of all, it allows us to establish reliable criteria to identify the forms that we want to call DFUs, independently of the actual linguistic expressions used in a particular language. If we understand that DFUs perform a certain function in discourse, the functional definition implies that ah those linguistic expressions able to perform such a function will be defined as DFUs. This type of definition requires of course an independent description of discourse functions as well as an independent description of the function of DFUs. This is why a discourse framework capable of providing an independent theoretical account such as the one presented in the preceding chapter was needed. In what follows I identify the major functions of these forms in discourse and suggest Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 91 4.2.1. The connectivity function. Probably the most widely accepted description of these forms from a functional point of view corresponds to Schiffrin’s (1987) definition of functions of DMs. Schifffin identified their main role in discourse as ‘sequential elements that bracket units of talk’ (Schifffin, 1987:31). This concise definition is complemented later by adding other functions to these forms: they locate the utterance in one or more planes of talk of her discourse model (Cf. 1.2.1.2.), they index the utterances to the speaker, the hearer, or both, and they index the utterances to prior and upcoming discourse. As Schifffin herself discusses, defining these forms as bracketing elements of units in discourse poses a fundamental problem: the necessity of having a good description of what is understood by ‘discourse unit’ (cf. Romera 1998 as well, where this problem is also pointed out). Relying on particular linguistic forms to determine units and at the same time defining units on the basis of the appearance of these forms creates a circular path which compromises seriously the analysis and the whole theory. It was also mentioned in the previous chapter that Schifffin presents other functional characteristics which could help overcome the circularity problem of this definition. The forms analyzed by her perform several indexical functions at different levels of analysis. First of all, they are capable of anchoring the utterance into one of the modules she establishes; these modules correspond roughly to the semantic, pragmatic and textual domains considered in our discourse model (cf. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 92 Ch.3), except for the interactional module (participation and information state frameworks), which I do not consider here. In addition to this indexical function, the nine forms she analyzes also have a deictic role, placing the utterance in relation to what has been said before and/or to what is coming next. And finally, they attach the utterance to a specific context, by virtue of which they relate the hearer and the speaker to what is being communicated. Nevertheless, although Schifffin offers a very detailed explanation of what she understands to be the functions of DMs, these seem more the result of the explanation of the functions of the 11 linguistic expressions she considers, rather than discriminating characteristics that can help us tell apart DMs from other linguistic expressions in discourse. The relations among the proposed characteristics and their role in the identification of DMs remains unclear. One aspect that may be mentioned is that some of these characteristics seem to point to very similar functions. For example, a DM indexes an utterance with respect to the speaker or to the hearer, or it may also establish a connection between the two of them. This does not seem very far from anchoring an utterance into the participation framework in Schifffin’s own model, which ‘captures both speaker/hearer relations, and speaker/utterance relations’ (op. cit.:27). So, it is not very clear whether it is enough for a linguistic form to represent one of these two characteristics or functions for it to be considered a DFU, or whether some functions are more essential than others for identifying a DFU. In any case, it is fair and necessary to point out that Schifffin’s was a pioneering study of the grammatical category of DFUs, and this author Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 93 proposed those characteristics only for a set of DFUs, probably without any intention of attributing these characteristics to DFUs in general, that is, as DFUs’ defining properties. Therefore, if we are looking for defining criteria to identify a general category of units in discourse, we may want to establish a more restrictive functional definition somehow. Two other very influential researchers in the field have proposed certain functions for some these expressions in discourse: Fraser (1990; 1999) and Redeker (1990; 1991). Although from different theoretical perspectives, both seem to rely on one important aspect: the connecting function as a defining property of these forms. In his latest work on DMs, Fraser (1999) offers a definition of some linguistic expressions from a grammatical point of view based on the notion of connectivity. Although he explicitly states that the notion of discourse relation is not critical to their definition (:937), he later identifies them as forms which share a common property: they impose a relationship1 between some aspect of the discourse segment they are part of, call it S2, and some aspect of a prior discourse segment, call it S1. In other words, they function like a two- place relation, one argument lying in the segment they introduce, the other lying in the prior discourse. I represent the canonical form as <S1.DM+S2>. (:938 ) Fraser’s definition of DMs is very much indeed tightened to the notion of relation. A DM is a lexical expression which primarily functions imposing a ‘relation’ between two segments in discourse. That is, a DM ‘connects’ two segments in a specific way. However, the use of the term ‘imposing’ in his definition is intended to represent a Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 94 difference with respect to other definitions that use the connection function as their basic notion. For Fraser, DMs must be compatible with the interpretation of the segments they relate, since not all DMs can appear in the same contexts. Additionally, in some cases they connect a non-explicit proposition with an explicit one. For example, in (1) (extracted from Fraser (1999)), it is obvious that the DM ‘but’ is joining segment B to a presupposed proposition triggered by segment A. (1) A: Here is a triangle. B: B u t it has four sides. (:942) These facts constitute evidence that DMs do not simply display the connection between the units related but they impose a specific type of relation: ‘a DM imposes on S2 a certain range of interpretations, given the interpretation(s) o f SI and the meaning of the DM.’ (: 942). He goes as far as saying that ‘a DM does not ‘display’ a relationship [...] any more than a verb displays a relationship between a subject and object.’ (:942). On the other hand, from a different perspective, Redeker (1990, 1991) prefers calling these forms discourse connectors, and she defines them as: a word or phrase [] that is uttered with the primary function of bringing to the listener’s attention a particular kind of linkage of the upcoming utterance with the immediate discourse context. (Redeker, 1991:1168). She also points out to a connective function, but from a more cognitive perspective; she sees these expressions as linguistic devices that indicate to the 1 Italics added. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 95 listener how to interpret the connection between two segments in discourse. So, in contrast to Fraser, who defines these units in relation to the internal interpretation of two segments (as imposing a relation between them), she defines them more as tools which help the listener choose a ‘particular kind’ of connection between two utterances. These two perspectives are not necessarily exclusive or incompatible. The fact that a linguistic expression helps the listener choose a specific kind of connection between two segments, to a certain extent is the same as saying that it ‘imposes’ internally a specific relation between them. Once a particular interpretation has been selected by means of a linguistic device, other alternative interpretations are eliminated; that is, one interpretation is ‘imposed’ over the others. However, the connection established by DFUS is not comparable to the one represented by a verb between its subject and its object, as Fraser suggests in his work as well. The main distinction between the relation holding between a verb and its arguments and the one existing between a DFU and its arguments relies on the different nature of the relations. On the one hand, the semantic relation established between the arguments of a verb and the verb cannot be recovered in the absence of the lexical item expressing the verb itself, unless that piece of information has been previously made obvious to the listener by other means. This is because a verb introduces part of the propositional meaning that makes possible the interpretation of the segments. Since DFUs only carry procedural value, the relations established by a DFU on the other hand remain underlying the interpretation of the two segments in such a way that it is Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 96 still possible to obtain a coherent interpretation even in the absence of the lexical DFU. In addition to this, the units related by the DFU are semantic representations of events or states of affairs, whereas the arguments of a verb are not. Consider for example, the difference between the utterance in (2), in which two nouns (possible subject and object of a verb, respectively) are placed together in the absence of the verb, and the utterance in (3), in which two sentences are joined by no explicit DM (the empty parentheses represent the absence of both lexical forms: a verb and a DM, respectively). (2) The tree () the leaf. (3) I want to go to the movies tonight. ( ) It’s my birthday. (Fraser:944) It is obvious that from the listener’s point of view, example (2) is an impossible combination in the absence of no additional information. No speaker would leave the listener to infer what she is trying to communicate putting two NPs side by side with no more information; the absence of the verb yields an incoherent interpretation. In contrast, (3) presents an interpretable example in which if no DFU is used there is a default relation (causal) that allows the listener to interpret these segments coherently: ‘I want to go to the movies tonight, because it is my birthday’. In addition to this argument, I agree with Knott and Dale (1994) in that we must assume that DFUs represent devices used by actual speakers; if we defined them just as devices which describe a piece of discourse, then there is not possibility of explaining why in certain cases there is the option of not using them. It is for these Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 97 reasons that we believe DFUs are better defined primarily as devices that help the listener choose one interpretation over other possible ones. 4.2.2. DFUs as linguistic expressions that make explicit a coherence relation. In this line of thinking, we find researchers who get one step further in the definition of the role of these forms in discourse. Within the Coherence Relation Theories (CRT), Sanders (1992), Sanders et al. (1993), Polanyi, (1995), Spooren (1997) among others consider that DFUs not only direct the listener to a specific type of linkage, but they do so ‘by making explicit the nature of the links among pieces of information’ (Polanyi, 1985:14). These researchers understand that listeners/readers achieve coherence of a piece of discourse by interpreting the relations that exist between different pieces of information2. As I pointed out earlier, these relations are the ones responsible for the coherent interpretation of discourse. DFUs are then seen as linguistic expressions used by the speaker to bring to the surface a specific relation among the set of possible ones which could yield a coherent interpretation. This definition of DFUs in terms of coherence relations is the one I am adopting in this dissertation as well. I understand DFUs precisely as devices that make explicit one particular kind of relation between the semantic representation of events or states of affairs. One important aspect of this definition is that the relations that make possible a coherent interpretation are always present as a possible choice 2 Polanyi’s understanding of discourse structure deviates from the generally accepted conception in Rhetorical Structure Theories that discourse relations are the only ones responsible for creating coherence. However, she accounts for the information these relations provide by means o f complex discourse units rules. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 98 for the listener; the mission of the specific linguistic expression that acts as a DFU is precisely to realize one of them. If we consider the sequence in (3) above, we could interpret different types of connections between the two members of the relation: causal, adversative, or justificative, among others. They are made explicit by the use of specific DFUs. (4) (a) I want to go to the movies tonight because it’s my birthday. (b) I want to go to the movies tonight, but it’s my birthday. (c) I want to go to the movies tonight, after all, it’s my birthday. (4a) yields a causal interpretation: T want to go to the movies tonight because it is my birthday’. On the other hand, in (4b) after a friend’s invitation to go to the movies I may respond that I want to go to the movies tonight but it is my birthday, letting her infer that T have other plans already’. Thirdly, (4c) illustrates a case in which I, as the speaker, may feel the need of justify my desire of going to the movies, which somehow I perceive as something that may not be well received by the listener (i.e., the second member presents a justification to the content of the first one). Thus, it is the explicit DFU in each of these sequences that allows us to obtain one of these possible interpretations. DFUs are hence defined in terms of their functionality in discourse; in fact, I define them exclusively on the bases of this property. I treat them as basic discourse functional expressions, which act as the expression of the coherence relations holding between propositional units. Both these units are defined in terms of the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 99 coherence relations that make discourse interpretable, but whereas propositional units are the arguments related in a binary predicate relation, DFUs are the linguistic expressions that make explicit the relation between the two arguments. Since both help build the structure of discourse, they are both considered as elementary units. So, in practical terms, I suggest defining DFUs as all those linguistic expressions which are functional categories, whose function is to make coherence relations explicit. As I mentioned at the beginning of this section, adopting this functional definition as the discriminating criterion to identify DFUs helps avoid some a prioristic problems other definitions face. It allows us to talk about DFUs independently of the actual language forms. For instance, in the example below all the linguistic expressions that can be identified as representing the relations that make the meaning of these units interpretable will be called DFUs. I have partitioned this example into units and bracketed them according to the relations established: (5) 1 [[[[we were strolling inside buildings,] 2 [() touching words engraved in marble,] 3 [() hearing voices,] 4 [and imaging ourselves in different areas]]] 5 [[[rather than merely reading about events] 5 [or watching a film.]]]]3 In the first place, there are 5 identifiable coherence relations and 3 linguistic expressions that mark 3 of them. The other two relations are left inexplicit, that is, they are left for the listener (the reader in this case) to infer them. Units 1 through 4 are connected by three instances of the List relation: 2 is connected to 1, 3 to 2 and 4 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 100 to 3 through the same relation. Only in 4 does a linguistic expression make explicit this relation, and. Then, 6 is connected to 5 through an inclusive disjunction relation made explicit by the linguistic expression or. Finally, these two units (5 and 6) connect to the previous sequence (units 1 to 4) by a comparison relation made explicit by the expression rather than. Therefore, the three linguistic expressions found to represent coherence relations (and, or, and rather than) will be considered DMs. 4.2.3, The expression o f coherence relations and the meaning o f DFUs. As we saw previously the differences between DPUs and DFUs are based primarily on their semantic representation. The meaning of DFUs constitutes one of the most interesting issues in the study of these forms. If we agree with the idea that DFUs represent procedural meaning, that is, they instruct the listener as to ‘how’ to connect a segment with a previous unit in discourse, we should be able to identify precisely how a DFUs can connect two segments, that is, what is the precise meaning it establishes. In the framework chosen in this dissertation, the answer to this question is quite clear: since a DFU makes a coherence relation explicit, the meaning of a DFU must correspond necessarily to the meaning of the coherence relation/s it represents. The interesting question that arises now is whether that particular DFU can be said to express a specific type of relation on its own, i.e., in the absence of any contextual information. For example, is it possible to say that the DFU and by 3 Example extracted from a literary magazine. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 101 itself means addition, or that the DFU because by itself does not mean contrast? Spooren (1997) presents interesting cases that seem to suggest that it is in fact possible to identify the specific type of procedural meaning a DFU expresses in the absence of contextual information. In his work he analyses cases in which the identifiable relation one would expect a DFU to represent is not the one that establishes coherence in a particular context. He considers examples such as the one in (5): (5) Since June 1st Ron Kaal (43) has been editor in chief of the monthly O. Kaal was approached last year by the publisher, Maurice Keizer, after he had written a critical article on the first issue of the magazine in NRC Handelsblad. (Op. Cit.: 150) It is obvious that in abstract terms we understand the DFU after as the prototypical representative of a temporal relation: we could say that after has a temporal meaning. However in this case, as Spooren indicates, there is a clear causal connection between ‘Kaal was approached by the publisher’ and ‘he had written a critical article’. The causal reading could be paraphrased as ‘Kaal was approached because he wrote a critical article’. So, the question is: can we say that the meaning of after is the expression of a temporal relation, or rather that it expresses temporality and causality? As I mentioned earlier in this dissertation (cf. 1.2.1.1.), this issue has led researchers to propose different alternatives regarding the meaning of DFUs. The fact that a temporal expression can appear in contexts with a causal interpretation has Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 102 been interpreted as a proof of the polysemy of DFUs. The maximalist approach argues that these are all different meanings of the form after, and therefore we should take DFUs to be polysemic forms capable of expressing different meanings in every context. On the other hand, the minimalist view and extended minimalist approaches (cf. Hansen 1998) would interpret the case of after as a DFU which represents a core meaning (temporal) but which is capable of having other interpretations in certain contexts, always as an extension of the core meaning. In my opinion, the answer to this question lies on the hypothesis of the acquisition of new meanings in language developed by Traugott (1982; 1989) and Konig and Traugott (1988). These authors propose that there is a tendency in language which accounts for the acquisition of new meanings by lexical and non- lexical items. Traugott (1982) in particular shows that a word initially has an inferential meaning, interpretable as a result of functional implicatures derived from the context in which it appears, but later on this meaning becomes a part of the meaning of the word itself. That is, after being used in specific contexts where it leads the listener to infer a meaning other than its usual one, a word ends up adding that new meaning to its original one, and then being able to express it in contexts other than the ones in which it acquired it. DFUs (‘conjunctives’ in Traugott’s terminology) are among the very clear cases this author presents. For example, she illustrates the case of again which is related to against. ‘Originally meaning ‘facing locally’, in the context of question and answer, again(st) came to mean ‘(reply) again, (answer) back’. [] And a later stage yet, again signals the speaker’s intention to extend the discourse with some further relevant information.’ (1982: 251). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 103 I claim that this tendency operates the same way for DFUs, that is, that these forms extend their use from expressing certain relations to expressing other relations. Of course, the process of incorporating new meanings is not an overnight process; like any diachronic change, it requires years, sometimes centuries, to develop and is in continuous evolution. So what we are able to observe and hypothesize now about the meaning of DFUs is no more than a synchronic cut in the evolution of that particular form. A DFU is capable of representing in the majority of cases a particular relation, which could be called the canonical meaning of the form, that is the identifiable procedural meaning of the form in isolation, but through the incorporation of inferential implicatures in certain contexts the same DFU is able to express additional relations which after enough frequency of use will end up being part of the DFU’s canonical meaning. Schiffrin (1992) illustrates this process by the case of then in English. As this author shows, the temporal then is capable of functioning in the expression of warrant-inferential sequences and in the connection of global meanings, and Schiffrin points out that ‘the way the temporal meanings are predicated in text (as global meanings) and in warrant-inferences (epistemically) seems to reflect [...] different historical stages’ (:786). However, there are also certain constraints on the range and type of new contexts in which DFUs may appear, proven by the fact that not any DFU may express any relation. For example, as Spooren (1997) points out, it is difficult to express a concessive relation using and or a contrastive relation using because (op. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 104 cit.:153). The correspondence of a DFU with certain types of relations is not produced in a random or arbitrary manner. The forms that we presently know as expressing discourse relations derive from lexical items which already had certain lexical meanings. In their evolution process, when these lexical forms extend their use to new contexts, such as the representation of coherence relations, the relation they come to express must necessarily be compatible with the original meaning of the forms, that is, the meaning of the relation must share some aspect with the DFU’s primitive meaning. The case of Spanish pues ‘so’ illustrates this point clearly, as it evolves from two Latin adverbial forms post ‘behind’ and postea ‘behind that, after that’ to become a textual form that expresses a ‘causal or resultative’ relation (Traugott 1982: 255). This change is absolutely explainable in the light of the primitive semantic features which characterize the meaning of the coherence relations. This is shown by Sanders (1992 ;1997); Sanders et al. (1993); and Spooren (1997), who claim that relations are organized into different categories according to certain primitive features they have in common. After different experiments, these authors identified several basic relational criteria when connecting two segments in discourse. Addition and Causality are the first two basic cognitive operations in which two segments may relate, because neither one is derivable from any other operation. Addition is a weak relation and causality is its strong counterpart. These authors also argue that the order of the units plays an important role in the complexity of the relations as well: segments may relate to each other in a basic or a non-basic order. Finally, it is claimed that polarity determines positive and negative Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 105 relations. Based on the combination of these primitive features, it is possible to establish relationships among the coherence relations: ‘Temporal and causal relations are close because they only differ in the primitive Basic Operation (distinguishing between additive and causal relations). List and Contrast relations are close because they differ only in the primitive Polarity (distinguishing positive from negative relations). On the other hand, List relations and Negative Cause relations [] are distant because they differ both in Basic Operation and Polarity (List: additive, positive; Negative Cause: causal, negative).’ (Spooren, 1997:153). So, returning to our example of pues, it is obvious that the Latin adverbs post ‘behind’ and postea ‘behind that, after that’ are perfect candidates to expand into the expression of causality in discourse. The original meaning of temporality of these lexical items and the meaning of the causality relation they come to express are compatible because they only differ in one primitive feature: the Basic Operation. I claim the same principle applies to the extension of a DFU into the expression of new relations. The same DFU is able to express only those relations that are close in their semantic categorization. So, going back to the original example in (5), for a DFU like after it is perfectly possible to appear in contexts in which it can be interpreted to express a causal relation. On the other hand, it would not be possible for after to become an expression of a relation of contrast such as the one established by but, because they differ in two primitives features: polarity and basic operation. In addition to this first constraint, there is a second constraint that allows us also to explain other phenomena regarding the direction of the extension of DFUs in the expression of different relations. A DFU expresses only equally or more Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 106 specific relations than its most frequent one. For example, we find that the DFU but, which has a prototypical function of representing contrast relation, may also express a semantically close relation: concession. On the other hand, the DFU which typically represents concession, i.e., although, is unlikely to represent a simple contrast relation. Observe the following two examples of the Spanish counterpart of but (i.e., ‘pero’) drawn from our corpus. In (6) pero expresses contrast, and in (7) it expresses concession. (6) esto caliente esta bien pero si no [esta caliente], no. (6’) this (the dish) is good when warm but if it’ s not [warm], no. F23L23 (7) la otra crema, yo la se hacer pero la tengo apuntada. (7’) the other cream, I know how to do it but I have it (the recipe) written down. F13L68 The same concession relation expressed by pero in (7) may be expressed by the corresponding concessive DFU, aunque ‘although’ (cf. (8) below). In contrast, the paraphrasis of (6) with aunque yields a totally different interpretation: (8) aunque la otra crema, yo la se hacer la tengo apuntada. (8’) although the other cream, I know how to do it I have it (the recipe) written down. F13L68 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 107 (9) *aunque esto caliente esta bien si no [esta caliente], no. (9’) * although this (the dish) is good when warm if it’ s not [warm], no. F23L23 Similar conclusions can be found in other DFUs. For instance, and may express the relations of contrast and concession or causality besides its prototypical function of representing addition. But it is very unlikely to find a causal DFU like because in the expression of temporality or addition (cf. Spooren, 1997). The explanation for these findings lies again in the restrictions posed by the particular categorization of coherence relations, such as the condition that a DFU can only represent or express new relations which are more specific than or as equally specific as the relation that they express most frequently. I stated above that relations which share common features belong to the same category. Additionally, within a category relations are also organized hierarchically around the principle of specificity. Specificity has been explained in different ways according to different researchers. Marcu (1999) and Marcu et al. (1999) suggest that relations share a dominance relationship within respect to one another. The meaning of some relations is contained within the meaning of others. For example, in the case of the category of contrastive relations the meaning of concession is contained in the meaning of contrast. That is, the expression of concession can be conceived as the expression of some type of contrast, as in concessive relations there is a contrast between two arguments: one of the arguments is presented in support of a certain Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 108 premise, and the other is presented against it (Oversteegen, 1997). However, the opposite cannot be maintained: not in every case in which we find a contrast relation is the expression of concession involved, as we saw above (cf. the impossibility of paraphrasing (6) with (8)). Therefore, these authors suggest that the asymmetrical relationships inside the categories reveal differences in specificity in the meaning of the relations. The relations which are part of another relation in their category are the ones which express more specific meanings. In contrast, the less specific ones (more general in meaning) are the ones capable of containing others. This would perfectly explain why a DFU such as pero, which expresses contrast, can also extend its functions into expressing a more specific relation such as concession. However, a DFU such as aunque, which primarily expresses concession, is not capable of expressing a more general relation like simple contrast. A different definition of specificity, complementary to Marcu’s4, is the one offered by Sanders and associates. Their notion of specificity is based on the complexity of a category of relations according to the number of primitive features the relations in it are composed of. In Spooren’s words: Cause relations are more specific than List relations, because a List relation involves mere conjunction of two states of affairs, whereas a Cause relation also involves an implication relation between the two states of affairs. A Contrast relation is more specific than a List relation because a Contrast relation also involves a negation relation between the related states of affairs. (Spooren, 1997: 154). 4 Sanders et al. focus on the explanation o f the differences among the 12 categories o f relations they establish, that is, differences between classes of relations, whereas Marcu (1999) explains the phenomenon with respect to the internal organization of each category. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 109 This notion of specificity could also explain the case of but and although. In terms of primitive features the relation of contrast and the relation of concession present a fundamental distinction with respect to their complexity. Contrast is a less complex relation, as it is a semantic relation, whereas concession is a more complex one, a pragmatic one5. Also, contrast involves the comparison of the propositional content of two units, that is two events or states of affairs; on the other hand, concession involves the opposition of two arguments in support of and against a non-explicit argument which must be inferred. So to summarize this far, I agree with defining the meaning of DFUs meaning as a procedural meaning. The specific procedural meaning of a DFU coincides with the meaning of the relation that they most frequently express or represent. We could call this primary relation the canonical meaning of a DFU. However, DFUs are open to extend their functions and represent additional relations, through the incorporation of inferential implicatures given by the context in which DFUs appear. However, these extensions are subject to certain constraints. In the first place, the meaning of the relation that a DFU comes to express must be compatible with the original meaning of the lexical form that became a DFU. I illustrated this point with the Spanish DFU pues, which evolves from two Latin temporal adverbs. As we saw, the meaning of temporality of these adverbs is compatible with the meaning of the relation of causality that pues expresses as a DFU. This tendency also applies to the extension of the relations a DFU represents. A particular DFU can only express 5 The lower degree o f complexity of semantic vs pragmatic relations has been supported by Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 110 relations which are compatible in meaning. Additionally, a second constraint applies to the expression of relations by a DFU. A DFU can only extend its functions into the representation of relations which are at an equal or lower level of specificity. 4.3. Discourse Functional Units and types of relations they represent. Besides the principles conditioning the extension of the relations expressed by DFUs, some researchers have proposed that there exist basic differences in the functions that DFUs may perform, which can lead to establishing a categorization of these forms in different types. I refer here to the distinction between semantic, pragmatic and textual aspects of language. These three terms and the distinctions they encode have been applied differently in different fields; however, the basic concepts behind them are agreed on by most researchers. The most widely accepted distinction in discourse studies is the one between the terms semantic and pragmatic. For some authors semantic/pragmatic are two notions that apply to the meaning of linguistic forms (Sweetser (1984); Traugott (1989)). The definition of these terms is based on the functional distinction between external vs. internal meaning of linguistic forms (cf. Halliday and Hasan, 1977; Martin, 1992). Semantic or external meaning is the one that has its referent in the concrete, experiential or real world domain. That is, semantic meaning is taken to represent the propositional content of the expression. Pragmatic or internal meaning acquisition studies, in which results show that semantic relations are acquired before pragmatic ones (Romera, 1998). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. I l l on the other hand refers to the expression of attitudes, beliefs and emotions towards what is being expressed. Thus, whereas the semantic meaning has its referent in external phenomena outside discourse, the pragmatic meaning refers to the internal development of the communication. For other authors however, the distinction between semantic and pragmatic applies primarily to the relations that provide coherence to a piece of text (Sanders 1992; 1997; among others). The principles behind semantic and pragmatic are the same, but since these authors are mainly concerned with the organization of discourse, they do not limit themselves to the explanation of individual words, but to the relations that operate in the coherent interpretation of a text. In particular, the distinction between semantic and pragmatic represents one of the primitive features to categorize relations: A relation is semantic if the discourse segments are related because of their propositional content (i.e. the locutionary meaning of the segments).... A relation is pragmatic if the discourse segments are related because of the illocutionary meaning of one or both segments. In pragmatic relation C(oherence) R(elation) concerns the speech act status of the segments. (Sanders, 1997:122) On the other hand, the term textual does not represent a characteristic of the meanings of linguistic expressions anymore. For authors like Traugott (1982; 1989), textual refers to a component of language which involves only aspects of the construction of a text, and basically has to do only with the ‘resources available for creating a cohesive discourse.’ (Traugott, 1982: 248). As we can see, this notion of textual could be understood as subsuming any aspect related to the creation of Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 112 coherence in discourse, and from this point of view, all types of coherence relations (semantic and pragmatic) and their lexical representation would necessarily be included in the textual component of language. In contrast to this definition of textual, in the context of the study of DFUs some other authors have applied this term more restrictively. For Schiffrin (1987) and Redeker (1987; 1990), textual is also a component of a model of language, but only in correspondence with the functions that DFUs may perform. Schiffrin talks about a plane in discourse (exchange structure) in which DFUs function as signaling transitions between speakers’ turns or adjacency pairs. In very similar terms, Redeker considers that there exists a textual or sequential discourse component in which DFUs perform a metalinguistic function, introducing new topics, facilitating the transition to an interruption segment, and so on. So whereas for Traugott textual refers to a component of language, for researchers in discourse studies the term textual represents a component of discourse in which DFUs perform mere sequential functions, where no semantic or pragmatic functions are involved. As I say, the concepts of semantic, pragmatic and textual functions have been used by some researchers to distinguish different types of DFUs. From a broad perspective on the notion of DFUs, Polanyi (1988) presents a classification of DFUs in three types depending on their function: assigners, connectors and discourse PUSH/POP markers. The first class is formed by assigners, which are forms that signal relevant social information in the interaction, such as who the participants are Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. or who is involved in the interaction. Proper names such as vocatives and expressions like hello would be included in this category. The second class, called connectors, is composed of those forms used to connect clauses and paragraphs together. In this category we find DFUs such as and or because. And finally, there is a third class of discourse PUSH/POP markers which ‘signal the embedding, continuation and returns to and from discourse constituents at the various levels’ (:604). Polanyi presents a broad classification of the functions of DFUs. The first category represents a function that deals with external aspects of discourse, and this makes this author consider forms such as hello and other vocatives as DFUs. On the other hand, the second and third categories represent the expression of discourse- internal functions: one class of DFUs connects clauses or longer sequences together, and the other signals connections among discourse constituents at different levels. In a somehow similar fashion, Fraser’s latest proposals (Fraser, 1999) distinguish two types of DFUs, somewhat close to Polanyi’s connectors and push/pop markers. The first group are DFUs which relate messages', a second group is the one composed of DFUs which relate topics. DFUs in the first group relate ‘some aspect of the messages conveyed by the segments S2 and S I’ (:946). Messages can be related in two different ways: involving the propositional content of the messages related, or on the other hand, involving their epistemic meaning. The second type of DFUs are those relating topics, and they involve exclusively the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 114 relations including different aspects of discourse management, such as topic return, topic introduction, or digression, among others. Fraser’s two groups resemble very much the last two classes of DFUs presented in Polanyi’s typology. In both cases, one type of markers connect individual clauses together {connectors in Polanyi’s terminology or message relating markers in Fraser’s), whereas the other type help manage the transitions in discourse {push/pop markers and topic relating markers, in these authors’ respective terminologies). Fraser is more specific in the case of message relations, following the semantic versus epistemic distinction proposed by Schiffrin (1987) and Sweetser (1990). 4.3.1. Types ofDFUs or types o f relations? With the exception of the first type of Polanyi’s categorization, which differs considerably from the functional definition of DFUs adopted in this dissertation, the other two types of markers clearly respond to the distinction between semantic, pragmatic and textual types of relations that I have already presented. However, unlike Polanyi and Fraser, I do not postulate the existence of different types of DFUs, but only of different types of relations. Within this framework, DFUs are only the lexical representations of coherence relations, so the properties and differences between semantic, pragmatic and textual domains correspond to characteristics of the relations themselves and not of the DFUs per se. In other words, the different characteristics observable in DFUs Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 115 are a reflection of differences in the relations they express. The obvious fact that DFUs present differences does not necessarily mean that there exist different types of DFUs; rather, it means that there are particular characteristics of the relations that the DFUs represent that make them behave in a certain way. One strong argument supporting this position is that a particular linguistic expression that functions as a DFU cannot be said to represent only one type of relation. If we claim that there are different types of DFUs, we should accept that there are different linguistic forms which correspond to each of these categories, that is, some forms would be used to represent semantic relations, others to represent pragmatic relations and others to represent textual relations. Thus, for example, we should expect a correspondence between a DFU like pero ‘but’ and the expression of only one type of relation: semantic or pragmatic or textual. However, most of the empirical studies performed across languages (Schiffrin 1987; 1992 and Redeker 1990; 1991 among others) suggest precisely the opposite, namely that the same linguistic forms are capable of representing semantic, pragmatic and textual relations. That is to say, we cannot establish an exclusive relation between particular forms and particular functions. So, the view defended here is that the same linguistic expressions are capable of representing all three types of relations, semantic, pragmatic and textual, and therefore we cannot single out one type of relation for one DFU. The empirical analysis on Spanish DFU presented in the next chapter should shed some light into this issue as well. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 116 In any case, it must be acknowledged that DFUs must be subject to different principles when representing different types of relations, precisely due to the differences in nature of each type of relations. I suggest that semantic and pragmatic relations contribute directly to coherence, whereas textual relations only do so in an indirect manner. A widely accepted idea in discourse studies on the interpretation of coherent discourse fragments (i.e., two DPUs connected by a coherence relation) is that the interpretation or content of the combined units is not equal to the sum of the individual contents of each unit. That is, the interpretation of the units is compositional and corresponds to the meaning structures underlying the discourse segments plus the meaning of the relation between those structures (Warner, 1985; Sanders et al. 1993). It is in this sense that we can say that depending on the source of coherence each type of relation contributes differently to the interpretation of the units. More precisely, I claim semantic and pragmatic relations contribute directly to the interpretation of the discourse fragments, whereas textual relations only affect the interpretation of the segments indirectly. Arguments in favor of the notion of constituency6 of discourse fragments can support this position. Warner (1985) suggests that the constituents of a discourse fragment show a high degree of interdependence. The first argument he uses in favor of this claim is precisely the notion of semanticity (Zwicky, 1978; Warner, 1985). The content of the combination of two segments connected by a relation is not equivalent to the sum 6 The notion o f constituency has sometimes been criticized of bringing a biased perspective into the notion of discourse structure in which only a semantic point of view is taken into account. (Cf. Martin, 1996). Here, we use the concept of constituency to show particular differences among types o f relations. No particular claims on the structure of discourse are intended. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 117 of the contents of each of the segments in isolation; that is, the truth value of the combined fragment is different from the truth values of its members. This seems to be true for semantic and pragmatic relations, but not for textual relations. In fragments related semantically or pragmatically it is possible to assign truth values to each of the members in the relation, but those values do not correspond to the truth value to the entire fragment. The truth value of the whole fragment is only partially dependent on the contribution of its segments. For example, cases like (10) or (11) presented by Warner show how the truth value of these sequences is only ‘contingently’ conditioned by the values of their members. In addition to them it is necessary to take into account the value of the relation as well: (10) Harry recently started dating Yetta. Not because all Mary’s hair fell out. (11) The peanuts weren’t delivered until midnight. So the elephants got restless. (Warner, 1985: 72) In (10), if the relation shown between the two members is true, that is, that Harry’s action is not caused by Mary’s hair falling out, then this fragment must be interpreted precisely in such a way that the truth value of the first member is realized regardless of the truth value of the second member. That is, Harry’s actions can be still true despite of Mary’s baldness not being true. And in (11) we find again that the truth value of the fragment does not correspond to the sum of the truth values of each member: (11) is true only if the relation between the two members is true (i.e., there is a direct cause-result relation between the elephants’ restlessness and their peanuts Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 118 not being delivered on time), that is, the first member must be true for the second member to be true. So, in semantically and pragmatically connected fragments what contributes to the entire truth value of the fragment is the values of the members related and the value of the relation and the conditions it imposes on the interpretation of the values of the members. This property, however, seems not to be true for textual relations. The values of the members related in a textual manner are not conditioned by the value of the relation itself. For example, in (12) (extracted from Fraser 1999) it is clear that the topic shift relation that joins both members does not affect their truth values. (12) I am glad that is finished. To return to my point, I’d like to discuss your paper. (:949) The interpretation of this fragment does not depend on the conditions posed by the relation of topic shift with respect to the values of units connected. The relation does not indicate an interdependence between them. Thus, it is possible to say that the sum of the values of each member alone is equal to the value of the whole fragment. The second most important characteristic in favor of the notion of constituency in discourse fragments which would again support the difference between semantic/pragmatic and textual relations is the property of inseparability, which states that the constituents connected by a coherence relation are not separable. In practice this property means that there is no possibility of introducing Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 119 any intervening material between the two constituents of a fragment unless this material is semantically interpreted as conjoined to the first member of the relation. Observe the examples proposed by Warner (1985), in which it is obvious that the insertion of a new clause between the members of (13) is only possible when its propositional content can be conjoined by and. (13) Professor Arid must stop assaulting coeds. Otherwise, he’ll be arrested. (14) Professor Arid must stop assaulting coeds. (and) he better stop brewing beer in his basement, too. Otherwise, he’ll be arrested. (15) " “Professor Arid must stop assaulting coeds. His wife is getting upset. Otherwise, he’ll be arrested. (Warner, 1985: 79). Another example is provided by (16) and (17) extracted from the epistemic and speech act categorizations of Sweetser (1990). (16) John loved her, because he came back. (17) What are you doing tonight? Because there’s a good movie on. In both cases only a unit semantically of the same type as the first member can be inserted within the two units as shown in (18) - (21). (18) John loved her and was willing to compromise, because he came back. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 120 (19) * John loved her, he must compromise, because he came back. (20) What are you doing tonight? Are you meeting your friends? because there’s a good movie on. (21) *What are you doing tonight? There is something I want to tell you, because there’s a good movie on. In (16) the first member of the relation is a conclusion based on the speaker’s knowledge. That is, to be able to obtain a coherent interpretation of the two units, this member can only have an epistemic reading. Thus only a unit that represents another conclusion based on the same speaker’s knowledge can be interpreted as conjoined to the first one, as in (18). This is the reason why (19) does not represent an interpretable sequence, as the unit inserted is not a conclusion of the speaker’s knowledge, but rather a description of John’s past love situation. In (20) and (21) again only a unit of the same type as the one in the first member of the relation can be inserted in between the two members, that is, a question speech act. The property of inseparability does not hold for units which are in a textual or sequential relation. A topic change, or the introduction of digressions or asides are relations that allow the insertion of new material with much fewer restrictions than the ones just seen for semantic and pragmatic relations. Observe the example in (22) extracted from Fraser’s typology: (22) I am glad that is finished. To return to my point, I’d like to discuss your paper. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 121 It seems that any type of new material can be inserted between two units connected by a topic shift relation such as the ones in (23)-(25). (23) I am glad that is finished because you and I were both tired of it. To return to my point, I’d like to discuss your paper. (24) I am glad that is finished. Now you must go on with your life. To return to my point, I’d like to discuss your paper. (25) I am glad that is finished. Do you know how many times I looked at it?. To return to my point, I’d like to discuss your paper. In (23) a unit joined by a semantic causal relation to the first member intervenes between them and the result is perfectly interpretable. In (24) and (25) two speech acts are inserted and still a coherent interpretation is possible. So again, the second characteristic that proves constituency of a discourse fragment can be said to be true for members that are connected through semantic and pragmatic relations, but not for units joined by textual relations. It is in this way that I suggest that semantic and pragmatic relations contribute directly to a coherent interpretation of a fragment, and textual relations only indirectly. The different status of the coherence relations must necessarily reflect on the behavior of DFUs. DFUs will be subject to some constraints posed by the specific relation they represent. I claim the most important constraint that types of relations impose on DFUs is their different behavior with respect to the property of extractability. I define Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 122 extractability in the following manner: For any DFU r representing some coherence relation R, r is extractable if and only if for any pair of discourse constituent units P and Q, the extraction of r yields exclusively R(P,Q). In other words, this property implies that a DFU is only extractable if the two constituent units can still be interpreted as connected only by the same relation and no other than the one the DFU expresses. Thus, in the example with pero previously introduced in (7) (repeated below in (26)), the impossibility of extracting the DFU without compromising the relation appears clear. As can be seen in (27) it is difficult to retrieve the same interpretation through the same coherence relation of concession after the DFU has been extracted. (26) la otra crema, yo la se hacer pero la tengo apuntada. (26’) the other cream, I know how to do it but I have it written down. F13L68 (27) la otra crema, yo la se hacer ( ) la tengo apuntada. (27’) the other cream, I know how to do it ( ) I have it written down. F13L68 What I propose here is that due to the different nature of the relations and their different contribution to coherence, DFUs may or may not be susceptible of extraction. In general terms I want to argue that since semantic and pragmatic Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 123 relations contribute to coherence directly, DFUs representing these relations will not be extractable in the majority of cases. Possible exceptions to this property will be those cases in which the linguistic and/or extralinguistic information provided in the discourse makes a relation the only possible one between two members. In contrast to semantic/pragmatic relations, DFUs representing textual relations will be susceptible of extraction more easily, since textual relations contribute to coherence only in an indirect maimer. Observe for instance the following example suggested by Fraser (1999) to illustrate a case of a topic relater DFU, where it is obvious that the extraction of the DFU does not cause a different interpretation: (28) This dinner looks delicious. Incidentally, where do you shop? (28’) This dinner looks delicious. () where do you shop? (Fraser, 1999: 949) Thus, to summarize, I propose that there are different types of DFUs, with different properties. Concretely, they behave differently with respect to the property of extractability in correlation with the type of source of coherence that the relation they represent is establishing. However, we suggest that these are differences caused by the specific type of relation holding between two segments of discourse, not by intrinsic properties of the DFUs themselves. A point that supports our position is the fact that the same linguistic expressions can occur expressing different relations, as we saw in the case of pero above. Therefore, unless we suggest that there are Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 124 homophonous forms capable of having different interpretations in discourse (i.e., two different types of pero, one for each type of relation it expresses, or two different types of pues), we must conclude that there are different types of DFUs only in relation to the type of coherence relation they represent. In the absence of any context it is not possible to establish whether a particular DFU is exclusively correlating with only one type of relation. For example, it is difficult to assert that pero represents only pragmatic relations, or that pues represents only semantic relations. 4.4. Summary. In this chapter I have argued that DFUs are to be defined as functional categories in discourse, which together with discourse propositional units are basic categories of discourse structure. DFUs are defined on the basis of the coherence principles and their functionality. DFUs are those linguistic expressions that act as the representation of the coherence relation holding between two constituent units. This definition relies primarily on a functional perspective which integrates DFUs as a part of an already well-established discourse framework. That is, this definition does not identify DFUs on the basis of their formal properties, but rather on the functions that these linguistic forms perform as hypothesized in a full-fledged theory of discourse. I argue that the category of DFUs is composed of all those linguistic expressions (regardless of their formal features) whose function is to make coherence relations explicit. By adopting this functional perspective we avoid some a prioristic Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 125 problems that other definitions face, such as the necessity of accommodating exceptional cases to the group of forms that share certain formal characteristics (cf. 2.1). Whereas DPUs represent the expression of information regarding events or states of affairs, or the speaker’s attitudes and beliefs with respect to those events, DFUs represent procedural meaning. Procedural meaning is understood as information on how those events and/or their discourse representation relate to one another (Blakemore, 1987; Fraser, 1997; 1999). Within this framework, the procedural meaning of a particular DFU corresponds to the meaning of the relation it most frequently represents. However, DFUs are not limited to the representation or expression of only one procedural meaning. In fact, as proven by the research carried out by Traugott and associates (Traugott 1982; 1989; and Konig and Traugott 1988), any linguistic expression is capable of incorporating additional meanings. In the case of DFUs, the expansion towards representing new relations is constrained by several principles. First, the meaning of the relation a DFU expresses must be compatible with the original meaning of the lexical form that became a DFU. Second, once a DFU is representing a certain relation, it can only express additional relations which are compatible in meaning. And finally, a DFU can only represent new relations which are equally or less specific than the one it is already representing. Besides these constraints that regulate the expansion range of DFUs in the discourse coherence relations they may express, there are additional restrictions of a different nature on DFUs still associated to the coherence relations DFUs express. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 126 The distinction between semantic, pragmatic and textual types of relations has been taken by some researchers as the basis of the categorization of DFUs into different types. The definition of these three terms refers to the different types of connections that can be established between two DPUs, which in larger scale correspond to different components of language (Halliday and Hasan, 1977; Martin, 1985; among others). The semantic domain includes the interpretation of the propositional content of the expression, which is also called external meaning. On the other hand, the pragmatic domain refers to the expression of attitudes, beliefs and emotions towards what is being expressed. Thus, whereas the semantic meaning has its referent in external phenomena outside discourse, the pragmatic meaning refers to the internal development of the communication act. Finally, the textual component can be understood first from a general language framework as the component which includes all the phenomena related to the construction of text, in other words all the resources available in the language to provide a cohesive and coherent interpretation of a text. And second, the textual component can be understood from a more discourse oriented perspective in which this component refers to all those coherence relations that do not express semantic or pragmatic meaning but only pure sequential transitions in discourse. For some authors like Polanyi (1985) or Fraser (1999) these concepts represent a valid principle to draw a categorization of DFUs, and they establish a typology of DFUs creating a correlation between particular linguistic forms and these three types of relations. In the view defended here DFUs can be divided in different groups corresponding with the type of relation they represent as Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 127 well, but I argue that it is not possible to establish an exclusive correlation between a particular form and only one type of relation. Any DFU is capable of representing semantic, pragmatic and textual relations. The differences we see in DFUs with respect to these three types correspond to differences in the relations themselves and not to intrinsic properties of the DFUs. As we saw, there are different features that characterize a relation according to their source of coherence; these features prove that semantic and pragmatic relations contribute to the coherence of the related units in a direct fashion whereas textual relations only do so indirectly. Segments connected semantically or pragmatically form constituency whereas textually connected segment do not, and thus, as stated above, this different contribution is directly responsible for the fact that only those DFUs which represent textual relations are susceptible of being extracted or omitted, i.e., not expressed with the relation still being interpretable. Finally, to conclude this chapter, what I presented up to this point constitutes the theoretical framework adopted in this dissertation, created on the basis of coherence relations and their implication in the definition of the basic types of units in discourse. I have outlined in theoretical terms the definition, the properties and the constraints DFUs undergo. Now, the necessary step to take is to test this theoretical model against natural data. In the next chapter I present the study I carried out on a sample of 68 natural conversations in Peninsular Spanish and the results we obtained in our empirical analysis. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 128 CHAPTER 5. Empirical Study of DFUs on Modern Spanish. 5.1. Introduction. In the last two chapters I have presented in detail the theoretical framework of discourse analysis adopted in this dissertation, grounded on the notion of discourse relations. With this framework in mind, the theoretical claims made needed to be tested empirically. In order to do that I set up a study on natural data in which I analyzed the interaction between actual DFU forms, the relations they expressed and their frequency of occurrence. Also the proposals on the properties of DFUs presented above were also empirically tested. This chapter presents the analysis of natural oral Spanish discourse samples extracted from the Corpus de Referencia del Espanol Actual (CREA). The chapter is organized as follows: section 5.2 presents the characteristics of the corpus analyzed. In section 5.3 I discuss the methodology used in the analysis of data. Section 5.4 presents the quantitative and qualitative results obtained. And finally in section 5.5 I provide a summary of the main conclusions derived from this analysis. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 129 5.2. Data Source. 5.2.1. The corpus. The corpus used for this study is based on oral discourse samples compiled in the CREA (Corpus de Referencia del Espanol Actual). The CREA is composed of a series of texts which include oral and written contemporary samples collected by researchers from the Universidad Autonoma de Madrid (Autonomous University of Madrid, Spain) in different Spanish speaking countries. The CREA contained a large selection of Spanish materials from very different sources, so certain parameters needed to be determined first before selecting the parts to be analyzed. As it is recognized by researchers who work with actual language data (Biber, 1994), it is not possible to talk about characteristics of a discourse without specifying the type of discourse we refer to. That is, different types of discourse reveal different characteristics. For example, the study of oral narratives, the study of arguments, or the study of interviews will show different properties. Some researchers have suggested that more subdivisions must be made even within each of these specific areas. Polanyi (1982) proposed to introduce further distinctions within the genre of oral narrative, based on common social variables such as speaker differences or turn-taking. The main idea here is that the validity of the conclusions extracted from the analysis of real language data is conditioned by the degree of control we have over the variables that characterize it. Certain homogeneity in the characteristics of the corpus is therefore required to obtain reliable results. It is not possible to draw generalizations from the analysis of a Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 130 set of data for which variables such as speaker age, speakers’ social and educational level or written/oral distinction have not been controlled for. In the same way, the conclusions of this analysis will be applicable only to the characterization of the data constrained by similar variables. Biber (1994) proposed several groups of parameters to define the characteristics of the corpus to be analyzed. He distinguishes seven situational parameters: Communicative characteristics o f participants', relation between addressor and addressee', setting', channel', relation o f participants to the text', purposes, intents and goals', and topic/subject (:40). Each of these parameters constitutes a first level of generalization which realizes concretely through particular values. For example, the participants characteristics are defined by the number of addressors/addressees* and the presence or absence of an audience; the relations between addressor and addressee include the types of social relations between the participants, the extent of shared knowledge, interactiveness and personal relationship. (Op.Cit.:40). For the purpose of my analysis, the first parameter I set to choose was the channel. Since this is mainly a study of oral discourse, only oral samples were extracted from the CREA. However, in order to obtain the desired level of homogeneity I mentioned before, further parameters needed to be set to classify these oral data. The oral samples in CREA included TV reports, news, documentaries, radio/TV interviews and conversations, among other types of discourse. Although all these registers belong to the same group of oral, a first distinction can be made between TV/radio programs and natural conversations. They Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 131 differ primarily in their mode o f transmission-, TV/radio programs use the mass media to transmit to communicate, whereas the natural conversation represents a form of face-to-face communication. Additionally, these two groups differ in three other fundamental aspects: purpose, relation among participants and relation between participants and text. First, the purpose of radio/TV news, reports and documentaries is to transfer information and/or entertain, whereas the purpose of natural conversations in general is the exchange of information. The fundamental distinction in purpose is also associated to the second aspect, the different relations the participants establish in each of these registers. In reports, news and documentaries there is a removed general audience, and no concrete individual represents the audience at the moment in which the actual act of communication takes place to offer a response in real time. In this respect and to a certain extent, TV/radio communication represents a type of monologue. In conversational settings on the other hand, there is an immediate audience who is able to interact with the other participants in real time; conversations are characterized by their dialogic mode. Finally, these two registers differ in the relation between the participants and the text, as the circumstances of production are very different. Namely, while natural conversations represent spontaneous communication, news, documentaries and reports show a higher degree of planning. All these differences must necessarily have an impact on the use of particular language structures. In the case of the analysis of DFUs, it is obvious that the expression of certain relations such as requests, confirmations, or interruptions, Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 132 which DFUs make explicit, may not be represented at all in non-interactional communication. In interactional contexts such as natural conversation, on the other hand, a frequent appearance of these relations could be found. However, besides the primarily dialogic mode that characterizes conversations, this type of oral communication may represent as well different modes of monologic discourse such as narrations, descriptions, explanations, justifications, and other interactional exchanges such as arguments, question-answer pairs, challenges, etc. (Schiffrin, 1987). Such richness in types of discourse is the main reason why natural conversations where chosen as the basis of the analysis. In addition to register characteristics, the other most important variable to control for in the analysis of language use is the speakers’ dialectal variety. In the case of Spanish, there are substantial lexical, grammatical, semantic and pragmatic differences among the varieties spoken throughout the Hispanic world. In particular, in the study of oral discourse, it is difficult that such variable phenomena as DFUs match in frequency and use in the different varieties. Therefore, the corpus must limit itself to a single variety. The one chosen for this study was the variety of Northern Peninsular Spanish spoken in Madrid and its surroundings. Finally, all the participants must show similar characteristics with respect to age and socio-economic background. Having into account all of these parameters, 74 conversations were extracted from the CREA. They corresponded to natural conversations in which in most of the cases the speakers were unaware of the presence of the tape recorder. The Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 133 conversations were held between 2 and 5 participants and were recorded in familiar settings, in Madrid and its surroundings. All the participants in the recordings were adults between 25 and 55 years of age, and belong to the ample socio-economic Spanish middle class (See Sawoff (1980) for appropriate definition of the Spanish middle class in modem Spain). The total amount of data selected from the corpus corresponds approximately to 16 hours of recording. The conversations analyzed differed in length from each other, but I calculated approximately a mean of 15 minutes for conversation. Out of the 74 conversations initially chosen, 4 conversations were excluded from the analysis; these were recordings of university talks, which apart from a brief question- answer period at the end represented long academic monologues. Additionally, the type of formal register this language represents contrasted greatly with the language used in familiar conversations, which represented the majority of the corpus. Finally, 1 more conversation had to be excluded. In this case, the interaction between the participants was conditioned by the requirements of the situation. The members of an electoral polling place were calling the voters’ names, and confirming one by one their personal information. The conversation presented no interaction either between the polling place members and the voters or among themselves. Therefore I excluded this recording as well. Consequently, the results of the data presented in this chapter correspond to a total of 68 conversations which counted 95,683 words. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 134 5.3. Method of analysis. The transcription of the data was already available in the CREA, along with some parts of the recording. A second transcription was made on those parts for which audio recordings were available to measure accuracy and reliability. The transcriptions were made paying careful attention to linguistic (textual and discourse) and contextual (situational) information. Prosodic information was also conveyed in the transcripts; pauses, lengthening and reduction of stress were marked. A second transcription by a second transcriber was contrasted with the first one in order to confirm reliability. 5.3.1. The coding process. To accomplish the coding of all the information I was interested in for each of the DFUs encountered, the data was first divided into DPUs, following the criteria presented in chapter 3 (Cf. Sect. 3.3.3.). Edus were identified as the minimum segments that relate to other segments through a coherence relation. As I explained in chapter 4, I consider DFUs to be functional categories that make these relations explicit. Concretely, DFUs were identified as those forms that appeared at the boundaries of discourse units expressing a coherence relation between them. Example (1) illustrates a case of the identification of the DFU pero ‘but’ expressing a pragmatic relation of concession: contrary to expectation. Context: Speaker 1 (SI hereafter1 ) is telling the story of how she lost track of her glasses for several days. She looked for a safe place to 1 In all the examples extracted from the Spanish corpus each speaker is identified by the letter H Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 135 leave them when she was about to take a shower and she put them inside her husband’s bathrobe pocket. Then she forgot about it. (1) <H1> 1 En el momento en el que vas a guardar, 2 que no sabes donde 3 ipum!, aprovechas aquello. 4 Pero no es una cosa que hayas pensado hacerla ni mucho menos, 5 y luego, no te vuelves a acordar... (1 ’) <S1> 1 The moment you are going to put it in a safe place, 2 that you don’ t know where 3 boom!, you take advantage o f anything. 4 B ut it is not something you thought o f doing , o f course not, 5 and then, you won’ t remember again... F15L1 SI introduces in line 4 a commentary about the nature of these mistakes; contrary to what one may think, i.e., that they involve some thinking, she states that ‘it is not something you thought of doing’, that is, they are thoughtless actions. Pero ‘but’ in this case is the DFU that expresses the relation between units 1, 2, 3 and unit 4. It establishes the contrast between the possible expectation the listener can make from the content of units 1, 2 and 3 (‘these are planned actions’), and the statement in 4 (‘they are totally spontaneous actions)’. The coherence between these units is of a pragmatic nature since what is related is not the content of the units, but rather the expectation created by the content of units 1, 2 and 3 and the denial of that expectation in 4. (hablante ‘speaker’) and a number assigned by the transcriber to every member in the conversation as they first intervene. I have kept the original participants numbers as they appear in the transcription. The English translation that I provide after each Spanish example identifies each speaker as S plus their corresponding number. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 136 Each instance of a DFU was coded alone if there was one identifiable coherence relation expressed by that form. However, in some cases a cluster of several DFUs appeared together and only one relation could he identified. In those cases, the cluster was coded as one instance of a complex DFU which expressed one relation. For example, in (2) the forms pues ‘so’ and entonces ‘then’ appear together and only one possible relation of consequence!result could be identified. These two forms were identified as one instance of the DFU cluster pues entonces, which expressed just one relation. Context: S3 is talking about the differences between the places she and her friends used to go when they were teenagers and where young people go now. (1) <H3> Estaba mas de moda entre la gente de tu edad que tu conocias que era (del barrio) pero ahora los ninos que tiene diecisiete o dieciocho anos <H5> Todos se van a <ininteligible> <H1> Si, <ininteligible>. Pues entonces debe ser... de generaciones. (2’)<S3> It was more fashionable among people your age, people that you knew, who were from your neighborhood. but now seventeen or eighteen year-olds <simultaneous> <S5> They all go to <unintelligible> </simultaneous> <S1> Yes, <unintelligible>. So then it must be generational. F27L68 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 137 For each of the forms encountered several variables were coded. First, each example was formally coded according to its file number, the line in which it appeared, the number of speakers in the conversation and the speaker who uttered it. Additionally, five other non-formal variables corresponding to the characteristics of DFUs described in chapter 4 were assigned. 1. Types o f units and relations. First, I was interested in coding the types o f units a DFU was relating, that is, the contexts in which a specific form appeared. For example, (3) illustrates a case in which the form pues ‘so’ appears in a very frequent context: a question-answer pair. (3) <H1> 1 f Y que vamos a hacer con la aDFUinistracion? <H3> 2 Pues lo que dispongais. (3 ’) <S1> 1 And what are we going to do with the administration? <S3> 2 So/well what you decide. F25L13 In this case, pues ‘so’ was coded as a DFU which relates the unit to its right -unit 2, the answer- to the unit to its left -unit 1, the question . Additionally, I coded the relationship the right unit bore with respect to the unit and/or units to its left. I considered that for the majority of the cases the member that 2 When I talk about right and left unit I refer to a hypothetical progressive representation o f the units. For example, in (I) below, given a representation that reflects the progression in production and/or writting, the unit introduced by pues ‘so’ is placed to the right o f unit 1, since 1 is produced first, and 2 is produced later in time. (I) <H2> 1 Yo no lo entiendo. ^Como lo hacen? <H3> 2 Pues con un... <simultaneo> con una operacion (I’) <S2> 1 I don’ t understand. How do they do that? <S3> 2 So/well with.. <simultaneous> with surgery F26L25 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 138 bears the relationship expressed by the DFU is the member headed by the DFU. Although there is still an ongoing debate in the literature on how DPUs are related to one another, for practical reasons I decided to adopt the Rhetorical Structure Theory principle (Marcu, 1998), which states that the direction of the relation between units in discourse is generally made to the left, that is the member headed by the DFU attaches to a previous unit. Nevertheless, adopting this decision does not imply a defense of the theory that discourse structure is built incrementally, that is, that each new unit is processed as soon as it appears and is immediately attached to the previous unit, adjacent to its left. Recent experiments in computational treatments of discourse (Marcu, Romera and Amorrortu, 1999) show that when human annotators try to attach upcoming material to previous known text, in at least 20% of the cases they have doubts about where to make the attachment. Additionally, this experiment showed that in general multiple partial sets of units are created first by the analyst and then assembled to one another. These conclusions seem to indicate that at least in the process of analyzing discourse an incremental model of assignment of relations may not be a correct approach. Despite these conclusions, however, Marcu, Romera & Amorrortu’s 1999 experiment also confirms the Rhetorical Structure Theory that in the majority of cases the pragmatic relation is made from right to left. In almost all cases studied, it is the upcoming material the one that has to be related to the previous content and not vice versa. For example, in (4) from the corpus, it is clear that the relation of Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 139 adversativity is established in the right-to-left direction between the complex unit composed of 5, 6, 7 and 8 and unit 3 throughpero ‘but’. (4) <H2> 1 Si lo van a arreglar todo, 2 si, pintar y todo... escaleras y todo 3 Yo queria haber pintado esta puerta ya como la de abajo, 4 estan dentro [5 pero parece ser que hay que... 6 vamos a acordar una norma de color en todas igual o tal 7 porque estos lo pintaron por su cuenta 8 los del piso de abajo] (4’) <H2> 1 yes, they are going to fix it all, 2 yes, they 1 1 paint and everything...stairs and all 3 I wanted to paint this door like the one downstairs, 4 they are inside [5 but it seems that we need to... 6 we are going to agree on a color norm the same for all 1 because these guys did the painting on their own 8 the guys downstairs} F31L182 The one exception that we find to the right to left directionality of the relations is represented by conditional relations. In general, the unit that establishes the relationship is the unit to the left with respect to the unit to the right. For example, in (5) it is clear that unit 2 expresses a condition relation with unit 3 and not with unit 1, as it would be expected if the right-to-left directionality principle applied here. (5) <H1> 1 Depende del horario del vuelo. 2 Si el vuelo sale por la manana, 3 adios mi... mi... mi plan Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 140 (5’) <S1> 1 It depends on the time o f the flight. 2 I f the flight leaves in the morning, 3 I can kiss my...my...my plan goodbye F14L65 The DFU si ‘i f appears in this case at the unit boundary, just like in the right- to-left relations, but it does not relate the unit it is heading with the unit to its left, but rather with the following unit. However, even conditional sentences may show the right-to-left pattern as well, for example, in question formulas such as (6), or in cases in which a speaker introduces a condition for the realization of the content of a previously said unit, as in (7). (6) <H2> jQuemal! Y si no... ^quete... que te puedepasar si te pillan en esas cosas? (6’) <H2> How bad! And if not... what can ... what can happen to you i f they catch you in those things? F19L17 (7) <H1> Si. Luego a la salida si quieres te la paso. <H4> Si no te la han roba<(d)>o... <H1> Si no me lo han manga<(d)>o. (7’) <S1> Yes. Later, when we leave, if you want, I ’ ll give it to you. <S4> I f they didn’ t steal it from you <S 1 > I f they didn't steal it from me. F15L14 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 141 In addition to coding the types of units that DFUs related, I also coded the role the unit headed by the DFU played with respect to other units, in the majority of the cases the right unit to the unit/units to its left, or in the exceptional case of conditionals the left unit with respect to the right3. For example, in (8) the unit to the right of the DFU entonces ‘then’ was coded as being in sequential relation with the previous units (2 and 3). Context: SI is explaining the confusion they experienced at the beginning of a friend’s wedding. (8) <H1> 1 Llegaba... 2 entraban ellos 3 pero todavia estaban saliendo los novios anteriores 4 entonces se quedaron parados aqui 5 hasta que salieron los novios y se marcharon (8’) <S1> 1 I was getting there... 2 they were entering (the church) 3 but the previous couple were still inside 4 then they stayed still there 5 until the couple exited and left. F12L77 2. Source o f coherence. The next variable coded referred to the nature of the relation created by the DFU, that is, whether the source of coherence was semantic, pragmatic or textual. In chapter 3 (Cf. Sect. 3.1.4.2) I defined semantic relations as those in which the propositional content of the units is connected; pragmatic relations as those which represent an involvement of the speaker by expressing her believes or attitudes towards the reality of the events expressed in the discourse; and 3 There are additional relations for which the right-to-left pattern does not seem to be the general behavior, for example the case o f pragmatic relations o f background, enablement, and some instances of justification. However, none of these relations have a clearly identified representative DFU, therefore those cases were not considered as exceptions to the principle adopted. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. textual relations as those that facilitate transitions to a commentary, correction, new topic, etc., that is, those that introduce argumentation mechanisms (See Chapter 3 for illustrative examples of each type of relation). 3. Extractability. Finally, I coded another important variable regarding the relation between the DFU and the coherence relation expressed: whether the DFU was extractable with respect to the relation it expressed. In Chapter 4 I provided the following definition of extractability for DFUs: (9) for any discourse marker r representing some coherence relation R, r is extractable if and only if for any pair of discourse constituent units P and Q, the extraction of r yields exclusively R(P,Q). For example, in (10) I considered that the extraction of the DFU pero ‘but’ did not preserve the relation of objection established between 5 and the speech of SI (units 1-4). Context: S2 is watching apartments for sale. She lives out of Madrid with her family and her parents are buying an apartment for her and her brothers and sisters now that they have started college in the city. (10) <H1> 1 que mira, ojala la hubiera cogido sin hacer obra 2 y la hubiera hecho a su gusto. 3 Y la habia salido tan barata. 4 Pero bueno. <H2> 5 Pero... pero hay que estar aqui tambien pendiente (10’) <S1> 1 look, I wish she had taken it before the reforms 2 and she had made them as she pleased Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 143 3 and it would have been really cheap for her 4 but it’ s OK. <S2> 5 But... but one needs to be here all the time and... F33L34 In 1-4, SI (the owner of the apartment) mentions that if her mother had bought the house before ordering for the reforms it would have cost her much less and she would have been able to reform it her own way. S2 then introduces an objection to S i’s mother having been able to take responsibility on reforms, as making reforms implies being on site constantly. The objection relation is established primarily through pero ‘but’. Observe that in (11) the objection relation becomes less clear once the DFU pero ‘but’ is taken out. 1 que mira, ojala la hubiera cogido sin hacer obra 2 y la hubiera hecho a su gusto. 3 Y la habia salido tan barata. 4 Pero bueno. 5 hay que estar aqui tambien pendiente y ... 1 look, I wish she had taken it before the reforms 2 and she had made them as she pleased. 3 And it would have been really cheap. 4 But it’ s OK. 5 one needs to be here all the time and... F33L31 Without pero ‘but’, unit 5 can be still interpreted as an objection to 1-4, but it can also be interpreted as a justification as to why her mother did not buy the apartment (11) <H1> <H2> (11’) <S1> <S2> Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 144 and make the reforms on her own. That unit 5 can also be a justification to 1-4 becomes more apparent when we replace pero by the DFU es que ‘that is’4. (12) <H1> 1 que mira, ojala la hubiera cogido sin hacer obra 2 y la hubiera hecho a su gusto. 3 Y la habia salido tan barata. 4 Pero bueno. <H2> 5 es que hay que estar aqui tambien pendiente y ... (12’) <S1> 1 look, I wish she had taken it before the reforms 2 and she had made them as she pleased. 3 And it would have been really cheap. 4 But it’ s OK. <S2> 5 ‘cause one needs to be here all the time and... On other occasions, however, in the expression of the relation of objection, pero can be extractable. Observe in (13) how the omission of pero does not affect the objection relation between 8 and 2-6. (13) <H3> 1 Y empieza el Coco, dice el Coco: 2 "Hostia, pues ahora venden unos machetes de puta madre 3 y con el mando desmontable 4 y que se abre... 5 y llevan..." <risas> "... 6 y llevan aguja..." <risas> "... y dedal." <H4> 7 jOstras! <H3> 8 <risas> j Hostia! "jPero, tio, ^como te vas a comprar eso?" (13’) <S3> 1 And Coco starts saying, Coco says: 4 I have not found an appropriate English translation for es que. I believe that the closest meaning in this case is represented by ‘because’. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 145 2 "man, so now you can buy some real good machetes 3 and with an off handle 4 and one that you can open... 5 and they come... <laugh>... 6 and they come with a needle. <laugh> and a thimble." <S4> 7 Wow! <S3> 8 <laugh> Man! “ but, man, how can you buy something like that? ” F28L181 In (13) even when pero has been extracted there is no other possible interpretation of the question expressed by S3 than an objection to the content of the indirect speech introduced in units 2-8. Notice in (14) that the substitution of pero by any other DFU (y ‘and’, pues ‘so’, entonces ‘then’) is not possible without compromising the coherence of the discourse. (14) <H3> 1 Y empieza el Coco, dice el Coco: 3 "Hostia, pues ahora venden unos machetes de puta madre 3 y con el mando desmontable 4 y que se abre... 5 y llevan..." <risas> "... 6 y llevan aguja..." <risas> "... y dedal." <H4> 7 jOstras! <H3> 8 <risas> jHostia! " i *y/*pues/*entonces, tio, ^como te vas a comprar eso?" (14’) <S3> 1 And Coco starts saying, Coco says: 2 "man, so now you can buy some real good machetes 3 and with an off handle 4 and one that you can open... Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 146 5 and they come... <laugh> ... 6 and they come with a needle. <laugh> and thimble." <S4> 7 Wow! <S3> 8 <laugh> Man! ‘‘*and/*so/*then, man, how are you going to buy something like that? ” The issue of the conditions under which a DFU can be extracted is discussed in detail later in section 5.4.3. In sum, each instance of a DFU was coded first by several formal variables such as file number, line of occurrence, number of speakers in the conversation, and speaker uttering the DFU, the type of unit to left and right of the DFU and the relation expressed by the right member, the type of source of coherence established: semantic, pragmatic or textual, and finally whether the DFU was extractable and the meaning of the DFU. (18) illustrates the coding of the DFU o sea ‘that is’ in example (17). (17) <H1> por eso te hablaba yo de que una habitacion, queria hacer yo una habitacion de estas y haber hecho un armario con, asi con la puerta. <H2> Si. <H1> <?,Sabes? Que tenia yo ya hecho el dibujo y todo. 0 sea, cogiendo de la puerta la... aquella de la habitacion de alia... (17’) <S1> that’ s why I told you that a bedroom, 1 wanted to make a bedroom like that and make a cabinet with, like that with the door. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 147 <S2> Yes <S1> You know? I even had the drawing made and everything. that is, beginning from the door... that one o f that room... F33L45 (18) FORM FILE LINE SPKR #SPKR OSEA 33 45 1 2 LEFT RIGHT COHER EX STATEMENT ELAB-SPECIFIC SEMANTIC EXTRACTABLE The coding was done by two different annotators. The first annotator coded all instances of DFUs found in the data. A second coder following the guidelines on coding as the first annotator coded 15% of the data to assure reliability. The agreement reached by the two coders was of 81.64%. 5.4. Results and discussion. In this section I present the results I obtained after the coding and analysis of the DFU instances identified in the data. I focus first (Sect. 5.4.1) in the number of DFUs found and the particular high frequency of reduced group of forms that account for 76.7% of the data. I explain what I believe to be the reason for this high frequency in the light of the interaction between contextual constraints and specificity in DFU’s meaning. Next I explore the particular relations each of these high frequency forms represents in the data within the semantic, pragmatic and Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 148 textual domains (Sect. 5.4.2). Finally, in Sect. 5.4.3 I undertake the issue of the extractability of DFUs as I account for the results encountered in the analysis. 5.4.1. Forms and frequency. In the 68 files analyzed a total of 2,292 instances of 108 DFUs were coded. Figure 1 shows the forms encountered and their frequency. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 149 Form Frequency Percent PERO 439 19.2 Y 298 13.0 PORQUE 256 11.2 PUES 224 9.8 ESQUE 174 7.6 OSEA 122 5.3 ENTONCES 121 5.3 BUENO 73 3.2 SI 51 2.2 CUANDO 48 2.1 YLUEGO 41 1.8 COMO 37 1.6 PARA 36 1.6 Y ENTONCES 33 1.4 ADEMAS 26 1.1 OYE 26 1.1 VAMOS 26 1.1 LUEGO 22 1.0 ES DECIR 19 .8 AUNQUE 13 .6 BUENO PUES 12 .5 LO QUE PAS A ES QUE 12 .5 Y ADEMAS 11 .5 SINO 9 .4 ASI QUE 8 .3 POR EJEMPLO 8 .3 PORQUE ES QUE 8 .3 TOTAL 8 .3 0 7 .3 QUE 5 .2 AHORA 4 .2 PERO ES QUE 4 .2 PUESENTONCES 4 .2 EN VEZ DE 3 .1 PORESO 3 .1 PRIMERO 3 .1 QUIERO DECIR 3 .1 YAHORA 3 .1 YBUENO 3 .1 YNADA 3 .1 AHORABIEN 2 .1 APARTE 2 .1 BUENO ENTONCES 2 .1 DEPENDE 2 .1 DESPUES 2 .1 HASTAQUE 2 .1 MIRA 2 PERO BUENO 2 PERO VAMOS 2 .1 SEGUNDO 2 .1 SIN EMBARGO 2 .1 TAMBIEN 2 .1 VAMOS A VER 2 .1 YA QUE 2 .1 Y ASI 2 .1 Y CUANDO 2 .1 Figure 5.1. Forms and frequency of DFUs Form Frequency Percent Y DESPUES 2 .1 Y EN FIN 2 .1 AL 1 .0 A NO SER 1 .0 ATODO ESTO 1 .0 CON LO CUAL 1 .0 CUARTO 1 .0 DE TODOS MODOS 1 .0 EINCLUSO 1 .0 ENCIMA QUE 1 .0 EN LUGAR DE 1 .0 EN PRIMER LUGAR 1 .0 EN SEGUNDO LUGAR 1 .0 ENTRE 1 .0 ESMAS 1 .0 LO QUE OCURRE- ESQUE 1 .0 LO QUE QUIERO DECIR- ES QUE 1 .0 LUEGO ADEMAS 1 .0 LUEGO DESPUES 1 .0 MASLUEGO 1 .0 MIENTRAS 1 .0 NO OBSTANTE 1 .0 OBIEN 1 .0 0 BUENO 1 .0 OYEBUENO 1 .0 OYE UNA COSA 1 0 PERO ADEMAS 1 0 PERO AHORA 1 .0 PERO LO QUE PASA- ES QUE 1 .0 PERO SI 1 .0 POR 1 .0 POR OTRO LADO 1 .0 PUESBUENO 1 .0 PUES ESO 1 .0 PUES ES QUE 1 .0 PUES MIRA 1 .0 PUESNADA 1 .0 PUES SI 1 .0 SEGUN 1 .0 SI ES QUE 1 .0 TERCERO 1 .0 UNA VEZ QUE 1 .0 YA 1 .0 Y A PARTIR 1 .0 Y ASI QUE 1 .0 YCLARO 1 .0 Y INCLUSO 1 .0 Y POR LO TANTO 1 .0 Y POR TANTQ 1 .0 Y SOBRE TODO 1 .0 Y UNA VEZ QUE 1 .0 Y YA LUEGO 1 .0 Total 2292 100.0 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 150 The first important characteristic that we notice in Figure 1 is that only 9 of the 108 forms account for 76.7% of the results. These are pero ‘but’, y ‘and’, porque ‘because’, pues ‘so’, es que ‘is that’, o sea ‘that is’, entonces ‘then’, bueno ‘well’, si ‘if. The other 99 forms account only for the remaining 23.3%. These results are not surprising. Some of these forms have been the focus of different studies in Spanish ever since the first attempts to analyze DFUs. Among them y, bueno, pero, pues have received special attention (Fuentes Rodriguez, 1993; Martin Zorraquino, 1994; Montolio Duran, 1991; Moya Corral, 1996; Porroche Ballesteros, 1993, 1996; Portoles, 1989), whereas other forms such as es que, porque, osea have remained less studied. Although not many of these studies incorporate quantitative data in their analyses, it seems that there is enough justification supporting the attention that has been paid to these forms, namely that they are by far the most frequent DFUs in conversational Spanish. On the other hand, forms such as sin embargo ‘although’, no obstante ‘in spite o f, al fin y al cabo ‘in the end’, bien ‘good’, es decir ‘that is to say’, esto es ‘that is’ which are usually studied along with bueno, pero, or pues (Casado Belarde, 1991; Montolio Duran, 1992 among others) are very infrequent in the corpus, if they appear at all. With respect to the relations expressed by these 9 forms, I also found that they are used to express a wide range of relations. Besides this, they are the most frequent forms by far for each of the relations analyzed. For example, pero is the most frequent form to express objection, adversativity, contrast, concession, denial o f expectation, request, and clarification. Figures 2-8 show the differences encountered in frequency between pero and other forms that express the same relations. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 151 form Frequency Percent Valid Percent Cumulative Percent Valid PERO 87 91.6 91.6 91.6 PEROAHOR 1 1.1 1.1 92.6 SINO 6 6.3 6.3 98.9 YENFIN 1 1.1 1.1 100.0 Total 95 100.0 100.0 Total 95 100.0 Fig.5.2. Adversative form Frequency Percent Valid Percent Cumulative Percent Valid PERO 77 70.6 70.6 70.6 PUES 12 11.0 11.0 81.7 Y 9 8.3 8.3 89.9 BUENO 3 2.8 2.8 92.7 LQPEQ 3 2.8 2.8 95.4 AHORA 1 .9 .9 96.3 ENVEZDE 1 .9 .9 97.2 LQOEQ 1 .9 .9 98.2 PEROESQU 1 .9 .9 99.1 PEROVAMO 1 .9 .9 100.0 Total 109 100.0 100.0 Total 109 100.0 Fig. 5.3. Contrast form Frequency Percent Valid Percent Cumulative Percent Valid PERO 37 59.7 59.7 59.7 AUNQUE 13 21.0 21.0 80.6 SINO 3 4.8 4.8 85.5 AHORA 2 3.2 3.2 88.7 SINEMBAR 2 3.2 3.2 91.9 VAMOS 2 3.2 3.2 95.2 NOOBSTAN 1 1.6 1.6 96.8 PEROVAMO 1 1.6 1.6 98.4 YENTONCE 1 1.6 1.6 100.0 Total 62 100.0 100.0 Total 62 100.0 Fig. 5.4. Concession Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 152 form Frequency Percent Valid Percent Cumulative Percent Valid PERO 37 97.4 97.4 97.4 PEROVAMO 1 2.6 2.6 100.0 Total 38 100.0 100.0 Total 38 100.0 Fig. 5.5. Contrast with expectation form Frequency Percent Valid Percent Cumulative Percent Valid PERO 22 84.6 84.6 84.6 ENTONCES 1 3.8 3.8 88.5 PEROESQU 1 3.8 3.8 92.3 SIESQUE 1 3.8 3.8 96.2 VAMOS 1 3.8 3.8 100.0 Total 26 100.0 100.0 Total 26 100.0 Fig. 5.6. Clarification fo r m Frequency Percent Valid Percent Cumulative Percent Valid PERO 88 83.0 83.0 83.0 ESQUE 9 8.5 8.5 91.5 LQPEQ 5 4.7 4.7 96.2 PEROESQU 2 1.9 1.9 98.1 ADEMAS 1 .9 .9 99.1 PEROLQPE 1 .9 .9 100.0 Total 106 100.0 100.0 Total 106 100.0 Fig. 5.7. Objection form Frequency Percent Valid Percent Cumulative Percent Valid PERO 34 81.0 81.0 81.0 OSEA 4 9.5 9.5 90.5 ESQUE 2 4.8 4.8 95.2 Y 2 4.8 4.8 100.0 Total 42 100.0 100.0 Total 42 100.0 Fig. 5.8. Request Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 153 As we can observe, pero in every single relation represents at least 75% of the cases appearing in the data. The other DFUs that compete with pero in many cases represent combinations of pero with other lexical items {pero ahora ‘but now’, pero es que ‘but is that, pero lo quepasa es que ; ‘but what is going on is that’, among others). In other cases they are forms that seem to be more restricted to specific contexts, for example sino ‘but rather’1 , which can only appear when the first member of the relation has been negated, aunque ‘however’, or no obstante ‘nevertheless’. This seems to be the case with most of the DFUs that appear competing with these 9 forms. In qualitative terms, it seems that in every case the less frequent forms often represent the same relation as the most frequent forms, but the speaker restricts them to certain contexts. Let us illustrate this point with a concrete example of the DFU y ‘and’. When the relation expressed by y is sequence, a competing form is despues ‘after’. One could think that despues should be the prototype of the sequence relation, but paradoxically despues is used only in 2.3% of the cases, against 81.8% ofy. Despues is only used in contexts in which the relation cannot be resolved by other means but using a specific DFU. A concrete example of this tendency is presented in (19). In this case, it is clear that the context is so restricted that despues is the necessary link to express sequentiality between units 1-8 and unit 9. 4 sino is considered in descriptive grammars (Cf. Alcina & Blecua, 1975) as an adversative/substitutive conjunction. It appears in coordinate constructions, when the first member introduces a negation or a question and the second member introduces an affirmative contrast, e.g. No la guerra sino la paz (not the war but/but rather the peace) (O.Cit.:l 180). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 154 Context: SI explains how things were getting more and more complicated when she was late for a wedding. (19) <H1> 1 La blusa es de sedita muy fina, muy fina, 2 me podia cargar la blusa, 3 mira, yo con unos nervios, 4 me temblaban las manos, 5 yo intentando por todos los medios coser la hombrera, 6 la hombrera no se ajustaba, 7 la... la... la gasa se me resbalaba por todos los lados, 8 mira, espantoso. 9 Despues no encontraba taxi, horrible. (19’) <S1> 1 The blouse is very fine very fine silk, 2 I could destroy the blouse, 3 look, I was so nervous, 4 my hands were shaking, 5 I was trying to saw that shoulder pad in any possible way, 6 the pad didn’ t adjust, 1 the... the... the...gauze slided everywhere 8 look, awful. 9 Then I wouldn ’ tfind a cab, horrible. F12L91 It seems that there is no possible substitution of despues byy without losing the temporal sequence relation established between these units. The reason for the impossibility of replacing one by the other lays on the interaction between the contextual information and the level of specificity the relation must show in order to establish coherence (Cf. Sect. 4.2.3). I suggest the impossibility of replacing one DFU by the other is due to fact that each of them represents relations with different degrees of specificity. More precisely I propose that the degree of specificity the relation must present with respect to the units it connects is in a proportionally inverse relation. The higher the unit’s level of information, the lower the relation’s degree of specificity needed. To explain more clearly the theory Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 155 of the interaction between contextual information and specificity in coherence relation, I will point out here that following Asher and Lascarides (1998), I consider discourse relations among units to be underspecified (Reyle, 1993); this means that initially they are empty variables which need to be resolved under certain coherence constraints. These constraints involve the amount of world knowledge and linguistic knowledge provided in the context, and also some textual conditions. For example, one of the principles proposed by Lascarides and Asher (1993) suggests that the resolution of the rhetorical relation of narration when little linguistic and world knowledge is available is given under textual conditions like the following: (20) Narration If the clause B currently being processed is to be attached by a discourse relation to the clause A that’s part of the text processed so far, then normally, Narration (A,B) holds. Axiom on Narration If Narration (A,B) holds, and A and B describe the eventualities el and e2 respectively, then el occurs before e2. (Op.Cit. :442-443). Having this framework in mind, I proposed that the higher the information based on linguistic and world knowledge provided in the context, and the higher the number of coherence constraints met by the units, the lower the need for a more specific relation. That is, the higher the amount of information provided in the units, the lower the level of specification of the relation. Going back to (19), the information provided in 1-8 and in 9 does not reveal the relation between them. The listener does not have enough informational resources to presuppose the chronological order in which the events that are being told may be happening. Additionally, without a temporal reference it is also Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 156 possible to think of additional events in between. We could assume that the action of ‘not finding a cab’ could have happened later in the day. Y is able to represent temporal sequential relation only in contexts in which the propositional and/or pragmatic knowledge provided by the units is rich enough to resolve the chronological direction or order of the events. Observe example (21), in which the sequential relation between these actions is established primarily by the contextual knowledge the listener possesses, in conjunction with the DFU. Context: S1 is telling a friend how his brother wanted to get an eye operation to reduce his myopia, heir parents went to a prestigious eye clinic in the country and the doctor disapproved of the surgery. (21) <H1> y entonces fueron mis padres a Barcelona a... <H3> i,A Barraquer? <H1> <ininteligible> a Barraquer, <H3> Si. <H1> Y les dijo que... que nada, 0 sea que... que ni hablar, (21 ’) <S1> and then my parents went to Barcelona to... <S3> To Barraquer? (Doctor’ s name) <S 1 > <unintelligible> to Barraquer, <S3> Yes. <S1> And he told them that.... no way, 1 mean that... that not even talk about it. F26L83 Y in this case is making explicit a sequential relation between the events related by SI. However, the level of contextual information, by means of linguistic and extra-linguistic knowledge is sufficiently high already. That makes less necessary the use of a highly specific relation to establish the proper sequential order between these events. F s general function as expressing ‘addition’ is sufficiently specific as to interpret that the action of visiting the doctor and the doctor’s response are immediately sequential. What I argue Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 157 here is that in (21) there is a piece of contextual knowledge between the two actions which is absent in (19), and therefore the need to specify the relation is avoided. But the fact that more specific relations must be used in less informational contexts does not offer an account as to why consistently the DFUs which express the least specific relations are the most frequently used. In this respect, I believe the answer lies on the concept of explicitness or rather of inexplicitness, and the use made of the speaker’s shared knowledge (world knowledge, linguistic knowledge). I have referred somewhere else (Romera, 1996) to the theory defended by pragmaticists and sociolinguists (Sinclair, 1991; Warren, 1993; Finegan & Biber, 1994) claiming that inexplicitness is a characteristic of naturalness in conversation. By inexplicitness I refer to the ‘use of language that is context dependent’ (Warren, 1993:37). In this particular case, ‘context’ stands for the knowledge assumed to be shared by both the speaker and the listener. Warren (1985) shows how participants in conversations modify their level of explicitness depending on the amount of shared knowledge. The higher the access to resources of common knowledge, the higher the level of inexplicitness. So, in conversational situations like the ones analyzed here, where the level of familiarity between participants is high and the mutual knowledge shared is great, the tendency is to rely on contextual information and to ‘produce utterances at their lower level of explicitness required to reach a successful outcome in a particular context’ (Warren, 1993:51). The inexplicitness theory can explain why the DFUs capable of expressing less specific relations are the most frequently used. The speaker is relying on contextual information, basically world and linguistic (semantic/pragmatic) knowledge and textual knowledge to reach the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 158 acceptable levels of coherence for communication, using the less specific DFUs to mark relations among units. Only in those cases in which not enough contextual information is given to the listener the marked option takes place, i.e., using a more specific DFU to ensure coherence. In summary, we have seen that the most important characteristic with respect to the DFUs encountered in Spanish conversation and their frequency of occurrence is directly related to the degree of specificity of the relation expressed by the DFU and the context in which the relation is given. Out of 108 DFUs found, only 9 of them account for more than 75% of the tokens. These 9 forms are also the most frequently used to express almost all the semantic and pragmatic relations found in the data. To explain this high frequency, I pointed out differences in the degree of specificity in the relations these DFUs express and in the contexts they are inserted in. I proposed that given a set of units among which a discourse relation holds, the degree of specificity in meaning to ensure the establishment of a semantic/pragmatic relation of coherence is directly inverse to the explicitness of the context. The higher the contextual information (word/linguistics/textual knowledge), the lower the need to make the relation explicit by means of a DFU. Therefore, DFUs which are capable of expressing more specific relations in meaning, such as the one in (21), are only used in contexts in which the degree of information provided by other means is not sufficient to achieve the necessary levels of coherence. To explain the conversational tendency to resort to contextual means (i.e. common knowledge) rather than inserting additional linguistic material, I argued that this is in agreement with the theory that claims that inexplicitness is a characteristic of naturalness in conversation (Sinclair, 1991; Warren, 1993; Finegan and Biber, 1994). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 159 According to this theory, the tendency in conversations is to rely on contextual information and to produce utterances at the minimal level of explicitness needed to reach coherence. In the next section, I will present in more detail the relations each of these most frequent forms in Spanish tend to make explicit. 5.4.2. Forms and relations. The semantic, pragmatic and textual source o f coherence. As we just saw, the majority of relations in the corpora are expressed by a few DFUs. In this section I explore the particular relations that each of the forms expresses in the corpus, focusing primarily on the most frequent DFUs. We observe that these relations are basically semantic and pragmatic for the most frequent DFUs, although they also express textual relations. The DFU bueno is the only form that represents textual relations primarily, but it may also represent semantic and pragmatic relations. These results show that the same forms are able to make coherence relations explicit within all three domains-semantic, pragmatic and textual-, from which we can conclude that the category ‘DFU’ is not an isolated class of “difficult” words to be distinguished from their homophonous forms at other levels of analysis. From the taxonomy of relations presented in Chapter 3, I found a total of 66 relations, the 10 most frequent being consequence, cause, explanation, elaboration, contrast, objection, continuation, adversative, temporal and additional. Each of the 9 DFUs mentioned above (pero, y, porque, pues, es que, o sea, entonces, bueno, si) express several of these relations, although they concentrate on specific types of relations, or relations of a certain domain, as I discuss next. With respect to the types of relations, in Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 160 general it is possible to say that these forms tend to represent primarily semantic relations, although very often they also represent pragmatic relations. Of the 9 forms only pero, entonces, pues and es que are found to be expressing pragmatic relations more frequently, whereas si, porque, o sea and y represent basically semantic relations. Finally, bueno represents essentially textual relations with lesser or no representation of semantic and/or pragmatic relations. In the following section I discuss the results obtained for each of these DFUs according to the types of relations expressed. 5.4.2.1. Forms most frequent in semantic relations: si, porque, osea, y. The forms that appear expressing semantic relations more frequently are as I said, si, porque, y and osea. Among them, si and y have the highest percentages in expressing semantic relations in comparison with pragmatic and textual representations. Si represents the ‘prototypical semantic case’: it expresses a semantic relation in 82.4% of the cases. The relation that si most often represents is conditional. It also has representation in other relations, and it is already possible to foresee the progression of si towards expressing other pragmatic relations. Y also has its highest percentage of representation in the semantic realm (71.2%), although it also appears as one of the DFUs that most often represents the textual mode of continuation. Porque expresses the semantic relation of cause in 67.7% of the cases, and it also conveys very frequently the pragmatic relation of explanation. Finally, although o sea represents semantic relations proportionately more frequently than other relations, it shows a lower percentage than the above mentioned DFUs (54%) and a higher frequency in the pragmatic and textual Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 161 domains. The discussion of results that follows is presented according to decreasing frequency of representation in the semantic domain. Si ‘if. Si expresses basically one type of relation: the condition relation. In the majority of the cases si (82.4%) is a content relator, that is, it relates the propositional contents of the two units in a conditional way. The unit introduced by si expresses a necessary condition for the realization of the content apodosis. (Sweetser, 1990). Then in 17.6% of the cases si expresses also condition but representing pragmatic cases of expression of agreement or disagreement with respect to the previous unit. Figure 9 presents the quantitative results. FORM SEMANTIC PRAGMATIC TEXTUAL si condition agreement/disagreem. % 82.4% 17.6% 0% Fig.5.9. Percentage of representation of s/in sem., pragm. and textual domains. The prototypical semantic condition relation expressed by si is illustrated in (22). (22) <H2> 1 A1 llegar a Barajas, 2 si vemos que sigue esta movida 3 podemos coger la "M-40" hasta Plaza Castilla... (22’) <S2> 1 When we get to Barajas 2..1f we see that this goes on the same 3..we can take the M-40 to Plaza Castilla... CREAF27L315 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 162 In this case the relation can be paraphrase as ‘If X then Y \ ‘ if we see that this goes on the same, then we can take the M-40 to Plaza Castilla Additionally, there are also other cases in which si may seem not to express such clear-cut conditional cases, or at least they cannot receive the content reading as in the previous case. Si in these other cases expresses condition as well, but the condition is embedded in a non-explicit structure that conveys two pragmatic relations: support and objection to what has been said in the previous unit. These cases of si represent the purely interactional ones, that is the pragmatic relations expressed by si. Let us see the following examples (23-26). (23) <H1> ^Pero como puedes tu tener eso, Irene? No lo entiendo. <H2> Pues porque tengo una cinta de cuando yo soy pequena y luego, pues sales. <H1> Pero es que es imposible, si nadie me filmo en video, nunca. (23’) <S1> But how can you have that, Irene? I don’ t understand. <S2> Well because I have a tape o f when I am a child and then you are on it too. <S1> But that is impossible, i f no one ever videotaped me. CREAF11L125 (24) <H5> ^Lo vas a llevar al piso de tu novio? <H2> No, no. Si lo necesito aqui, porque necesito muchas... estar muchas horas diarias, o sea que... (24’) <S5> Are you going to take it to your boyfriend’ s apartment? <S2> No, no. I f I need it here, because I need many., to be many hours a day, so that... CREAF42L139 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 163 (25) <H3> Y hubo un escandalo porque dijeron que lo cerraron y tal, una temporada, ^eh? Porque dijeron que habian puesto came de perro. <H5> Si tiene que ser mas cara la came de perro que la came de.. (25’) <S3> And there was a scandal because they said that they closed it down and all that, for a while, ah? Because they said they had used dog’ s meat. <S5> I f dog’ s meat must be more expensive that the meat of... CREA F27L308 (26) <H1> Oye, pues enlaotra... que estabahaciendo de... <H3> Si me acuerdo que la repitio, porque sabia que estaba haciendo el primo. Esta haciendo lo de... miralo. (26’) <S1> listen, in the other... which he was making as if. <S3> I f I remember that she repeated it, because she knew that he was making faces. He is making that of... look at him. F21L63 In all these examples si seems to be the expression of something else than a conditional relation. In fact the proper translation of these examples into English should in some cases omit the DFU. In Spanish these examples can be found only in interactional settings. Their frequency is not very high (they amounted only to the 17.6% of the total cases), but they represent what I understand as the other end of the continuum from semantic to interactional functions of DFUs. Si represents here an interactional type of relation. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 164 The first thing to notice about these examples is that the si unit could be interpreted as the cause and/or the explanation why the speaker is in agreement or disagreement with the content of the previous unit. Some traditional grammars have mentioned these uses of si as causal uses (Esbozo, 1973). However, I suggest that these cases still represent the conditional si, which together with the apodosis and sometimes without it expresses an objection or an agreement not with the content of the previous units, but rather with the speech act performed in it. (23) and (24) are the clearest cases. In (23), SI is posing an objection to the statement of S2, in which she asserts that she has a videotape in which SI appears when she was 6 years old. The objection is made explicit through the apodosis: ‘it is impossible' and then the protasis poses a condition which validates the conclusion of the apodosis. Notice that the objection is posed on the speech act of asserting, not on the realization of the apodosis: I f no one videotaped me then what you just said is not possible. Example (24) presents an even more interesting case. As in other cases of questions we will see later, the conditional unit introduced by si establishes a disagreement not with the question per se, but with the relevance of the question. SI has asked about whether S2 would take her computer to her boyfriend’s place. The answer of S2 is a straight ‘no’, and then she introduces the condition that justifies her negation and invalidates the assumption made by asking the question, that is, if I need my computer here, then the assumption in your question about taking it to my boyfriend’ s is not relevant. To make my point clearer, consider the following constructed example in which it is obvious that the introduction of si changes the relation between the question and the answer: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 165 (27) HI ^Os vais a ir a vivir juntos ahora? H2 No, no estamos casados todavia. No, si no estamos casados todavia. (27’) S1 Are you going to live together now? S2 No, we are not married yet. No, i f (but) we are not married yet. In the first case, in which si is not present, the interpretation of the unit we are not married yet can be treated as the justification or explanation to the negative answer to the previous question. However, the unit introduced by si is not an explanation for why they are not living together, but a condition that invalidates the relevance of the question. In the present circumstances in which we are not married there is no possible assumption that we would move in together, therefore the question loses its relevance. Finally, the last two examples (25 and 26) are somehow different; they present a protasis attached to the previous unit without an apodosis. In 25 and 26 we have again two cases in which the units introduced by si represent the conditions under which the objection or the agreement with what has been said are validated. In (25) if dog’ s meat is more expensive than other kinds o f meat, then what you say about the restaurant selling dog’ s meat must not be right. And in (26), if I remember that this person had to take another picture, then what you say about the funny nature o f the person in the pictures must be right. The difference between (23)-(24), and (25)-(26) is that in the former examples the apodosis is present. In (23) for example the apodosis is made explicit: ‘ it is impossible ’. Nevertheless, its absence would not cause a change in the relation of objection or support towards what has been said by the interlocutor. The apodosis offers only minimal Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 166 information; what really bears the force of the support or of the objection is the protasis, introduced by si. So, I suggest that the apodosis is not necessary, since it can be easily presupposed. The apodosis in these cases of objection and support responds always to the same pattern: (28) IF X, THEN what you say must be right. (29) IF X, THEN what you say must not be right. Notice that as in all conditionals the causal relation is present. (28) and (29) could be also paraphrased as: (30) what you say must be right BECAUSE X (31) what you say must not be right BECAUSE X Thus, it is not surprising that they can be interpreted as the cause or of the reason why the speaker agrees or disagrees with what has been said. To summarize, si in conversational Spanish represents two types of relations, both of them deriving from the conditional function of si. In the most protoypical case, si is a content relator. The unit it heads expresses a necessary condition for the realization of the content of the other unit. These are the majority of the cases encountered in the data. On the other hand, si also represents pragmatic relations of objection or agreement with what another speaker has uttered before. In these cases si still represents a conditional relation, but this time the condition is embedded in a non-explicit structure that works in the expression of two pragmatic relations: support or objection to what has been said in the previous unit. These cases of si represent the purely interactional ones. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 167 y ‘and’ Impressionistically, y has been considered one of the most general and ‘multi- relational’ DFUs that we can find in conversational Spanish. It has been believed that this DFU may appear in several contexts expressing different semantic, pragmatic and textual relations. However, as I am about to present, I believe that y only represents two basic types of relations: addition and continuation. In the corpus y expressed primarily addition at the semantic level. The total percentage found was 72.1%. Additionally, I also found that y was used to connect units through a continuation relation in 14% of interactional contexts. Iconsidered this to be a pragmatic type of continuation since in all cases it involved the request for new information on the part of the speaker. Finally, in addition to this type of pragmatic continuation, I also found that y in 13.9% of the cases is able to express sequential or textual continuation. FORM SEMANTIC PRAGMATIC TEXTUAL y addition request/continuation, textual continuat. % 72.1% 14% 13.9% Fig.5.10. Percentage of representation of yin sem., pragm. and textual domains. It is generally accepted that y as well as its English counterpart and expresses a basic relation of addition. When this happens the members joined by the conjunction may appear in reverse order without essentially altering the meaning of the conjunct. These are the so-called symmetric uses of y/and. For example, ‘King Tsin [a Chinese Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 168 restaurant] has great mu shu pork and China First has good dim sum’ (Sweetser, 1990: 87). Additionally, it has also been suggested that the conjunction appears in asymmetric relations, in which the order of the members is decisive to understand the events being described (Lakoff, 1971). For example, in cases such as the ones in (32), the alteration of the position of the members related by y obviously produces either an impossible interpretation or a change in the coherence relation. (32) a. Casate y veras b. He trabajado toda la noche y tengo sueno. a’ * Veras y casate. b ’ Tengo sueno y he trabajado toda la noche. (32’) a. Marry and you ’ 1 1 see. b. I have worked all night and I ’ m tired. a’ *You ’ 1 1 see and marry. b ’ I ’ m tired and I have worked all night. (Mederos Martin, 1988: 218) The conjunction in (46a-b) is said to represent conditional and consecutive values. These values are not accounted for only by the general meaning of addition of this DFU, but they also require for their interpretation a pragmatic maxim that summons the speaker to express the actions in the order in which they occur (Mederos Martin, 1988:219). Given this principle, the listener would understand that the events or states expressed occur in temporal order, and would not need any further reference to infer a possible cause-result and/or a temporal relation between the first and the second member in the relation. In (32a) one must get married first to be able to experience the results of marriage. And in (32b) the state of ‘being tired’ is understood as a result of ‘having worked all night’. So at Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 169 least apparently y could be said to express not only addition, but also temporal and consecutive relations. According to the definition of DFU I proposed earlier in chapter 4, if y appears in contexts where the relations between the members can be understood as addition, conditional and consecutive, and temporal we could assert that y is making explicit these relations and therefore it is able to represent them all. However, I do not think this is the case: I propose that in these cases y is representing or making explicit only one semantic relation of addition. In both symmetric and asymmetric cases y expresses the same and only one relation of addition. In the symmetric type, the addition relation appears alone, that is, there is only one coherence relation between the members. In contrast, in the asymmetric type there is an additional second relation (temporal or consecutive) which is not being made explicit by any DFU. This is confirmed by the fact that in every asymmetric example along with y, it is possible to introduce another DFU which makes this second relation explicit. For example, observe in (33) how the insertions of entonces ‘then’ and por eso ‘that’s why’ do not modify the relations holding between the members, but they only make the temporal and consecutive relations explicit: (33) casate y entonces veras. he trabajado toda la noche y por eso/entonces estoy cansado. (33’) Marry and then you ’ 1 1 see. I have worked all night and that’s why/then I ’ m tired.2 2 The translation of this example may not sound totally appropriate for an English native speaker. However, this example constitutes a perfectly regular utterance in Spanish. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 170 What y expresses in every case is the need of the speaker to ‘put things side by side’ at the same level. That is, presenting two events or states with the same status from the speaker’s point of view. For example, in (34) both assertions of SI indicate what she believes to be two equally important qualities of a dress. (34) <H1> este es un vestido muy lindo y tiene su chaquetita que es una monada. (34’) <S1> this is a very nice dress and it has its little jacket that is very cute. F3L123 Notice that in this case SI states that the dress is nice, but she also believes that the jacket is nice. Through the use ofy she is connecting two equally important states. Ify were to be omitted, then the relation of equality between both events could be lost. (34a) <H2> 1 este es un vestido muy lindo 2 i tiene su chaquetita que es una monada. (34a’) <S2> 1 this is a very nice dress 2 i it has its little jacket that is very cute. In (34a), although unit 2 can still be interpreted as the statement of another quality of the dress -the dress is nice and it has its little jacket, another interpretation is that 2 adds an explanation as to why SI finds the dress nice, -the dress is nice ‘because it has its little jacket’. Thus, the absence ofy makes the relation non-explicit and ambiguous. In the asymmetric cases of (32), as well as in the symmetric one in (34), y is only expressing the equality of the events or states in the mind of the speaker. The reason why a second relation is present in the asymmetric utterances (temporal or consequence) but generally not made explicit by the use of a DFU is, as Sweetser (1990) explains, because Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 171 ‘the order of clauses parallels the real-world order of events described in the clauses, so that it becomes unnecessary to add further specification of the temporal ordering of events being narrated’ (Sweetser, 1990: 87). In the corpus y expressed primarily addition at the semantic level. The total percentage found was 72.1%, out of which 49.7% of the cases were examples of the symmetric relations of y, and 22.4% of asymmetric relations (temporal sequence in 15.1% of the cases, and consequence in 7%). Additionally, I also found that y was used by the speaker to connect units through a continuation relation. For example, in 14% of the cases y headed a unit which expressed the continuation of a speaker interaction through the utterance of a request. (35) illustrates this case. (35) <H1> Y vamos a ir el Miercoles a ensayar con la orquesta. Y el Aleluya de Hendel, jay, que precioso!, el Aleluya. <H2> Y la orquesta £de donde viene o...? <H1> Es de Madri<(d)>. No se, no se ni que orquesta es ni nada [•••]• <H2> fY cuantas corales cantan? (35’) <S1> And we ’ re going on Wednesday to rehearse with the orchestra And the Alleluia by Hendel, how beautiful, the Alleluia. <S2> A nd the orchestra, where do they come from or...? <S1> They ’ re from Madrid. I don’ t know who they are or anything.[...] <S2> And how many choirs are singing? F43L121 In this example, S2 is directing the conversation through continuous turns of questions. The questions represent requests for information that SI must provide as background for the main topic to follow (SI is going to perform with her choir some time in the future). Y Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 172 is relating every new request for new information to the previous interaction. The DFU is introducing an initiation of a conversational move composed by an initiation, in this case in the form of a request for new information, an answer which follows the request, and in some occasions a third turn or a follow up (Berry, 1981; Tsui, 1994). When y expresses this type of continuation, I consider that the relation is of a pragmatic nature, since what y is relating is a speech act -a request, which represents the initial move of a conversational interaction. This speech act bears the speaker’s intention and planning towards the discourse. Finally, I also found that y in 13.9% of the cases is able to express sequential or textual continuation of the speaker’s own discourse through different turns, as in (36): (36) <H1> lo de la universidad yo nunca lo he comprendido eh... el ambiente universitario [...] y los odios y los c-... tal, y la... cosas son impresionantes entre... entre... entre la gente, £no? Es, es una cosa. <H2> <ininteligible> bastante, £eh?. En eso te aDFUiro. <H1> jY... y ademas viven un mundo de superioridad! (36’) <S1> I have never understood academia ah... the academic atmosphere [...] and the hate and the th-... and so on, and the...things are terrible among among among them, y ’ know? I t ’ s, it’ s a terrible thing. <S2> <unintelligible> a lot, ah?. I aDFUire you for that. <S1> And...and besides, they live in a world o f superiority! F18L227 This use of y is also illustrated in a few cases in which y also expresses a textual relation of topic shift, as in (37): Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 173 La otra es la de </simultaneo> cangrejos de verdad. esa estaba bien. Ahora, Esa estaba muy bien, Esa esta muy bien. Y creo que querfa pedir tambien el... la receta del... del dulce, <simultaneo> ^como se llama?, del helado aquel. The other is the one o f real </ simultaneous> crabs, that was good. Now, that was really good, that is really good. And I think I wanted to ask you also the... the recipe of... o f the sweet, <simultaneous> what is it called? that ice cream. F13L70 I defined this type of continuation as sequential or textual since the coherence is provided through the connection of different conversational sequences neither in a semantic nor in a pragmatic way. So, y also follows the pattern of other DFUs in being capable of expressing relations in the three different domains: semantic, pragmatic and textual. The majority of the cases found were examples of the expression of the semantic relation of addition. Also, y expresses the pragmatic relation of continuation, mainly introducing a tripartite conversational move: an initiation request, an answer and an optional follow-up. Finally, in a low percentage, y is capable of expressing textual relations as well, introducing a shift in the topic of the conversation, or indicating speech continuation through different turns. Porque ‘because’ . The next most frequent DFU in expressing semantic relations is porque ‘because’. Porque presents three major relations in discourse. The first one is the semantic relation (37) <H1> <H2> <H1> <H2> (37’) <S1> <S2> <S1> <S2> Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 174 of cause, which represents the majority of the cases encountered in the data (67.6%). The propositional content of the unit bearing the DFU is understood as the cause of the propositional content of the previous unit. Additionally, like the other DFUs in Spanish, porque is able to express also pragmatic relations. In 32% of the cases porque expresses explanation. This is a pragmatic or interactional relation in nature, because it involves the inclusion of the speaker’s intentions with respect to what is being said. And finally, porque also represents in an extremely low percentage (0.4%) a textual or sequential relation: aside. Thus, porque presents the property of DFUs defended in this dissertation, namely that the functional categories known as DFUs can make explicit semantic, pragmatic and textual relations. Figure X presents the results obtained for porque: FORM SEMANTIC PRAGMATIC TEXTUAL porque cause explanation. aside 67.6% 32% 0.4% Fig.5.11. Percentage of representation of porque in sem., pragm. and textual domains. As mentioned above, the most frequent relation expressed by porque is cause, which is considered a semantic relation since a real world event is presented as the cause of another real world event. Example (38) illustrates a case of causal relation established by porque: (38) <H1> 1 ^Has hecho fotos? 2 has hecho fotos a Montse? <H2> 3 No. <simultaneo> Eso fixe por la... <H3> 4 A Montse, no, 5 porque </simultaneo> no tenia flash. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (38’) <S1> 1 Did you take pictures? 2 and did you take pictures o f Montse? <S2> 3 No. <simultaneous> That was in the... <S3> 4 O f Montse, no, 5 because < 1 simultaneous> I didn’ t have flash F24L319 The unit introduced by porque presents the reason why S3 did not take pictures of Montse: because ‘he did not have a flash’. So the unit in 5 ‘not having a flash of a camera’ is presented as the direct cause of the speaker’s action of not taking pictures. Porque is also able to express one type of pragmatic relation: explanation. This relation is similar to the pragmatic meanings proposed by Sweetser (1990) for causal conjunctions. Sweetser presents two types of pragmatic readings for English causal connectives like ‘because’: the epistemic and the speech-act readings. In both cases the main characteristic is that a content interpretation of cause is not possible. She proposes that examples such as: (39) John loves her, because he came back (40) What are you doing tonight, because there’s a good movie on. cannot be interpreted in the real causal sense. For example, in (40) the fact that he came back was not the likely cause for his love. Rather, the interpretation that should be made of this utterance is that the knowledge of the fact that John came back causes the speaker to get to the conclusion that he loves her. This is what she calls the epistemic reading. In (40) also, the unit introduced by ‘because’ cannot be interpreted as the real cause of the event or situation in the preceding unit. Rather, ‘because’ expresses the cause of the speech act. That is, the fact that there is a good movie tonight causes the speaker to ask Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 176 her interlocutor about her plans. This second type of reading is what she calls the speech act reading. In my analysis I have found cases of porque expressing both readings, but I considered that both could be understood as cases of the rhetorical relation of explanation. In terms of argumentative theory, both types of interpretations respond to the same speaker’s intention: offering an argument for the explanation of the first unit. In the case of epistemics, an argument is offered to explain the conclusion that comes after the knowledge of something, and in the case of speech-acts an argument is offered to explain the fact of having said something. Examples (41) and (42) represent parallel cases to the ones presented by Sweetser and illustrated in (39)-(40). For us, they.illustrate two cases of the explanation relation expressed by porque: (41) <H5> ^Si? Joe, eso es ‘marketing’ 0 sea, eso es publicidad y no es otra cosa, porque no son tan buenos. (41’) <S5> Yeah? Woe, that’ s marketing 1 mean, that’ s publicity and nothing else, because they are not that good. F27L342 (42) <H1> ^Que te iba a decir? ^Tienes que presentar un plan de estudios en el extranjero? ^Tienes <simultaneo> que decir...? <H2> Depende </simultaneo> que beca. O sea, ^lo dices por alguna en concreto o...? <H1> No. Porque no se que tipo de becas hay. (42’) <S1> What was I going to say? Do you have to present a curriculum abroad? Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 177 Do you need <simultaneous> to say...7 <S2> It depends on </simultaneo> the type o f grant. I mean, you say it for some type in particular or...7 <S1> No. Because I don’ t know the types o f grants there are. F47L17 However, as Sweetser also acknowledges, not all the cases of the causal DFU porque are so clearly distinguishable between the semantic and pragmatic domains. Sometimes the real-world cause of an event may also be the cause of the speaker uttering the speech act. For example, in (43) SI is giving instructions to S2 on how to proceed to duplicate DNA chains. (43) <H2> 1 Entonces si repites el ciclo muchas veces 2 lo que obtienes es millones de copias, 3 pero siempre de un trozo de A.D.N. pequeno, 4 porque para que la cosa funcione 5 tienes que quitar todo lo que sobra. (43’) <S2> 1 then if you repeat the cycle many times 2 what you get is millions o f copies, 3 but always from a small piece o f DNA, 4 because fo r the thing to work 5 you must remove all the extra material. F2L116 In cases like this, there are two possible interpretations. First, we could understand that the reason why the DNA piece to reproduce must be small is that everything that is not necessary must be removed for the process to work. Therefore the units introduced by porque (4-5) are presented as the real cause of the event in 3. This would be a semantic reading. But a second interpretation is also possible: the knowledge that the speaker has about how the process works (units 4-5)leads to the conclusion that the DNA piece to be Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 178 reproduced must be small. It seems then that both the causal and the explanatory interpretations are possible. In this case the real-world event is not only the direct cause of another real-world event, but it is also the cause of the speaker’s conclusion. In these cases I almost always favored the semantic interpretation. Although the pragmatic relation of explanation is possible as well, for practical purposes I decided that the real- world causation always prevailed. However, I believe that these are the cases that enabled the transition of semantic relations to the pragmatic relations expressed with porque. Finally, as mentioned above I also encountered a case of a textual relation expressed with porque. The case I present here could be argued to be also a case of explanation. However, I believe that it represents an extension of the pragmatic cases of porque into the textual domain. Let us examine the example: (44) <H2> 1 Y entonces me dijo: 2 "Yo he esta<(d)>o en Moscu... se... ^sabe usted?", 3 [porque yo tenia poquisisimos anos, 4 pero... me... llamaba de usted,] 5 y eh... "he esta<(d)>o en Moscu, 6 y... y la verdad, aquello no me ha gusta<(d)>o nada; (44’) <S2> 1 And then he said to me: 2 “ I have been to Moscow... you{formal) know?, 3 [because I was very very young, 4 but... he... called me... ‘ usted’ (formal you)] 5 and ah... “ Ihave been to Moscow, 6 and... and the truth is, Ididn ’ t like that at all; F18L149. In (44) the units introduced by porque cannot be interpreted as the real cause of a real- world event. First of all there is no real-world event to relate to; rather, the speaker is Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 179 reproducing the speech of another speaker, i.e., the words that someone else had said to him. Therefore, ‘Because I was very young, but he called me usted’ cannot be interpreted as the reason for any event mentioned before. So, the question is what relation is porque establishing with the context. We could argue that the unit headed by porque is an explanation of what this person said, and that therefore it could be a case of a pragmatic relation as the one seen before. However, the paraphrases that worked out in the cases of epistemic and speech-act interpretations would not convey the meaning of this relation. Units 3 and 4 do not constitute the information necessary in order for the speaker to reach his conclusion in Unit 2, and neither are they the reason why the speaker is reporting the speech of somebody else. So in this case, porque is not introducing a relation of explanation like the ones seen before. Rather, what the speaker is doing is introducing a meta-comment on the nature of the language used in the reported speech, a comment that does not provide a cause or an explanation for any of the speech acts of topics at hand, but that introduces additional information on the characterization of the language of the person reported: the speaker elaborates on the strange fact that this person called a child ‘usted’, when in Peninsular Spanish this form is reserved for older or more powerful addressees. So porque is introducing an observation from the part of the speaker, totally unrelated with the main argument of the previous units, hi 6, SI returns to the main topic again by repeating the same utterance as in 3, introducing it with a continuative y ‘and’. So, although porque still introduces an explanation, (this is why porque is used instead of some other DFU), it is also introducing a comment that should be considered aside from the main argument. Therefore I considered this case as one of the incursions of porque into the expression of textual relations. More precisely, porque helps to make coherent Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 180 units 4-5 whith the rest of the units. This relation has been characterized as aside, given that the comment introduced by porque does not provide a personal evaluation, but only an elaboration of the facts being stated. Thus, porque has its main function in the expression of the semantic relation of cause. Porque also expresses the closely related but less frequent pragmatic relation of explanation. Finally, I also found that porque is used in the expression of one textual relation: aside. Although a textual case like this was only found in one occasion in this corpus, I believe that it constitutes evidence of the thesis defended in this dissertation, namely that we should consider all the appropriate contexts or domains in which a DFU is able to express a relation: semantic, pragmatic and textual. A DFU provides different sources for coherence in all three domains. o sea ‘that is’ O sea is one of the least studied DFUs in Spanish. The few studies devoted to o sea coincide in recognizing that this form is able to provide primarily a semantic connection between the utterances it joins (Casado Velarde, 1991; Cortes Rodriguez, 1991; Schwenter, 1996). They point out that o sea can express apposition, causal explanation, continuation and correction. Additionally, Schwenter identifies some epistemic uses of o sea when the DFU is in co-occurrence with ‘first person singular subjects and belief verbs’ (Op. Cit.: 864). For him, in these cases o sea functions ‘as an ‘epistemic parenthetical’ that indicates speakers’ degree of commitment to what they say’ (:865). The results of my analysis basically confirm the proposals of these authors: o sea is primarily expressing semantic relations. It represents semantic relations in 54% of the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 181 cases encountered. Additionally, I also identified cases in which o sea expresses pragmatic relations, the epistemic uses proposed by Schwenter (1991) and others. These counted 37% of the cases. Finally, I consider as well that o sea is also able to express a textual relation: restart, in which the DFU signals the transition between a false start and a new one. These cases constitute the remaining 9%. FORM SEMANTIC PRAGMATIC TEXTUAL osea restatement conclusion restart % 54% 37% 9% Fig.5.12. Percentage of representation of osea in sem., pragm. and textual domains. Although I agree with the classification presented by the authors mentioned above, I believe that the semantic relations expressed by o sea all belong to the general category of relations of restatement. When o sea creates coherence in the semantic domain, the principal characteristic of the relation is that the unit introduced by this DFU expresses a re-elaboration of the propositional content of the first unit. The re-elaboration can take different forms depending on the connection the speaker desires to create. More precisely, the unit introduced by o sea may introduce a neutral restatement, a specific elaboration of a concept previously mentioned, or a correction. In this section I present some examples of each of these types of relations. The first case, ‘neutral restatement’, involves a reformulation of the propositional content expressed in the previous unit through an alternative expression (Casado Velarde, 1991: 108). This type of restatement constitutes the neutral mode of the relation because Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 182 the semantic content of the two units is presented as evoking the same referent. Observe the examples in (45) and (46), in which both of the units refer to the same entity and/or the same concept: (45) <H1> £Y despues que vas a hacer? (-.Solo aprendes pintura al oleo? <H2> Es que el ano que viene, 0 sea, despues de Navidades, pues va a poner de... grabado, de... (45 ’) <S 1 > And then what are you going to do ? you just learn oil painting? <S2> Es que next year, 1 mean, after Christmas, well they are going to teach... engraving, of... F10L97 (46) <H5> ^Si? Joe, eso es ‘marketing’ 0 sea, eso es publicidad y no es otra cosa, (46’) <S5> Yeah? Woe, that’ s marketing 1 mean, that’ s publicity and nothing else, F27L342 In (45) the time reference that S2 has in mind when she talks about ‘next year’ is the same as the time referent of ‘after Christmas’. The same can be said of example (46), in which the concept of ‘marketing’ is the same as the concept of ‘publicity’ for the speaker. Hence, the most prominent characteristic of these two cases is that the referents or concepts which appear in the units joined by o sea are the same in the mind of the speaker. Of course, it is possible to argue that expressing the same referent by alternative formal expressions is not a mere artifact of repetition, that there must be a reason why the speaker feels the need to express the same referent by alternative forms. There is in fact a Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 183 reason for this. I believe that the referent evoked by the two alternative expressions is the same, but only in the mind of the speaker, not necessarily in the mind of the listener. When the speaker feels or believes that the listener might not evoke the same referent through the linguistic expression uttered first, then the speaker believes that a second expression is necessary to direct the propositional content towards what she has in mind. For instance, in (45) the expression ‘next year’ could be understood as referring either to the ‘next academic year’ or to the ‘next calendar year’. In fact, given the topic of the conversation (the painting classes that S2 is taking), the most likely or default referent for ‘next year’ would be ‘next academic year’, although what S2 has in mind is ‘the calendar year that starts in January, after Christmas’. The speaker is probably aware of the potential default interpretation by the listener, and thus the need for an alternative expression which directs the listener to the right referent arises. In (46) the same need arises when a foreign expression ‘marketing’, which may not evoke a clear referent for the listener, gets re-elaborated into the Spanish term ‘publicidad’. The second case of restatement is one in which the unit introduced by o sea expresses an elaboration of the propositional content of the first unit. In these cases the speaker may elaborate to different degrees the referent(s) she has mentioned in the previous unit. In contrast with the neutral mode of restatement explained above, in this case new information is introduced which does not necessarily get evoked by the mere utterance of the referent. (47) and (48) illustrate two cases of an elaboration-specific relation expressed through o sea. Each of them represents different degrees of elaboration: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 184 (47) <H4> Lo que pasa que yo este cable no se. i,Esto para que es? ^,Tu sabes para que es esto? <H3> Para la bateria, 0 sea, es otra forma de alimentation. (47’) <S4> What happens is that this cable, I don’ t know. This is for what? Do you know what this is for? <S3> For the battery, 1 mean3 , it is another form o f feeding. F38L207 (48) <H1> Entonces lo que haces es derivarlo, eso es una tecnica de derivation de A.D.N. o sea, coges el A.D.N. problema, lo calientas y al calentarlo se abre la <ininteligible>, entonces echas la sonda y lo dejas enfriarse. (48’) <S1> Then what you do is derive it, that’ s a technique o f DNA derivation that is, you take the problematic DNA, you heat it up and by doing that the <unintelligible> opens up, then you put the probe in and you let it cool down. F2L89 It is clear that each of these examples represents a specific elaboration of the referents presented in the previous unit. In (47), o sea introduces the unit that explains in easy terms what a battery is, i.e., ‘it’s another form of feeding’. In (48) however, the elaboration gets to a very high degree of specificity, that is, the whole process of DNA derivation is introduced in the sequence after o sea. 3 As Schwenter points out, translating o sea into English is not always an easy task. Sometimes when this DFU expresses a pure reformulation relation, it is possible to translate it as ‘that is’. However, this is not always the case. In examples (47) and (48) above I thought that ‘I mean’ captures the relation established by o sea better than ‘that is’. In other cases, I could not find a proper translation for this DFU. On those occasions I simply leave the original Spanish form o sea and try to convey the sense o f the relation through a gloss. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 185 Finally, in the analysis of the semantic relations expressed by o sea I also found a third type which could be also characterized as a type of restatement. However, in contrast with the previous two types of restatement, the unit introduced by o sea does not express an equivalence in the content of the two units by neutral restatement or elaboration, but rather it presents a change in referent4. The referent mentioned in the first unit gets modified or substituted by a different one. See for example in (49) how the speaker introduces through o sea not only a different expression but also an entirely different referent from the one uttered in the previous unit: (49) <H1> Y luego investigar que demonios es una... una cadena de protei-.... de... aminoacidos... 0 sea, perdon, de nucleotidos. (49’) <S1> And then to investigate what the hell is a... a chain o f protei-... o f amino acids... 1 mean, sorry, o f nucleotides F2L141 Here it is obvious that the intention of the speaker is to substitute the referents just mentioned (chains of proteins / amino acids) by the new one (nucleotides), so the relation established by o sea can be called ‘correction’ or ‘substitution’. Regarding the frequency of these semantic relations expressed by o sea, there is a big difference in the percentages of use of each of them. Elaboration-specific is the most frequent by far (43.4% of the cases), followed by neutral restatement (9%) and finally correction has a very low percentage of occurrence (1.6%). 4 Casado Velarde (1991) also discusses the non-equivalent meaning of o sea and identifies it as a meaning of rectification. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 186 O sea is also able to express certain pragmatic relations apart from the epistemic uses proposed by Schwenter and others. In particular, I consider that the relation of conclusion that o sea can fulfil and which is classified by these authors as a semantic relation represents in fact a case of a pragmatic relation. This is because the conclusion relation expressed by o sea presents especial characteristics that distinguish it from a semantic relation. Let us examine the following examples: (50) <H2> No tiene nada, o sea que en principio no es configurable la tag eta ni nada de eso. (50’) <S2> It has nothing, o sea that in principle the card is not configurable or anything like that. F32L286 (51) <H4> "Bueno, ya sabes... una broma..." O sea que... hay que controlar un poco. (51’) <S4> " Well, y ’ know ...it is a joke..." O sea that... we need to control it a bit. F28L153 In both of these examples the unit introduced by o sea expresses a conclusion that the speaker herself derives from the information presented in the first member of the relation. In example (50) SI is telling S2 about the possibility of buying a computer card that would solve the problems S2 is having right now. The card is empty, and this information leads the speaker to reasonably conclude that ‘in principle it is not configurable’. In (51), S4 has told the other participants in the conversation how a Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 187 friend’s girlfriend discovered a note that he had written to some male friend with obscene language in it. After telling the story, S4 reasonably advises the other participants to be careful about what they write to their male friends: ‘o sea that we need to control it a bit’. The conclusions derived in (50) and (51) do not fall logically from the premises presented, or at least in principle they need not do so in the mind of the listener. Rather, they constitute inferential knowledge that appears to be clear primarily in the mind of the speaker. The meaning of this relation could be paraphrased as: ‘the knowledge of premises X and Y leads the speaker to conclude Z’. It is in this respect that I suggest that the conclusion relation introduced by o sea seems to be closer to the pragmatic domain rather than to the semantic one. That is, the events or concepts put in relation in the premises do not logically trigger the conclusion introduced by o sea. Rather, the knowledge of the events and/or concepts leads the speaker to present a reasonable conclusion, not a direct logical one. The second important characteristic that makes this relation different from the semantic relations examined above is that the linguistic expression itself presents formal differences with respect to the regular o sea. In all the cases in which the relation expressed is one of conclusion (22.1%), the DFU is followed by the conjunction que ‘that’. I suggest that the presence of que makes the relation change immediately from restatement to conclusion. Observe for example the cases presented above in (46) and (48), repeated here in (52) and (53): (52) <H5> ^Si? Joe, eso es ‘marketing’ O sea, QUE eso es publicidad y no es otra cosa, Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 188 (52’) <S5> Yeah? Woe, that’ s marketing so, THAI6 that’ s publicity and nothing else, F27L342 (53) <H1 > Entonces lo que haces es derivarlo, eso es una tecnica de derivacion de A.D.N. o sea, QUE coges el A.D.N. problema, lo calientas y al calentarlo se abre la <ininteligible>, entonces echas la sonda y lo dejas enfriarse. (53’) <S1> Then what you do is derive it, that’ s a technique o f DNA derivation so THAT you take the problematic DNA, you heat it up and by doing that the <unintelligible> opens, then you put the probe in and you let it cool down. F2L89 In these examples the introduction of que immediately triggers a different relation: whereas in the cases in which o sea appears alone the equivalence/inequivalence relation is the predominant one, the meaning of conclusion prevails in the presence of que. Finally, o sea also establishes textual relations in addition to the semantic and pragmatic relations discussed so far. The predominant one is restart, in 9% of the cases analyzed. The prototypical restart relation shows up when the speaker introduces the first part of an utterance, and at any point she feels the need to redirect her speech into a different direction. On some occasions the speaker realizes that there is other information 5 For the English speaker these contrasts may not translate the difference I try to point out between o sea and o sea que. Mainly o sea could convey the meaning of ‘that is’, whereas o sea que is closer to ‘so’. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 189 she needs to introduce first for the proper organization of her discourse. In contrast, some other times she just needs to restate differently the information she is trying to convey. What distinguishes the restart relation from the ones previously seen is that the speaker never gets to complete the utterance preceding o sea, there being always a break down at some point before the completion. Examples (54) and (55) illustrate the most typical cases of the two types of restart relations: (54) <H1> cuando queria entrar en la oposicion para...controlador, pues... tenia que ser... o sea, le pedian 3... o sea, 3 dioptrias se admitian como maximo, £no?, (54) <S 1 > when he wanted to get into the exams fo r ... controller, well...it had to be... o sea, they asked him 3... o sea, 3 diopters were admitted as the maximum, y ’ know? F26L73 (55) <H1> Tu lo que estas buscando es... [o sea, lo... lo... lo que suele suceder es que cuando tienes el marco corrido eh... pues te aparece un... un nucleotido, o sea, una... un triplete de final de proteina eh... cuando llevan muy pocos aminoacidos.] Entonces tu lo que buscas son los lazos mas largos, antes de llegar a un final de proteina. (55’) <H 1 > You what you ’ re looking for is... [o sea, what what what usually happens is that when you have a frame eh... then you have a., a nucleotide, o sea, a... a triplet o f a protein’ s end eh... when it carries very few amino acids,] then what you look fo r is the largest links, before getting to the protein’ s end. F2L159 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 190 In (54) the speaker needs to find the right structure to convey the desired meaning; she tries in two occasions and she seems not to be content with neither the first nor the second, and o sea helps her make the transition between the false starts and the new one. (55) on the contrary presents a case in which o sea introduces a whole sequence which corresponds to the background information (the steps along the process that the listener will have to go through) that is necessary to understand the content of the main unit. The speaker realizes that without this information there is no understanding of the instruction he is trying to pass on to the listener, namely to look for the largest links of the protein. Hence, o sea also presents the pattern already found in the other DFUs. This form is capable of making explicit relations in the semantic, pragmatic and textual domains. The most frequent cases are represented in the semantic or content domain. Here o sea expresses different realizations of the general relation of restatement: neutral restatement, elaboration-specific and correction. In lower percentage o sea expresses also the pragmatic relation of conclusion, in which the speaker introduces through this DFU the logical conclusion to a series of premises given previously. And finally, o sea expresses sequential relations as well, helping the transition between false and new starts in the organization of discourse or introducing information that is necessary to convey the message. S.4.2.2. Forms more frequent in pragmatic relations: pero, es que, entonces, pues. The next group of forms have their primary representation within the pragmatic domain. They are pero ‘but’, entonces ‘then’, pues ‘so’ and es que ‘it’s that’. One of these forms(pero) also represents semantic relations in a high frequency. In contrast, pues Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 191 and entonces represent pragmatic relations much more frequently than their semantic function as causal and temporal adverbs. I also discovered one form which has representation only in the pragmatic realm: es que. Its origin is different from entonces, pues and pero, which arose originally as semantic connectors. Es que arises from the need of making explicit a pragmatic relation of justification, which is possible only in the interactional context. Additionally, all these forms have representation in the textual domain as well, entonces and pues being the most frequent. In this section I present the distribution of the functions of these four forms in the data. pero ‘but9 Pero has a clear distribution in the pragmatic, semantic and textual domains. It is predominantly more frequent in the representation of several pragmatic relations: concession, contrast with expectation, clarification, objection, and request (57.6%), but it also expresses quite frequently the semantic relation of opposition (39.5%). Less frequently (2.9%), pero is used as well in the textual realm to represent return to a previous topic (topic return). FORM SEMANTIC PRAGMATIC TEXTUAL pero opposition concession/ contrast/ clarification/obj ection/ request topic return % 39.5% 57.6% 2.9%% Fig.5.13. Percentage of representation of pero in sem., pragm. and textual domains. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 192 The semantic relation established by pero is that of opposition. By semantic opposition I understand ‘a contrast which does not involve contradictions’ (Spooren, 1989; Oversteegen, 1997). (56) illustrates a prototypical example of this contrastive relation expressed by pero. (56) <H3> decia Javi que no tenia ni puta idea... No sabia conjugar verbos... Hablaba y tal... mas o menos fluidamente pero conjugar verbos, ni puta idea. (56’) <S3> Javi said that he had no idea... He didn’ t know how to conjugate verbs... He spoke and so on... more or less fluently, but conjugating verbs, not a clue. F28L52 In (56) it is clear that through pero the speaker intends to express the contrast between the language skills of this foreign person. On the one hand he was able to speak the language fluently, but on the other hand he did not know the conjugations. In addition to the clear examples such as the one in (56), semantic opposition is also expressed in other cases in which the idea of contrast may not be that evident. I refer here to the adversative cases traditionally proposed for pero in Spanish. (See Porroche, 1993 for a review of the literature). In the adversative relation the contrast presented by the unit introduced by pero may not necessarily be established against the totality of the propositional content of the first unit, but rather it may represent only an opposition with a part of the first unit. Nevertheless, as well as in the ‘total’ contrast or opposition cases, the adversative relation does not involve contradiction between the two members. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 193 (57) <H4> no me puedes? por ejemplo, yo vengo £no? y me decis lo que hay y me dais el pre- presupuesto que va a ser, para yo saber cuanto dinero me va a costar antes de arreglarlo. <H3> Si, pero de todas formas hay que revisar todo eso. (57’) <S4> And couldn ’ tyou? fo r example, I come here, right? and you tell me what there is and you give me the es- estimate, to know how much money is going to cost me before fixing it. <S3> Yes, but in any case we have to revise all this. F38L258 In (57) a client and a service representative of an appliances store are discussing how much money repairing a camera will cost. Previous to (57), the salesman has indicated that only the inspection of the camera will cost the client some money, so the client wants to know if they can give him an estimate before the inspection. The answer of the representative is not entirely affirmative: ‘they can give the customer an estimate but they have to do the inspection as well’. In this case the contrast is not established against the whole propositional content of the first unit. The client’s unit involves two suggestions: the first one ‘give him an estimate’ and the second ‘do not do the inspection’. The salesman’s response is positive to the first part of the suggestion ‘they will give him an estimate’, but negative to the second ‘they will need to inspect the camera anyway in spite of the estimate. Pero is then establishing an opposition to the client’s second suggestion by the salesman’s second part of the answer. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 194 In the corpus, the total and partial opposition relations expressed by pero show the same distributional frequency, with the adversative relation being only 2 points more frequent than the contrast relation (adversative 19.9% and contrast 17.6%). But the most frequent relations expressed by pero are the pragmatic ones. In this domain pero is one of the most polyvalent DFUs in Spanish. Several researchers have identified different pragmatic values of pero, such as introduction of objections, clarifications and questions among others (Porroche, 1996). The typology presented here abounds in the proposals of these studies but in addition to them it also establishes other pragmatic relations expressed by pero. I have divided these relations in two types: the epistemic ones, which involve a contrast between inferential contents, and the interactional ones, which can only be found in interactional contexts. The epistemic relations expressed by pero represent the introduction of a contrast against an expectation created by the first member of the relation. That is, the contrast is not established against the propositional content of the first units, but against an inference the listener may have drawn after it. For example, observe how in (58) the unit introduced by pero cannot possibly be interpreted as a contrast to the propositional content of the first unit. (58) <H1> la otra crema, yo la se hacer pero la tengo apuntada. (58’) <S1> the other cream, I know how to do it but I have it written down. F13L68 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 195 Unit 1 states that the speaker knows how to do a certain dish, but it also sets the inference that ‘if the speaker has the knowledge of a certain procedure, she can follow that procedure relying on her memory without the need for a written recipe’. What the pero unit introduces then is the contrast between the inference created by 1 -there is no need to follow a written recipe-, and an opposite argument. These are the cases defined as concessive cases. Concession is understood as the expression of two contrasting arguments in favor and against a thesis. That is, the first unit expresses an argument against a certain position, whereas the second unit provides an argument in favor of the same thesis. (Oversteegen, 1997:62). To explain the notion of concession semantically, researchers have adopted the concept of topoi: ‘conventional scales underlying communication’ (Ducrot, 1980; McKeown and Elhadad, 1991; Kehler, 1994; Oversteegen, 1997). For example, in the analysis of (59) the topoi underlying units 1 and 2 could be paraphrased as: (59) 1. if I know the recipe, I can follow the procedure by memory. 2. if I have it written down, I cannot follow the procedure by memory. The thesis presented in 1 -knowing a recipe implies being able to follow the procedure by memory- is opposed or argued against by the argument presented in 2 -having a recipe written down is not a good argument to support this thesis. This relation of opposition is indicated through the use of a DFU. Concession with pero in Spanish has not been quite accepted by traditional grammars, which reserve the expression of this relation to other forms such as aunque Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 196 ‘although’ and sin embargo ‘however’. Traditional grammars define concession cases expressed through pero as adversative inversions (RAE, 1973; Alcina and Blecua, 1975). By adversative inversion these grammars refer to the relation of opposition between two members, in which a primary unit expresses an opposition with respect to another main unit. ‘en esta construction es el miembro primario el que introduce la reserva que ha de permitir entender, enriquecida, la enunciation adversativa’ (Alcina y Blecua, 1975:1176). For traditional grammarians, the relation of concession is established by a subordinate unit, which stands in opposition to the primary member. As I mentioned above, they reserve this relation to adverbial forms such as aunque, as in aunque no me conoce, lo llamare (‘although he does not know me I’ll call him’). However, the concessive relation that connects these two clauses in this sentence can be also expressed by a coordinated structure with pero: no me conoce, pero lo llamare (‘he doesn’t know me, but I’ll call him’), where the unit headed by the connector is in coordination with the previous one, in contrast to the aunque ‘although’ unit, which is in subordination relation with the primary clause. In my opinion, the relation established in both cases is absolutely the same: concessive. The distinction between subordination and coordination does not seem to affect the type of connection established between the two members in the relation. What we are facing here is simply a syntactic issue that does not modify the type of relation expressed, so I will not concern ourselves with this issue here. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 197 Within the percentage of pragmatic relations of pero (57.5%), concession can be considered only as a relatively frequent relation in the data, as only 9.3% of the total cases with pero represent instances of concession. In addition to concession, I also identified another epistemic relation which also responds to the pattern of expressing opposition to an inference created by the first unit. In contrast with the concessive relation, however, the inference created from the first member does not follow from its content, i.e., the inference does not hold a causal relation with the content expressed. In the case of concession the topoi underlying the relation can be expressed through an ‘if-then’ proposition (cf. (59)). In contrast, in this second type of epistemic relation the inference drawn from the first unit represents a common knowledge associated with the unit through the participant’s extra-linguistic experience. Observe the example in (60), in which SI is telling a friend the conditions to apply for a certain fellowship: (60) <H2> Mira, es que la de la Caixa esa sale en mayo... ju- en junio, es cuando hay que entregar los papeles, mayo y junio pero es para dos anos adelante, o sea que si tu por ejemplo la pides en mayo que viene... es para irte dentro de dos anos a hacer un master. (60’) <S2> Look, the La Caixa one comes out in May...Ju- in June, is when you have to submit the paperwork, May and June but it is for two years or more, so that if you for example apply next May... it is to go in two years to do a master’ s degree. F47L21 The topic of the conversation is about applying for fellowships. Usually, fellowship applications are submitted the year before the planned academic activities, but in the case Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 198 of the fellowships awarded by the institution the speaker is talking about, the applications have to be submitted two years in advance. Thus, the unit introduced by pero presents a contrast between the assumption based on common knowledge provided by the world experience and a fact contradicting that assumption. So what the pero unit does is present evidence against an expectation created not by the content of the previous unit but also by the world knowledge associated with it. Notice in this case the impossibility of representing the inference through an ‘if-then’ topoi as in the concession examples just seen, because the inference is not a direct result of the propositional content of the first member of the relation. (61) *if the deadline to apply is June, then the tenancy will start the next year. I have called this relation simply contrast to expectation, since it differs from the concession relation but it still represents a contrast with an inference created by previous context. This relation is almost as frequent as concession in the corpus (8.5%). Both together represent 17.8% of the total relations established by pero. The relation ‘contrary to expectation’ conveyed by pero may seem very similar to another pragmatic relation: clarification. However, clarification also involves cases in which there is no inferential knowledge drawn on world experience to be contrasted with the content of the pero unit. On many occasions, the speaker’s clarification prevents the listener from making a particular inference based on a particular knowledge that only she has in mind. For example, in (62) the clarification introduced by pero prevents the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 199 listener from making any assumptions in advance about the reasons that caused the event expressed in the first unit to happen. (62) <H1> Fuimos a “Gerardo” Pero no porque lo obligara yo sino porque me dijo mi hermana: “hemos ido a Gerardo” (62’) <S1> we went to “Gerardo ” but not because I forced him rather because my sister told me: “ w e’ ve gone to Gerardo ” F22L1 The event of the speaker and her fiance going to “Gerardo” (a top designer’s cloths store in Madrid) was caused by the fact that the speaker’s sister recommended her this place, and not because SI forced her fiance to go there. The introduction of pero not only clarifies the reasons why this event happened, but also blocks any possible assumptions the listener can make about SI having forced her fiance to shop in such a store. Additionally, in other examples the clarification introduced by the pero unit represents only additional information the speaker utters in order to address possible erroneous conclusions. Observe in (63) how the speaker through the pero unit adds important information that she did not include in the first member of the relation. (63) <H1> anades mi- leche... mitad leche, mitad agua. Pero hay que ir anadiendo sin dejar de mover. (63 ’) <S 1 > you add ha- milk... half milk, half water. But you have to keep adding it without stopping stirring. F13L17 It is clear that when giving out a recipe one must provide the listener with all the pertinent instructions. In this case, to an expert cook it may seem obvious that the step of mixing cream, water and milk involves stirring the mixture constantly while doing it, but Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 200 for a person with no cooking experience this is an important step she needs to know. SI has not provided her listener with this information in her first utterance. Therefore she clarifies that the right way to mix these ingredients is by stirring them constantly. So in cases like this the clarification is meant to complete or repair the content of the units just uttered to avoid possible confusions. The fact that in the cases just seen there is no necessary inference to be made before the introduction of the clarification mademe keep the relations ‘contrary to expectation’ and ‘clarification’ in two separate groups. The former implies the presentation of a contrast with a possible inference created by the first member of the relation, whereas the latter involves either blocking a possible assumption to be made later or adding supplementary information to avoid wrong interpretations. In the corpus these two cases of the clarification relation accounted for 5% of the total number of pero instances. The second group of pragmatic relations conveyed by pero is the one formed by what I called interactional relations, because they are only possible in interactional contexts. Concretely, I am referring to the relations of objection and request. They represent the majority of the pragmatic cases of pero (20.1% and 14.7%, respectively). Two units are in an objection relation when the second unit presents a disagreement or a challenge with respect to the content expressed in the first unit. For example in (64), through the pero unit S2 introduces an impediment for the realization of the event presented in the first member of the relation. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 201 (64) <H1> 1 que mira, ojala la hubiera cogido sin hacer obra 2 y la hubiera hecho a su gusto. 3 Y la habia salido tan barata. 4 Pero bueno. <H2> 5 Pero... pero hay que estar aqui tambien pendiente y. (64’) <S1> 1 look, I wish she had taken it before the reforms 2 and she had made them as she pleased 3 and it would have been really cheap fo r her 4 but it’ s OK. <S2> 5 But... but one needs to be here all the time and.. F33L34 In 1-4 SI expresses his belief that having bought an apartment before any reforms were made would have saved the buyer some money. S2 then challenges this opinion introducing in 5 an argument against it. If one buys a house as is then she must make the reforms on her own adding to the initial cost the time and energy spent at the apartment while the reforms are being made. So the condition presented in 5 constitutes a counter argument against the thesis proposed previously in 1-4. The objection relation is the most frequent of the pragmatic relations expressed by pero. I found 20.1% of objection relations in my analysis. Close to objection there is still another type of pragmatic relation: request. Under this label we find cases that in general present a request for confirmation or clarification of certain information that the listener understands as being incomplete or contradicting a piece of information received previously. All these cases represent the second term of an interactional/conversational move, that is a response or a challenge to an informative unit (See Tsui, 1994 for a taxonomy of conversational units). Additionally, in all instances these responses take the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 202 form of a question or a statement followed by a tag question. Examples (65), (66) and (67) illustrate cases of request for clarification, request for confirmation in the form of a question, and request for confirmation in the form of a statement plus tag question, respectively. (65) <H1> 1 Y se ha pillado un apartamento que era super cutre por 45.000 pelas. 2 No se... creo que le pagan mas, 3 pero aun no tenia ni cama todavia. <H2> 4 Hostia. Pero, ^quien, Agustin? (65’) <S1> 1 and he’ s gotten an apartment that was super rundown for 45,000 pesetas. 2 I don’ t know.. .1 think they pay him more, 3 but he didn’ t even have a bed yet. <S2> 4 Shit. but who, Agustin? F16152 (66) <H4> 1 jA h.J jOstras! ^.Esa que no pasa nada en... toda la pelicula? <risas> <H4> 2 Es para que no la veas, tio. <H2> 3 Pero, ^la has visto? (66’) <S4> 1 oh...! wow! that where nothing happens in... the whole movie? <laughs> <S4> 2 It is fo r you not to see it, man. <S2> 3 But, have you seen it? F28L309 (67) <H1> 1 en primer lugar, mi marido tenia que haber dicho: "Oye, mira, he quedado con Angelines para manana 2 porque vamos a cambiar impresiones sobre un trabajo 3 <borrado involuntario> ... trabajo de ningun tipo. <H4> 4 Pero en principio era para., proponer un trabajo, £no? Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 203 (67’) <S1> 1 first o f all, my husband should have said: "look, I ’ m meeting with Angelines tomorrow 2 because we are going to discuss a project 3 <involuntary erasure> ... no project at all. <S4> 4 But in principle it was to for... a project proposal, right? F8L25 In (65), given that several referents have been mentioned in the previous context of this conversation and that Spanish is a pro-drop language which allows for non insertion of a pronominal subject, S2 is confused about the information received, and in 4 he requests clarification of the referent of the actions expressed in 1-3. In (66) and (67) the information being challenged has either been mentioned in the previous context or it constitutes known information for the listener. In (65), before making his negative remarks about the movie, S4 has said that he saw it recently; after his comments which seem to contradict the opinion that SI has about the same movie, SI needs confirmation that S4 has in fact seen it; this is why in 3 she introduces her request for confirmation. In (66), SI is complaining about her husband not being truthful about his relation with another woman. She states that he should have been more straight about it. On the other hand, the listener (S4) seems to have contradictory information, that is, that originally the encounter between S i’s husband and the other woman was truly job related. Since this seems to be in opposition with S i’s complaints, S4 demands confirmation of the veracity of the information she had. So in both (65) and (66) the response introduced by pero expresses a request for confirmation of certain information already known to the listener (which may have being learned throughout the course of the conversation or by other means) that seems to conflict with the information being given Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 204 by the speaker. As I said above, in the case of request of confirmation the format can vary between a question, as in (65), or a statement with a tag question attached to it, as in (66). Both relations are quite frequent as well, as they represent 14.7% of the total instances of pero in the data. Finally, pero also represents textual relations in a few number of cases (2.9%). The sequential relation introduced by pero is that of topic return. It acts as a push-pop marker, that allows the speaker to go back to the main topic after a commentary, a digression or a statement about a different topic as a way of contrast with the material previously inserted. (68) illustrates a prototypical example of this relation. (68) <H3> 1 Ya. 2 Yo desde luego, vamos, me miro asi con cara extrana 3 y no... <H2> 4 Si es... si es simplemente eso. <silencio> 5 Te miraria con cara extrana 6 porque claro, un tecnico <ininteligible> brillante <ininteligible>. 7 No, pero ya te digo que la solution es esa, simplemente. (68’) <S3> 1 yeah. 2 I o f course, I mean, he looked at me like this with weird face. 3 and not... <S2> 4 Yes it’ s... simply that. 5 H e’ d look at you with weirdface 6 because o f course, a technician <unintelligible> brilliant <unintelligible>. 7 No, but I ’ m telling you that the solution is simply that. F52L330 In this case, S3 has told a group of friends her problems in fixing her tape recorder at a local electronic store. S2 has suggested an easy way to solve it. In 1-3 S3 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 205 mentions the strange look in the repairman’s face when she explained her problem to him. After this utterance, S2 repeats that the solution is easy, and then he goes on to comment about the possible reasons of the repairman’s strange look in the face (5 and 6). After the comment he returns in 7 to the main topic of how to repair the appliance by repeating his earlier statement about the solution, through the use of pero. This DFU helps the speaker make the transition between a digression -his personal interpretation of the repairman’s surprise-, and the main topic. However, the use of pero indicates more than a neutral mode of topic return, otherwise a neutral mode of topic return could have been expressed by y. In fact S2 re-introduces his way back to the original topic as a contrast with his remark. The appearance of ‘no’ before pero indicates that the speaker is disregarding his own remark or digression, and through pero he is indicating the focusing on the content of the next unit. Pero confirms again the property of DFUs of being capable of representing different relations in the semantic, pragmatic and textual domain. Pero is one the most polyvalent DFUs in Spanish. It is predominantly strong in the pragmatic domain, being able to make explicit epistemic and interactional relations, although it is very powerful as well in the representation of semantic opposition. Throughout all these relations it is evident that the notion of opposition (semantic, pragmatic and/or textual contrast) is the common denominator for all of them. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 206 Entonces ‘ then \ Entonces ‘then’ also has a clear distribution in the tree domains studied. Although not very frequently, semantically it still maintains its original meaning of the representation of temporal relation between two units (8.5% of the cases). In contrast, it is in the pragmatic and the textual domains where entonces finds its primary strength; in 63.2% of the instances, entonces represents the pragmatic relation of consequence, and in 28.3% of the cases it represents continuation within the same speaker’s discourse. FORM SEMANTIC PRAGMATIC TEXTUAL entonces temporal consequence continuation % 8.5% 63.2% 28.3% Fig.5.14. Percentage of representation of entonces in sem., pragm. and textual domains. Entonces as a DFU has its origin in a temporal adverb that later becomes a connector. This adverb relates two events or states establishing a time reference between them. One of the events/states is placed either within the same time frame set up by the first one, or right immediately after it6. For example, in (69a) below the action presented in the clause introduced by entonces is placed within the time set by the first clause. In contrast, in (69b) the action introduced by the entonces clause takes place immediately after the occurrence of the event in the first clause. 6 See Schifirin (1990, 1992) for a detailed account o f the continuing time reference and shift time reference values of the English counterpart of entonces, ‘then’. Although entonces and ‘then’ do not have equivalent uses in English and Spanish, they share the values of expressing time reference in a continuing and/or consecutive manner. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 207 (69) a. Maria estaba trabajando en su casa y entonces se acordo de que debia llamar a Luis. b. Maria entro en el apartamento y entonces llamo Luis. (69’) a. Maria was working at home and then she remembered she had to phone Luis. b. Maria got in her apartment and then Luis called. This temporal relation between two events or states is the one that entonces still represents in the semantic domain. This relation however, is the least frequent in my analysis. The results show that entonces establishes a temporal semantic relation only in 8.5 % of the cases. Two typical examples of this relation are illustrated in (70) and (71): (70) <H1> 1 Y luego, separas las yemas de las claras, 2 entonces en un bol, pues pones la- 3 para... para batir a punto de... nieve, las claras. (70’) <S1> 1 And then, you take the egg yolks apart from the whites, 2 then in a bowl, you put the- 3 to... to beat until... stiff, the whites. F13L84 (71) <H1> Y [me dice]: "Te voy a ensenar". Y entonces es cuando he empezado a ver que salias tu en dos o tres. (71’) <S1> And [she says]: "I’ ll show you". And then is when I've started seeing that you were in two or three. F21L53 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 208 In (70) the new recipe step presented in 2 (beat the egg whites) must occur necessarily once the first step (separate the egg yolks from the whites) has been completed. So entonces introduces a shift in the time reference, that is, one action must be completed before the second takes place. We can call this relation temporal after. In (71) on the other hand we find a case of continuation of the time frame set in the previous unit. In this example SI is talking about some pictures she and her family took at a recent wedding. SI is telling a friend that her mother started showing her the wedding photos (her mother is reported to say: “I’ll show you”), and that then she realized that her friend was in several of them. So the time reference set by the first utterance -while seeing the pictures- constitutes the reference frame to place the second event: the realization that her friend was in several of the pictures. In this case the relation expressed by entonces can be considered as temporal meanwhile. Nevertheless, the most frequent relation that entonces establishes is not a semantic temporal relation, but the pragmatic relation of consequence. In 63.2% of the cases this DFU was found to express an inferential relation between two units, in which the unit introduced by entonces is inferred as a logical consequence of the event or state presented before. This type of relation has been identified by other researchers as well (Schiffrin 1992 among others) as an epistemic relation7.1 call it pragmatic because it does not relate the propositional contents of the units only; rather, this relation forces the speaker to infer and interpret that two states or events which in the real world would not necessarily 7 For Schiffrin (1992) the concept of epistemics represents a more general notion than for other researchers (Traugott, 1989; Sweetser, 1990). It is not uniquely related to knowledge and speaker’s beliefs, but it also represents the attitude of the speaker towards the proposition expressed. “I am viewing epistemic meaning very broadly, as having to do not just with knowledge and belief about possibilities, but with commitment to the truth of a proposition and the way different sources of information (compare: different modes of Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 209 imply a cause-consequence connection must be interpreted that way. Observe for example in (72): (72) <H1> 1 Esta muy arregladito y muy curioso, si. 2 Si, si, si. 3 No, si le digo la verdad que tiene luz, que tiene luz, 4 entonces aunque aqui no haya ventanas esta bien... (72’) <S1> 1 I t’ s very well arranged and very nice, yes. 2 Yes, yes, yes. 3 Yes, I ’ m telling you the truth, it’ s bright, it’ s bright, 4 then although here there are no windows it’ s fine. F31L121 Here SI is visiting an apartment for sale. Previous to (72), the owner has mentioned that one of the rooms has no windows, but that he has managed to compensate this possible drawback by opening extra windows in the room next to it, thus providing more brightness to the whole apartment. In unit 3 SI agrees with the fact that the apartment is bright enough. Then in 4 she uses entonces to connect her expression of agreement and her opinion that the lack of brightness is consequentially not a problem. 3 and 4 could be paraphrased as: ‘given what I said before -that the apartment is bright enough-, then the lack of windows in this room is not a problem for me’. In this example, there is some extra-linguistic knowledge the interlocutors must retrieve to obtain a proper interpretation. Apartments by themselves do not have intrinsic properties which make them right or wrong. However, the world knowledge indicates to us that usually apartments increase their value according to certain parameters, among others whether they are spacious or bright. The lack of brightness would imply that the apartment is not knowing) alter such commitment” (Schiffrin, 1992:774). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 210 right in the eyes of the buyer, and the opposite consequently implies that the apartment is fine. Entonces is in this case indicating a consequential knowledge established between the two units: the apartment is bright enough and since this is a characteristic that makes the apartment more valuable then the apartment/the situation is fine. In other cases entonces indicates or instructs the listener that she must make an inferential connection between the two units joined by entonces', that is, there is no need for the listener to possess extra-linguistic knowledge to interpret the consequential relation. Observe the case in (73). The listener must infer that given the step taken in 1, the event expressed in 2 is a consequence of it. (73) <H1> 1 Esa sonda la tienes marcada para reconocerla 2 entonces puedes coger los trozos que sabes que tienen ese.. ese trozo de cadena que tu sabes que es la proteina. (73’) <H1> 1 that probe you have it marked to recognize it, 2 then you can take the piece that you know is the one that has that... that piece o f chain that you know is the protein. F2L93 In the context of the conversation (the chemical process SI is talking about) the fact that a probe is marked has as a consequence that one can later identify the particular pieces of DNA one is interested in. The use of entonces is the one that brings in the context the consequential relation between one event and the other. The semantic temporal relation and the pragmatic consequential relations expressed by entonces may seem unrelated at first sight. However, there are some cases in which the link between them is still clearly visible. In these cases, it is difficult to interpret the relation one way or the other. Unfortunately, in the corpus very few cases of Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 211 ambiguous temporal/consequential relations (3.3%) were found. One of the examples is presented in (74) below: (74) <H1> Llegaba... entraban ellos pero todavia estaban saliendo los novios anteriores, entonces se quedaron parados, aqui, hasta que salieron los novios y se marcharon. (74’) <S 1 > They arrived... they were getting in but still the previous couple were getting out, then they stayed still, here, until the couple got out and left. F12L77 SI is telling a friend that when she arrived at a wedding the church was very crowded because several weddings were celebrated on the same day. We can see here the double nature of entonces. The relation between the two units can be interpreted as temporal continuation ‘event 1 meanwhile event 2’, or as consequential ‘given event 1, in consequence event 2’. The temporal interpretation would indicate that at the time the first couple SI is talking about were about to enter the church they saw the other couple coming out, and in that particular time reference they waited until the church got empty. The situation of the couple exiting the church sets the time reference for the event of waiting expressed with respect to the couple that were about to enter. That is, one couple exits and at that time the other couple waits. The consequential interpretation on the other hand, would express that since the first wedding was still finishing up, in consequence the second wedding waited until they left. As I mentioned before in the previous section, Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 212 cases like this confirm the hypothesis that the representation of one DFU in one domain should not be seen in isolation from its representation in the other fields. Finally, entonces is also able to express a textual relation of continuation in 28.3% of the cases, in which neither the consequential nor the temporal relations are traceable in the use of entonces. We could say that this DFU is used as a device that ensures the continuation of different sequences (a speaker’s interrupted turn, for example) or the simple flow of sequences in one speaker’s discourse. es muy facil se rompe el... el pivotito este de las cintas, ademas que yo creo que es una cosa que deberiais ir haciendo a medida que... que las fuerais grabando enteras. Claro. Claro. entonces para borrarlas tienen que poner un papelito... it’ s very easy you break the... this little thing o f the tapes, besides I think it’ s something you should be doing as you...you are recording them entirely. Sure. Sure. then/so to erase them you have to put a little paper... F52L404 (75) represents a case in which through entonces S2 manages to retake his turn after SI and S2’s short interventions in the conversation. S2 is a professor in Linguistics and is giving SI and S3 (his students) instructions on how to record natural data. In his initial discourse S2 instructs them in blocking the tapes once they are finished to prevent involuntary erasing. At that point SI and S3 break in the conversation agreeing to S2’s (75) <H2> <H1> <H3> <H2> (75’) <S2> <S1> <S3> <S2> Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 213 advice. S2 then retakes his turn keeping on talking about how to erase the tapes in case they need to. He does so by using entonces. In this case it is difficult to obtain neither a temporal nor a consequential reading of entonces. Rather, this DFU is expressing only a textual relation of continuation. In some other cases, it seems that entonces is able to express two possible relations at the local and global levels of discourse. This confirms the double nature of entonces'. textual continuation marker and consequential. Observe the example in (76): (76) <H1> 1 Yo por ejemplo tengo un horario de diez a dos y de cuatro a ocho. 2 pero la persona que entra a las ocho y media, evidentemente sale mucho antes. 3 Si coge dos horas para comer 4 sale mas tarde, 5 si coge una 6 sale antes. 7 Entonces mas o menos intentamos que la libreria no se quede sola a medio dia 8 y por eso nos acoplamos unos a otros. (76’) <S1> 1 Ifo r example have a schedule o f ten to two and from four to eight. 2 but the person that begins at eight thirty, evidently leaves much earlier. 3 I f she takes two hours fo r lunch 4 she leaves later, 5 if she takes one (hour) 6 she leaves earlier. 7 Then more or less we try not to leave the bookstore unattended at lunch time 8 and that’ s why we accommodate (our schedules) to one another. F59L380 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 214 In this case, entonces can be interpreted either as a continuation marker or as consecutive marker. A first interpretation could be the textual one. In units 1 to 6, SI (the manager of a bookstore who is recruiting new assistants) talks about how the schedule can be flexible; as long as the employees work for eight hours they can choose their own times to begin work and to have lunch. In units 7 and 8 then, she goes on to say that there is a need for this flexible schedule since the bookstore keeps open at lunch time and someone must be there, so not everybody can have lunch at the same time. The relation between the first units 1-7 and the units introduced by entonces could be interpreted as mere joining of two units, the first unit stating that there is a flexible schedule and the second unit stating that employees accommodate their schedules to keep the bookstore open at lunch time. In fact, the absence of entonces would not prevent us from obtaining this interpretation. However, a possible second interpretation could be the consequential one: previous to (76), SI has responded to one candidate’s question about the schedule of the library. SI mentions that it is very flexible depending on two things, personal preferences and bookstore constraints. Then she explains how the bookstore is open all day from 8:30 a.m. to 9:00 p.m. So after all this information, the sequence introduced by entonces constitutes a logical consequence of the premises established before. The bookstore must be open at all times, therefore the employees accommodate their schedules to cover the whole day, including lunch time. It seems that entonces is capable of expressing both consequential and textual relations, but at different levels of analysis. In (76), at a local discourse level involving the speaker’s own discourse entonces could be interpreted as expressing a textual relation Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 215 such as the one seen above in (75). At a global level of discourse, this DFU could also be interpreted as relating larger pieces of discourse with a consequential connection. In the case of entonces we have enough evidence to support the theory that none of the cases found in the semantic, pragmatic and textual domains are to be taken in isolation from one another. The representation of temporal, consequential and continuation relations by entonces reflects the tendency of speakers to use the same forms to represent different relations in discourse in response to a need of expressing interpersonal and textual meanings. Pues ‘so’ Spanish pues belongs to the category of DFUs that can be considered to have an original connective function. Like other connective markers such asy ‘and’ orpero ‘but’, pues still maintains its original semantic relation, the causal relation. However, the expression of cause seems to have been taken over primarily by other DFUs such as porque, leaving only a small representation for pues. In my analysis I found few cases of pues expressing cause (6.3%). In contrast, pues has enriched its repertoire of pragmatic and textual relations to a great extent. In the pragmatic domain the most frequent relation expressed by pues is result or consequence (50.9%). In the textual domain, pues is almost as frequent as in the pragmatic domain (42.8%), representing a relation of continuation in cases which some researchers have called ‘pure illative cases’ (Martin Zorraquino, 1991; Porroche, 1996 ) or in other cases what Paez (1982) has called ‘procedural pues’; this is a marker that ’signals a manner of process in oral conversation Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 216 but not a judgement or a manner of reasoning’ (:333). In what follows, I will discuss the distribution of these relations in the data. FORM SEMANTIC PRAGMATIC TEXTUAL pues cause consequence continuation % 6.3% 50.9% 42.8% Fig.5.14. Percentage of representation of pues in sem., pragm. and textual domains. First, in the semantic domain, pues seems to keep its original function as a form that expresses causal connection. However, this relation is represented in a small proportion. Only 14 cases out of 224 instances of pues corresponded to this relation. Let us see two of these examples. (77) <H2> puede ser que tu tengas el trozo que empieza codificando un aminoacido, pero puede ser que tengas... que te falte uno, pues como empiezas un aminoacido pa arriba las par. los tripletes son completamente diferentes porque empiezas cambiao. (77’) <S2> it may be that you have the piece that begins coding the amino acid, but it may be that you have... that you ’ re missing one, because since you start an amino acid up the pa-... the triplets are completely different because you start the other way around. F2L150 (78) <H 1 > Ella se la ha mandao, no la ha contestao pues no quiere saber nada de el. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 217 (7 8 ’) <S 1 > She returned it, she didn’ t answer to it became she wants to know nothing o f him. F5L24 In both cases the pues unit presents the cause of the event uttered in the previous unit. Notice that the translation of pues in both these cases is necessary equal as the one of porque ‘because’. In fact, in both examples pues can be replaced by porque without changing the relation between these units. The expression of cause is the least frequent relation in the corpus. These 14 examples represent 6.3% of the total number of occurrences of pues. The next group of relations expressed by pues belongs to the pragmatic domain. They are much more frequent than the semantic causal relation. The most prominent pragmatic relation represented by pues is consequence or result. This type of relation has been identified as pragmatic because it does not relate the propositional contents of the units only, but rather it forces the speaker to infer that two states or events which in the real world would not necessarily imply a cause-consequence connection must be interpreted that way. The result or consequence relation is represented by pues in several different contexts, which has led some researchers to attempt an identification of different functions of this DFU. However, I consider that even though the contextual environment, changes, the relation of consequence or result between the units stays constant. This explanation, I believe, offers a more homogeneous treatment of the DFU pues. Let us see several examples in which pues establishes the consequence relation between two units. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 218 (79) <H2> se esta tratando con... tambien que de unos fondos a fondo perdido que da este... el ayuntamiento, eh... para... <H1> para arreglar... <H2> para arreglar casas de estas porque como esta ahi el Reina Sofia, pues parece que... que... que estan dispuestos por esta zona a dar [dinero] (79’) <S2> we are trying with... also that they give some non- refundable funds that this ...the city hall gives ah...to... < S1 > to refurbish... <S2> to refurbish houses like these because since the Reina Sofia is here, then it seems that... that...they are willing to give [some money] around this area F31L250 (80) <H1> Bueno, de todas maneras, con los problemas que da, pues habra que ir. (80’) <S1> Well, in any case, with all the problems it is giving us, then we ’ 1 1 have to go. F32L278 (81) <H3> Yo me voy en esquina Garcia Paredes esquina Miguel Angel. Es que vengo de hacer el came de Identidad. Y <simultaneo> toa la manana. <H2> jHuy! </simultaneo> pues habra <simultaneo> estao... (81 ’) <S3> I get off at the corner o f Garcia Paredes the comer o f Miguel Angel, es que I ’ m coming from the ID Issuing Office and <simultaneous> the whole morning <S2> Puff!! </simultaneous> so you may have been there.... F30L69 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 219 In these three examples, pues is establishing a connection between two units in which the second member of the relation must be taken to be a consequence of the first one. The contexts in which pues appears are very different, but in all of them the relation represented remains the same: consequence. In (79) we find a structure in which pues is used very frequently. This structure in Spanish takes the form of a conditional structure with a first term preceded by como ‘since’ (literally ‘how’) and a second term generally introduced by pues. The first member of the relation introduces a premise under which the result presented by the second member becomes a logical consequence. In (79), the fact that City Hall now gives some money to restore the area of the city they are talking about is due to the proximity of an important art museum: Museum Reina Sofia. This relation could thus be seen as a causal relation, the reversal of the consequential or resultative relation, as is well known. In fact, the example in (79) could be paraphrased either as a resultative-conditional construction as in (82), in which the circumstances expressed in the first term set the conditions necessary for the second member to follow as a logical result: (82) given that the R.S. is there then City Hall gives money or as a causal one as in (83): (83) City Hall gives money because the R.S. is there. Pues appears expressing the consequence in the second member of the structure in (82), where the first member is introduced by the conditional como. In some other cases, however, there is no need for another DFU to introduce the member expressing the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 220 condition. (84) illustrates one example of this situation. The condition that leads to the result presented by the pues unit is introduced by a different structure. SI has asked a friend to help him repair his computer. After his friend tells him he cannot do much this time, finally he realizes that he needs to take it to a professional repairman. The condition that leads him to this conclusion is introduced by a construction formed by con ‘with’, a noun phrase, and an adjectival or relative clause which modifies the nucleus of the NP. This construction has the same value as a condition. The relation could be paraphrased again as one of condition: (84) given all the problems it is giving us then we should go (and repair it) Again, pues appears introducing the member expressing consequence in this conditional relation. Finally, (81) presents a clear example of an inferential consequence based on the content of another speaker’s utterance. In this case there is no need for the occurrence of a linguistic form expressing the conditioning circumstances. S3 states that she has gone to the ID Issuing Office to renew her license. Immediately, based on common world knowledge -one always must stand long lines at the ID Issuing Office, S4 infers that she must have been waiting for a long time. The underlying conditioning circumstances are so obvious that introducing additional linguistic material to express them would result in redundancy (violating as well the economy principle of discourse). The paraphrasis in this case would be: (85) given that you have been at the ID Issuing Office then you must have waited a long time. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 221 The use of pues makes the relation of result or consequence explicit in this sequence8. Still, within the group of contexts in which pues establishes a consecutive relation I found several examples that could seem at first sight to be expressing a relation of contrast between the content of two units. Let us see the following examples. (86) <H4> j Jode, mira por donde va la cola, cono! <H1> jJode! [Que pilila! <H4> Pues yo no he visto anuncia<(d)>o el concierto, macho. (86’) <S4> Shit, look where the line is getting to, <S 1 > shit! what a line! <S4> but/so I have not seen the add fo r the concert, man. F152L62 (87) <H3> Si. Estos van a Madrid. La gente va con paraguas. Es que amenaza... <H1> Pues esta aclarando ^eh? (87’) <S3> Yes. Those are heading to Madrid. 8 I believe some o f the cases presented by other researchers as ‘replies’ {replicas, Porroche 1996) correspond to the same type of inferential consequence relation established by pues. Porroche (1996) gathers several cases from Portoles (1989), in which S2 responds to the content o f S i ’s utterance. (i) Me duele la cabeza. - Pues tomate una aspirina I have a headache. Then take an aspirin. (ii) Me gustan los toros -jPues eres un tercermundista! (sic) I like bullfighting, Then you are a third world country person! (Porroche, 1996:80 from Portoles, 1989) These cases offer clear instances o f inferential consequences prompted by the content of the first unit. They are equivalent to the example presented in (81), the only difference is that instead of uttering a statement, here the speaker utters a speech act. They could be paraphrased as: (ia) given that you have a headache then I am telling you to take an aspirin. (iia) given that you like bullfighting then I am telling you that you are a third world country person. Following Sweetser‘s (1990) distinction between epistemic and speech act cases in discourse We could identify (81) as an epistemic case and (i)— (ii) as cases of speech acts. In my opinion all o f them belong to the pragmatic domain and are connected by the same relation. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 222 People have umbrellas. It is going to rain. <S1> But/so it is clearing up, you know? F35L198 (88) <H2> ^El que? ^El que hemos heredado? <H1> El sentido de la responsabilidad en el... en el trabajo y.. tratar de mejorarlo y <ininteligible> no conformarme con cualquier cosa; si <ininteligible> esto y os sale muy bien, ir a... otra cosa, ir a mas... Eso es lo que os habeis llevao <H2> Pues tu lo Arenas. Tu dices: "bueno, ya esta bien, £no?, ya esta bien”. (88’) <S2> what? what have we inherited? <S 1 > the sense o f responsibility in the... in the work and., to try to improve everything and <unintelligible> not getting by with just anything', if <unintelligible> this and it comes up right go for... something else...go for more... that’ s what you have gotten. <S2> but/so you stop that (in us). you say: “ well, that’ s enough, OK?, that’ s enough”. F20L12 In each of these examples, pues seems to be introducing a contrast between a statement uttered previously and a new one. If this was the case then we should be able to replace pues by pero ‘but’, which is the prototypical DFU to express contrast, as we saw above. However, the exchange between pues and pero does not work out, as replacing pues by pero changes the relation from a contrast into an objection, and this does not seem to be the connection in these sequences. I believe that in these cases pues is introducing the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 223 second part of an argument that comes from some inferential knowledge commonly understood by speaker and listener. In (86) SI has established a long argument about the values he is transmitting to his children. In (87) SI and S2 are talking about the huge amount of people standing line at the doors of a concert stadium. In (88), while passing by through a town close to Madrid, S3 states that people are carrying their umbrellas because it is going to rain. Each of these statements creates certain inference that seems contradicted by reality in the eyes of the listener. As in the consecutive cases seen before, these inferences are inserted in an underlying resultative-conditional structure. (89) GIVEN the previous statement THEN (86) you should enforce these values in us (87) I should have seen the concert’s add (88) It shouldn’t be clearing up AND reality is such that: (86) you stop us (87) I didn’t see the concert’s add (88) It is clearing up Thus, what pues is signaling here is that this apparent contradiction to the statement previously uttered is the product of an inferential knowledge created after it. Therefore, the contradiction is not established between what the speaker has said and what the listener replies, but rather between the inference and the replica. In all these examples the underlying answer with pues is: (90) pues [tu deberias aplaudir estos valores en nosotros y sin embargo] tu los frenas. (91) pues [deberia estar oscuro y sin embargo] esta aclarando. (92) pues [yo deberia haber visto el anuncio y sin embargo] no he visto el anuncio. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 224 (90’) then [you should enforce these values in us and nevertheless] you stop them. (91’) then [it should be darkening and nevertheless] it’ s clearing up. (92’) then [I should have seen the add and nevertheless] I did not see the add. So it is clear now why pero is not a possible replacement for pues in these examples. The insertion of pero would definitely alter the type of relation, as pero establishes a direct contrast between the propositional contents of two statements, whereas pues establishes contrast through the inference created by the statement. The presence of pues is precisely what indicates that the contrast must not be established directly between two statements, but instead through the inference created by the unit to which pues attaches. Similar to other consecutive structures with pues, this relation is inserted in an underlying resultative-conditional structure. The two uses of pues to convey a relation of consequence (the first kind indicating clear consequence and the second kind I call consecutive-contrastive relation) constitute the majority of the occurrences of pues found in the corpus. The first kind represents 45.5% out of the whole number of occurrences of pues and the second kind amounts to 5.4% of all instances of pues. All together these pragmatic relations represent 50.9% of all the occurrences of pues. Finally, in the textual domain, I found that pues appears most frequently in the conversational ‘question-answer’ pair, where pues introduces the second term of this relation, i.e., answer (21.4% of all cases). I basically agree with other researchers (Portoles, 1989; Martin Zorraquino, 1991; Porroche, 1996) that at this point the use of pues in these sequences has become a sequential instrument with the main function of Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 225 helping the transition between a question and an answer. See for example the following fragment in which two cases of pues appear after their respective questions. (93) <H2> ^que te... que te puede pasar si te pillan en esas cosas? <H1> jPff! Pues un multazo pero... de ordago. <H2> <simultaneo> Y entonces, £por que lo haces? <H1> Y a ellos... </simultaneo> Y a ellos... Pues porque no habia mas manera de demostrar quien... (93 ’) <S2> what can... what can happen to you if they catch you in those things? <S1> Pff. well a huge fine but...a huge one. <S2> <simultaneous> And then, why do you do it? <S 1 > and to them...</simultaneous> to them... well because there was no other way to demonstrate who... F19L17 In (93), after each of the questions formulated by S2, SI introduces his answer with pues. The relation between both members would still remain the same even in the absence of pues. The connection between this textual pues and the pragmatic cases just seen may seem unrelated at first, but in some cases it is possible to find a link between the consecutive relation represented by pues and the relation expressed in the question- answer pair. Observe for example in (94) the way this sales woman answers her client’s request, first repeating her client’s request in the form of a question and then answering consequently to her petition. (94) <H1> Queria ver que tenia de faldas... largas principalmente. <H2> < [,D e faldas largas? pues mira, te voy a ensenar esta. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 226 (94’) <S 1 > I wanted to see what you had in skirts... long mainly. <S2> in long skirts? so look, I ’ ll show you this one. F3L3 S2’s response to S i’s request could be interpreted as a question and answer pair. The question is repeating S i’s request, and the answer with pues is uttered as a response, as a way to comply with the request. The use of pues here could be interpreted as introducing the result of the request. Stretching a bit the pattern found in other consecutive cases with pues, we could also rephrase this sequence in resultative-conditional terms: given that you request me to show you long skirts then look at what I am going to show you. Hence, we could suggest that the actual use of pues may have been initially expressing the compliance towards the implicit speech act of request that every question represents, that is: given that you ask X, then I answer Y. Pues in the textual domain is also able to represent the general relation of continuation with a very high frequency of occurrence (21.4%). The speaker uses pues to mark the continuation of her own speech as a way of topic/turn return, after an interruption, or after the insertion of some unrelated material. The length of the intervening material inserted within the part of the discourse corresponding to the same speaker’s turn may vary from a single utterance (a comment or some extra information), to an entire sequence of utterances. Pues is used then to recuperate the main argument and to link the speaker’s new units with the previous ones. Observe the following examples: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 227 (95) <H2> es el presidente del comite al... de la TEI, al que yo pertenezco y se esta encargando ahora de... del corpus oral para la TEI, o sea que lo que nosotros podamos decir para eso... <H1> Huy, que bien. <H2> Pues sera muy tenido en cuenta (95 ’) <S2> he is the president o f the... TEI committee, to which I belong and he’ s in charge now o f the...the oral corpus for the TEI, so that what we can say for that... <S1> Wow! how good!. <S2> Pues it will be very much taken into account F52L364 (96’) <H1> La he acaba<(d)>o todita, todita. <risa> <H2> Parece que estes embarazada. <risa> <H3> Pues digo que se pueden poner unos automaticos... pegando la almohada a la sabana bajera. (96’) <S1> I ’ ve finished it all, all <laugh> <S2> it seems like you are pregnant. <laugh> <S3> so I ’ m saying the we can put some buttons... sticking together the pillow to the lower sheet. F44L19 In (95) the intervening material is just the interruption SI makes to introduce a small comment of only one utterance. In (96) however, the conversation had turned from talking about pillows and beds to what one of the speakers, SI, is eating. Then after a whole sequence about an unrelated topic, S3 retakes her turn, accompanying the introduction of the unit with pues to the expression ‘I’m saying’, to remind the other participants that she is retaking her previous turn. Some other times the same speaker feels the need to come back to a part of her own argument after she has introduced unrelated information. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 228 (97) <H1> 1 Y luego este trozo, 2 que era lo que hemos metido de la cocina, 3 pues aqul se pue aprovechar para armario, 4 para meter escobas, 5 para cualquier cosa. (97’) <S1> 1 And then this part, 2 that was what we have put extra in the kitchen, 3 pues here we can use it for a cabinet, 4 to put the brooms, 5 for any other thing. F34L268 In this case SI has initiated his discourse introducing only the subject of his utterance. Then immediately he introduces additional information that intervenes between the subject and what is going to be predicated about it. To retake the structure again he introduces his next unit through pues, indicating the linkage between the first and the third units as thematic roles of the same structure. The results of pues show again that DFUs clearly represent relations in several domains. In this case in particular, pues has moved away from its initial representation of semantic relations (cause) and is able to express pragmatic and textual connections. However, it is possible to establish a common notion of result or consequence that underlies all of these functions or relations. Es que ‘it’ s that’ . Finally, the last DFU studied in detail in this section is es que ‘it’s that’. The relations that this DFU expresses are tightly related, as I will show next. In the vast majority of cases it expresses a pragmatic relation: justification (90.1%). In a much lower frequency, in the semantic domain this DFU establishes an elaboration-specific Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 229 relation between two units (5.6%). The remaining 4.3% corresponds to cases which could be interpreted as pertaining to either the pragmatic or the textual domains. FORM SEMANTIC PRAGMATIC TEXTUAL es que elaboration justification continuation/j ustific. % 5.6% 90.1% 4.3% Fig.5.16. Percentage of representation of es que in sem., pragm. and textual domains. Es que is one of the least studied DFUs in Spanish. In fact, some of the most prominent researchers in discourse analysis in Spanish hardly include this form in their classifications (Fuentes Rodriguez, 1987; Mederos Martin, Portoles, 1998; Zorraquino and Portoles, 1999). I considered es que as a DFU for two reasons. First, according to my definition, a DFU is an expression that makes explicit the semantic, pragmatic or textual relation established between different discourse units. I have found that the form es que adjusts to this definition and is capable of expressing semantic, pragmatic and textual relations. Second, the absence of the form results in the loss or under-specification of the relation holding between the units. Observe for example in (99) how the absence of es que transforms the utterance into a different type of unit, from representing a justification of the content presented by the same speaker in the previous units to an objection to the other speaker’s opinion: (99) <H1> 1 Y estuve una temporada sin hacer nada 2 yvi que se vivia muy bien ^eh? <H2> 3 Ya lo creo. 4 Estoy esperando yo tambien ese momento. <H1> 5 No, bueno, hija, hasta que tu llegues a mi edad. 6 Pero oyes, es que yo tenia entonces ya 68 anos, fijate. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 230 (99’) <S1> 1 and I was doing nothing for a while 2 and I saw that one lives very well, ah? <S2> 3 I know that. 4 I ’ m also waiting for that moment. <S1> 5 No, well, kid, until you get to my age. 6 But look, es que9 by then I was 68 already, you know? F32L324 Compare it with (100), identical to (99) with the omission of es que: (100) <H1> 1 Y estuve una temporada sin hacer nada 2 yvi que se vivia muy bien £eh? <H2> 3 Ya lo creo. 4 Estoy esperando yo tambien ese momento. <H1> 5 No, bueno, hija, hasta que tu llegues a mi edad. 6 Pero oyes, yo tenia entonces ya 68 anos, fijate. (100’) <S1> 1 and I was doing nothing for a while 2 and I saw that one lives very well, ah? <S2> 3 I know that. 4 I ’ m also waiting for that moment. <S1> 5 No, well, kid, until you get to my age. 6 But look, by then I was 68 already, you know? F32L324 Although (100) presents a complex case of the use of two DFUs in a row, pero and es que introducing the same unit, it is clear that the absence of es que does affect the relation that unit 6 establishes with the previous sequence. The two DFUs present different relations. On the one hand, pero connects unit 6 to units 3 and 4 and makes explicit the objection that unit 6 constitutes to them. On the other hand, es que expresses the fact that unit 6 is a justification of what SI has said in 1 and 2: that he decided not to do any work for some time after he retired. The presence of es que allows us to paraphrase this relation 9 The translation o f es que is a difficult one. I am not aware of an exact equivalent o f this DFU in English. The bilingual speakers I consulted accept in some cases ‘the thing is that’ as the closest interpretation. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 231 as “I did nothing for a while because I was old enough to do that”. In contrast, the absence of es que leaves only the DFU pero leading the unit and presenting it as an objection to S2’s utterance in unit 4 { ‘ I ’ m also waiting for that moment’ ). This relation could be paraphrased as: ‘you are waiting for that moment, but you are still too young in comparison with the age I had when I did it’. So there seems to be a genuine reason to identify the form es que as a DFU. As well as the other DFUs seen so far, es que is able to express rhetorical relations in discourse, and its absence may result in the loss of the relation. Therefore, I will treat this form as a regular DFU. As I said before, not many researchers have considered studying es que. Among the studies consulted in Spanish, only Cortes Rodriguez (1991) refers to es que as a possible DFU. He treats this form initially as an expletive form, that is, a ‘conceptually empty’ form with the basic function of being a tool to avoid silence, disconnection, etc. (op. cit.:29)1 0 . Later he also mentions that due to the discourse use of es que, it acquires certain senses and becomes useful for other functions (op. cit.:31). More precisely, he points out to a possible causal value of es que. In my analysis I found that es que apparently seems to express causal relation. However, a closer look to the contexts in which it appears reveals that in fact what es que establishes is a relation of justification rather that a cause for a real event. I agree with Cortes Rodriguez that es que is acquiring functions in discourse that may lead us to think that it is establishing a causal relation between two units. However, I 1 0 It is not exactly obvious to me what Cortes Rodriguez means by ‘expletive’. In his work, there is not a clear definition of this term at any point. Nevertheless, I believe that ‘expletive forms’ for Cortes Rodriguez are unable to represent relations comparable to what I understand as semantic and/or pragmatic relations. In Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 232 do not think that this is an entirely true causal relation comparable to the one established by the prototypical causal DFU porque. Cortes Rodriguez states that es que was originally part of the causal expression porque es que. Later the use of es que independent from porque would end up carrying the same meaning or property of expressing causality. Although I do not have a diachronic explanation of the origin of es que available at this point, I am more inclined to believe that this form may be a clipped part of other expressions also identified in the analysis and which introduce justification: lo que pasa es que ‘what happens is’ and its variants lo que ocurre es que / lo que sucede es que ‘what’s going on is that’. The form es que appears in these three expressions as well as in conjunction with porque (cf. Fig.l in 5.4.1.). In my opinion, the distinction between all these constructions lies on the fact that in porque es que and pero es que it is possible to identify two different relations, whereas in the case of lo que pasa es que, lo que ocurre es que, there is only one single identifiable relation. Observe the following examples in which the distinction between porque and es que that I suggest here can be clearly perceived: (101) <H1> 1 Claro. Lo deje ya preparao. 2 Es que no me gusta poner mucho ahi en el este 3 porque es que produce una cantidad de calor. (101’) <S1> 1 Sure. I left everything prepared. 2 Es que I don’ t like putting a lot there in this one 3 because es que it produces much heat. F35L210 other words, they represent sequential or textual relations. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 233 (102) <H1> 1 Claro. Lo deje yapreparao. 2 Es que no me gusta poner mucho ahl en el este 3a porque produce una cantidad de calor/ 3b es que produce una cantidad de calor. (102) <H1> 1 Sure. I left everything prepared. 2 Es que I don’ t like putting a lot there in this one 3a because ( ) it produces much heat/ 3b () the thing is it produces much heat F35L210 In (101), SI is talking about the illumination system he uses for his store. He has been away for several days and now he’s telling some friends that before leaving he left only a few lights prepared to switch on at night. In 3 he introduces a unit preceded by porque es que. Notice in (102) how the alternative between the two DFUs porque and es que yields different interpretations. In line 3 a, when es que has been removed from the unit, it is clear that porque presents a reason to explain the content of 2: ‘SI does not like putting many lights in a certain part of the store because it produces excessive heat’. So we could say that 3a is the cause of 2. In contrast, in the case of 3b, the presence of es que involves also the introduction of the speaker’s attitude towards the message presented in 2. The unit preceded by es que is intended to justify the information presented before, that is, to increase the listener’s readiness to accept the speaker’s right to present the information (Mann and Thompson, 1996:251). Thus, unit 3a is in a relation of justification with unit 2: SI presents information that prompts the listener to accept S i’s dislike towards putting excessive light in certain part of his store. Notice also for example, that the same explanation is valid for unit 2 with respect to unit 1. In unit 2, SI is also introducing a justification for the fact that he left things prepared in a certain way, that is, that he Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 234 decided to leave only ‘a few lights to be switched on’ and not all of them. The listener has not objected to what SI has said at any point, but not leaving all the lights prepared could suggest for instance that SI is trying to save some money while away from his business, which could be seen as a sign of stinginess in the eyes of the others. Thus, the appearance of porque and es que heading the same unit involves two different relations. Porque on the one hand establishes a causal relation between the two members; on the other, es que presents a justification for the speaker’s right to give certain information. The fact that both DFUs appear together and that both seem to be expressing the same relation can be explained because in many occasions the cause that makes a certain real-world event or state happen can be also used as a justification, that is, it can be used to make the listener accept the information just given. This is precisely the case we just saw in (101). In other cases however, the argument presented as a justification does not necessarily coincide with the real cause of the event. See for example in (103) how S i’s response to S2 can be only interpreted as the justification of S i’s right to pose a question: (103) <H1> 1 estos?, ^no son vestidos? o... <H2> 2 ^Estos? <H1> 3 Si. <H2> 4 No, <simultaneo> para ti son muy serios. <H1> 5 jAh! </simultaneo> 6 Es que me habi-... me habia llamado la atencion <simultaneo> el color. (103’) <S1> 1 And these?, aren’ t they dresses? or... <S2> 2 these? <S1> 3 yes. <S2> 4 no, <simultaneous> they are too serious fo r you. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 235 <S1> 5 oh! </simultaneous> 6 es que the color had-... had caught my eye. F3L111 SI is in a clothing store shopping for dresses, and while looking around the store she is interested in knowing if a stack of clothes she just saw are dresses. When the saleswoman answers no to her question, she immediately feels the need to justify her previous intervention, introducing in 5 an argument that validates her question. Obviously the propositional contents of 5 and 6 do not represent the real cause of the content of the question: the fact that the color of those clothes caught her eye is not the cause of them being dresses; nor even the reason why she thought they may be dresses. Rather, the statement headed by es que represents the justification to her asking the question. The relation could be paraphrased as: ‘the fact that I liked the color of those cloths is what prompted me to ask that question.’. Finally, an additional reason that makes me believe that es que could be more related to other expressions such as lo que pasa es que than to porque es que is that in almost every instance of a relation introduced by es que, the full form lo que pasa es que can be inserted instead without altering substantially the relation. Observe for example the fragment presented in (103) reproduced again in (104); the introduction of lo que pasa es que seems to have no effect on the nature of the relation between the two units. In contrast, when we introduce porque instead of es que the relation changes into a different one. (104) <H1> 1 i Y estos?, < ;,n o son vestidos? o... <H2> 2 ^Estos? <H1> 3 Si. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 236 <H2> 4 No, <simultaneo> para ti son muy serios. <H1> 5 jAh! </simultaneo> 6 Lo que pasa es que me habi-... me habia llamado la atencion <simultaneo> el color. 6 *Porque me habl-... me habla llamado la atencion <simultaneo> el color (104’) <S1> 1 And these?, aren’ t they dresses? or... <S2> 2 these? <S1> 3 yes. <S2> 4 no, <simultaneous> they are too serious for you. <S1> 5 oh! </simultaneous> 6 what happens is that the color had-... had caught my eye. 6 ^because the color had-... had caught my eye. F3L111 In this case, the substitution by lo que pasa es que seems to be able to maintain the same relation of justification between 6 and 1. SI is still offering an explanation as to what prompted her to ask the question. In contrast, the substitution by porque does not really preserve the relation between 6 and 1; it makes this sequence rather incoherent, it is difficult to find the type of relation porque is expressing in that position. The most prominent relation expressed by es que in the pragmatic domain is justification. However, this DFU also has some representation in the semantic and textual domains. In the semantic domain es que establishes a relation of elaboration with respect to the unit to which it attaches. The speaker provides additional and more elaborated information on the content of a previous utterance which apparently did not provide sufficient information for the purpose of communication. Observe the example in (105), in which S2 gives additional elaboration on the types of fellowships available, expanding Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 237 more specifically on her previous general statement about the different conditions for each type of grant: (105) <H2> Depende que beca. Es que mira, la de la Caixa esa sale en mayo... ju- enjunio, es cuando hay que entregar los papeles, mayo yjunio pero es para dos anos adelante, o sea que si tu por ejemplo la pides en mayo que viene... es para irte dentro de dos anos a hacer un master. (105’) <S2> It depends on the grant. Es que look, the La Caixa one comes out in May...Ju- in June, is when you have to submit the paperwork, May or June but is fo r two years or more, so that if you for example apply next May... it is to go in two years to do a master's degree. F47L21 The elaboration-specific relation has been classified as a semantic relation, since it provides additional information on the propositional content of a previous unit. However, under this relation it is also possible to understand that the speaker is providing more specific information with the attitudinal intention of increasing the listener’s willingness to accept the information, i.e. of justifying his elaboration. Here is where the link between elaboration and justification meet. And here is where es que acquires its double semantic and pragmatic nature. If and when the elaboration on certain material already presented in the discourse can also be identified as an instrument to influence the attitude of the listener towards the message, we can talk about justification. In the following example the fine line that divides these two relations is clear. On the one hand we can interpret that S2’s explanation of how she found about their meeting cancellation constitutes only specific additional information on her previous statement (‘we just found out’). On the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 238 other hand, another plausible interpretation is to consider S2’s explanation as a justification on the fact that they learned about this event before S3. (106) <H3> ^Sabiais que se ha anulao la...<simultaneo> la reunion? <H2> No </simultaneo>. Lo acabamos de saber, vamos. es que me ha llamado a mi Antonio para decirme que anulaba la reunion y a los veinticinco minutos o asi ha subido vuestro hermano Luis a decirme que... que le acababa de decir el portero que me dijera que no habia reunion <risa>. <H3> Jo<(d)>er como te tratan i,eh? (106’) <S3> Did you know the meeting <simultaneous> was canceled? <S2> No </simultaneous>. we just found out, I mean. es que Antonio called me to tell me that he was canceling the meeting and around twenty-four minutes later your brother Luis came up to tell me that... that the doorman just told him to tell me that there was no meeting <laugh>. <S3> Wow, how well they treat you! ah? F42L57 So, cases like this offer some evidence on the close links that enable the same form to represent different relations in different domains. The distinction between elaboration- specific and justification lays on the introduction of the expression of an attitudinal component. Through an elaboration relation two units are connected by their external or propositional meanings. In contrast, through a justification relation two units relate primarily by the internal (evaluative/perceptual) meaning. To sum, justification is the most prominent relation expressed by es que. My analysis shows an overwhelming majority of cases in which es que represents this relation between two units (90.1%). The relation of justification can be established Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 239 between units of the same speaker’s discourse, that is, to justify an statement that she has made previously (see examples (101), (102) and (103) in this section) , or it can also be established between the statement of one speaker and a replica, that is, presenting an argument that validates another speaker’s statements. Below I offer another example in which S2 validates what SI is saying. (107) <H1> Un buen puericultor es un tio que sabe dominar todos los aparatos del niflo. <H2> Es que es un medico de medicina general pero de ninos. (107’) <S1> A good child care specialist is a guy that knows how to treat all the child’ s organs <S2> Es que she is a primary care physician but for children. F20L1 In (4) S2 presents an argument that makes S i’s statement valid in the context of the conversation, namely that there are good and bad child care specialists. Besides a justification relation which I understand is situated in the pragmatic domain, I also found that es que in few occasions is capable of expressing a textual relation of topic return. Although this use presents a rather low frequency of appearance in the corpus (5.6%), I believe that it presents the link between the pragmatic cases of es que and the textual ones. Observe in (108) how SI, when prompted to retake a previous topic of conversation, initiates her turn with es que. (108) <H2> j Ah! Y ^que me ibas a decir de la chica esa de tu clase? <H1> jAh! Eso, es que, es que es una chica que esta muy tonta, muy tonta y va... (108’) <S2> oh! and what were you going to tell me about that girl in you class? Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 240 <S1> ah! yeah! es que, es que she is a girl who’ s very silly, very silly and she goes... F43L29 The units introduced after es que could be perfectly understood as continuing with the topic of S i’s friend’s story, and by placing these units right after S2’s request for a topic return, SI is doing exactly that, using es que to retake her turn and introduce again her main argument: that her friend is very silly. However, if we look at the units that follow this sequence, we could also interpret that es que is introducing a justification for S2 to tell a story about her friend; the story is interesting and worth telling because it shows how silly she acts. (109) below offers the sequence following (108). (109) <H2> estamos un dia en el recreo con unas chicas de clase y se pone: "Leticia, ^me acompanas a la ‘toilet’?" Yyo: W [ . . . ] y: "^Que si me acompanas a la ‘toilet, si'l vous plait’? y digo: "^eh? No, es que no me apetece". Ahi por la puerta, todas a reimos, tronchandonos. (109’) <S2> we are one day at the break with some classmates and she says: “ Leticia, can you come with me to the ‘ toilet’? And I: “ to?” [...] and: if yo u ’ d come with me to the ‘ toilet s ’ il vous plait? ” and I say: “ what? No, I don't feel like it ”. By the door, everybody laughing and laughing. S2’s story is in fact no more than one episode of her friend using French words in a regular conversation among a group of 17 year old classmates. This is seen by S2 and her classmates as a sign of stupidity that provokes their laughs. The story would not be Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 241 interesting per se in the context of the conversation, and knowingly S2 justifies her right to tell it by introducing in the beginning the reason why she is going to do so, namely that she thinks that her friend is stupid. Hence, this example shows the double nature of es que in certain sequences. It may represent a sequential relation of topic return or continuation and at the same time keep the pragmatic representation of the relation of justification. To sum up, es que displays the same feature or property of other DFUs of being able to represent semantic, pragmatic and textual relations. All the coherence relations established by es que are closely linked to each other. The separation between the relations of elaboration and justification lies on the attitudinal character imposed on the relation. Whereas elaboration implies the connection between the external or propositional meanings of the units, justification implies the introduction of the speaker’s attitude in the discourse. Finally, the examples of textual relations expressed by es que found in the data could be also interpreted as involving a relation of justification, thus suggesting that this DFU seems to be still in the process of extending its functions from the pragmatic to the textual domain. S.4.2.3. Forms more frequent in textual relations: bueno. So far we have seen forms whose more prominent functions were representing semantic or pragmatic relations. In this section I present the last DFU analyzed in detail: bueno ‘well’. It represents primarily textual relations, that is, it helps the speaker make sequential transitions in discourse. But in addition, contrary to what some researchers have suggested (Bauhr, n.d; Porroche, 1992; Briz, 1993; Fuentes Rodriguez, 1993), I found that bueno is capable of establishing pragmatic and semantic relations as well. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 242 Bueno 6 w ell, Much of the literature on bueno as a DFU (Bauhr, n.d; Porroche, 1992; Briz, 1993; Fuentes Rodriguez, 1993) as well as the literature on its English counterpart well (Schiffrin, 1987) have concentrated on the values that this form represents in a pragmatic domain, thus leaving aside the possible issue of whether bueno may also perform a semantic function, that is whether bueno may connect the propositional contents of units in any way. My analysis on data from Peninsular Spanish confirms the general idea that the predominant function of bueno belongs in the textual domain. Bueno establishes what I call parenthetical relations (commentary, digression and aside) and sequence shift in 42.5% of the cases. However, in addition to this, bueno does also represent the semantic relation of partial correction in 31.2%. Finally in the pragmatic realm, bueno introduces a partial disagreement against another speaker’s opinion (26.3%). All these relations expressed by bueno stand very closely related. They all introduce additional information which may modify to some extent the content of a previous unit, but always on the basis of the acceptance of what has already been said. FORM SEMANTIC PRAGMATIC TEXTUAL bueno partial correction partial disagreement parenthetical % 31.2% 50.9% 42.5% Fig.5.17. Percentage of representation of bueno in sem., pragm. and textual domains. Initially, bueno seems to differ from other DFUs seen this far in a fundamental way. It is not an intrinsic DFU, that is, whereas the other DFUs are connectors in origin, bueno does not have this property. For example, pero, y and porque may not appear as Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 243 anything else but as connectors. Bueno's primitive function is that of being an adjective, qualifying a noun, and this relation can be considered semantic since “it entails the speaker’s commitment to the existence of that relation in the world” (Redeker, 1990:369). This is illustrated in examples (110) and (111): (110) Este cafe es muy bueno. this coffee is very good (-fm, -pi). (111) ella dio buenas conferencias durante su carreraprofesional she gave good(+fm +pl) talks during her career. In (110) and (111), bueno refers to the good quality of the object predicated, i.e., the coffee and the talks, respectively. However, in establishing connections between two units bueno is probably in disadvantage with respect to other expressions. In the case of entonces for example, which may also not be considered a connector in nature since originally it is a temporal adverb, We see how the presence of the adverb may initially establish a reference in time between two events expressed in two separate units, that is, it may place the two events within the same time reference or in a later time frame. Bueno, however can only establish a connection between two units through anaphoric reference. Observe in (112) how buena is interpretable only in reference to the noun that appears in the previous unit la nueva pelicula de Almodovar. (112) Ayer vimos la nueva pelicula de Almodovar. Nos parecio bastante buena. (112’) Yesterday we saw Almodovar’ s new film. It seemed very good to us. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 244 I believe that this anaphoric reference is the same mechanism that operates in cases in which bueno appears as an autonomous expression indicating consent. Observe the following example extracted from Portoles (1999): (113) A. ^Quieres un poco mas de paella? B. Bueno (113’) A. Do you want some more paella? B. OK In cases like this, bueno is considered what Portoles calls an ‘assertive adverb’1 1 . An adverb that stands for the old consensual phrase bueno esta ‘good it is’, which was used to express consent to a petition, a suggestion or an offer (Martin Zorraquino, 1992). This phrase expresses an agreement or approval on the part of a speaker of an offer or proposal made in the first unit. That is, through the mechanism of anaphoric reference the adjective bueno is attributed to the content of the previous unit. The cases of bueno that I interpreted as being used as DFUs are closely related to bueno as an expression of consent, but they stand one step further. In the expression of consent alone, bueno acts as an autonomous form, that is, it does not establish any relations between two units in discourse, and is therefore a DFU. However, when bueno acts as a functional unit that connects two discourse units (i.e., when bueno is a DFU) bueno establishes the acceptance of what has been said before and at the same time introduces a unit that modifies to some extent that acceptance. Compare for example (114) below with the 1 1 I am unclear about the reasons why Portoles treats bueno as an adverb. In other parts of his work he classifies bueno as an inteijection (Op.Cit:76-77). One possible explanation is that by calling bueno an adverb he might be referring to the invariable nature of this form; that is, in contrast with its use as an adjective, this bueno does not present overt morphological agreement in gender and number with the noun, much like adverbs derived from adjectives in Spanish. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 245 previous example in (113). It is obvious that S4 in (114) is allowing the introduction of a modification to his previous utterance without an explicit rejection of it. <H4> 1 Javi vino ayer. <H1> 2 ^Fernandez? <H3> 3^Ah, si? 4 Anda. No lo sabia. <H1> 5 No lo sabiamos, no. <H4> 6 Bueno, vino hoy por la manana, a las... 7 que hora me ha despertao a mi?, 8 que hora ha entrao en mi habitation? 9 A las... ocho menos cuarto... <S4> 1 Javi came yesterday. <S1> 2 Fernandez? <S3> 3 oh yeah?, 4 Well, I didn’ t know. <S1> 5 We didn’ t know, no A 0 0 V 6 Well, he came today in the morning, at. 7 at what time did he wake me up? 8 at what time did he get in my room? 9 at...seven forty-five... F19L89 This case differs considerably from the example in (113). In (113) bueno is the only answer to the question and the only expression of consent, and it does not express any relations between unit 1 and any other units. In (114), however, bueno is not a lexical item anymore, but rather a DFU that establishes a relation of self-correction between what S4 has said in 1 and a new statement in 6 intended to modify it. We can safely say that in (113) bueno is not a DFU but a lexical item (an adverb following Portoles 1999) that expresses consent to a request, suggestion or petition, whereas in (114) it is a DFU which represents a semantic relation of correction. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 246 In fact, correction is the most frequent semantic relation expressed by bueno-, in 31.2% of all occurrences, bueno is used to represent a partial correction of the content of a previous statement. Another typical example of correction is illustrated in (115): (115) <H1> jAno y medio viendo enfrente siempre la misma cara! <H3> Y el mis-... Y el mis-... <H1> Y la misma manera de hablar. Bueno, este habla ahora un poquito mejor. (115’) <S1> Year and a half seeing always the same face before you! <S3> and the sam-.., and the sam-... <S 1 > and the same way o f taking. Well, this one speaks a bit better now. F29L42 It is clear in this example that the use of bueno introduces an instruction to the listener: she must reconsider part of the content of the unit she has uttered previously (y la misma manera de hablar), and modify it by the new statement headed by bueno. In these cases the introduction of bueno establishes the notion that the previous statement must be partially modified or corrected, although not completely. I also considered as cases of expression of correction when the unit introduced by bueno presented a more global concept which covers or includes the content of several previous referents. The bueno unit completes the speaker’s previous statements by the introduction of a global term which makes her discourse more clear and concise. I call this relation an inclusive correction. Observe for example in (116) how the unit introduced by bueno presents a global concept that includes all the previous ones. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (116) <H2> Y todo esto lo.. .lo arreglaremos bien, todo esto, todo esto, estas bisagras, todo esto, bueno, todo lo de la puerta. (116’) <S2> And all this we...we 1 1 fix it well, all this, all this, these hinges, all this. Well, all these things in the door. F31L341 The last unit introduced by bueno ‘all these things in the door’ presents a referent in which to include all the other parts of the door that the owner mentions as needing to be fixed. Bueno is indicating that the last unit must be taken as a possible replacement for all the referents just uttered. Again, notice that as in the other cases of correction just seen, in this example the type of correction established by bueno does not indicate that the content of the previous statement is to be disregarded. Very closely related to the cases of semantic correction, I found that in interactional exchanges the speaker uses the connection with bueno to express a different perspective on the content of another speaker’s utterances, creating this way a challenge to what has just been said. In these cases, we can observe how the notion of counter position between two propositional contents which underlies the semantic relation of correction as seen above transposes into the pragmatic domain. Here the relation acquires an attitudinal component. The speaker presents a different perspective on the other speaker’s message but without disregarding it in its totality. Observe the example in (117), where SI introduces a new statement with bueno which can be interpreted as a challenge to S2 conclusion, although he does not reject the content of the utterance of S2: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 248 (117) <H2> [estas harto de] aguantarle todo el dia... £no? <H1> Si... bueno, al fin y al cabo nos llevamos bastante bien, <simultaneo> pero... (117’) <S2> [you ’ re tired of] standing him all day ...right? <S1> yes... well, actually we get alone pretty well, <simultaneous> but... F29L21 The topic of the conversation is about S2’s complaints about his job. Among other things, he sees only the same people all the time, especially his closest coworker, with whom he has said he spends all the time. Thus, S2 concludes that S1 must be tired of his coworker. SI agrees only in part with this conclusion, and then he introduces additional information that could challenge this conclusion. These are the cases which from an interactional point of view can be understood as representing a disagreement or rejection to what has been said. The introduction of the contrastive unit with bueno always indicates the opposition must be made on the basis of the acceptance of part of the other speaker’s message, i.e., it is not a complete rejection or opposition to a previous statement. In (118), for instance, S3 introduces a contrastive opinion against S4’s utterance, but the contrast is intended to reject the content of S4 message only partially. (118) <H4> En Haiti creo que... [...] todo lo de la magia negra y los zombis, [...] Debe haber un ambiente alii entre ocultismo, religion oscura y... y fanatismo... Debe ser la hostia, macho. <H3> Bueno, pa mi que tampoco tanto. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 249 (118’) <S4> In Haiti I think that... [...] all the black magic and the zombies, [...] There must be such an ambiance with occultism, dark religion and... and fanaticism... It must be incredible, man. <S3> Well, to me [I think] it is not that much. F28L275 In (118) S3 is presenting a different perspective on the social situation in Haiti than the one described by S4; in S3’s opinion what S4 is saying must be modify in its dimensional value: ‘the situation in Haiti is not that incredible’. But notice that S3 at the same time is not rejecting S4’s position in its entirety, that is, there is an implicit acceptance that there is occultism, dark religion and fanaticism, although the situation must not be as incredible or bewildering as S4 is describing. The pragmatic cases in which bueno introduces partial disagreement constitute quite a high percentage of the data, as 27.3% of bueno instances are found to represent this relation. Nonetheless, as stated in the beginning of this section, bueno is predominantly frequent in the textual domain. The most frequent group of relations established by bueno in the textual domain are the ones I identify under the general label of parenthetical. They constitute 26.3% of the occurrences of bueno. This type of relation includes commentary, digression and aside. What all these relations have in common is that the information introduced by the use of the DFU constitutes a deviation from the main line or argumentation, although it is related to the topic at hand. With the introduction of the new unit through bueno, the speaker is instructing the listener to accept the main Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 250 argument that is being presented, and to hold on to it while she (the speaker) adds new information. Sometimes the new information takes the form of a commentary, introducing the personal speaker’s opinion about some issue, but other times the speaker may go into digressions or asides, which may not need to be narrowly related with the main topic. In all cases however, the speaker returns to the main sequence; this is why I called the introduction of such information and the relation established by bueno 1 j ‘parenthetical’ . Examples (119), (120) and (121) illustrate cases of commentary, digression and aside, respectively. (119) <H 1 > Que esta muy tonta, muy tonta y en clase no hay nadie que la aguante na mas que yo, que esta a mi lao, siempre tengo que aguantarla, en fin, y bueno, hija, tambien en el fondo me da igual, pobrecilla, pero es que hay otras veces que se pasa de plasta. Pero bueno, total, que estamos un dxa en el recreo con unas chicas de clase... (119’) <S1> she is very crazy, very crazy and in class there’ s no one that stands her but me, cause she’ s next to me, I always have to stand her, OK, and well, actually I don’ t care that much, poor girl, but sometimes she’ s too dumb, but it’ s OK, so, one day we were in the break with some other girls from our class... F43L35 1 2 On some occasions, however, the speaker does not return to the original topic and the new topic becomes the central topic of the conversation. In this case I do not consider this relation a commentary hut a topic shift. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 251 (120) <H1> (he pensado) te llamo desde abajo y si no para decirte que bajamos a y media. Total que vo- estoy abajo i,no?, bueno, a todo esto ya sabes que Antonio cuando tiene mas de una llamada se pone nervioso, tarda media hora en ponerte. Y a mi lo que mas me hierve la sangre porque es que eso, es una lastima pero bueno, es que no puede ser. Total que mira, luego llaman a Nicolas (120’) <H1> {I thought) I would call you from the reception and if not to tell you that we ’ 1 1 be downstairs at halfpast. So I go- I ’ m downstairs, right?, well, by the way you know that Antonio gets nervous when he has more that one phone call, it takes him half an hour to connect you. and what makes my blodd boil is that, it’ s a shame but OK, it’ s that it cannot be. So then look, they call Nicolas. F42L32 (121) <H2> Mira yo he visto bueno, ya te lo conte, el chico ese, es que me impresiono. (121’) <S2> Look, I ’ ve seen well, I told you already, that guy, it really shocked me. F32L130 Finally, bueno establishes a third type of textual relation which is almost as frequent as the parenthetical ones (26.1% of all occurrences of bueno). These relations represent fundamentally the return to a main topic after one of the previous parenthetical units or after the introduction of another topic or a topic shift, especially the latter. (122) illustrates a typical example of topic shift. A mother and her daughter are talking about Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 252 why they put a heater in the middle of the entrance. Suddenly, the mother (S3) changes topics through the use of bueno. (122) <H2> Estaba de barrera, mama, para que cuando vengais por el pasillo salteis. <H3> Bueno, ^alguien quiere sopas de ajo? yo... (122’) <S2> It was there as a barrier, mom so that when you come in the hallway you jump. <S3> Well, someone wants garlic soup? I... F43L350 (123) illustrates an example of the use of bueno to return to the main topic of the conversation after some interactions dealing with a different topic. (123) <H2> Si... eh... no te ha... ^no te hable el ano pasado...de... en los Camavales que. hay un lunes... 0 sea, empiezan un fin de semana <H1> Me tengo que pintar los labios, aunque este muy feo, pero es que los tengo sequisimos. <H2> Nada, limpiate. [Long discussion about lipsticks and perfumes] <H1> Bueno, dime... dime, “el lunes de Camaval de resaca”, digo “el lunes de resaca del Camaval”. (123’) <S2> Yes... ah... haven’ t I... didn't I talk to you last year... about the Carnivals that... there are on Monday...? 1 mean, they start on a weekend <S 1 > I have to color my lips, although it is not appropriate, but I have them dead dry. <S2> No problem, clean yourself. [Long discussion about lipsticks and perfumes] <S 1 > Well, tell me...tell me, “ ’ Monday o f hangover Carnival’, I mean “hangover Monday o f Carnival". F12L256 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 253 In this example, SI and S2 begin talking about a typical holiday in S2’s hometown. The conversation is taking place while the two friends are having lunch at a restaurant. Before the topic of the Carnivals gets started, SI feels she must color her lips after the meal (she is feeling uncomfortable because they are very dry). After a long discussion on lipsticks and perfumes (not included here), SI comes back to the old topic again, that is, she makes the transition from the recent topic to the old one through the use of bueno. To some extent, one can argue that the relation expressed by bueno in (122) and (123) is the same, namely topic change. The only difference would be that in (122) the topic introduced by bueno is totally new, not mentioned before in the context of the conversation, whereas in (123) the topic has already been introduced or initiated before. Other than that we could consider the two cases as illustrating the same type of relation, which could be called more generally sequence change (Coulthard and Brazil, 1981). To sum up, bueno follows the pattern of all DFUs studied here so far. From a lexical item capable of establishing anaphoric reference, bueno transposes to the expression of coherence relations. As a DFU it is able to represent semantic, pragmatic and textual relations. All of the relations established by bueno seem closely related; as we saw, the expression of partial correction, partial disagreement and introduction of new sequences through bueno imply the introduction of additional information on the basis of the acceptance of what has already been said. To us, this is again a clear indication of the need to consider all the cases, not only the pragmatic ones, in relation to each other. In addition, it confirms the idea that the same historical changes that lexical items undergo provoked by the need to express textual and interpersonal meanings (i.e., Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 254 grammaticalization processes, cf. Traugott, 1982, 1989, among others) are reproduced at the level of discourse, in the expression of discourse coherence. Thus, some forms are able to express not only semantic but also pragmatic and textual relations. 5.4.3. Summary. I have presented results that show that DFUs are forms capable of expressing semantic, pragmatic and textual relations. Of the 9 forms that I focused on, my analyses showed that some presented their highest frequency in the representation of semantic relations, that is, relations established between the propositional contents of the units connected. These were si, porque, y and o sea. Others represented pragmatic relations more frequently, that is, relations that linked the attitudes and beliefs of the speakers towards the concepts or ideas expressed in the units related (pero, entonces, pues and es que). Finally, only one DFU represented textual relations more frequently: bueno. That is, in the majority of the cases this DFU expressed relations that help establish transitions between discourse sequences. Although these forms are more frequent in one domain, there are also capable of establishing coherence relations in every one of them. Although I have not investigated DFUs diachronically, the results seem to be in accordance with proposals of development of meanings in discourse which suggest that many lexical items and language constructions undergo changes provoked by the necessity of expressing interpersonal and textual meanings (Traugott, 1982; 1989). The way the forms seen in this section distribute their representation in each of the three domains (semantic, pragmatic and Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 255 textual) seems to be a perfect reflection of these diachronic tendencies. Eight out of these nine DFUs show quite clear representations in each domain. Si is the only form that still has not made its way into the textual domain, but I showed clearly how from the semantic representation of condition, si becomes representative of the interactional relations of agreement and objection. Additionally, the fact that I was also able to find some cases which were not clearly distinguishable between two domains (the case of the expression of relations of cause and explanation by porque, in the limit between a semantic and a pragmatic relation, or the intersection between the expression of temporality and causality in the case of entonces) gives us evidence that pragmatic and textual cases of DFUs should not be seen in isolation from their representation in the semantic domain. 5.4.4. Extractability. A commonly assumed characteristic of DFUs is the fact that these forms in some contexts may be omitted with supposedly no loss of any ‘meaning’ in the discourse sequence. Several researchers, especially within the Spanish tradition, have suggested that in many cases the forms I call here DFUs can be considered no more than instruments to fill up possible holes in the argumentative line or to buy some time while the processing of the forthcoming discourse takes place (Vigara, 1980; 1990; Hernando Cuadrado, 1988; Cortes Rodriguez, 1990; Beinhauer, 1992, among others). The results I am about to present demonstrate that this idea is wrong. DFUs in the vast majority of instances cannot be extracted from the context they appear in without placing at risk the coherence of the sequence. DFUs can be extracted just in those cases in which the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 256 relation expressed by them is inferable from the information provided by the propositional content of the units or by the contextual knowledge (linguistic and/or extra- linguistic). And even in those cases the speaker may have additional reasons to ensure coherence by making explicit the connection between two units. One of the ideas defended in this dissertation is that a DFU is extractable only if the relation it represents between two discourse units remains the same in the absence of the DFU. That is, given two discourse units related through a DFU which represents a specific coherence relation, the DFU is extractable if and only if the same relation and no other relation can still be identified as the one creating coherence between those two units. There could also be situations in which apart from the relation holding between the units and which the DFU helped make explicit, in the absence of the DFU other relations may be interpretable between the two units. In that case, I understand that the functionality of the DFU is lost in its absence. As stated in chapter 4, a DFU is a basic function-unit in discourse which precisely works to make explicit the relation that the speaker finds more suitable for her purposes. If we consider that the use of DFUs is to make explicit one relation out a possible set of relations, then the fact that two members could be interpreted as connected through more than one relation in the absence of the DFU defeats the purpose of the DFU itself. As I mentioned in chapter 4 as well, the necessity of representing a relation through a DFU is determined or conditioned by the number of possible relations that can be established between the two members and the type of relation. After having discussed the examples found in the analysis this issue can be explained here in more detail. Two members in discourse may bear a stronger or a weaker relation depending mainly on two Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 257 factors. First, the relation established between the propositional contents of the units, and second, the information already known by the speakers through a previous context or by extra-linguistic knowledge. For example, in (124) it is the propositional content of the two members related which makes them coherent through a clear contrast: (124) <H1> El fucsia... tiene un escote normalito el verde, no, el verde necesita ir sin sujetador. (124’) <S1> The fuchsia one... has a normal cleavage the green one doesn’ t, the green one requires that you wear no bra F3L119 A saleswoman is comparing two dresses and establishing a comparison between them: the fuchsia dress has a normal cleavage whereas the green dress has a larger one. The listener may immediately establish coherence between these two units through a relation of contrast, which is clear from the propositional content of the speaker’s discourse. Thus, the need to make the relation explicit does not arise and hence no DFU is inserted. In addition to the propositional content of the members of the relation, the contextual information, linguistic background and/or extra-linguistic knowledge may also provide important cues to establish coherence. In these cases, the only information represented in the propositional content of the units may not be sufficient to establish a coherent relation, and the contextual information may provide important clues as to how the units relate. Observe for example the case in (125), where in the absence of a DFU, the propositional content of the units alone may not allow for a possible interpretation to establish coherence between the units. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 258 (125) <H1> 1 ya sabes lo que hay 2 ya has fabricado la sonda... 3 entonces ^para que quieres multiplicarlo? (125’) <S1> 1 you know what is in there 2 you have made the probe... 3 then why do you want to multiply it? F2L60 Initially, units 1 and 2 may have different interpretations. The absence of a DFU may lead us to think that both events are part of a list of steps in the process of developing a DNA experiment. However, previously in the conversation, another speaker has made clear that the purpose of constructing and using a probe is to discover DNA structure. Given this information, another possible way of establishing coherence between 1 and 2 is through a causal relation. That is, 2 could be the cause of 1 (you know what is in there, because you have made the probe). In fact, this is the most likely interpretation after the background information provided by the context. So, the insertion of a DFU is in the first place intended to avoid possible problems in cases in which the range of possible relations is wide enough to allow for different interpretations. This is the case illustrated in (125). A similar example to the one in (125) is presented in (126). In this case, neither the propositional content nor the contextual or the extra-linguistic information could give us the kind of coherence the speaker is trying to convey. (126) <H1> 1 tiene una salon muy grande, todo a la calle. tres habitaciones, dos cuartos de bano, 2 y muy bien, todo muy independiente, 3 ( ) hay un pasillo por medio y a los lados las puertas tambien. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 259 (126’) <S1> 1 it has a very large living room, all to the street, three bedrooms, two bathrooms, 2 and very well, everything very independent, 3 ( ) there is a hallway in the middle and on the sides the doors as well. F12L14 In principle the propositional and contextual content of units 1, 2 and 3 allows for different interpretations. The first interpretation is that the introduction of information about the hallway constitutes another part of the description of the apartment: ‘three bedrooms, two bathrooms, and a hallway and doors on the sides as well’. However, another possible interpretation is that the speaker intends to convey a causal relation between 2 and 3: ‘everything is very independent, because there is a hallway and door on the sides’. Also, less likely but possible as well, the introduction of a DFU like entonces ‘so’ could lead us to interpret that the speaker finishes one sequence in 2 and she is ready to go on the next sequence, to talk about a different part of the apartment. All these alternative interpretations are possible, but in fact, the speaker places the DFU o sea ‘that is/I mean’ before unit 3, to introduce a restatement which gives specific information and clarification on the content of the previous unit (unit 2), which states that ‘everything is very independent’. So the speaker has used a specific form (o sea) to make explicit the relation she judges to be more convenient to establish coherence. That is to say, among a range of possible relations that would make these units coherent, she has picked out the most appropriate one to convey the information in the way she wants and she has made this specific relation explicit through the use of a DFU. Without the presence of this DFU, other interpretations of the relation between those two units would have been possible. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 260 The other case in which the presence of DFU is essential arises when the range of relations between two units is too small or non existent, so that the speaker needs to make explicit what sort of coherence she is establishing. In these cases, the propositional content of the members related does not establish a strong relation and the linguistic and extra-linguistic contexts fail to provide enough information to make those units cohere. Observe for example the case in (127), where the actual DFU expressed is omitted for the sake of illustration, its place of occurrence being represented in brackets. In the absence of a DFU, apparently there seems to be no connection between units 2 and 3. (127)<H2> 1 Yo no me meto en nada. 2 Yo no me meto en nada 3 () este senor estahaciendo laguerradesdehace anos. (127’)<S2> 1 I ’ m not saying anything.. 2 I ’ m not saying anything. 3 () this man has been making the war years ago. F31L220 The content of these units seems to be unrelated at first sight. That is, one could think that units 1 and 2 correspond to different sequences in discourse. However, this is not the case: the speaker related units 1 and 2 and 3 through the DFU pero, which introduces a concession relation between two speech acts, the first of them explicit and the second one omitted. This relation could be paraphrased as: I ’ m not saying anything, but [what I will say is that] this man has been making the war for years. (128)<H2> 1 Yo no me meto en nada. 2 Yono me meto en nada 3 pero este senor esta haciendo la guerra desde hace anos. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (128’)<S2> 1 I ’ m not saying anything.. 2 I ’ m not saying anything. 3 but this man has been making the war years ago. In cases like this, then, the presence of the DFU is absolutely necessary to achieve coherence out of this sequence. In fact, not surprisingly, in the analysis I found that the overwhelming majority of DFUs were not extractable. That is, they needed to stay in place for the relation to hold between the two members. Overall, DFUs are not extractable in 85.5% of the cases. As predicted, due to the different contribution of each type of relation to the coherence of the fragment, the content relators or DFUs which express semantic relations are non-extractable in almost any instance: 96.3% of non- extractable cases vs. 3.7% of extractable cases. This proportion is almost identical in the pragmatic domain: pragmatic relators are non extractable in 93.1% of the cases and the extractable cases augments in 3 points: 6.9%. Finally, in the textual domain the reversed proportion is observed: DFUs are extractable in the majority of the cases (58.4%), but still a great number of cases are non-extractable: 41.6%. Figure 18 shows the results. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 262 C O N T R I B U T I O N T O C O H E R E N C E * E X T A C T I V I T Y C r o s s ta b u la tio n EXTACTIVITY NON-EXTRACTIBLE EXTRACTIBLE Total CONTRIBUTION PRAGMATIC Count 863 64 927 TO COHERENCE % within CONTRIBUTION TO 93.1% 6.9% 100.0% COHERENCE % of Total 37.7% 2.8% 40.4% SEMANTIC Count % within 955 37 992 CONTRIBUTION TO 96.3% 3.7% 100.0% COHERENCE % of Total 41.7% 1.6% 43.3% TEXTUAL Count % within 155 218 373 CONTRIBUTION TO 41.6% 58.4% 100.0% COHERENCE % of Total 6.8% 9.5% 16.3% Total Count % within 1973 319 2292 CONTRIBUTION TO 86.1% 13.9% 100.0% COHERENCE % of Total 86.1% 13.9% 100.0% p <.000 Figure 5.18. Percentages of extractable DFUs according to source of coherence. Within the semantic and pragmatic domains there is little doubt that in the vast majority of the cases a DFU was needed in place to express a relation which otherwise would be lost or under-specified. However, in these domains there are also a few cases in which I judged that the DFU at hand was extractable, that is, the identification of the relation was not at risk even in the absence of the DFU for it was the obvious candidate to connect the two members. This meant that the speaker chose to make this relation explicit anyway. So the natural question is why the speaker makes this choice when apparently there is no need for it. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. If we have a closer look at each of the cases we see that there are several reasons why an otherwise ‘obvious’ relation is made ‘explicit’. Notice that the terms ‘obvious’ and ‘explicit’ are not synonymous words. I understand that a relation is obvious when it can be recognized as the default candidate to establish coherence between two discourse units. However, a relation will be explicit when the relation has been made visible to the listener by the wording of a DFU. Thus, a relation can be: a) obvious and not explicit, when that relation is the only candidate to suit the coherence connection between two units and it is not worded; b) it can be not obvious and explicit, when the relation has to be worded to be made evident; c) obvious and explicit, when the relation is the only interpretable candidate and the speaker still utters a word that shows evidence of it; d) it can be not obvious and not explicit, which would simply yield in mi scommuni cati on. The most likely cases to happen are (a) and (b), namely when the relation is obvious and therefore no need for the wording of a DFU arises, and when the relation is not obvious and thus the speaker uses a DFU to make it evident to the listener. These are the two cases I found in the analysis of the data. The other two alternatives (i.e., (c) and (d)) are predicted to be the less likely or unlikely cases to be found, due either to redundancy in (c) or miscommunication in (d)). In this work, however, I am not in the position of making conclusions about the cases in which a speaker failed to use a DFU, since what I set up to investigate was precisely the opposite pattern: the actual occurrence of a DFU. Nevertheless, I am in the position of analyzing possibility (c), i.e., that a DFU was used even when the relation was obvious. In what follows I discuss the findings of this analysis. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 264 Extractable DFUs in the semantic domain. First of all, in the semantic domain the results obtained show that the very few cases of extractable but nonetheless expressed DFUs correspond only to instances of y and o sea, and more specifically to cases in which y represents the neutral addition relation and o sea the relation of neutral restatement. The absence of the DFU in these cases leads to a juxtaposition relation which only indicates that two units are placed side by side, and the expression of the DFUs conveys the same relation. So it should have been very easy for the speaker to have avoided the use of the DFU and leave the listener to understand the connection. However, I believe the reason why she inserts them has to do with her attitude towards the form in which she wants the message to be transmitted. By making explicit an obvious relation the speaker is increasing the listener’s readiness to accept or believe what she is saying. In other words, it serves an attitudinal purpose. Observe the examples in (129) and (130), which illustrate this point: (129) <H1> Y les dijo que... nada, osea que... que ni hablar, osea que... que para nada. (129’) <S1> And he said to them that...nothing, I mean that... don’ t even think o f it, I mean that... no way. F26L83 (130) <H1> 1 y hay momentos que yo la miro a usted 2 y la veo asi, ta, ta, en dos.. .dos. 3 Yo trazo, y yo trazo una linea 4 y...y veo dos lineas, 5 y aqui puedo leer un ratito 6 pero ya lo tengo que dejar. (130’) <S1> 1 and there are times that I look at you Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 and I see you like this, ta, ta, in half... half. 3 I draw a line, and I draw a line 4 and ...and I see two lines, 5 and here I can read for a little while 6 but soon I have to stop. F31L163 It is clear that making explicit the relation between the units by the insertion of a DFU which could have been omitted altogether has a clear effect of providing extra information to the listener as to how the speaker is intending to communicate her message. In (129), SI wants to make clear that the eye operation she was willing to undergo was absolutely disapproved by her doctor. In (130), the effects of a progressive sight condition SI is suffering from are to be said one by one; SI makes sure of that by and introducing every one of this symptoms withy: correcting and restarting his utterance in 3 withy, and again introducing line 5 withy as well. Extractable DFUs in the pragmatic domain. In the pragmatic domain, the phenomenon goes along the same lines. The only cases of extractable DFUs that were expressed corresponded to instances of pero and pues. Like in the semantic domain, the explicitness of the relation through these DFUs marks an attitudinal characteristic on the part of the speaker towards the way she wants the message to be interpreted. Observe for example the case in (131), where although the relation of objection is obvious it is made explicit through pero. (131) <H3> 1 Y empieza el Coco, dice el Coco: 4 "Hostia, pues ahora venden unos machetes de puta madre 3 y con el mando desmontable 4 y que se abre... Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 266 5 y llevan.< risas> "... 6 y llevan aguja..." <risas> "... y dedal." <H4> 7 jOstras! <H3> 8 <risas> jHostia! "jPero, tio, como te vas a comprar eso!" (131’) <S3> 1 And Coco starts saying, Coco says: 2 "man, so now you can buy some real good machetes 3 and with an off handle 4 and one that you can open... 5 and they come...<laugh>... 6 and they come with a needle. <laugh> and a thimble." <S4> 7 Wow! <S3> 8 <laugh> Man! “ but, man, how can you buy something like that! ” F28L181 Marking explicitly the relation has the effect of making specifically evident S3’s attitude or position towards S4’s actions. S3 wants to make clear his position of absolute objection or opposition towards his friend’s intention of buying a machete with a needle and a thimble enclosed. Another factor that may lead to the expression of a DFU in the presence of an otherwise obvious relation is that the speaker feels the need of making explicit a pretty obvious relation when there are circumstances that appear to impede a correct interpretation. This is the case when the speaker considers that she has not received an appropriate response from other participants in the conversation, which may indicate to her that there is need for more explicitness. For example, observe in (132) how after the first objection of SI towards S2’s message, SI takes S2’s justification or explanation as an unsatisfactory answer and formulates the same objection again twice, with the introduction of pero in order to make the relation explicit. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (132) <H2> si, y estas tu, asi <H1> un momenta, ^una cinta de video y estoy yo? <H2> si <H1> i,como es eso posible? <H2> de pequena <H1> £,pero como es eso posible? <H2> y haciendo asi <H1> pero di ^como es posible? (132’) <S2> yes, and you are in it, like this <S1> one moment, a videotape and I am in there? <S2> yes <S1> how is that possible? <S2> when you were little <S1> but how is that possible? <S2> and doing like this <S1> but say it, how is that possible? F11L95 It is pretty obvious sthat S2 is not responding to S i’s objection about the reasons why S2 has a videotape of her. SI then feels the need to make explicit the relation that connects her utterance with S2’s initial message, first through the use of pero and finally introducing a command after making the relation explicit. The insertion of explicit material has the purpose of making S i’s requests absolutely clear, which seem to be obviated by S2. Procedural constraints may also be another factor which plays a role in triggering the need for explicitness andence in the expression of extractable DFUs. Something that has become obvious tome and other scholars (Taboada, forthcoming) is that although the ideal analysis of interactional data should consider all the aspects surrounding the processing of discourse that each speaker goes through in her turns and interactions, in general the analyst is only able to work with the finished product. This fact may prevent us to some extent from understanding correctly the factors and the patterns behind the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 268 extractability of a form. When we look at the data as a finished product, what may seem to us like an obvious relation between two units may not have been that clear for the speaker in real speech time, while she is organizing her discourse. Some elements may occur that the speaker feels that distract the flow of ideas on the main topic of the interaction, such as a pause or an interruption, for instance, or the introduction of information concerning another topic. Then the speaker may feel that there is a need for an explicit connection with the idea that was being stated. Observe the example in (133): (133) <H1> 1 Y luego como el lunes alii, el lunes que viene es fiesta 2 porque es lunes de resaca, 3 ... pues yo le dije a Lola: 4 Mira, yo prefiero quedarme ese lunes en Madrid. (133’) <S1> 1 And then since Monday there, next Monday is a holiday 2 because it is ‘ hangover Monday 3 ... then I told Lola: 4 Look, I prefer to stay that Monday in Madrid. F21L37 The relation of consequence between unit 1 and units 3 and 4 may seem obvious to us, and still the DFU pues is expressed. Notice however that in unit 2 the speaker introduces additional material to elaborate the first member of the relation (i.e., why the following Monday is a holiday), which creates some distance in the flow of ideas from the second member of the relation, the consequence (she told Lola she wanted to stay in Madrid, expressed in units 3 and 4). This is what may make the speaker feel the need to make explicit the introduction of the second member of the relation, through the expression of pues. This is then an instance of a DFU making explicit a relation which may seem obvious to us when reading the interaction transcribed but which could have been seen Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 269 otherwise by the speaker, as she thought that the connection between the members of a relation may have been distracted or detoured somehow, after the insertion of elaborating material, for example. Extractable DFUs in the textual domain. Finally, the majority of extractable instances of DFUs were found in the textual domain. This is due to the nature of the textual relations themselves. The interpretation of some sequential relations like topic return or topic shift for example may not require as much degree of explicitness as the interpretation of a semantic and/or a pragmatic relation. The range of relations that can intervene between two sequences connected by a topic shift relation may be very small, so in many cases the connection between them does not offer possibilities for wrong interpretations. Besides this, as I pointed out in Ch. 4, the presence of DFUs in the expression of textual relations is seen as a way of helping the transition to a new topic or a new issue in discourse, or ‘leading into or out of a commentary, correction, paraphrase, aside, digression, or interruption segment.’ (Redeker, 1990:369). So the fact that the speaker makes these relations explicit responds in some cases to the desire of marking this transitions in a smoother way. In every instance of an extractable DFU in the textual domain I found that in the absence of the DFU the underlying textual relation was not really at risk, but the textual transition was perceived as rough and sudden. The example in (134) illustrates this point. Notice that the difference between (134) and (135) lies on how smoothly the transition is made. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 270 (134) <H1> Y ya cuando me dijo eso el otro dia digo: "Pues no se... <H2> Ya <silencio> <H1> Bueno, ya son las 10. Me voy a ir porque esta me va a llamar a las 10 y media... (134’) <S1> And when he said that to me the other day I say. "I don’ t know..." <S2> aha <silence> <S1> Well, it’ s 10 already. I ’ m going to leave because she is going to call me at 10 thirty... F21L5 (135) <H1> Y ya cuando me dijo eso el otro dia digo: "Pues no se... " <H2> Ya <silencio> <H1> () ya son las 10. Me voy a ir porque esta me va a llamar a las 10 y media... (135’) <S1> And when he said that to me the other day I say. ''Idon’ t know..." <S2> aha <silence> <S1> () it’ s 10 already. I ’ m going to leave because she is going to call me at 10 thirty... The relation established between these two sequences is topic shift and sequence shift, that is, not only does S1 initiate a new topic but she also indicates she is ready to finish the interaction (i.e., she is ready to leave). This relation in fact is not at risk when not made explicit by a DFU. However, the transition between these two sequences is much rougher when it is not made explicit by the presence of a DFU Thus, to summarize, with respect to the issue of the extractability of DFUs the first conclusion my results showed is that, due to their direct contribution to coherence, in the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 271 semantic and pragmatic domains DFUs are overwhelmingly not extractable; they must be used in order to make explicit the link that makes two discourse units cohere. There were however a small percentage of extractable DFUs in these two domains. Given my definition of extractability, namely that a DFU is extractable only if the relation it represents is the default candidate to establish coherence between two units, the question that remained to be answered was why a speaker would choose to use extractable DFUs, that is why the speaker chooses to make a relation explicit when this relation is obvious. We find three possible cases. First, the relation is not strictly necessary to establish coherence, that is it establishes coherence only in an indirect way. This was the case for the majority of the textual relations. Topic shift, sequence shift or topic return are relations easily inferred, as the range of relations that may appear between sequential units joined as topic shift or topic return is much smaller that in the case of semantic or pragmatic relations. The fact that the speaker decides to make them explicit anyway may respond to her desire of making these transitions in a smoother way. Second, in the semantic and pragmatic domains where relations contribute directly to coherence, we find that by making explicit an obvious relation the speaker incorporates and modifies the way the message must be interpreted. That is, the need of explicitness responds to the speaker’s desire of providing the listener with additional information as to how she is intending to communicate her message. Finally, in some cases what may seem like an obvious relation to us (as analysts dealing with a transcribed product) may not have been that evident to the speaker at the time of the interaction. I illustrated this point with a case in which the insertion of extra material made the speaker express the coherence relation explicitly to ensure that the listener would make the correct attachment. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 272 5.5. Conclusions. In this chapter I have presented the empirical study on Spanish natural data. The sample analyzed corresponded to 68 conversations extracted from the Corpus de Referencia del Espanol Actual (CREA). The total size of the sample was of 95,683 words which amounted to approximately 16 hours of recordings. All the conversations were held between 2 and 5 adult (25-55 years old) speakers of Peninsular Spanish in familiar settings. Following the criteria set in the theoretical framework followed in this dissertation, the coding process consisted first in dividing the samples into elementary discourse units. These were identified as the minimal segments that relate though a coherence relation. Then DFUs were identified as those forms that appeared at the boundaries of the discourse units expressing a coherence relation between them. Each instance of a DFU identified was then coded into different variables: formal variables such as file and line in which the DFU appeared, speaker who uttered the DFU and number of speakers in the conversation; and the DFU’s characteristics such as relation it expressed, source of coherence (semantic, pragmatic and textual), extractability and meaning of the DFU. In the 68 files analyzed a total of 2,292 instances of 108 DFUs were found. The results showed that out of the 108 types identified, only 9 accounted for 76.7% of the results. These were pero ‘but’ y 'and', porque ‘because’, pues ‘so’, es que ‘the thing is’ o sea ‘that is’, entonces ‘then’, bueno ‘well’, and si ‘if. To explain the high frequency of these forms with respect to the others I suggested that given a set of units among which a Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 273 discourse relation holds, the degree of specificity in meaning required for the DFU is directly inverse to the explicitness of the context. The higher the contextual information (propositional/extra-linguistic/ textual knowledge) available to the listener, the lower the need to make the relation explicit by means of a DFU. The natural tendency in conversation to rely on contextual information and to produce utterances at their lower level of explicitness (Warren, 1993) explains the speaker’s preference to resort to contextual information instead of using the most explicit linguistic material. A detailed study of the 9 most frequent forms in the corpus revealed important conclusions with respect to the types of relations DFUs may make explicit. Each of these forms with the exception of si were able to represent semantic, pragmatic and textual relations. We saw that for each of them there was a close connection among the relations expressed. We saw how the introduction of certain evaluative, perceptual and cognitive characteristics into a relation could transpose it from the semantic domain into the pragmatic one, that is, from relating the propositional contents of two members to establishing a connection between the attitudes or beliefs of the speaker. The relations of condition or opposition are good examples of the connection I tried to show. In the semantic domain when two units are in conditional relation, one of them represents a necessary condition for the realization of the content of the other unit. The DFU capable of representing such a relation in Spanish is si. In the pragmatic domain the conditional relation still exists, but the same conditional structure used to connect propositional contents is able to express the speaker’s beliefs towards a statement as well. For example, in (1) it is clear that the conditional member presents a statement that works against the implicit assumption made by the speaker when asking the question: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 274 (1) <H5> ^Lo vas a llevar al piso de tu novio? <H2> No, no. Si lo necesito aqui, porque necesito muchas... estar muchas horas diarias, o sea que... (1’) <S5> Are you going to take it to your boyfriend’ s apartment? <S2> No, no. I f I need it here, because I need many., to be many hours a day, so that... F42L139 Thus the old semantic conditional relation has become a pragmatic one. And the DFU si which was able to express the semantic relation of condition is able to represent the pragmatic relation as well. As for the relation of opposition, we observe the same pattern. In the semantic domain two units could be in an opposition relation when their propositional contents implied a contrast between them which did not lead to contradiction. In the pragmatic domain the same relation of opposition may be used to introduce contrasting arguments in favor or against a particular thesis or inference. In this case again the incorporation of the speaker’s attitudes towards a statement transforms the semantic relation into a pragmatic one. Then we are not dealing with semantic contrast but rather with pragmatic concession. One step further, the speaker may use the same relation of opposition in order to put in contrast a whole sequence against the one she is about to introduce. In this case we talk about a textual relation of topic return. Pero is the prototypical DFU that expresses a relation of opposition, and pero is able to represent all the relations associated with it in the semantic, pragmatic and textual domains. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 275 So, it seems reasonable to say that there is a strong connection between all the relations expressed by a DFU. This takes us to a very important conclusion, which is that DFUs are forms that must be considered in its totality, that is, that claiming that a DFU is only a pragmatic or a textual operator presents an incomplete view of the issue. A DFU is a functional unit able to perform its duties not only in the pragmatic and textual domains, but also in the semantic realm. Excluding this part from their definition neglects an important characteristic of the function of DFUs. In addition to these conclusions the present results also showed that DFUs are not be taken by ‘expletive’ forms, that is DFUs in the bast majority of the cases are not extractable. Only in the minority of cases I found that they could be extracted but the speaker makes the choice of using them. In this cases the need to make a relation explicit through the use of a DFU responds to other reasons such as the speaker’s desire of making smoother textual transitions or to provide the listener with additional information as to how she is intending to communicate her message. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 276 CHAPTER 6. CONCLUSIONS In this final chapter, I briefly summarize the major points of this dissertation and discuss the theoretical implications of the findings of this study. 6.1. Summary. I began my doctoral research with the main goal of providing a characterization of the category ‘Discourse Marker’ that would overcome the shortcomings of prior definitions, whose basis often had an ad-hoc nature. In chapter 2, I reviewed a possible methodological approach in which a set of DMs could be defined primarily on the basis of some shared formal properties, such as their distributional properties, or the prosodic characteristics they frequently seem to display Then, the conclusions extracted from the analysis of these partial groups of forms have been used to define a general category of DMs. I demonstrated that this methodological approach was unsatisfactory in that criteria were not well-defined and apply only to a subset of the set of forms classified as DMs. On the other hand, if our objective is to obtain reliable criteria capable of making general predictions Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 277 about the conditions under which a linguistic expression is to be considered a DM, we need to propose a theory of DMs that can survive with independence of the analysis of the particular forms themselves. I thus concluded that there is a need for a more independent discourse framework, one in which these forms could be defined in terms of some intrinsic (not circumstantial) features. Consistent with these premises, I claim that the most reliable way of defining these forms is by integrating them as a category within a general discourse framework, on the basis of independent principles. I propose a definition of categories in discourse on the basis of functional roles instead of structural features. Based on the types of meanings segments in discourse express I suggest distinguishing two types of units, DPUs and DFUs. DPUs are those segments that express propositional content, defined in terms of semantic and phonological features. DFUs are units that bear no propositional content but only procedural one; they transmit information on how to interpret the DPUs in order for them to be coherent. With respect to the particular discourse model adopted to account for the functionality of these units, I focused my attention on theories based on the notion of discourse coherence, and in particular on those that define discourse as a series of compositional units organized around certain principles that hold them together and make them cohere. I grouped these theories under the generic name of Coherence Relation Theories (Grosz and Sidner, 1986; Mann and Thompson, 1988; Sanders et al., 1992; Hobbs, 1990; 1995). These theories assume that there are certain cognitive Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 278 resources and inferences a listener/reader uses to obtain a coherent interpretation of a piece of discourse, and these are the coherence relations. The theoretical definition of a coherence relation adopted here is ‘a binary predicate whose arguments are discourse events or states of affairs.’ On the basis of these theoretical tenets, I then define DFUs in more detail as those linguistic expressions that make explicit the specific coherence relation between two DPs. Discourse is thus understood as a series of DPUs defined and interconnected by a series of coherence relations, and some functional units (DFUs) whose mission is precisely to be the linguistic expressions of those coherence relations (Cf. Figure 1 below). COHERENCE RELATION Figure 6.1. Discourse constituency. In this framework DFUs form a functional category with a particular role in the discourse model: making explicit the coherence relations between two propositional units. This model is supported by independent principles, namely the coherence relations, which stand on their own as primitive features in the theory and not just as tools designed specifically to account for the behavior of DFUs. This is precisely what enables us to define DFUs independently of the formal properties of Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 279 the linguistic expressions themselves. From a functional point of view, within a global discourse framework, DFUs are identified as all those forms capable of representing coherence relations in a discourse fragment. It is important to stress that these coherence relations between DPUs exist regardless of the presence of a lexical expression that realizes them overtly, as they are rhetorical relations holding between two units of discourse. Thus, in some cases, even if no DFU were expressed, the relation between two DPUs would be interpreted correctly. However, in other cases more than one relation could be interpreted to hold between two DPUs in the absence of a DFU, and thus a DFU is needed in order to ensure interpretability. That is, when the information presented in the propositional content of the DPUs and/or in the linguistic and extra-linguistic context is not sufficient to ensure an unambiguous interpretation on the kind of discourse coherence relation that holds between two DPUs, the need arises for a linguistic expression to make the coherence relation explicit. Since the DFUs’ main function is to represent coherence relations, a critical issue to examine is the interaction between linguistic expressions that act as DFUs and the coherence relations they represent. I propose that the match between DFUs and coherence relation is not made in a random manner. There are certain constraints that make certain linguistic forms capable of representing some relations but not others. It is at this point that the controversial issue of the meaning of DFUs comes into play. The most important formal feature attributed to DFUs is the proposed procedural meaning. I support the notion that DFUs only represent information Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 280 concerning ‘how’ real world events, states of affairs, and/or their discourse representation relate to one another, in contrast to DPUs, which provide actual information on those events or states of affairs and/or on their discourse representation. Within the framework adopted in this dissertation, it is absolutely clear that how a particular DFU relates two units (i.e., the procedural meaning that that DFU encodes) corresponds to the meaning of the relation it represents. But the problem faced here is that certain expressions may appear as the representation of not only one, but of several different relations depending on the context. For some authors this fact constitutes proof of the polysemy of DFUs; however, for others it can be interpreted as a sign that DFUs only have one central meaning that is capable of acquiring different interpretations in different contexts. I argue that we must look at this issue from an evolutionary point of view. As shown by Traugott (1982; 1985, among others), DFUs are capable of acquiring new meanings in the same manner that other linguistic expressions do: by the incorporation of inferential implicatures obtained from the contexts in which they appear. Thus, what we understand by a DFU’s meaning must coincide with the meaning of the relation it most frequently represents in the majority of the contexts, but we must also bear in mind that the same linguistic form is capable of extending its functions into the representation of additional relations in other contexts. In Chapter 4 ,1 illustrate this point with the case of the DFU after: in the majority of the contexts after represents a temporal relation, but as Spooren (1997) shows, it also appears in other contexts expressing causal meaning (Cf. 4.2.3). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 281 However, the incorporation of new relations by the same form is not realized in a free manner; it is only produced under certain constraints: 1. The first constraint involves the compatibility o f the meaning o f the lexical form and the meaning o f the relation it comes to represent in discourse, in accordance with the assumption in the literature on DMs that all such linguistic forms have evolved from lexical forms with specific lexical meanings. This constraint is that both meanings must be compatible. I demonstrated how this could be proven under a theory that considers the meaning of coherence relations to be composed of a combination of primitive features (Sanders et al., 1992). The meaning of the relation and the meaning of the linguistic form are compatible if they share the majority of their basic features. Thus, a linguistic expression like a temporal adverb is an excellent candidate to become a functional unit in discourse and represent a relation like causality, as, according to this typological characterization of the meanings of relations, temporal and causal meaning differ only in one primitive feature (Spooren, 1997:153). It is important to note that the meaning compatibility constraint applies not only to content forms evolving into functional units, but equally to actual DFUs extending to new relations: that is, any particular DFU can express only relations that are compatible in their semantic categorization. 2. The second constraint in the DFU-coherence relation interaction concerns the categorization of relations in terms of their specificity: the same DFU can only extend its functions into the representation o f relations which have equal or lower levels o f specificity. The term specificity in general refers to the degree to which the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 282 meaning of a relation is subsumed within the meaning of other relations. For example, the relation between contrast and concession is hierarchically organized from more general to more specific. That is, contrast is a more general meaning than concession. According to Marcu et al. (1999), this can be explained by the fact that a concession relation can be understood as a type of contrast relation (a contrast between two arguments, one in favor and another against the same common thesis), whereas a simple contrast relation does not necessarily imply the notion of concession between two units. Therefore, under this constraint we predict that a DFU like pero ‘but’, the prototypical representative of contrast relation, should be easily extendable into the representation of concession. However, it is more difficult for a DFU like aunque ‘although’, which represents concession, to express simple contrast. 3. The third and final constraint deals with extractability: a DFU should be more easily extractable when expressing textual relations than when expressing semantic or pragmatic ones. The correspondence between a DFU and the type or source of coherence it represents has led some researchers to propose different types of DFUs, those which represent semantic and pragmatic relations, and those which represent textual relations. My position in this regard is that we can only talk of different types of DFU as long as there exists a correspondence between a particular linguistic expression and one and only one type of relation. For example, we could say that a DFU like pero ‘but’ would be a semantic/pragmatic DFU (a message relator, in Fraser’s terminology) if this DFU limited its functions to the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 283 representation of semantic and pragmatic relations. However, this is not the case; the empirical study of DFUs presented in this dissertation shows that linguistic expressions (and pero in particular) are capable of representing relations in the three domains: semantic, pragmatic and textual. Therefore, it is difficult to establish an exclusive correspondence between a linguistic form and only one particular function. Nevertheless, I argued that the type of coherence each relation represents must have a direct influence in the extractability (rather than the non-extractability) of DFUs. Semantic and pragmatic relations contribute directly to coherence, whereas textual relations only do so in an indirect manner. The semantic interpretation of fragments which are connected by a semantic or a pragmatic relation does not correspond to the sum of the interpretation of the fragment members in isolation. Similarly, these fragments cannot be interrupted by any intervening material which is not semantically conjoined to the first unit. These two properties are not true of textual relations. This different behavior of textual versus semantic and pragmatic relation has a direct impact on the behavior of the DFUs, particularly with respect to the property of extractability. A DFU which represents a semantic or a pragmatic relation must be less susceptible of extraction than a DFU which expresses a textual relation, due to the different nature of each of these types of relations. After a detailed definition of the theoretical framework adopted in this dissertation and the constitution of the category of DFUs and their properties, I tested (in Chapter 5) this framework against real data, based on an empirical analysis of 68 natural conversations between adult native Spanish speakers, extracted from the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 284 CREA (Corpus de Referencia del Espanol Actual). These conversations represent an excellent sample of Contemporary Peninsular Spanish; they were collected in Madrid and its surrounding metropolitan area between 1990 and 1994 by researchers of the Universidad Autonoma de Madrid and are reliably transcribed. The data were coded and analyzed following the premises of the theoretical framework. The results of these analyses confirmed predictions about the behavior of DFUs, thus bearing crucial implications for the conceptualization and development of these forms. In what follows I present the general conclusions of the empirical study and their implications for a theory of DFUs. 6.2. Conclusions from the empirical study and implications for a theory of DFUs. 1) The first major finding of the analysis is that out of a total of 108 types of DFUs found in the 68 conversations analyzed, only 9 of them accounted for the vast majority of the results (76.7%). That is, 9 DFUs are used extremely frequently, and the other 99 appear very infrequently (23.2%). A well-known tendency in language can explain these results: the inexplicitness tendency. This general tendency in language affects the use of DFUs in the same way it does the use of any other linguistic expression (Sinclair, 1991; Warren, 1993; Finegan and Biber, 1994) and can be defined as the speaker’s tendency to use the lowest level of explicitness required to ensure communication, relying as much as possible on contextual Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 285 information. The 9 most frequent DFUs found are the ones capable of representing the least specific coherence relations. By the inexplicitness theory, if the successful interpretation of a fragment can be resolved by either a DFU which represents a less specific relation or a DFU that represents a more specific relation, speakers prefer to use a linguistic expression that can express the least specific relation and rely on contextual information (linguistic and extra-linguistic). A linguistic form that expresses a more specific relation is chosen only when there is a higher need to ensure coherence. This explains why in the data in the majority of cases speakers choose to use those DFUs that represent the least specific relations versus the more specific ones. The next two findings are intimately related and together provide the most important implications for a theory of DFUs. 2) The first of these two is the finding that 8 of the 9 most frequent DFUs are capable of expressing relations in each of the three possible domains: semantic, pragmatic and textual. With the exception of si ‘if, all these forms (pero ‘but’, y ‘and’, porque ‘because’, pues ‘so’, es que ‘the thing is’, o sea ‘that is’, entonces ‘then’, and bueno ‘well’) were found to represent semantic, pragmatic and textual relations. This result is extremely important for two reasons. First, it argues against the assumption that each DFU bears only one type of coherence relation. That is, it supports the proposal defended here that the same linguistic forms are capable of having representation in different domains. In the second place, this result also helps us give a more unified and consistent notion of what a DFU is. The traditional Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 286 proposal among discourse studies that these forms are primarily ‘pragmatic’ operators unclassifiable in other levels of analysis is now replaced by a different understanding of these forms. A DFU is a discourse functional unit capable o f expressing not only pragmatic relations, but also semantic and textual ones. A linguistic form does not necessarily have to have exclusively a pragmatic function to be considered a DFU. For example, a form like pero ‘but’ is a DFU as long as it constitutes the representation of a coherence relation. I have also argued here against the notion that DFUs can only be found at an intersentential level. A discourse analyst is typically presented with a stretch of discourse in which only some percentage of the units correspond to the abstract notion of ‘sentences’. As I showed in Chapter 3, in most cases this division into sentences is difficult to maintain. Therefore, it seems unreasonable to establish a particular status for a linguistic form such as pero in certain contexts, (such as the intersentential context, joining two clauses (cf. line 3 below)) and a different status when pero appears in another context (such as at the beginning of a turn of speech, without joining two sentences in the same turn (cf. line 7 below)). Unless there are other intrinsic reasons why the same linguistic form should be considered differently at different levels of analysis, I proposed to respect the functional criteria and consider a form like pero a DFU in every context in which it develops the same function. For example, in the following fragment all the pero cases (and in general any other expression that functions in the representation of coherence relations) should be considered DFUs. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 287 (1) <H2> 1 y si no... 2 ^que te... que te puede pasar si te pillan en esas cosas? <H1> 3 jpffl pues un multazo pero... de ordago. <H2> 4 <simultaneo> Y entonces, £por que lo haces? <H1> 5 y a ellos... </simultaneo> Y a ellos... 6 pues porque no habia mas manera de demostrar quien... <H2> 7 pero, en realidad, ^quien tuvo la culpa? <H1> 8 el. (1’) <H2> 1 and if not... 2 what can happen to you if they catch you in those things? <H1> 3 jpjf! well a big fine but... a huge one. <H2> 4 <simultaneous> and then why do you do it? <H1> 5 and to them... and to them... </simultaneous> and to them... 6 well because there was no other way to prove who... <H2> 7 but actually who was at fault? <H1> 8 him. (CREA. F19L55) In the two cases above, pero carries the lexical representation of two different coherence relations. In line 3 pero expresses a relation of contrast to expectation; in line 7, it expresses the speaker’s request for clarification. Therefore, in both cases this form should be considered a DFU, regardless of the structure of the units connected -sentential or not. As mentioned above, si ‘i f was one of the 9 most frequent DFUs analyzed, but the only of this group to represent exclusively semantic and pragmatic relations (and not textual). I give an account for this fact in association with the next finding. 3) The third important result obtained from the analysis presented is that there is a narrow fit between the meanings o f all relations represented by the same Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 288 DFU. All the relations expressed in the three different domains or sources of coherence (semantic, pragmatic and textual) by the same linguistic form seem to share a common basic relation. For example, to mention only a few cases, I demonstrated that all the relations expressed by si shared a common notion of condition, y represented in all three domains the same basic relation of addition, a common notion of opposition was found in all cases of pero, and a common relation of consecutivity was evidenced for all instances of entonces ‘then’, etc. It seems that, in general, the relation expressed by a DFU in the semantic domain is maintained and carried through in the two other domains with the addition of some additional features. In this respect, I believe that this finding is explainable in the light of Traugott’s proposals of semantic shift in language. The framework that Traugott (1982; 1988; 1989) presents for the analysis of semantic change in language can be applicable to the case of the representation of new relations by DFUs. However, some of the concepts of her proposal must be adapted in some ways for them to be compatible with this model. Traugott utilizes a slightly modified version of the Hallidaian functional-semantic model of language (Halliday 1977 and following). As explained in Traugott (1982), she conceives a linguistic system of three components: semantic, textual and expressive. Semantic or ideational ‘involves the resources of the language for making it possible to talk about something’ (:248); the textual component involves those ‘resources available for creating a cohesive discourse’ (:248); and finally, the expressive or interpersonal component involves the resources to express ‘personal attitudes to what is being Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 289 talked about, to the text itself, and to others in the speech situation’ (:248). These three components are in fact very similar to the ones I have proposed for the analysis of DFUs here. Her first and third components, semantic and expressive, are roughly equal to the semantic and pragmatic domains. However, there is one more component, the textual one, that for Traugott represents a more general concept. Whereas in this dissertation textual involves only those purely sequential aspects of discourse, in Traugott’s terms all devices that the language disposes of to create cohesion are included in the textual domain. In this view, the entire present study, that is, the study of linguistic expressions that represent coherence relations, should be included in the textual domain. These different perspectives on a model of analysis may seem at first difficult to reconcile, but I suggest a plausible approximation between the two views: the present study would be in fact included in the textual component of language, but I propose to understand this component as a separate component in which the three domains are reproduced again, that is, the textual component understood in the sense of Traugott would be a small scale representation of a larger model of language, as illustrated in Figure 2. SEMANTIC DOMAIN PRAGMATIC DOMAIN TEXTUAL DOMAIN Fig. 6.2. Textual component of language. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 290 Within the textual component we can again find devices that are capable of establishing cohesion through the semantic, pragmatic or textual aspects of the units connected. Once these modifications are introduced, I believe that the tendencies that Traugott argues that operate over semantic change are also applicable to the explanation of DFUs’ representations of new relations. Traugott (1989) basically points out three tendencies which can account for a theory of semantic shift in language. I focus only on two of them. The first and more important for the present study is the one that proposes that ‘meanings tend to become increasingly based on the speaker’s subjective belief state/attitude toward the proposition’ (:35). This can clearly explain how all the forms analyzed in this corpus are capable of expressing a semantic relation and a pragmatic relation, and both are narrowly related. For example, the DFU entonces ‘then’ represents a temporal relation in the semantic domain, and it represents a relation of consequence in the pragmatic domain. This DFU (‘then’) offers the same distribution in English as well (cf. Schiffrin, 1992). The explanation for this particular shift in the representation of relations by DFUs can be accounted for by a very powerful notion in pragmatics and in linguistics in general: the inferences created by conversational implicatures. Koning and Traugott (1988) explain precisely the case of ‘connectives’ like hence or since in English, which initially express temporal relations and due to their inferential interpretation in conversational contexts, end up adding the expression of causality to their other meanings. We can say that this is exactly the case we find with entonces. This DFU Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 291 represents the semantic relation of temporality, but it is also able to express the pragmatic relation of consequence. The link between these two relations seems to be in the conversational implicature that infers that the expression of precedence in time can also refer to the sequence of cause-consequence (Koning and Traugott, 1988:112). The inference is understandable when we observe cases like (2) below, in which it is clear how entonces is easily interpretable as expressing either consequential temporality or consequence of an action as a result of another action: (2) <H1> Llegaba... entraban ellos pero todavia estaban saliendo los novios anteriores, entonces se quedaron parados, aqui, hasta que salieron los novios y se marcharon. (2 ’) <S 1 > They arrived... they were getting in but still the previous couple were getting out, then they stayed still, here, until the couple got out and left. F12L77 So, it seems that DFUs represent one example of this general tendency of incorporation of pragmatic meanings. A DFU always expresses a basic coherence relation, and this relation in certain contexts incorporates the expression of aspects related to the speakers’ beliefs or attitudes; then what initially began by being an inferential use of the DFU eventually becomes a common use of it. But in this entire process the basic meaning of the relation (what we might call its core meaning) is Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 292 always present throughout all the relations expressed by the same linguistic form. In the particular case of entonces it is easy to see how the basic idea of ‘consecutive actions’ is always at the basis of both temporal and consequence relations. In the case of temporality, this is a relation identifiable in the real world: if event 1 happens before event2, then event2 happens after event 1; it is a semantic relation. In the case of consequence relations, there has been the incorporation of the speaker’s beliefs that two consecutive actions are also in a relationship of cause and consequence, although in the real world it might not necessarily be the case. We have then a pragmatic relation. As I mentioned above, this aspect can be considered a general tendency, as it was observed for all of the 9 DFUs analyzed in the corpus. The second tendency of semantic shift pointed out by Traugott and which can be applicable to DFUs concerns the expression of textual relations by DFUs. In this respect, Traugott (1989) also suggests that there is a general tendency in language by which ‘meanings based in external or internal described situation(s) [tend to become] meanings based in the textual and metalinguistic situation.’ (:35). However, as I explained above, the term ‘textual’ in this case represents a different notion than the one used in my work. In Traugott’s framework this tendency accounts for those lexical items that eventually become linguistic devices to express coherence in the construction of texts (i.e., DFUs). For example, she mentions precisely the case of connectives like while, that from signifying ‘the time that’ changed into the expression of a textual relation like ‘during’ (:35). In this case, I are not trying to offer an explanation for the shift of lexical expressions into DFUs, but rather for the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 293 ability of a linguistic expression that already functions as a DFU to express textual relations {sequential relations, to distinguish them from Traugott’s terminology), in addition to semantic and pragmatic relations. In this respect, these data show that there is in fact a general tendency for the same DFUs that express semantic and pragmatic relations to express textual relations as well. In addition to this, the meaning of the textual relation incorporated into the DFU is closely related to the meaning of the semantic and pragmatic relations. For example, returning to entonces, we can say that this DFU is capable of representing also the textual relation of continuation, understood as consecutivity of two sequences in discourse: after having said something, the speaker is going to add something else. Thus, the initial meaning of placing two actions, events, states and textual sequences consecutively is maintained in all the relations represented by this DFU. There remains at least one open issue with respect to the representation of semantic, pragmatic and textual relations by DFUs, and this is the direction in which they make the incorporation of new relations. In other words, the question that arises would be: is it possible for a DFU to express semantic and textual relations only? Or on the contrary, does a particular form always have to express a pragmatic relation in order for it to be able to represent textual relations? The data considered here do not provide enough information to reach any conclusions on the direction that the extension into new relations can follow. Without a diachronic study, it is impossible to predict whether the expression of textual relations by DFUs becomes a direct extension of their representation of semantic relations (i.e. semantic > textual | Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 294 semantic > pragmatic), or whether the path of the extension of relations must always be semantic > pragmatic > textual. In the data analyzed, the only finding that offers some suggestion with respect to this issue is the result obtained for the DFU si. In this case, as I mentioned above, I found that si only represented semantic and pragmatic relations, but not textual. This fact could lead us to conclude that in general all DFUs express semantic and pragmatic relations and only some of them express textual relations. However, this conclusion would be risky for two reasons: first, 1 DFU out of 9 provides too little evidence to guarantee the reliability of the results. Second, according to Traugott, not all forms must necessarily evolve in the same way; that is, some forms may be used by the speaker to express pragmatic or textual relations and not others, but this only indicates than some linguistic expressions are used by the speaker to express different relations at different points in time: To say that textual or expressive meanings are newer than propositional ones does not mean that a language can at some stage have only propositional meanings, or only textual and expressive ones. It means simply that, given a form X, the textual meanings associated with it, if any, will be later than the propositional meanings associated with it1 , etc. (Traugott, 1989:34). For these reasons, the only safe conclusion we could draw from the analysis of these data is that in principle all DFUs are capable of representing semantic, pragmatic and textual relations. These relations always exist in the language and are Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 295 available for the construction of discourse. When the need for the explicit expression of some of these relations arises, certain forms already in the language are extended to cover these needs. This means that some forms extend their functions into the representation of textual relations (as well as pragmatic) sooner than others according to expressive requirements; the fact that at certain stage in the language a linguistic form does not represent textual relations does not mean that this form is never going to be able to do so. So in the case of si, the fact that no cases of representation of textual relations were found in the data does not imply that si will not incorporate this function in the future; only that the right conditions must arise for it to happen. In sum, the second and third findings mentioned in this section can be considered the most important ones of this analysis. DFUs are linguistic forms capable of having representation in the three domains of language, not only in the pragmatic field. The relations they express are constrained not only by requirements of compatibility and specificity, but also by sharing basic meaning. This basic meaning, that in general corresponds to the meaning of the semantic relation, incorporates pragmatic and textual aspects according to needs that arise in the construction of discourse. Finally, all DFUs in principle are capable of incorporating pragmatic and textual relations in their functions, and the fact that some forms do not have representation in one domain (pragmatic or textual) does not imply that at some point they are not capable of doing so. 1 1 Bold face is mine. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 296 4) The fourth and final major finding of this study deals with the issue of the possible extraction of DFUs. For many researchers these forms can be considered linguistic expressions whose role in language is to fill holes in the argumentative line or to help the speaker in the processing of discourse; therefore they can be extracted with no loss of meaning in the content of discourse. I have argued in the present theory that DFUs are not expletive forms and their function is more relevant than that of space fillers. The results of the present analyses support this position, in that the vast majority of cases DFUs are not extractable without seriously compromising the coherence of discourse. As predicted, when representing semantic and pragmatic relations, DFUs are less susceptible of being extracted, due to the direct contribution of these relations to coherence. In contrast, when the same forms represent textual relations they are more easily extractable, again because textual relations do not contribute directly to coherence. My analysis focused primarily on the study of those cases in which the DFU had been used even when the relation was apparently obvious. I found three possible reasons for making a relation explicit even when the correct interpretation of the coherence relation was not at risk (Cif.5.4.3). In the case of textual relations which do not contribute directly to coherence, making them explicit by the use of a DFU could respond to a speaker’s desire to smooth transitions in discourse. On the other hand, the very few cases in which semantic and pragmatic relations could have been interpreted unambiguously in the absence of a DFU responded to a desire by the speaker to insist on the relation Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 297 that connects two units, making clearer the relation that is being conveyed in the message. In summary, all the results and conclusions presented in this thesis, support a different perspective from the traditional consideration of these forms, as expressions with little linguistic value in terms of their contribution to language. I argue that DFUs are functional units in language that deserve a profound and serious analysis, not only by themselves but also in correlation with other possible functional units in language. DFUs represent or express a specific type of relations, which I understand as rhetorical relations. However, there are very likely other functional units that express other types of relations. In this respect, the work of some researchers within relevance theory points to interesting connections. Escandell and Leonetti (1997), for example, suggest considering determiners as linguistic forms that at a different level represent the explicitation of other types of relations. The connection between different functional units in language is an interesting issue that awaits further research. This work contributes to a deeper understanding of discourse structure in general and of the Spanish language in particular. This linguistic knowledge represents as well the basis for the study of discourse in other related areas, such as computational linguistics. An adequate description of the linguistic properties of the units in discourse is a necessary step towards the development of automatic methods of discourse interpretation and production. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 298 REFERENCES Abraham, W. (ed.) (1991). Discourse Particles. Descriptive and theoretical investigations on the logical, syntactic and pragmatic properties o f Discourse Particles in German. Pragmatics and Beyond New Series, 12. Amsterdam and Philadelphia: John Benjamins. Alcina, J. and J.M. Blecua (1975). Gramatica espanola. Barcelona: Ariel. Andersen, E. (1991). Speaking with style: The sociolinguistic skills o f children. London and New York: Routledge. Andersen. E. (1996). A cross-cultural study of children’s register knowledge. In Slobin, D., J. Gerhardt, A. Kyratzis, and G. Jiansheng (eds.), Social interaction, social context, and language: Festschrift for Susan Ervin Tripp, 125-142. Hillsdale, NJ: Erlbaum. Andersen, E., M. Brizuela, B. Dupuy, and L. Gonnerman (1996). The acquisition of discourse markers as sociolinguistic variables: a cross-linguistic comparison. In E. Clark (ed.), Proceedings o f the 27th annual Child Language Research Forum, 61-70. Stanford: CSLI Publications. Andersen, G. (1998). The pragmatic marker like from a relevance-theoretic perspective. In Jucker, A. and Y. Ziv (eds.), Discourse Markers. Descriptions and theory. Amsterdam and Philadelphia: John Benjamins. Anscombre, J.C. and O. Ducrot (1977). Deux maix en franqjais. Lingua 43:23-40. Anscombre, J.C. and O. Ducrot (1983). L ’ argumentation dans la langue. Brussels: Mardaga. Aoun, Y. (1985). A grammar o f anaphora. Cambridge, MA: MIT Press. Asher, N. (1993). Reference to abstract objects in discourse. Norwell, MA: Kluwer Academic Publishers. Asher, N. & Lascarides, A. (1998). The semantics and pragmatics of presupposition. Journal o f Semantics 15: 239-299. Austin, J. (1962). Flow to do things with words. Cambridge, MA: Harvard University Press. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 299 Bainhauer, W. (1929). El espanol coloquial. Madrid: Gredos. Bauhr, G. (1994). Functiones discursivas de bueno en espanol modemo. Linguistica Espanola Actual 16: 79-124. Berry, M. (1981). Systemic linguistics and discourse analysis: A multi-layered approach to exchange structure. In Coulthard, M. and M. Montgomery (eds.), Studies in discourse analysis. London: Routledge and Kegan Paul. Biber, D. (1984). An analytical framework for register studies. In Biber, D. and E. Finegan (eds.), Sociolinguistic perspectives on register. New York and Oxford: Oxford University Press. Biber, D. and E. Finegan (1984). Sociolinguistic perspectives on register. New York and Oxford: Oxford University Press. Blakemore, D. (1987). Semantic constraints on relevance. Oxford: Blackwell Publishers. Blakemore, D. (1992). Understanding utterances. Oxford: Blackwell Publishers. Brinton, L. (1996). Pragmatic markers in English. Berlin and New York: Mouton de Gruyter. Brizuela, M. (1995). Discourse Markers as indicators of register shifts. Ms., University of Southern California, Los Angeles. Casado Velarde, M. (1991). Los operadores discursivos es decir, esto es, o sea y a saber en espanol actual: Valores de lengua y funciones textuales. Linguistica Espanola Actual 13: 87-116. Chafe, W. (1980). The deployment of consciousness in the production of a narrative. In Chafe, W. (ed.), The pear stories: cognitive, cultural and linguistic aspect o f narrative production, 9-50. Norwood, NJ: Ablex Press. Chafe, W. (1992). Prosodic and functional units of language. In Edwards, J. and M. Lampert (eds.), Talking data: transcription and coding in discourse research. Hillsdale, NJ: Erlbaum. Coulthard, M. and D. Brazil (1981). Exchange structure. In Coulthard, M. and M. Montgomery (eds.), Studies in discourse analysis. London: Routledge and Kegan Paul. Cortes Rodriguez, L. (1991). Sobre conectores, expletivos y muletillas en el espanol Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 300 hablado. Malaga: Editorial Libreria Agora. Dahlgren, K. (1996). Discourse coherence and segmentation. In Hovy, E. and D. Scott (eds.), Computational and conversational discourse: burning issues, an interdisciplinary account, 111-138. Berlin: Springer Verlag. Ducrot, O. et al. (1980) Les mots du discours. Paris: Minuit. Escandell Vidal, V. and M. Leonetti Jungl (1997). Categorias funcionales y semantica procedimental. In Actas del Congreso Intemacional de Semantica: Universidad de La Laguna. Fraser, B. (1990). An approach to discourse markers. Journal o f Pragmatics 14: 383- 95. Fraser, B. (1996). Pragmatic markers. Pragmatics 6: 167-90. Fraser, B. (1999). What are Discourse Markers? Journal o f Pragmatics 31: 931-952. Fuentes Rodriguez, C. (1993). Comportamiento discursivo de bueno, bien, pues bien. Estudios de Linguistica de la Universidad de Alicante 9: 205-221. Grosz, B. and C. Sidner (1986). Attention, intentions and the structure of discourse. Computational Linguistics 12: 175-204. Haiman, J. (1978). Conditionals are topics. Language 54: 564-589. Haiman, J. (1980). The iconicity of grammar: isomorphism and motivation. Language 56: 515-540. Halliday, M.A.K. and R. Hasan (1976). Cohesion in English. London: Longman. Halliday, M.A.K. (1978). Language as a social semiotic: the social interpretation o f language and meaning. London: Edward Arnold. Halliday, M.S.K. (1985). Spoken and written language. Geelong: Deakin University Press. Hansen, M-B. (1998). The semantic status of discourse markers. Lingua 104: 235- 260. Harder, P. (1991). Linguistic meaning: Cognition, interaction and the real world. Nordic Journal o f Linguistics 14: 119-140. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 301 Harris, Z. (1988). Language and information. New York: Columbia University Press. Heine, B., U. Claudi and F. Hiinnemeyer (1991). Grammaticalization: A conceptual framework. Chicago: University of Chicago Press. Helbig, G. (1988) Lexicon Deutscher Partikeln. Leipzig: Enzyklopadie. Hernando Cuadrado, L. (1988). El espanol coloquial en ‘‘ El Jarama ”. Madrid: Playor. Hobbs, J. (1979). Why is discourse coherent? In Gemsbacher, M. A., and T. Givon (eds.), Coherence in spontaneous text, 20-70. Amsterdam and Philadelphia: John Benjamins. Hobbs, J. (1990). Literature and Cognition. Stanford, CA: CSLI Lecture Notes 21. Hobbs, J. (1996). On the relation between the informational and intentional perspectives on discourse. In Hovy, E. and D. Scott (eds.), Computational and conversational discourse: burning issues, an interdisciplinary account, 139- 157. Berlin: Springer Verlag. Hopper, P. J. (1991). On some principles of grammaticalization. In Trauggott, E. C. and B. Heine (eds.), Approaches to Grammaticalization, Yol 1: 17-35. Amsterdam and Philadelphia: John Benjamins. Hopper, P. J. (1996). Some recent trends in grammaticalization. Annual Review o f Anthropology 25: 217-36. Hopper, P. J. and E.C. Traugott (1993). Grammaticalization. Cambridge: Cambridge University Press. Hovy, E. and E. Maier (in press). Parsimonious or profligate: How many and which discourse structure relations? To appear in Discourse Processes. Hovy, E. and D. Scott (1996). Computational and conversational discourse: burning issues, an interdisciplinary account. Berlin: Springer Verlag. Hymes, D. (1972). Modes of interaction of language and social life. In J. Gumperz and D. Hymes (eds.), Directions in sociolinguistics: the ethnography o f communication, 35-71. New York: Holt, Rinehart and Winston. Jucker, A. and Y. Ziv (1998). Discourse markers. Descriptions and theory. Amsterdam and Philadelphia: John Benjamins. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 302 Kheler, A. (1994). Common topics and coherent situations: Interpreting ellipsis in the context of discourse inference. In Proceedings o f the ACL-94. Knott, A. and R. Dale (1994). Using linguistic phenomena to motivate a set of coherence relations. Discourse Processes 18: 35-62. Knott, A. and C. Mellish (1997). A data-driven method for classifying connective phrases. Ms., University of Edinburgh. Konig, E. (1991). The meaning o f focus particles. London: Routledge. Konig, E. and E. Traugott (1988). Pragmatic strengthening and semantic change. The conventionalizing of conversational implicature. In Hiillen, W. and R. Schulze (eds.), Understanding the lexicon: meaning, sense and world knowledge in lexical semantics, 10-24. Tubingen: Niemeyer. Kroon, C. (1995). Discourse particles in Latin. Amsterdam: Gieben. Labov, W. and D. Fanshel (1977). Therapeutic discourse. New York: Academic Press. Lakoff, R. (1971). I f s, and’s, and hut’s about conjunction. In Fillmore, C. J. and D. T. Langendoen (eds.), Studies in linguistic semantics, 114-149. New York: Holt, Rienhart and Winston. Lascarides, A. and N. Asher (1993). Temporal interpretation, discourse relations and commonsense entailment. Linguistics and Philosophy 16: 437-493. Lehmann C. (1985). Grammaticalization: synchronic variation and diachronic change. Linguistic Style 20: 303-18. Maim, W.C. and S. Thompson (1988). Rhetorical structure theory: toward a functional theory of text organization. Text 8: 243-281. Marcu, D. (1998). Instructions for manually annotating the discourse structure of texts. Ms., Information Sciences Institute, University of Southern California. Marcu, D., M. Romera, and E. Amorrortu (1999). Experiments in constructing a corpus of discourse trees: Problems, annotation choices, issues. In Proceedings o f the workshop on levels o f representation in discourse, 71-78. Edinburgh: University of Edinburgh. Martin, J. R. (1992). English text. System and structure. Amsterdam and Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 303 Philadelphia: John Benjamins. Martin, J. R. (1996). Types of structure: deconstructing notions of constituency in clause and text. In Hovy, E. and D. Scott (eds.), Computational and conversational discourse: burning issues, an interdisciplinary account, 1-72. Berlin: Springer Verlag. Martin Zorraquino, M.A. (1994a). Sintaxis, semantica y pragmatica de algunos adverbios oracionales asertivos en espanol actual. In Demonte, V. (ed.), Gramatica del espanol, 557-590. Mexico: El Colegio de Mexico. Martin Zorraquino, M.A. (1994b). Bueno como operador pragmatico en espanol actual. In Alonso, A. (ed.), II Encuentro de linguistas y filologos de Espana y Mexico, 403-412. Salamanca: Universidad de Salamanca. Martin Zorraquino, M.A. (1998). Los marcadores del discurso desde el punto de vista gramatical. In Martin Zorraquino, M. A. and E. Montolio Duran (eds.), Los marcadores del discurso. Teoria y analisis, 19-55. Madrid: Arco Libros. Martin Zorraquino, M.A. and E. Montolio Duran (1998). Los marcadores del discurso. Teoria y analisis. Madrid: Arco Libros. Martin Zorraquino, M.A. and J. Portoles (1999). Los marcadores del discurso. In Bosque and V. Demonte (eds.), Nueva gramatica descriptiva de la lengua espanola. Madrid: Fundacion Ortega y Gasset. McKeown, K. and M. Elhadad (1991). A contrastive evaluation of functional unification grammar for surface language generation: A case study in choice of connectives. In Paris, C. L., W. R. Swartout, W. C. Mann (eds.), Natural language generation in artificial intelligence and computational linguistics, 351-396. Dordrecht: Kluwer. Mederos Martin, H. (1988). Procedimientos de cohesion en el espanol actual. Tenerife: Cabildo Insular. Meillet, A. (1912). Linguistique Historique et Linguistique Generate. Paris: Champion. Miller, J. and R. Weinert (1995). The function of like in dialogue. Journal o f Pragmatics 19: 435-452. Montolio Duran, E. (1991). Asi pues entonces, lo mejor sera que pienses bien lo de casarte. Acerca de los procondicionales en espanol. Foro Hispanico 2: 43-53. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 304 Montolio Duran, E. (1992). Los conectores discursivos: acerca de, al fin y al cabo. In Martin Vide, C. (ed.), Lenguajes naturales y lenguajes formales VIII, 43-53. Barcelona: PPU. Moya Corral, J.A. (1996). Los mecanismos de interordinacion: A proposito de pero y aunque. Granada: Universidad de Granada. Oliveira e Silva, G. and A. Tavares de Macedo (1992). Discourse Markers in the spoken Portuguese of Rio de Janeiro. Language Variation and Change 4: 235- 249. Oversteegen, L. (1997). On the pragmatic nature of causal and contrastive connectives. Discourse Processes 24: 51 -85. Polanyi, L. (1988). A formal model of the structure of discourse. Journal o f Pragmatics 12: 601-638. Polanyi, L. (1996). The linguistic structure of discourse. Technical Report CSLI-96- 200. Stanford, CA: Center for the Study of Language and Information. Polanyi, L. and Scha, R. (1983). The syntax of discourse. Text 3: 261-270. Pons Borderia, S. (1998). Conexion y conectores. Estudio de su relacion en el registro informal de la lengua. Valencia: Universidad de Valencia. Porroche Ballesteros, M. (1993a). Aspectos de la sintaxis del espanol conversacional (con especial atencion a y). In G. Hilty (ed.), Actes du XXeme Congres International de Linguistique et Philologie Romane, II, 83-93. Bema: Francke. Porroche Ballesteros, M. (1993b). El componente discursivo-textual en las relaciones oracionales (las oraciones adversativas). In Seminario-coloquio La oracion. Zaragoza: Universidad de Zaragoza. Porroche Ballesteros, M. (1996). Las llamadas conjunciones como elementos de conexion en espanol conversacional: pueslpero. In Kotschi, T. et al. (eds.), El espanol hablado y la cultura oral en Espaha e Hispanoamerica, 72-94. Madrid: Iberoamericana. Portoles, J. (1989). El conector argumentative pues. Dicenda 8: 117-132. Portoles, J. (1998). Marcadores del discurso. Barcelona: Ariel. Real Academia Espanola. (1973). Esbozo de una nueva gramatica de la lengua espanola. Madrid: Espasa Calpe. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 305 Redeker, G. (1986) Language use in informal narratives. Effects o f social distance and listener involvement. Ph.D. thesis, University of California at Berkeley. Redeker, G. (1990). Ideational and pragmatic markers of discourse structure. Journal o f Pragmatics 14: 367-381. Redeker, G. (1991). Linguistic markers of discourse coherence. Linguistics 29: 1139- 1171. Reichman, R. (1978). Conversational coherency. Cognitive Science 2: 283-327. Romaine, S. and Lange (1991). The use of like as a marker of reported speech and thought: a case of grammaticalization in progress. American Speech 66: 227- 229. Romera, M. (1998). The acquisition of Spanish discourse markers. Ms., University of Southern California. Rudolph, E. (1989). Partikeln in der Textorganisation. In Weydt, H. (ed.), Sprechen mit Partikeln, 489-510. Berlin: Mouton de Gruyter. Sanders, T. (1997). Semantic and pragmatic sources of coherence: On the categorization of coherence relations in context. Discourse Processes 24: 119- 147. Sanders, T., W. Spooren, and L. Noordman (1992). Towards a taxonomy of coherence relations. Discourse Processes 15: 1-35. Sanders, T., W. Spooren and L. Noordman (1993). Coherence relations in a cognitive theory of discourse representations. Cognitive Linguistics 4: 93-133. Sawoff, A. (1980). A sociolinguistic appraisal of the sibilant pronunciation in the city of Seville. Grazer Linguistische Studien 11-12: 238-262. Schegloff, E. (1987). Between micro and macro: contexts and other connections. In Alexander, J.C. et al. (eds.), The macro-micro link. Berkeley: University of California Press. Schegloff, E. and H. Sacks (1973). Opening up closings. Semiotica 1: 289-327. Schiffrin, D. (1987). Discourse Markers. Cambridge: Cambridge University Press. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 306 Schiffrin, D. (1994). Approaches to discourse. Cambridge, MA: Blackwell Publishers. Schiffrin, D. (1997). Theory and method in discourse analysis: what context for what unit?. Language and Communication 17: 75-92. Schwenter, S.A. (1996). Some reflections on o sea: A discourse marker in Spanish. Journal o f Pragmatics 25: 855-874. Searle, J. (1969). Speech acts. Cambridge: Cambridge University Press. Silva-Corvalan, C. (1997) Gramatica y pragmatica discursiva. Keynote lecture presented at the Congreso Intemacional de Analisis del Discurso, Universidad Nacional de La Plata and Universidad de Buenos Aires, Argentina, 1997. Sinclair, J. (1988). Naturalness in language. In McCarthy, M. (ed.), Naturalness in language. English Language Research Journal Monograph 2, 11-20. Soria, C. and G. Ferrari (1999). Lexical marking of discourse relations: Some experimental findings. Ms., University of Pisa. Spooren, W. (1989). Some aspects of the form and interpretation of global contrastive coherence relations. Ph.D. thesis, University ofNijmegen. Stubbs, M. 91983). Discourse analysis. The sociolinguistic analysis o f natural language. Chicago: University of Chicago Press. Sweetser, E. (1990). From etymology to pragmatics, metaphorical and cultural aspects o f semantic structure. Cambridge: Cambridge University Press. Taboada, M. (forthcoming). Collaborating through talk: the interactive construction of task oriented dialogue in English and Spanish. Ph.D. dissertation, Universidad Complutense de Madrid. Taylor, T. and D. Cammeron (1987). Analyzing conversation. New York: Pergamon Press. Thagard, P. and K. Verbeurgt (1998). Coherence as constraint satisfaction. Cognitive Science 22: 1-24. Traugott, E. (1989). On the rise of epistemic meanings in English: An example of subjectification in semantic change. Language 65: 31- 55. Traugott, E. et. al. (eds.) (1986). On conditionals. Cambridge and New York: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 307 Cambridge University Press. Tsui, A. (1994). English conversation. Oxford and New York: Oxford University Press. Van Dijk, T. (1972). Some aspects o f text grammars. The Hague: Mouton. Van Dijk, T. (1985). Introduction: discourse as a new cross-discipline. In Van Dijk, T. (ed.), Handbook o f discourse analysis, Volume 1: Disciplines o f discourse, 1-10. New York: Academic Press. Vigara, A. (1980). Aspectos del espanol hablado. Madrid: SGEL. Vigara, A. (1990). La funcion fatica del lenguage. Actas del congreso de la Sociedad Espanola de Linguistica. XXAniversario, II: 1088-1097. Vincent, D. (1993). Les ponctuants de la langue, et autres mots du discours. Quebec: Nuit Blanche. Vincent, D. and D. Sankoff (1992). Punctors: A pragmatic variable. Language Variation and Change 4: 205-216. Warren, M. (1985). Discourse analysis and English language teaching: A contrastive study of discourse-based and communicative activities. M.A. thesis, University of Birmingham. Warren, M. (1993). Inexplicitness. A feature of naturalness in conversation. In Baker, M., G. Francis and T. Bonelli (eds.), Text and technology: In honour o f John Sinclair, 37-53. Amsterdam and Philadelphia: John Benjamins. Zubizarreta, M.L. (1998). Prosody, focus and word order. Cambridge, MA: MIT Press. Zwicky, A. (1978). Arguing for constituency. In Farkas, Ch., W. Jacobsen and K. Todrys, Papers from the fourteenth regional meeting o f the Chicago Linguistics Society, 503-511. Chicago. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

A variationist study of relative clauses in Spanish

PDF

Factors influencing the interpretation of novel words as adjectives in 4-year-old Spanish -speaking children

PDF

An anaphoric approach to clitic position in Spanish

PDF

Acquisition of Spanish verb morphology by bilingual children. A longitudinal study between the ages of 2;9 and 3;3

PDF

Grammaticalization and the development of functional categories in Chinese

PDF

Beyond words and phrases: A unified theory of predicate composition

PDF

Asymmetry of scope taking in wh -questions

PDF

Issues in the syntax of resumption: Restrictive relatives in Lebanese Arabic

PDF

"Master of many tongues": The Russian Academy Dictionary (1789--1794) as a socio -historical document

PDF

Connectionist phonology

PDF

Age -limited learning effects in reading and speech perception

PDF

Gossip, letters, phones: The scandal of female networks in film and literature

PDF

Generals and particulars in Thucydides

PDF

Form and meaning: Negation and question in Chinese

PDF

A descriptive grammar of San Bartolome Zoogocho Zapotec

PDF

Ellipsis constructions in Chinese

PDF

A Descriptive Syntax Of King Alfred'S Soliloquies

PDF

A Sociolinguistic Study Of Selected Vowel Changes In Los Angeles English

PDF

Educational development aid and the role of language: A case study of AIT -Danida and the Royal University of Agriculture in Cambodia

PDF

Darwin contra Rousseau: Evolutionary narrative and the discourse on the social bond in nineteenth-century France

Asset Metadata

Creator Romera, Magdalena (author)

Core Title Discourse functional units: A re-examination of discourse markers with particular reference to Spanish

School Graduate School

Degree Doctor of Philosophy

Degree Program Linguistics

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag language, linguistics,OAI-PMH Harvest

Language English

Contributor Digitized by ProQuest (provenance)

Advisor Andersen, Elaine (committee chair), Finegan, Edward (committee member), Silva-Corvalan, Carmen (committee member)

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c16-114834

Unique identifier UC11326981

Identifier 3027770.pdf (filename),usctheses-c16-114834 (legacy record id)

Legacy Identifier 3027770.pdf

Dmrecord 114834

Document Type Dissertation

Rights Romera, Magdalena

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA