~- ‘Fq:£ MICH GAN ST RSIE IIII II IIIIIIIIIIIIIII IIIIIII.I ' 3 12 300078 6701 II This is to certify that the dissertation entitled A Speech Act Theory of Metadiscourse presented by Paul Jude Beauvais has been accepted towards fulfillment of the requirements for Ph . D . degree m Curriculum and Instruction DMe August 19, 1986 MS U is an Affirmative Action/Equal Opportunity Institution 0-12771 RETURNING MATERIALS: MSU ace n 0 mp to remove this checkout from w your record. FINES will be charged if 550E is returned after the date stamped below. #__— pd m‘-‘ lea I III. up?! l :I." A~§I Int} v«.-, A SPEECH ACT THEORY OF METADISCOURSE BY Paul Jude Beauvais A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Teacher Education 1986 Copyright by Paul Jude Beauvais 1986 ABSTRACT A SPEECH ACT THEORY OF METADISCOURSE BY Paul Jude Beauvais Metadiscourse commonly is defined as "discourse about discoursing.” In its brief history, the term has appeared in several models of text structure; however, theorists disagree concerning the range of metadiscursive structures and the role of metadiscourse in a larger theory of text linguistics. This dissertation provides a detailed history and a critical analysis of the existing metadiscourse theories, and it offers an alternative theory that defines metadiscourse as a component of speech act theory. The first chapter surveys the history of metadiscourse from Zellig S. Harris' early use of the term to recent studies by Joseph M. Williams, Avon Crismore, and William J. Vande Kopple. The second chapter introduces four criteria for evaluating the utility of theoretical models. The existing metadiscourse models are analyzed in light of these criteria and are found to contain imprecise definitions of key terms. The models also are found to be collections of disparate structures instead of principled systems. The third chapter provides an overview of important works on speech act theory by J. L. Austin and John R. Searle. Particular attention is devoted to the distinction between illocutionary acts and prOpositions, the differences Paul Jude Beauvais between explicit performative structures and implicit ex- pressions of illocutionary intent, and the types of il- locutionary acts that are possible. In the fourth chapter, metadiscourse is defined as those illocutionary force indicators that identify ex— positive illocutionary acts. A taxonomy of metadiscourse types is provided, and canonical forms using performative or near—performative structures are identified for each type. Partially explicit forms of metadiscourse that do not provide an attributive subject also are identified. The dissertation concludes with suggestions forex- perimental studies using the proposed metadiscourse model. ACKNOWLEDGMENTS Support for this study was provided by the National Endowment for the Humanities. I gratefully acknowledge the assistance of Joseph M. Williams and the participants in the Summer Seminar on Style and the Structure of Discourse, conducted at the University of Chicago in 1985. I also wish to thank Lucille C. Bruch, James A. Drake, and Richard C. Gebhardt. Their support of release time from my teaching duties at Findlay College enabled me to complete this study. Thanks go out to Avon Crismore and Frank Parker for their cogent comments concerning the early drafts of this dissertation. I would like to express special thanks to Lois A. Bader for her support and guidance throughout the course of this project. And words are inadequate to express my gratitude to my parents and to Diane Kendig, without whose support I would have abandoned this project long ago. TABLE OF CONTENTS LIST OF TABLES.‘ ........................................ viii LIST OF FIGURES ........................................ ix INTRODUCTION........... ........... ....... .............. 1 CHAPTER ONE: A COMPREHENSIVE HISTORY OF METADISCOURSE THEORY .................. .. ............... 7 Metadiscourse in Zellig S. Harris' Linguistic Model.... 8 The Reader-Centered Text Studies of Joseph M. Williams......................... ............ 15 Avon Crismore on Metadiscourse in Textbooks ............ 27 Some Exploratory Discourse by William J. Vande Kopple. ......................... ...... 53 Other Works of Interest to Metadiscourse Researchers ............. ..... ............ 60 CHAPTER TWO: A CRITICAL ANALYSIS OF EXISTING METADISCOURSE MODELS ...................... . ............ 64 Criteria for Analyzing Metadiscourse Models ............ 64 An Analysis of Zellig S. Harris' Model ................. 68 An Analysis of Joseph M. Williams' Comments on Metadiscourse. ..... . .......... . ..... ... ............. 70 An Analysis of Avon Crismore's Metadiscourse Studies............ ........ . ......... ...... ............ 76 An Analysis of William J. Vande Kopple's Studies... ....... .. .............. .... .................. 86 The Prospects for an Adequate Theory of Metadiscourse..... ........... . ....... . ........ . ..... 89 CHAPTER THREE: BASIC CONCEPTS IN SPEECH ACT THEORY ............... ... .................... 91 J. L. Austin and the Origins of Speech Act Theory ..... . 91 vi John R. Searle on Speech Act Theory .................... 102 Toward a Speech Act Theory of Metadiscourse ....... .....110 CHAPTER FOUR: A SPEECH ACT THEORY OF METADISCOURSE... ............................... q ........ 112 A Speech Act Definition of Metadiscourse ............... 112 A Taxonomy of Illocutionary A Taxonomy of Illocutionary An Evaluation Primary Expositive Acts ........ ..... ................. I ....... 113 Secondary Expositive Acts ..... ....... .................. .‘ ...... 122 of the Proposed Metadiscourse Model........ ............................ 136 CONCLUSION........ ..................................... 142 BIBLIOGRAPHY....................................;......145 vii Table Table Table Table LIST OF TABLES Informational Metadiscourse in Textbooks ........ 38 Informational Metadiscourse in Nontextbooks ..... 38 Attitudinal Metadiscourse in Textbooks... ....... 39 Attitudinal Metadiscourse in Nontextbooks ....... 39 viii Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure 11: 12: 13: 14: 15: 16: 17: 18: ’LIST OF FIGURES Simple Expositive Performative Clauses. ....... 114 Relational Expositive Performative Clauses.... 115 Partially Explicit Relational Adverbs... ...... 116 Evaluative Expositive Performative Clauses.... 117 Canonical Commissive Expositive Clauses ....... 120 Canonical Relational Commissive Clauses. ...... 120 Canonical Reiterative Expositive ClauSes ...... 121 Third Person Simple Expositive Clauses........ 123 Third Person Evaluative Clauses.. ...... . ..... . 126 Third Person Commissive Clauses...............127 Third Person Relational Commissive Clauses....128 Third Person Reiterative Clauses.... ......... .129 Second Person Simple Expositive Clauses ...... . 131 Imperative Second Person Expositives.. ........ 131 Second Person Relational Expositives... ....... 132 Second Person Evaluative Clauses.............. 133 Second Person Commissive Clauses.. ........... .134 Second Person Reiterative Clauses ...... ....... 136 ix INTRODUCTION If a scholar in some future century were to survey the history of text linguistics, she would be justified in labeling the latter half of the twentieth century ”The Age of Expansion." As computer analysis of texts and other sophisticated research methods have been developed, researchers from a variety of disciplines have migrated to the burgeoning field. The result is that textual studies now are conducted not only by linguists, but also by reading specialists, composition researchers, literary critics, cognitive psychologists, computer scientists, and communications theorists. As interest in text linguistics has expanded in recent years, a problem common to many fledgling disciplines has developed. Because no consensus exists concerning an appropriate framework for studying text structure, theorists have advanced competing models that reflect differing views of the parameters and internal structure of the field. These models are based on the theoretical assumptions of the various fields that contribute to textual studies, and the models often employ terminology borrowed from these fields. Although theorists use many of the same key terms in their textual models, the same term may denote different elements in each model. Even the term "text" is defined 1 . 2 in differing ways by theorists from various disciplines. An inevitable result of disagreements concerning terminology and theoretical assumptions is a "babelogue" of voices, each speaking a language that the majority of others cannot understand. Research into text structure plods forward in several directions, but advances on each front go unannounced in the other chapters of action. If the Age of Expansion is to segue into an Eranf Consolidation, theorists of all persuasions will have to redouble their efforts to develop a unified program of study and a common vocabulary. In this dissertation, I hope to make a modest contribution to the development of a common vocabulary for textual studies. The topic of my study is I'metadiscourse," a term that has appeared in several models of text structure. Variously defined as ”writing about writing” (Williams, Style 226) and "discourse about discourse" (Vande Kopple, "SEDoM" 83), metadiscourse is used in text models to refer to elements that guide a reader through a text without themselves presenting propositional content. However, theorists remain divided concerning the range of constituents that should be grouped under the rubric of metadiscourse and how metadiscourse should function within a larger theory of text structure. Because competing theories of metadiscourse currently are vying for consideration, the future of the term in text linguistics remains to be determined. In the study that follows, I will examine the major 3 competing theories of metadiscourse, and I will argue that none of the existing theories can serve the needs of text theorists. After providing a detailed critique of the flaws in the important metadiscourse theories, I will offer an alternative theory of metadiscourse--one that defines metadiscourse within the larger context of speech act theory. I then will conclude with suggestions for experimental studies using the proposed metadiscourse theory. In the first chapter of this dissertation, I will provide a comprehensive history of metadiscourse theory. My survey of metadiscourse research will begin with Zellig S. Harris' first use of the term in 1959, and it will extend to the recent research on metadiscourse produced by Joseph M. Williams, Avon Crismore, and William J. Vande Kopple. The second chapter of this dissertation will present a critical analysis of existing metadiscourse models. I first will discuss the following criteria for evaluating theoretical models: (1) whether the definitions of key terms in a model are sufficiently precise to enable theorists to agree concerning the specific entities that the terms describe, (2) whether each category in the model possesses properties sufficiently similar to those of other categories in the model so that useful generalizations can be offered concerning the subject under consideration as a collective entity, (3) whether the constituents grouped under a category heading are sufficiently similar to allow generalizations that are 4 true for every constituent, and (4) whether the terms of analysis used in the model are able to reveal important properties of the subject under consideration. I then will analyze the existing models of metadiscourse in light of these criteria, and I will show that all of the models possess flaws that diminish their utility. In the third chapter, I will establish a foundation for my own metadiscourse theory by surveying important works on speech act theory by J. L. Austin and John R. Searle. I will devote particular attention to their discussions of the distinction between illocutionary acts and propositions, the differences between explicit performative structures and implicit expressions of il— locutionary intent, and the categories of illocutionary acts that are possible. I will conclude the chapter by explaining why an analysis of metadiscourse within the context of speech act theory not only is possible but also is advantageous. In the fourth chapter of this dissertation, I will offer an alternative theory of metadiscourse--one that treats metadiscourse as a component of speech act theory. I will begin by defining metadiscourse as those il- locutionary force indicators that identify expositive illocutionary acts, a definition that utilizes Austin and Searle's terminology. I then will introduce a taxonomy of metadiscourse types that includes primary and secondary classes of simple expositive acts, evaluative expositive acts, commissive expositive acts, and reiterative 5 expositive acts, with relational forms possible for each of the sub—categories. I will identify canonical explicit forms that use performative or near-performative clauses for each of the sub—categories, and I also will identify partially explicit forms that do not provide an attributive subject for an expositive illocutionary act. I will con- clude the chapter by showing how my proposed model conforms to all four of the criteria for good design that are pro- vided in Chapter Two. In the conclusion of the dissertation, I will offer recommendations for experimental studies of metadiscourse that utilize the model proposed in Chapter Four. By treating metadiscourse as a clearly delineated category within the larger context of speech act theory, my dissertation will suggest a means of subsuming a fragmented area of textual studies into a framework that is familiar to researchers from most of the disciplines that are involved in studying text structure. These researchers will be able to avail themselves of the work that speech act theorists have been conducting for over thirty years, thus providing a unified program for future metadiscourse studies. However, I believe that the contribution this dissertation can make to future textual studies extends beyond its analysis of one element in a comprehensive model of text structure. The greater significance of this study is that it illustrates the benefits that can result from the precise definition of carefully delineated categories 6 within the context of an already familiar theory. If this dissertation suggests the wisdom of using methods of inquiry that consolidate advances in textual studies while reducing the confusion that permeates the field, then it may serve as a model for future studies of text structure._ CHAPTER ONE A COMPREHENSIVE HISTORY OF METADISCOURSE THEORY The etymology of the word ”metadiscourse” is brief. The word was coined by Zellig S. Harris in 1959 to describe text elements which comment about the main information of a text, but which themselves contain only inessential information. Metadiscourse constitutes a minor category within Harris' linguistic model of information retrieval, and he uses the word only four times in his published work. Harris' ideas diminished in influence as Chomsky's transformational grammar ascended to prominence in the 19605, and the word "metadiscourse" disappeared from use for nearly fifteen years. Metadiscourse resurfaced in 1981, when Joseph M. Williams published his seminal book, Style: Ten Lessons in Clarity and Grace. Although he was unaware of Harris' previous use of the term, Williams' definition of metadiscourse as writing that guides readers without informing them (Style 81) is fairly consistent with Harris' earlier definition. Since Williams resurrected the word, metadiscourse has been studied by two researchers who were influenced by his work. From 1982 to 1985, Avon Crismore published a series of studies that examine the use of metadiscourse 7 8 in textbooks. In these studies, Crismore establishes links between metadiscourse and speech act theory, and she presents a metadiscourse model that is similar in several aspects to the one that I will advance in Chapter Four. And in 1985, William J. Vande Kopple published a detailed examination of metadiscourse in College Composition and Communication. In his essay, Vande Kopple enlarges the body of constituents that are grouped under the rubric of metadiscourse. Because the ideas in the studies mentioned above are central to my own work, I will summarize each of the studies at considerable length in this chapter. I will reserve my critical analysis of the studies for Chapter Two. Metadiscourse in Zellig S. Harris' Linguistic Model As I already have noted, the first use of the word ”metadiscourse" is by Zellig S. Harris in 1959. In an essay concerned with the role that linguistic transformations play in retrieving information from scientific articles, Harris presents a model for dividing sentences into smaller units of information and for categorizing those units according to the types of information they contain. The central element in Harris' model is the ”kernel sentence,” a device that also figures prominently in Chomsky's early work. Harris defines kernel sentences as elementary sentence structures that can be syntactically transformed and still retain their original meaning. His examples are all simple, active, positive, declarative structures: In English . . . kernel sentences have a few simple structures, chiefly N V and N V P N, N V N and N V N P N, N is N, N is A, N is P N. Every sentence, therefore, can be reduced by— transformation to one or more of these; and combinations of transformations of these generate every sentence. (”LTfIR” 461) (In the quotation, N stands for noun, 2 for verb, 2 for preposition, and N for adjective.) According to Harris, every sentence is composed of information kernels, each of which is a transformed version of a kernel sentence. As an example of this principle, he presents the sentence, "The optical rotary power of proteins is very sensitive to the experimental conditions under which it is measured, particularly the wavelength of light which is used” ('LTfIR" 463). In this sentence, ”experimental conditions" is a kernel derived by transforming the kernel sentence, ”Conditions are experimental.“ Further analysis reveals eight additional kernel sentences that provide the basis for the sample sentence ("LTfIR' 463). After dividing a sentence into kernels of information, Harris attempts to identify the constituent structure of the kernels themselves. He suggests that each kernel can be divided into a maximum of five sections, each of which he assigns a section number: 0. Connectors (binary operators) to other kernels (e. g., or, because, however, Lf . . . then); unary operators on the kernel (e. 9., not, perhaps, surprisingly enough). 1. Subject noun phrase: center N and its adjuncts. 10 2. Verb phrase, including its adjunct Q and g N (preposition plus noun phrase). 3. Post—verb ('object') N or N or g N phrase: center and its adjuncts. I 4. Adjuncts of the kernel as a whole (usually E‘N or connected sentence). (”LTfIR” 465) (In the quotation, 2 stands for adverb.) Although he acknowledges the difficulty of using his system to analyze the kernels of some sentences that have undergone transformations, Harris does not Specify which sentences and transformations are likely to cause difficulty. Although Harris asserts that his system is useful for categorizing the types of information an article contains, he also notes the need to supplement his analysis by identifying the functions of particular classes of informative kernels: The various sentences of an article differ in informational status, and even certain sentences which may be of interest to readers of the article may not be requested or useful in retrievals. The distinction is sharper in the case of kernels, because transformations usually separate out what we might consider side remarks, comments about methodology or prior science of the article, and so on, from the kernels that carry the central material of the article. This happens because in many cases the different types of material necessarily occupy different grammatical subsections of the original sentence. ("LTfIR” 464) Harris identifies four types of kernels, each of which conveys a different type of information. In a scientific article, some kernels contain the main material which is most likely to be included in an abstract of the article. Other kernels contain adjunct information which reports 11 conditions and detailed operations that apply to the main kernels. In yet another class are those kernels that comment about the methodology and underlying science of the article ('LTfIR' 464-65). And in a fourth distinct class are metadiscourse kernels: There are also metadiscourse kernels which talk about the main material (e.g., discussing the problems of the investigators). These contain words entirely different from those of the main kernels, except that.they often contain one word from a main kernel or a pronoun referring to a main kernel. ('LTfIR" 464—65) ' Harris then provides one example of a metadiscourse kernel that contains a more informative kernel: If as often happens the containing kernel is of a metadiscourse type (e.g., Ne have found that . . . ) it would be recorded (in full or in summary form) in the 0 section of the kernel which it contained, or else omitted altogether. ("LTfIR" 466) Harris' terminology is somewhat confusing, but the procedure he is describing can be expressed in simple terms. Harris is attempting to isolate the portions of a scientific article that should be included in an abstract. To do this, he divides the sentences in the article into simple kernel sentences. He then assumes that the sections present in the kernel and the patterns of word repetition in the sentences indicate the informative function of each kernel. After analyzing the form and word content of each kernel, Harris assigns it to one of his four categories for classifying information. He then constructs an abstract of the article, drawing most of the information for the abstract from the class of kernels that contain the most 12 important material. Because the metadiscourse kernels do not contain the main material of an article, they are of little importance to Harris, and he would omit them from an abstract. Harris' lack of specificity concerning procedures and the paucity of examples in his essay make it difficult to determine how well his information retrieval system would work. However, his description of metadiscourse is significant, because it introduces several traits that are preserved in later metadiscourse models. In Harris' system, metadiscourse contains information that is distinct from the significant content of a text; although a reader may be interested in the metadiscourse, it really does not express what the article is about. Harris' classification of metadiscourse as secondary information is shared by most contemporary metadiscourse researchers, and it is reflected in the assertion that metadiscourse does not convey propositional content (Vande Kopple, 'SEDoM' 83). It also is significant that Harris categorizes metadiscourse in the 0 class of connectors, because the connecting function of metadiscourse is explored in greater detail by Vande Kopple, whose metadiscourse taxonomy includes a class of ”text connectives" (”SEDoM” 83). Although Harris' system of dividing and categorizing information has been abandoned in recent studies, the influence of his pioneering work remains evident. Harris' only other use of the term "metadiscourse” in his published work appears in a paper on algebraic 13 operations in linguistics. In attempting to develOp a mathematical system to characterize natural language, Harris notes that one subset of a language is its "metalanguage," which includes those sentences that refer to the language itself: A[n] . . . important property of language involving a special subset of its sentences is that the metalanguage of a language is itself a set of sentences and a subset of the whole language. This includes the axiomatization [of word symbols, kernel, strings, and transformation rules], every grammatical description of the language, and all such individual sentences of the metalanguage as'Si'i§_g_sentence g£_English. ("AOiLS' 608) — Harris then observes that metalanguage can be conjoined with other natural language sentences to form sentences that are self—referential: Certain impredicative sentences, the syntactic core of the paradoxes, which name a sentence within itself (This sentence is . . . ) can be separated off from other sentences. An important result is the fact that metadiscourse sentences can be inserted into the discourse about which they speak: if some discourse occurs only in particular circumstances, we can conjoin this statement about the discourse to the discourse itself. (”AOiLS” 608) In the quotations above, Harris draws a distinction between ”metalanguage” and "metadiscourse": metalanguage includes all of the apparatus of a linguistic model, including the grammar rules of a language and those words that refer to linguistic structures; on the other hand, metadisCourse is defined by example as a subset of metalanguage, and it includes only those sentences that can be conjoined to the discourse about which they speak. While metalanguage is 14 an autonomous descriptive system, metadiscourse occurs only in particular circumstances-—it is context specific. In essence, metalanguage is language about a language, while metadiscourse is language about a language used for a particular discursive purpose. In his last use of the term ”metadiscourse," Harris establishes a precedent for other researchers' future definitions of the term. His characterization of metadiscourse as a statement about the discourse conjoined to the discourse itself anticipates the definitions of metadiscourse as “discourse about discoursing',(Williams, Style 81), ”discoursing about the discourse" (Crismore, ”TRoT:M" 280), and ”discourse about discourse" (Vande Kopple, ”SEDoM' 83). In his two studies that use metadiscourse as a linguistic category, Harris lodges the term within complex theoretical frameworks that are accessible only to trained linguists. Because both the information retrieval system and the algebraic operations discussion are not sufficiently developed to suggest immediate applications, even linguists, who were devoting most of their attention to Chomsky's transformational grammar throughout the 19605, did not use Harris' terminology. As a consequence, ”metadiscourse" disappeared from use until Joseph M. Williams resurrected the term in 1979. 15 The Reader-Centered Text Studies of Joseph M. Williams In the two decades following Chomsky's publication of Syntactic Structures in 1957, the grammatical models advanced by generative linguists became increasingly sophisticated. These models captured the attention of composition theorists, who pondered the applications of generative grammar in writing instruction; however, their efforts produced disappointing results. Writing in 1979, Joseph M. Williams observed that ”those attempting to exploit linguistics in teaching writing could have pointed to little that was promising and to less that had been delivered, particularly on any of those occasionally extravagant predictions made in the '50's and '60's' ("N—LLatToS” 24). In exploring the reasons for the failure of linguistic models in writing instruction, Williams resurrected the term "metadiscourse,” and he lodged it within the new context of a reader-centered model of text structure. Williams begins his critique of linguistic models in writing instruction by noting the important role that theory plays in determining our classroom practices: When we teach writing, what we teach about style depends on what we think we can verify and publicly demonstrate. And what we are able to verify and demonstrate usually depends on the categories and processes and relationships in our theory. If our theory includes the category ”word,” and we can agree on how to apply the definition, then we count words. If the system includes the category ”clause,” we count clauses. If it includes ”transformation,” we count transformations. (”N-LLatToS" 26) Because classroom practices depend on the theoretical 16 assumptions that inform them, it is essential that theories be appropriate for the purposes to which they are applied. However, Williams observes that most formal linguistic models are inappropriate for describing the communicative interaction of a writer and reader mediated by a text: Neither [descriptive nor generative grammars] directly addresses how the structure of a I sentence is experienced, either from the point of view of the speaker/writer or reader/listener. Both are, to this degree, text—centered. And yet if style is a component of rhetoric, as it must be in the teaching of composition, and rhetoric is the art of moving audiences, such text-centered theories must in some critical respects, miss the point. ('N—LLatToS' 26) If a text grammar is to be useful in writing instruction, its components must possess what Williams calls "affective salience” ('N-LLatToS” 30); that is, the terms of analysis must reflect the features of a text that are significant to a reader. However, Williams notes that most text models use the sentence as their highest level of analysis, and the categories used for sentence analysis often do not coincide with the features of a discourse that are salient to a reader: Although there has been a good deal of work recently into the functional analysis of sentences and into text grammars, just about all of those investigations begin with terms drawn from a theory of sentences in isolation. Our ignorance about the structure of discourse has prevented us from formulating a theory of sentences whose categories and relationships would interact with, or at least impinge on the categories and relationships constituting that theory of discourse. ('N-LLatToS" 27) If a model of discourse is to reflect a reader's experience of a text, the model cannot be built bottom-up from words 17 to sentences to larger units of discourse; as Williams observes, to build a model in this way is to violate a basic principle of theory construction: One principle in constructing theories of systems composed of hierarchically interlocking sub-systems is that boundary conditions are set from without rather than from within, and that to some degree parts take their character from the whole they constitute. In the case of sentences, they are constituent parts of and therefore shaped by a larger whole, the discourse they occur in. ('N-LLatToS' 27) Because sentences are shaped by the discourse in which they occur, an adequate sentence model must be constructed top-down, with the categories for sentence analysis reflecting a larger theory of discourse. However, most sentence models are not constructed in this manner. In the remainder of his essay, Williams offers a partial model of sentence structure as perceived by a reader. Noting that "a reader who senses a specific and clearcut intention reads and integrates all the details of the text to that intention” ("N-LLatToS' 31), Williams suggests that the rhetorical structure of a sentence can be divided into two categories: topic and comment. The topic of a sentence reveals what the sentence is about, while the comment contains information that modifies the reader's understanding of the topic. At another level of analysis, a sentence possesses a grammatical structure that includes the categories subject, verb, and object, and at yet another level, the semantic structure of the sentence is described by the categories agent, action, and patient, which reveal the functions that are served by 18 the grammatical structures. In a typical or "canonical” sentence, the categories at each level of analysis will occur in the sequences given above; that is, the topic will be followed by the comment, the subject by the verb and object, and the agent by the action and patient ("N—LLatToS" 32-34). However, the three levels of analysis are autonomous, in that the topic of a sentence is not always the sentence's grammatical subject, and the functional category ”agent" also need not fill the grammatical subject slot. As an example of topic/subject autonomy, Williams provides ”About style, we can say little with certainty” ('N-LLatToS' 38), in which "About style“ is the topic, while ”we" is the grammatical subject. Williams does not provide an unambiguous example of agent/subject autonomy, but examples are easy to find; consider the sentence, "Mary was struck by a train,” in which "the train” is an agent in indirect object position, while 'Mary' is the grammatical subject. In an afterword to his essay, Williams makes some adjustments to his model, changing the category "patient" to "goal,” and adding the categories ”theme” and ”rheme' as a variable level of analysis that corresponds to a fixed "topic/comment” level, thus paralleling the- distinction between a variable "agent/action/goal" level and a fixed "subject/verb/object" level. However, what is more important than Williams' particular categories is that the grammatical structures in his model constitute neither the sole nor the primary level of analysis. 19 As Williams himself observes, What is most important in all of this, I think, is that the grammar of the sentence need not be the first level of analysis. When we begin with grammar, with subjects, verbs, and objects, those categories become the presiding terms of our analysis and all other terms and levels of analyses flow from them. But if we begin where the experience of the sentence is, with the rhetorical and semantic structure and then integrate a necessary grammatical description with those levels, we have the beginnings of a theory which would allow us to deal with the experience of a discourse rather than the solipsistic terms of a theory of sentences. ('N—LLatToS" 35) Although one could argue that a multi-level analysis of sentence structure is itself a bottom-up approach to text analysis, since the sentence remains the largest unit of study, Williams at least has shifted the terms of analysis from the vocabulary of grammar to a new vocabulary, one that, in his own words, 'impinge[s] on [a] theory of discourse" ('N-LLatToS' 27). By asserting that the sentence has a rhetorical structure, Williams moves the study of sentences into the domain of rhetoric--a field that is concerned with authorial intention and reader reaction. This shift in context broadens the possibilities for studies of sentence structure. Within this expanded context for sentence studies, Williams resurrects the term ”metadiscourse," suggesting that it is a necessary category in an adequate theory of sentence structure: We shall certainly need a category in the analysis that we might call metadiscourse: all the elements in a sentence that refer to the process of discoursing, as opposed to the specific reference of the discourse. 20 I have in mind what goes on in a sentence such as I believe that in regard to the American pharmaceutical industry, we can say that there seems to be excessive federal government regulation. On its most ordinary interpretation, the sentence is not about me as a believer or about us as sayers, but about the American pharmaceutical industry. That is the topic of the sentence, that for which the rest of the sentence exists. The comment, the state— ment about the topic, is excessive regulation by the federal government. Everything else: I believe that in regard to and we can say that there seems to be is metadiscourse, discourse about discourse, elements referring not to the referents external to the discourse but to the act of discoursing, to how we should take the truth value or probability of the proposition about those external referents. ("N-LLatToS' 33) Williams' use of "metadiscourse" retains much of the sense of Harris' original definition; like Harris, Williams uses the term to refer to sentence elements that comment about the primary discourse in the sentence without presenting primary information of their own. However, Williams' definition specifies a role for metadiscourse distinct from the role that Harris recommends. While Harris' category includes sentence elements that are largely informative in their own right (e.g., comments about a researcher's problems in a scientific article), Williams limits the term to sentence elements that direct a reader's perception of a text. In Williams' model, metadiscourse acts as the writer's voice, whispering instructions to the reader concerning the writer's own judgement of the ideas he is advancing. For Williams, metadiscourse truly is a rhetorical device that allows a writer to mediate a reader's 21 experience of a text. In that regard, he succeeds in offering a text model that is writer/reader-centered instead of text—centered. I In his suggestion that metadiscourse indicates a writer's judgement concerning the truth value or probability of propositions, Williams anticipates two sub-categories of metadiscourse that will appear in his later work: hedges and emphatics. And his definition introduces a distinction that is retained in several later studies of metadiscourse: the categorical separation of metadiscourse and propositional content. What Williams' discussion does not include are an explanation of exactly how metadiscourse affects a reader and a listing of all the types of metadiscourse. Williams begins to address both of these problems in an essay titled ”Literary Style: The Personal Voice." Noting that "every text, regardless of its ideational content, is produced by one of the personae in [a] speech event, the speaker or writer" ("LS:TPV" 195), Williams observes that ”sometimes, the author speaks in the first person and refers directly to the discourse as he constructs it, sometimes including even the audience as a specifically mentioned ygg" ("LS:TPV" 195). These references to the participants in a discourse and to the discourse itself are metadiscourse, and Williams offers a familiar definition: But [the] primary discourse is embedded in metadiscourse, discourse about discourse, words and phrases and clauses-~even sen- tences--that refer not to the subject "out there" but to the act of discoursing, to the speech event that the discourse and its reader create. (”LS:TPV" 195) 22 Williams then suggests that writers differ in the types and amount of metadiscourse they use, and he provides a brief list of metadiscourse forms: Many writers stay out of their text almost entirely, relying on shorter discourse signals such as those that indicate cause: therefore, consequentl , as a result, s9, for, and so on; or contrast: on the other hand, instead, though, but, however, and so on; or the continuation of the discourse: indeed, moreover, in fact, in any event, at least, and, also, too, neither, second, then, next. An even less obvious presence is felt in words that comment on the probability of the proposition expressed in a sentence: probably, possibly, undoubtedly, certainly, surely, seemingly, apparently, clearly; it would seem that, 15 appears that, it is obvious that, it could be that. Other words express an attitude toward an event: It is fortunate that, inevitably, it is odd that, interestingly, etc. (”LS:TPV” 197) Concerning the appropriate use of metadiscourse, Williams observes that essays in the Federalist Papers begin almost one out of every three paragraphs with metadiscourse that refers to the argument itself or to the writer or reader's understanding of the argument, and he notes that ”this is the style of rational men arguing, debating, discussing with other rational men the great questions of a new form of government" (”LS:TPV” 197); however, Williams also notes that ”excessive metadiscourse can become mechanical and obtrusive, drawing too much attention to the process of discoursing" ("LS:TPV' 197). Three aspects of Williams' expanded discussion of metadiscourse are notable. First, Williams' list of metadiscourse types includes not only terms that refer to the truth value or probability of primary discourse; it also 23 includes causal connectors, obversative connectors, and sequencers. This increase in the entities categorized as metadiscourse anticipates the wholesale expansion of metadiscourse categories that is recommended in William J. Vande Kopple's 1985 study. Second, Williams notes that too much metadiscourse can make a text difficult to read. In observing that metadiscourse can have either a positive or a negative impact on a reader's ability to comprehend a text, Williams establishes readability as a major consideration in metadiscourse use, thus anticipating the readability studies of Avon Crismore. And finally, in suggesting that metadiscourse calls attention to the personae participating in a "speech event,” Williams establishes a precedent for considering metadiscourse within the context of speech act theory. Williams' most thorough discussion of metadiscourse appears in his seminal book, Style: Ten Lessons in Clarity and Grace. Again operating within the context of a text model designed for use in writing classrooms, Williams offers a detailed analysis of metadiscourse in a chapter titled ”The Grammar of Concision.” In the glossary of his book, Williams provides the following definition: Metadiscourse: Writing about writing, whatever does not refer to the subject matter being addressed. This includes all connecting devices such as therefore, however, for example, in the first place; all comment about the author's attitude: I believe, in my opinion, let me also pgint out; all comment about the writer's confidence in his following assertion: most people believe, it is widely assumed, allegedly; references to the audience: as you can see, you will find that, consider now the problem of . . . . (Style 226) 24 Noting that ”we need a term to distinguish writing that guides the reader from writing that informs the reader” (Style 81), Williams again asserts that metadiscourse is ”discourse about discoursing“ (Style 81). In the pages that follow, he examines in detail six categories of metadiscourse, dividing them into three groups, each of which contains two members. The first group includes hedges and emphatics, both of which express the certainty with which a writer presents material. Hedges are qualifying terms like "possibly," "apparently," "might,” and "seem," while emphatics include terms like ”it is obvious that,” "of course,” “invariably,” and "essential.” Concerning the functions of these terms, Williams makes the following observations: Hedges, or qualifications, let us sound small notes of civilized diffidence. They leave us room to backpedal and to make exceptions. An appropriate emphatic, on the other hand, lets us underscore what we really believe-—or would like our reader to think we believe. (Style 83) However, when used to excess, hedges ”become less hedges than meaningless modifiers,” while emphatics ”sound arrogant or at least defensive” (Style 83). Williams' second group of metadiscourse categories includes sequencers and topicalizers, both of which lead a reader through a text (§£ylg 83). Williams does not define sequencers, nor does he provide a list of examples in the chapter, but his discussion and the examples in his glossary definition allow us to infer that sequencers are terms that explicitly indicate relationships among passages of a text. 25 This class includes causal connecting words like ”therefore," obversative connectors like "however,” illustration markers like ”for example," temporal sequencers like "next" and ”after,” and numerical sequencers like ”in the first place," "second,” and "my third point is.” Topicalizers are terms that "focus attention on a particular phrase as the main topic of a sentence, paragraph, or whole section" (Sgylg 84), and Williams cites as examples such phrases as "in regard to,” “in the'matter of," and ”turning now to" (§£ylg_84). Concerning the appropriate use of these terms, he notes that overelaborate introductions are the least useful kind of sequencers, and he provides the following example: In this next section of this report, it is my intention to deal with the problem of noise pollution. The first thing I want to say about this subject is this: Noise pollution is . . . (§£ylg 83) Williams then notes that a single topicalizer often can substitute for a long string of sequencers and topicalizers. For example, in place of the above introduction, ”The next problem is noise pollution" would suffice (Style 83). Williams concludes that "there is/are" is the most common topicalizer, but he suggests that this phrase be used only when the topic it introduces is sufficiently important to be developed in several sentences (Style 84). Williams' final metadiscourse couple contains narrators and attributors, both of which tell a reader the sources of ideas, facts, or opinions. His examples of narrators include "I was concerned,“ "I have concluded," and I'I think” (Style 85), all of which have first person subjects. Unlike 26 narrators, attributors do not use first person subjects; in fact, Williams' examples do not use explicit subjects at all, because they all are agentless passives--his examples include ”high divorce rates have been observed to occur in parts of the Northeast that have been determined to have especially low population densities” (Style 86). Are we to assume that the corresponding agentive forms (”X has observed" and ”X has determined”) also are types of attributors? If so, then we can infer that attributors use either implicit or explicit third person subjects while narrators use first person subjects, but the two forms are similar in other regards. Concerning the importance of metadiscourse in written communication, Williams observes that virtually every text needs some metadiscourse: We need some metadiscourse in just about everything we write. Without it, we can't announce that we're changing the subject or coming to a conclusion, that what we're asserting is or is not certain, that our ideas are important. We can't define terms or acknowledge a difficult line of thought, or even note the existence of a reader. We can't outline what we intend to say. (Style 81) However, Williams does note that ”some writers use so much metadiscourse that they bury their ideas” (Style 81), thus suggesting again that metadiscourse can have either a positive or a negative effect on the readability of a text. Williams' discussion of metadiscourse in Style is the most influential treatment of the topic to date; his book is cited in most metadiscourse studies that have been 27 published since 1981. Among those that acknowledge their indebtedness to Williams is Avon Crismore, who herself has conducted several important studies of metadiscourse. Avon Crismore on Metadiscourse in Textbooks Following the publication of Style in 1981, the term ”metadiscourse” appeared in a series of studies by Avon Crismore on the characteristics of readable textbooks. Crismore's interest in textbook readability is first noted in a paper on improving the readability of a poorly written sixth grade social studies book; in that paper, she notes that explicit meaning cues such as headings, logical I connectors, and transitions can enhance the readability of a textbook (SfIaPWSGSSC 12). Although the paper does not use the term ”metadiscourse,” it does introduce several of the concerns that characterize Crismore's later work, including her interest in identifying the structural characteristics of an easily readable textbook. Crismore's first detailed study of metadiscourse is an ERIC document titled The Metadiscourse Component: Understanding Writing about Reading Directives. After citing Williams' comments on metadiscourse in Style, Crismore notes that it is important for readers to understand the functions that metadiscourse serves: With metadiscourse awareness and strategies for using it, readers will better understand the author's text plan. They will know whether they are reading the introduction, the body or conclusion of a text. And they'll know when the author has shifted to a different topic or defined a term. Readers will 28 understand that an author believes what he is asserting is reliable, that he is conceding this point and pointing out this point or that he considers certain ideas more important than others. (TMC 4) Having suggested the range of functions that metadiscourse can serve, Crismore then asserts that the general purpose of metadiscourse is to serve as a system of textual relevance cues for a reader. For her discussion of textual relevance cues, Crismore borrows from the work of several theorists. First she cites Teun A. van Dijk's 1979 essay on relevance, and she defines textual relevance as "the result of an operation by which a reader assigns some degree of importance to some property of the discourse" (TMC 4). She then suggests that two of van Dijk's categories, lexical relevance cues and pragmatic relevance cues, are forms of metadiscourse, and she provides a list of sub-categories and examples. Under lexical cues she lists direct relevance expressions such as ”important” and ”critical,” theme indicators such as ”the subject . . . is," summarizers such as ”in other words," concluders such as ”the conclusion," connectives such as ”so” and "thus," and superstructure signals such as "our premises are" and "the conclusion is." Under pragmatic cues, she lists global illocutionary force indicating devices such as "I (hereby) warn you” (2N9 6). Continuing her listing of metadiscourse types, Crismore next mentions four categories from Williams' §E1l23 hedges, emphatics, sequencers, and topicalizers (TMC 7); however, she does not indicate whether she considers these categories 29 to be forms of textual relevance cues, nor does she attempt to integrate them into the taxonomy that she already has presented. Instead, she introduces yet another classifying scheme, this onerorrowed from Randolph Quirk and Sidney Greenbaum's A Concise Grammar of Contemporary English. Noting that ”linguists would call most metadiscourse adverbials" (TMC 8), Crismore considers three classes of adverbials: adjuncts, disjuncts, and conjuncts. Defining adjuncts as adverbials that ”are integrated within the structure of the clause to at least some extent” (TMC 8), Crismore divides them into two sub-categories: limiters and additives. Limiters include exclusives such as ”alone” and ”only” and particularizers such as ”chiefly” and "mainly,” while additives include such terms as ”also" and ”too” (ENE 8). Unlike adjuncts, disjuncts and conjuncts are not integrated within clauses; instead, they occur before or between them. Crismore notes that "the function of disjuncts is to express how the writer evaluates the form of the content or what is said in the content" (ENE 11). and she divides them into two classes: style disjuncts such as ”seriously" and ”very frankly" and attitudinal disjuncts such as "probably" and ”certainly" (TNQ 9-10). Conjuncts serve a different function; Crismore notes that ”they indicate the connection between what is being said and what was said before" (TNQ 11), and she lists numerous sub-categories: enumeratives such as ”first" and ”secondly," reinforcers such as "also” and "furthermore," equatives such as ”likewise” and "similarly," transitionals such as 30 ”by the way” and ”incidentally," summatives such as "then" and ”in conclusion," appositions such as “in other words” and "for example,” result markers such as ”consequently” and "therefore,” inferentials such as “otherwise“ and "in that case,” reformulators such as "better" and ”rather,” replacives such as ”alternatively" and ”on the other hand,” antithetics such as ”instead" and ”on the contrary," concessives such as ”yet" and "in spite of that,“ and temporal transitions such as ”meanwhile” and "in the meantime” (INN 12—14). Following her lengthy listing of metadiscourse types, Crismore expands upon the importance of metadiscourse processing skills for a reader: Processing metadiscourse is as important as processing the primary topic discourse. Learning to process and understand meta- discourse is like learning to process and understand directions. And just as many readers find reading and following directions a problem, so also many readers find the same problem with metadiscourse. More attention must be given to giving students metacognitive awareness of metadiscourse and strategies for its use so that they may understand how to take the author, maintain schemas by connecting sentences . . . recognize the author's attitudes and whether he is being subjective or objective, and recognize the relevance signals. (ENE 14-15) She then offers thirteen suggestions for helping students to master metadiscourse. These include surveying texts used by the students and pointing out metadiscourse in them, giving students sentences, essays, and text chapters with the metadiscourse deleted and having them insert it, giving writing assignments that require metadiscourse of different types, having students bring in examples of 31 metadiscourse from their out—of—school reading, having students practice using metadiscourse in_both oral and written discourse, and holding students accountable for appropriate metadiscourse use in reading and writing tests (TNQ 16—17). In an appendix to her esSay, Crismore provides numerous exercises to develop students' knowledge of metadiscourse. These include cloze tests in which students must insert appropriate metadiscourse, multiple choice questions in which students must identify forms of metadiscourse in sample sentences, and paragraphs for which students must circle and label all the metadiscourse forms. Crismore's first treatment of metadiscourse is significant for several reasons. In suggesting that study of metadiscourse is an appropriate activity in reading instruction, Crismore expands upon Williams' concern with reader-centered text structure. While Williams addresses the issue of readability by providing suggestions concerning metadiscourse use for writers, Crismore acknowledges that it also is possible to train readers to recognize metadiscourse cues as useful devices that reveal text structure. This attention to instructing readers constitutes a significant advance in the practical application of metadiscourse theory. In the realm of theory, the most significant aspect of Crismore's study is her expansion of the categories grouped under the rubric of metadiscourse. By extending the term ”metadiscourse" to encompass the work of theorists who themselves do not use the term, Crismore produces a model that contains nine categories and twenty-three sub-categories. 32 Whether this increase in categories constitutes progress is questionable; in any event, some of the categories are superfluous, because Crismore does not eliminate the overlap among the categories in the three theories that she discusses.‘ In a 1983 publication titled Metadiscourse: What It Is and How It Is Used in School and Non-School Social Science Texts, Crismore presents a simplified metadiscourse model. Beginning by defining metadiscourse as ”the author's intrusion into the discourse, either explicitly orI non-explicitly, to direct the reader rather than inform" (M:WII 2), Crismore then expands upon her definition, noting that ”metadiscourse is the directives given to readers so they will understand what is said and meant in the primary discourse and how to 'take' the author” (M:WII 2). In the remainder of the paper, she offers a typology of metadiscourse and examines the types and amounts of metadiscourse used in eighteen social science texts. After advancing the premise that ”not only primary discourse but also metadiscourse is used for both referential and expressive ends" (M:WII 11), Crismore presents a typology that contains two general categories of metadiscourse: informational and attitudinal. She then elaborates on the types of informational metadiscourse: An author can give several types of information about the primary discourse to readers for better comprehension. He can explicitly or implicitly signal his goal or goals for the primary discourse; the topic or subject matter; the topic shifts; his main assertion about the topic (the thesis or controlling idea); the 33 significance or rationale; and the sequence, organization, discourse type and development methods he plans to use. (M:WII 12) Crismore notes also that connective signals are a form of informational metadiscourse, but she excludes them from her study because so much recent research already has dealt with them (M:WII 12): After observing that metadiscourse can be in the form of either cataphoric preliminary statements or anaphoric review statements (M:WII 12), Crismore presents four sub-categories of informational metadiscourse: goals, which are global preliminary or review statements such as ”The purpose of this unit is to enrich the way readers think about American Indians” and "We have in this book attempted to say something about American politics at the beginning of the 1970's," pre—plans, which are global preliminary statements concerning content and structure such as "This chapter is about Indians"-and/fWe can trace the development and change in that pattern of lifestyle,” post-plans, which are global review statements concerning content and structure such as "We have looked so far in this chapter at the history of one Indian tribe, the Mohawks' and ”We have argued earlier,” and topicalizers, which are indicators of local shifts in topic such as "Let us now turn to" and "So far as strategic planning was concerned" (M:WII 12—13). Crismore then shifts her attention to attitudinal metadiscourse. Concerning the various types of attitudinal metadiscourse, Crismore provides the following observations: 34 An author can also eXplicitly or implicitly signal his attitude toward the content or structure of the . . . discourse and toward the reader. He can give directives to readers about the importance or salience of certain points or parts of his primary discourse from his perspective, about the degree of certainty he has for his assertions and beliefs, about how he feels about the content of the meSsage, and about the distance he wishes to put between himself and the reader. The author commentary here is evaluative and expressive rather than referential and informational. (M:WII 13-14) She then identifies four sub-categories of attitudinal metadiscourse: saliency markers to indicate the importance of ideas, such as ”still more important" and ”equally important,” emphatics to indicate the author's certainty concerning an assertion, such as ”of course“ and "in fact," hedges to indicate an author's uncertainty, such as ”perhaps and "probably," and evaluatives to indicate an author's attitude toward a fact or idea, such as ”unfortunately,” “luckily," and "I think it is interesting” (M:WII 14). Crismore concludes her taxonomy by noting that metadiscourse can consist of words, phrases, or clauses, and it can be expressed from different points of view as reflected by the first person ”I think,” the second person ”remember that," and the third person ”this chapter is about" (NiNll 15). She suggests that ”the larger metadiscourse phrases and sentences indicate more explicit author intrusion into the primary discourse while shorter metadiscourse words such as.luckily or clearly indicate a more subtle intrusion on the author's part" (M:WII 15). Crismore views metadiscourse as a stylistic variable, and she notes that ”the amount and kind of metadiscourse and 35 person used for it in a text can be viewed as an index of author intrusion, author personality, and the author/reader relationship” (M:WII 15). For her examination of the types and amount of metadiscourse used in social science texts, Crismore selected eighteen sample texts, nine of which were intended for classroom use and nine of which were not intended for use in classrooms. The classroom texts were designed for use in grades 4, 5, 6, 7 8 8, 9—12, and in undergraduate college classes, and their topics included a variety of social science issues. The other texts included articles written for widely—read periodicals and specialized academic journals as well as chapters from books for both general and specialized audiences, and they also addressed topics in the social sciences. For her study, Crismore poses the following questions: (a) Are there differences in the amount and types of metadiscourse used by social science writers in materials used for school and non—school purposes? (b) Are there differences in the amount and types of metadiscourse used in social science textbooks across grade levels? (c) Are there differences among publishers of social science textbooks on the same grade level? Or for the same publisher on different grade levels? (d) Are there differences in the amount and types of metadiscourse used by non—textbook social science writers who write for different audiences? (M:WII 18) In calculating the amounts of metadiscourse used in the various texts, Crismore counts the units of metadiscourse in entire articles or chapters; however, she compensates for the differences in text length by reporting her results as frequency of metadiscourse instances per 1,000 words. 36 Concerning the problem of identifying what constitutes an instance of metadiscourse, Crismore notes that the diversity of metadiscourse types and "the vagueness of the boundaries between metadiscourse and other functions of language [make] it . . . difficult to find a set of empirical linguistic indicators for metadiscourse" (NiNll 22). She then proposes a list of indicators that includes modal verbs such as ”may" and "might," verbs "that name acts of speech or speech events" (M:WII 22) such as ”say,” "assert," and "argue,” verbs that refer to internal states such as ”think,” "know,” and "realize,” adverbial disjuncts such as ”clearly” and "fortunately,” constructions such as "It is interesting that" and ”It is true that,” words such as ”chapter," "section,” ”topic,” and ”maybe,” phrases such as ”in my opinion" and "it seems to me,” tense markers such as ”will,” "have,” and ”ed,” imperative sentences, and pronouns such as ”I” and "you” (M:WII 22-23). Utilizing these somewhat arbitrary guidelines, Crismore proceeds to count the instances of metadiscourse in the sample texts. Crismore's results are repotted in the Tables that follow. In brief, she found that textbooks and non—texts both showed some use of all four types of informational metadiscourse, but there were qualitative differences in the types of metadiscourse used. She notes that typical textbooks used more third person expressions and tended to concentrate on subject matter for pre— and post—plans, while non-textbooks tended to use more first person metadiscourse and to use less metadiscourse that identified 37 subject matter (M:WII 32). A greater difference between textbooks and non-textbooks was present in the use of attitudinal metadiscourse, with non—textbooks using about twice as much attitudinal metadiscourse as textbooks did (M:WII 36). In response to her second research question, Crismore notes that the amounts and types of metadiscourse used across grade levels did vary, with early elementary grade texts using no informational metadiscourse at all. Informational metadiscourse was present in later elementary and secondary texts, but not in the college texts—-a perplexing fact that Crismore tries to explain by suggesting that perhaps the cOllege textbooks she chose were not typical (M:WII 33). For attitudinal metadiscourse, she reports that use of all four sub—categories tended to increase with grade level; however, textbooks used less evaluative metadiscourse than any of the other forms of informational or attitudinal metadiscourse (M:WII 45). Concerning differences in the use of metadiscourse by the same publisher for textbooks written for different grade levels, Crismore reports that Harcourt Brace Jovanovich tended to use more pre- and post-plans in social science textbooks for grades 7—8 than for the grades above andI below (M:WII 34-35). And in response to her fourth question, she notes that non—school texts written for specialized audiences used more informational and attitudinal metadiscourse than was present in non-school texts written for general audiences, with the most notable difference present in the category of saliency markers (M:WII 50). 38 Table 1: Informational Metadiscourse in Textbooks Metadisc0urse Type V (with freouency per 1000 words) School Grade Total6 ' Levels Level Words Goal Preplan :Postplan Tooicalizer Early Elementary h 3.000 .00 .CO .00 . .00 Intermediate . Elementary 5-6 22 000 .CO .55 .23 .00 Junior High 7-8 11,000 .00 .55 .73 .00 High School 9-12 23,000 .00 .35 .26 .Oh College Under- grad. 23.000 .00 .00 .00 .00 Source: Crismore, M:WII 74 approximate Table 2: Informational Metadiscourse in Nontextbooks Metadiscourse Type (WIth frequency per 1000 words) Audience Totala Type Words Goal Preplan Postplan Topicalizer General 55.000 .04 .C9 .00 .00 Specialized 33.000 .18 3.09 .79 .09 Source: Crismore, M:WII 77 a approximate 39 Table 3: Attitudinal Metadiscourse in Textbooks Metadiscourse Type (with frequency per 1000 words) School Grade Total3 Level Level Words Salience Emphatic Hedge Evaluative Early Elementary 3-h 13,000 .00 .00 ' .15 .00 Intermediate Elementary 5-6 22,000 -.00 .05 .00 .00 Junior High 7-8 33,000 .18 .55 . .2h .06 High School 9-12+ ho,ooo .08 .50 .15 .18 College Under- grad. 23,000 .13 .61 .35 .08 Source: Crismore, M:WII 81 , approximate Table 4: Attitudinal Metadiscourse in Nontextbooks Metadiscourse Type (with frequency per 1000 words} Total Audience Type words Salience Emphatic Hedge Evaluative General 55,000 .22 1.73 1.11 .38 Specialized 33,000 .98 2.00 1.30 .hS Source: Crismore, M:WII 85 approximate 40 Because my own concerns are limited to providing a useful definition of metadiscourse and a systematic treat— ment of its constituent structure, Crismore's statistical analysis of texts is of no direct importance to my own research. However, her results do lead to two questions that other researchers may wish to consider. The first question is whether quantifiable differences in amounts and types of metadiscourse result in qualitative differences in the readability of texts. The fact that measurable differences in metadiscourse exist does not mean that readers necessarily will perceive the differences; to paraphrase Williams' term, the measurable differences may not be ”affectively salient." Crismore herself addresses this question in her 1985 doctoral dissertation, and she reports that differences in metadiscourse had little impact on students' ability to read and remember information from texts ("So'MaRA'" 33); however, more research is needed in this area. A concomitant question is whether the ability to recognize metadiscourse is a developmental phenomenon-- whether children acquire the ability to comprehend metadiscourse at a particular stage of neurological or mental development. An answer to this question would be useful to authors of textbooks, in that it would provide guidance concerning the wisdom of using metadiscourse in textbooks designed for various grade levels. However, any study of metadiscourse use is dependent upon an adequate definition of metadiscourse and a useful typology of metadiscourse forms, so these aspects of 41 Crismore's work merit particular attention. Although I am reserving my detailed critical analysis for Chapter Two, several significant features of Crismore's revised model deserve passing comment. The first of these features is the simplified listing of metadiscourse categories in Crismore's typology. Gone are the nine categories and twenty-three sub-categories of her earlier model in 1N3 Metadiscourse Component; in their place are two categories, each of which contains four sub-categories. This revision is necessary in order to render metadiscourse manageable for actual analysis of texts. However, problems remain in identifying the array of linguistic structures that are surface representations of each category, and Crismore's list of empirical indicators provides limited assistance in distinguishing metadiscourse from primary discourse. In presenting the list, Crismore resorts to definition by example, and the inclusion of such indicators as pronouns and tense markers is more confusing than helpful. In order to make her limited list of categories useful for text analysis, Crismore needs to offer precise rules for identifying the constituents in each category, but she fails to do this. Consequently, other researchers who attempt to count the instances of metadiscourse in her sample texts may arrive at significantly different figures for each metadiscourse form. This problem reduces the validity of Crismore's results. Crismore's next study of metadiscourse is reported in an article titled "A Message to Authors about Metadiscourse 42 Use in Instructional Texts.” Defining metadiscourse as "the rhetorical act of discoursing about the spoken or written discourse” (”AMtA” 66), Crismore notes that metadiscourse is "a level of discourse that adds nothing to the propositional content" ("AMtA" 66). She then repeats the distinction between informational and attitudinal metadiscourse that she first introduced in Metadiscourse: What It Is, but she provides slightly different typologies of the two forms. After Observing that "informational metadiscourse consists of explicit rhetorical devices found in initial or final summaries on either a global or local level or inserted in the primary discourse" (”AMtA" 66), Crismore lists the following types of informational metadiscourse: (1) a superordinate or macroproposition made up of a single discourse topic and its predication (the thesis, controlling idea, topic sentence, or main idea sentence), (2) a justification statement for the macroproposition, (3) purpose and goal statements, and (4) state- ments naming the discourse type or act and the organizing strategies. ('AMtA” 66) She then notes that "the function of such devices is to make the author's message and structure plans explicit” (”AMtA” 66). Concerning attitudinal metadiscourse, Crismore notes that this form ”consists of rhetorical devices that signal the presence of an author in the discourse and the author's attitude toward the truth conditions of the content or attitudes toward the content itself and toward the reader” ("AMtA” 66), and she then elaborates on the forms in this category: 43 These [attitudinal metadiscourse] devices are the use of pronouns of self—reference or audience reference, use of mental state or process verbs, and sentential adverbial constructions for conscious self-projection. Metadiscourse operators can be on the word, phrase, or clause level. ('AMtA' 66) Crismore observes that texts using metadiscourse "have a warm, conversational quality but at the same time are not overly informal" ("AMtA” 68), while texts that do not use metadiscourse "are written from the third person point of view and contain no self—reference, emphatics, hedges, or words expressing opinions" ("AMtA" 68). Noting that textbooks tend to use less metadiscourse than non-school books do, Crismore provides a list of considerations for authors to ponder as they write instructional texts. These considerations include whether authors should use a point of view that is subjective or objective, whether they should use first person pronouns to project themselves into a text, and whether authors should include explicit perspective statements ("AMtA' 68). In order to assess the effect of metadiscourse on students' attitudes towards texts and their retention of information, Crismore conducted an experimental study. Using 120 sixth grade social studies students as her subjects, she divided the students into two groups based on their scores on the Social Studies Comfort Index measure, which rated the students' attitudes toward social science readings. Those students whose scores suggested that they were interested in social studies were placed in a ”high comfort” group, while those whose scores suggested little 44 interest in social studies were placed in a "low comfort” group. The groups then were stratified using standardized reading comprehension scores, and students within each stratified group were assigned randomly to eight condition groups for the experiment ("AMtA" 70-73). For reading material, Crismore chose three passages about the Middle Ages from a Ginn social studies textbook. She then developed Seven variations for each of the original texts, with each variation using different metadiscourse forms. Her variables included the presence or absence of metadiscourse that indicated information, attitude, and voice. Students in each of the eight condition groups were given versions of the three passages that represented one of the eight possible combinations of metadiscourse forms (including the original versions that contained no metadiscourse). To assess the effect of varying the forms of metadiscourse, Crismore administered a variety of tests to measure students' comprehension of information, awareness of authorial attitude, the students' own attitudes toward the author and text, and their interest in the subject matter (”AMtA" 70—73). Using analyses of variance to determine the effect of variant forms on several measures of comprehension, Crismore found several significant effects of altering metadiscourse forms. Although her statistical report on the experiment is incomplete, she suggests that adding informational forms of metadiscourse improves comprehension, especially for low- vocabulary, low-ability students ("AMtA" 73). She also 45 notes that adding attitudinal metadiscourse helps students understand and remember the author's attitudes, feelings, and opinions. The high-comfort group recorded higher comprehension scores when informational metadiscourse was presented alone instead of in combination with voice forms of metadiscourse; however, the opposite was true for the low-comfort group, whose comprehension scores decreased when informational metadiscourse was presented alone (”AMtA" 73). Crismore interprets these results by suggesting that "adding voice (I, you, etc.) to informational metadiscourse apparently hurts the high—comfort, relaxed children, perhaps by distracting them from attending to the message; on the other hand, it definitely helps the low—comfort, nervous children,“ perhaps because they ”perceive a supportive guiding teacher—like author” ('AMtA” 74). Crismore does not provide any detailed information concerning how the variant forms affected the students' attitudes towards the author and text, nor does she say how the forms affected the students' interest in the subject matter. Concerning the design of her experiment, Crismore notes that "as the passage materials were being prepared, it became evident that metadiscourse requires a well-written primary discourse in order to be effective" (”AMtA' 74). She observes that numerous problems were present in the Ginn text from which she chose her passages, and she concludes that ”just adding metadiscourse to a poorly written text is not enough to enhance student performance and attitudes” ("AMtA” 74), thus casting doubt on the 46 reliability of her own findings. Among the problems that Crismore mentions in adding metadiscourse to a text are determining ”how much metadiscourse to put in—-the optimal level--and where in the sentence to put it” ('AMtA" 74). Despite the obvious shortcomings in Crismoreis report--the most notable of which is her incomplete statement of her findings-—and despite the fact that the results she does report are inconclusive, her study is important in that it introduces questions that researchers should consider in future studies. These questions include whether any general dicta concerning metadiscourse use can be determined or whether all instructions must be limited to specific, narrow audiences, and whether the placement of metadiscourse within a sentence is a significant factor in determining the affective salience of the metadiscourse. However, the aspect of Crismore's report that is most significant for my own study is her definition of informational metadiscourse as including "statements naming the discourse type or act" (”AMtA” 66), a definition that anticipates my own analysis of metadiscourse within the context of speech act theory. In her next study, reported in an article titled "The Rhetoric of Textbooks: Metadiscourse," Crismore addresses in passing the question of whether general dicta concerning metadiscourse use are possible. Noting that ”because of societal pressures, the nature of rhetorical communities changes over the years and so do the norms” ("TRoT:M" 279), Crismore observes that ”the content included and emphasized 47 in today's textbooks is quite different from that of the early part of the century” ('TRoT:M" 279), and she adds that these differences also extend to the style of textbooks. Citing H. P. Grice's conversational maxims as important considerations for textbook authors, Crismore notes that text structure must vary to meet the needs of different types of readers: Writers of considerate texts are aware that readers who are unfamiliar with the subject matter or the conventions of a particular genre may need more explicit guidance and information; or they may need a text that requires fewer higher-order inferences or a text that establishes an interpersonal relationship between writer and reader. In other words, writers might decide to use a rhetorical style different from one suitable for older, more experienced, or more knowledgeable readers. (”TRoT:M' 279) Crismore then asserts that a rhetorical style using metadiscourse may be more appropriate for textbooks than is an unelaborated, anonymous style ("TRoT:M' 279). The remainder of Crismore's article discusses the textbook survey reported previously in Metadiscourse: What l£_l§ and adds little to the findings of that work; however, Crismore does provide a slightly different definition of metadiscourse, noting that it is "contentless writing about writing [that] includes comments about the discourse plans, the author's attitudes, the author's confidence in his following assertion, and the use of self—references and references to the readers-—the interpersonal part” ('TRoT:M" 280). She also notes that researchers should consider ”how much of which type [of metadiscourse] is needed by which 48 students for which tasks under what conditions" ("TRoT:M" 296), thus establishing a broader context for future empirical studies of metadiscourse. Crismore continues her exploration of metadiscourse use in textbooks with a 1985 paper titled The Case for a Rhetorical Perspective on Learning from Texts: Exploring Metadiscourse. After noting the problems that many students encounter when making the transition from basal readers to social studies textbooks, she suggests that "some fundamental changes may need to take place concerning the notion of what a content-area textbook should be” (TCfaRP 7). Among the changes she recommends is a shift to ”rhetorical textbooks" that communicate "both the desired content information and the author's attitudes toward it” (TCfaRP 9), and she notes further that rhetorical textbooks would reflect ”a concern not only for the message but also for how it is presented, the message source (the author), and the message receivers (the readers)” (TCfaRP 9). In essence, Crismore is calling for the use of meta- discourse in textbooks. Once again, Crismore adjusts her definition of metadiscourse, noting that ”metadiscourse is a level of discourse that adds another proposition to the subject matter" (TCfaRP 10) and that "metadiscourse calls attention to the communicative speech act itself, seeks to engage the reader as an active human being, and signals the presence of the author” (TCfaRP 11). In offering this definition, Crismore follows Vande Kopple (”EEfFSP" 51) in using the 49 term ”speech act” when referring to metadiscourse. She then suggests that metadiscourse manifests the interpersonal function of language identified by M. A. K. Halliday and Ruqaiya Hasan, who note that the interpersonal component expresses the speaker's ”attitudes and judgements, his encoding of the role relationships into the situation, and his motive in saying anything at all” (Cohesion 27). After again asserting that metadiscourse divides into informational and attitudinal forms, Crismore proceeds to examine a variety of texts, including conventional social studies textbooks and unconventional textbooks. She notes that the conventional textbooks use a scientific, fact—based approach to social studies, present no author perspective or stance, and use no metadiscourse, while the textbooks that are unconventional "appear to have real authors . . . who make their presence andeersonality felt in the text" (TCfaRP 20) and use metadiscourse to inform readers of the attitudes and intentions of the authors (TCfaRP 20—21). She then repeats many of the questions for textbook authors that she posed in ”A Message to Authors," asking again what the effects are of writing a text from a subjective or objective point of view, of using first person pronouns, and of having an explicitly—stated author perspective (TCfaRP 22). Crismore concludes than an experimental study is needed to determine which text characteristics are most comprehensible and appealing to students, but she notes also that such a study must be interdisciplinary in nature, a fact that poses a problem for researchers: 50 The problem in carrying out such a study, however, is that although reading researchers have benefited a great deal from different disciplines such as educational psychology, linguistics, and pragmatics, the kinds of issues, questions, and text characteristics discussed previously have not been looked at comprehensively by any one‘discipline. (TCfaRP 24) Because few researchers are well—versed in all of the fields that are concerned with text structure and learning theory, few are capable of conducting the research that Crismore proposes. In her 1985 doctoral dissertation, Crismore makes use of most of the studies reported in the articles summarized above. Following a useful survey of metadiscourse research, including research in speech communications that presents concepts similar to metadiscourse without actually using the word, Crismore introduces the study of metadiscourse in social science texts reported in Metadiscourse: What It lg, the textbook chapter examination reported in Suggestions for Improving a Poorly Written Sixth Grade Social Studies Chapter, and the experimental study reported in ”A Message to Authors.” In reporting on the experimental study, Crismore provides more detailed information concerning results than she did in her earlier article. She again notes that few significant effects concerning comprehension were obtained by adding metadiscourse to the Ginn passages, and she adds that the effects that were obtained "related to students scoring higher on a subtest when the information being tested was directly stated or repeated in the added metadiscourse" (”So'MaRA'" 32), a fact that suggests the 51 improvement may be due to the repetition of the primary discourse rather than to any properties of the metadiscourse itself. She notes that "there was little evidence that the informational or attitudinal metadiscourse raised or lowered the scores for retention of information that was not manipulated (i.e., stated, repeated, or emphasized) by the added metadiscourse” (”So'MaRA'” 32—33) and that ”the effects of interpersonal voice were minimal" ("So'MaRA'" 33). Concerning two-way interaction effects of various forms of metadiscourse in combination, Crismore observes that use of "interpersonal voice helps the low comfort students when informational metadiscourse is present, and hurts the high comfort students under the same conditions. Further, interpersonal voice reduces the liklihood that low comfort students will remember such things as author attitudes” ("So'MaRA'" 33). Concerning students' attitudes toward texts with metadiscourse, she reports the following results: The three significant main effects indicated that the presence of informational metadiscourse decreased the students' rated preference for social studies texts which used a large number of first and second person pronouns and opinion words, the presence of attitudinal metadiscourse increased the high comfort group's interest in the Middle Ages somewhat. There were no main effects for interpersonal voice. ("So'MaRA" 34). Crismore also notes that students also viewed negatively the text versions that combined attitudinal and informational metadiscourse, and most students preferred the version that had no metadiscourse at all (”So'MaRA'” 34).I Regarding her analysis of reading time, she reports that “lengthening 52 the text in the manner required by including informational metadiscourse produced longer reading times” (”So'MaRA" 34), and she expands on this by noting that ”the addition of informational metadiscourse increased the length of the passage from approximately 3,000 words to approximately 6,000 words, an increase of 100 percent [that] yielded a reading time increase of only 40 percent" (”So'MaRA'” 35). Although the doubling of text length constitutes a design flaw that casts doubt on the value of Crismore's findings (and that itself is sufficient to explain the students' preference for the text version without any metadiscourse), it is important to place this flaw and other imperfections in Crismore's work within a context that includes a paucity of knowledge concerning a theory of metadiscourse. In the next chapter, I will offer detailed critical comments of Crismore's studies; however, I should note now that her repeated modifications of her definition of metadiscourse and her numerous adjustments to her metadiscourse typology reflect the fledgling state of metadiscourse research. Without a sound theoretical base, it is exceedingly difficult to conduct experimental research. Despite the significant flaws in Crismore's work, her studies constitute the most substantial body of research on metadiscourse to date, and her work contains insights that I will explore further in presenting my own metadiscourse model. 53 Some Exploratory Discourse by William J. Vande Kopple Another researcher who was influenced by Joseph M. Williams' work on metadiscourse is William J. Vande Kopple, whose interest in metadiscourse is first noted in a 1980 doctoral dissertation prepared under Williams' guidance at the University of Chicago. In his dissertation, Vande Kopple offers the following explanation of metadiscourse: Many discourses have at least two levels. The primary level is made up of the propositional content. But often there is also discourse about the act of discoursing, discourse which does not add propositional information but which signals the presence of the author. This kind of discourse calls attention to the speech act itself, often marking stages in the‘ development of the primary discourse, displaying the author's position on the primary discourse, or molding the reader's attitude about the primary discourse. This is metadiscourse. ("EEfFSP” 50-51) As an example of metadiscourse, Vande Kopple provides the sentence, "I would now like to assert that stealing is wrong,” in which "I would now like to assert that" is metadiscourse, and the rest of the sentence is primary discourse (”EEfFSP” 51). In an attempt to measure readers' retention of information from paragraphs that use a variety of patterns for distribution of old and new information, Vande Kopple begins several sentences in each of his sample paragraphs with five-or-six-word strings of metadiscourse such as "It is my firm conclusion that" and ”I can say without hesitation that“ ("EEfFSP" 51). Regardless of the primary information distribution patterns in the sample texts, Vande Kopple found that the high school sophomores to 54 whom he administered a recall test were unable to remember much metadiscourse, recalling a mean of 5.7 words of the mean 23.5 words of metadiscourse present in the four sample passages. This rate of recall is lower than the students' recall rate for primary discourse, which leads Vande Kopple to conclude that the results ”are tentative evidence that we process metadiscourse on a level different from the level we use for primary discourse" ("EEfFSP" 52). Of the nineteen subjects that Vande Kopple tested, thirteen recalled no metadiscourse at all on at least one of the four sample texts. Of these thirteen, nine had a mean SRA comprehension score of 49, which indicates medium reading ability. Vande Kopple observes that ”whether extremely good and bad readers process and store metadiscourse differently than average readers must remain a problem for future research” (”EEfFSP" 54). Several aspects of Vande Kopple's study are worth noting. The first is that in saying metadiscourse calls attention to an author's speech act ("EEfFSP" 51), Vande Kopple establishes a link between metadiscourse and speech act theory. Also significant are the structures that Vande Kopple identifies as metadiscourse; although he does Inot provide a detailed taxonomy of metadiscourse forms, all of the examples of metadiscourse in his sample passages are clauses that indicate an author's certainty concerning a following statement, and Vande Kopple does not include in his metadiscourse total adverbs such as ”probably" that appear in the passages. Although Vande Kopple may have 55 ignored adverbs because he considered them inconsequential in his study, their exclusion is curious, especially in light of their importance in Williams' work. However, what is most significant in Vande Kopple's study is his tentative assertion that metadiscourse is processed at a different level from primary discourse. Although more evidence is needed to verify this assertion, a distinction in the memorability of metadiscourse and primary discourse would constitute empirical evidence to justify treating metadiscourse as a discrete category of text constituents. Vande Kopple provides a more detailed discussion of metadiscourse in an essay titled "Some Exploratory Discourse on Metadiscourse." In that essay, he offers a definition that again opposes metadiscourse to propositional content, but this new definition makes no reference to speech act theory: On one level we supply information about the subject of our text. On this level we expand propositional content. On the other level, the level of metadiscourse, we do not add propositional material but help our readers organize, classify, interpret, evaluate, and react to such material. Metadiscourse, therefore, is discourse about discourse or communication about communication. (”SEDOM" 83) Vande Kopple then asserts that "there are at least seven kinds of metadiscourse, the boundaries and internal characteristics of which will probably have to be more closely surveyed in future work" ("SEDoM‘ 83). His seven categories include (1) text connectives, which ”try to guide readers as smoothly as possible through our texts and to help them construct appropriate representations of them in 56 memory" ("SEDoM' 83), and which include sequencers such as “first" and ”next,” temporal or logical relators such as ”at the same time" and "as a consequence, reminders about material presented earlier like “as.I noted in Chapter One,” statements of what material one is about to present such as "what I wish to do now,” and topicalizers such as "in regard to” and "as for” ('SEDoM" 83), (2) code glosses, which explain the meaning of words in the text, including definitions of foreign words, (3) illocution markers, which identify speech acts, such as "I hypothesize" and "to sum up” (”SEDoM" 84), (4) validity markers, which express our confidence in the propositional content we convey, and which include three of Williams' categories-- hedges, emphatics, and attributors (”SEDoM" 84), (5) another of Williams' categories, narrators, such as ”the principal reported" ("SEDoM' 84), (6) attitude markers such as "I find it interesting" and "surprisingly” ("SEDoM" 85), and (7) commentary, those remarks that are addressed directly to the reader, such as "most of you will note" and "you might wish" ("SEDoM" 85). Regarding the specific constituents that belong in these categories, Vande Kopple notes that some words and groups of words can fulfill the functions of more than one category. As examples he cites ”I hypothesize that,” which he notes "probably functions in most texts as both an illocution marker and a validity marker” ("SEDoM‘ 85), and phrases such as "to conclude this section,“ which "probably function as both text connectives and illocution markers" (”SEDoM" 85). 57 Elaborating on the functions that metadiscourse serves, Vande Kopple attempts to situate his seven categories within the network of three language macro-functions identified by M. A. K. Halliday in Explorations in the Functions of Language. The first of Halliday's macro-functions is the ideational set, which is ”concerned with the content of language, its function as a means of the expression of our experience, both of the external world and of the inner world of our own consciousness” (EitFoL 58), and Vande Kopple suggests that primary discourse belongs in this set (”SEDoM” 86).. Halliday's other macro-functions are the interpersonal set, which is concerned with “language as the mediator of role, including all that may be understood by the expression of our own personalities and personal feelings on the one hand, and forms of interaction and social interplay with other participants in the com- munication on the other hand” (EitFoL 58), and the textual set, which has ”an enabling function . . . [allowing] the speaker to organize what he is saying in such a way that it makes sense in context and fulfills its function as a message" (EitFoL 58). Vande Kopple asserts that ”the kinds of metadiscourse can convey either interpersonal or textual meanings” ('SEDoM” 86); into the interpersonal set he places illocution markers, validity markers, narrators, attitude markers, and commentary, while into the textual set he places metadiscourse forms that "help us show how we link and relate individual propositions so that they form a cohesive and coherent text" ('SEDoM" 87), by which he 58 means the various sub-categories of text connectives. Noting that much research on metadiscourse remains to be conducted, Vande Kopple poses a series of questions for researchers to consider: Are some kinds of metadiscourse more appropriate than others——or even necessary-—in some kinds of texts? Are some kinds of metadiscourse more appropriate in longer rather than in shorter texts? What conventions govern the uses of kinds of metadiscourse, and how are these learned? How do uses of various kinds of metadiscourse correlate with the knowledge, values, attitudes, and reading skills of readers and with the difficulty of propositional material? How do various kinds of metadiscourse affect the comprehension and recall of texts? Finally, might heavy or light uses of some kinds of metadiscourse in certain situations be based upon writers' understanding of their subjects or upon their rhetorical, cognitive, or emotional deVelopment? ("SEDoM' 88) However, he also notes that classroom applications of metadiscourse theory are possible even before the above research is conducted, and he suggests that numerous benefits could result from acquainting students with the appropriate use of metadiscourse insofar as we are able. These benefits include sensitizing students to the overuse and underuse of metaidiscourse, thus improving the clarity of their writing, making students aware of the needs of readers, and helping students to understand the concept of tone in writing. The final benefit that Vande Kopple suggests might result from instruction in metadiscourse use is a refining of students' ethical sensibilities ('SEDoM' 90-91). He notes that ”some writers intentionally omit illocution markers that would be necessary to indicate that their messages are not full and accurate reports 59 but speculations, wishes, predictions, or partial reports" (”SEDoM” 91), and that ”some writers sometimes use validity markers to disguise biases, unsupported assertions, slight distortions, or outright falsehoods as facts” ("SEDoM' 91). If students examine inappropriate uses of these forms of metadiscourse and compare them to appropriate uses, they ”might become more concerned about how they announce their discourse actions” ("SEDoM' 91) and "might become more sensitive to the relationship between claims of probability and truth, the actual truth value of information, and the effects of information on those who regard it as true" ('SEDoM” 91—92). Because his work appears in a recent issue of the influential journal, College Composition and Communication, Vande Kopple's analysis of metadiscourse is perceived by many researchers as the most authoritative treatment of the subject to date. In several regards, his work typifies the present state of metadiscourse research—-his binary division of a text into propositional content and a metadiscourse component represents the most common definition of metadiscourse, his seven metadiscourse categories include forms identified by other researchers, and his questions for researchers to consider reflect the interests of both Williams and Crismore. For these reasons, I will subject Vande Kopple's work to careful scrutiny in the next chapter. 60 Other Works of Interest to Metadiscourse Researchers Because my study is concerned with the etymology of the word ”metadiscourse” and the problems posed by existing definitions and typologies of metadiscourse and its con- stituents, I have provided fairly detailed summaries of all the works that make use of the term. However, numerous other studies examine concepts similar to metadiscourse without actually using the word. While these studies are of limited importance to my own research, some brief comments concerning several of the studies are in order. Those desiring a more detailed account of these works are advised to consult Crismore's doctoral dissertation. Among those who have studied concepts similar to metadiscourse is Bonnie J. F. Meyer, whose examinations of text structure use the term "signaling.“ Meyer defines signaling as 'a noncontent aspect of prose which gives emphasis to certain aspects of the semantic content or points out aspects of the structure of the content“ ('WIRfP' 313), and she adds that “signaling in passages shows an author's perspective on the relative importance of the content related in his passage“ ('WIRfP' 313). Meyer notes that ”words of signaling are not included in analysis of the structure and content of a passage since they do not add new content and relations, but simply accent information already contained in the content structure of the passage" ("WIRfP' 313), thus paralleling the distinction between metadiscourse and propositional content that several researchers have suggested. In 61 two experimental studies, reported in 1975 and 1980, Meyer found little evidence to suggest that signaling has an effect on readers' ability to recall information from a text. Another concept similar to metadiscourse is “non— topical linguistic material," a term used by Liisa Lautamatti, a researcher from Finland. Lautamatti views non-topical material as distinct from the content material of a text, and she identifies five categories of non-topical markers: meta~textual markers such as "in the following chapter we shall examine” (24), modality markers, a category similar to Williams' attributors, and attitude markers, commentary markers, and illocutionary markers, three categories that Vande Kopple has borrowed for his taxonomy. Cited in works by Crismore as well, Lautamatti's study has exerted considerable influence on recent studies of metadiscourse. Also comparable to metadiscourse is ”alignment talk," a term used by speech researchers to describe speakers' attempts to establish communicative roles and a context for understanding oral discourse. Alignment talk has been studied in a variety of settings; among the works on this subject are a study by Sandra L. Ragan and Robert Hopper of alignment talk in job interviews, and Fred Donaldson's study, "Metacommunication in Rough and Tumble Play.” Under the rubric of "meta-talk,” the phenomenon has been studied by Deborah Schiffrin, whose work has influenced Crismore's ideas in several ways. Schiffrin 62 notes that I'meta-talk allows a speaker to exercise control over the principal discourse at specific junctures during its production by projecting an animator who will bracket the expressive implications of what is being said" (231), ”a and she adds that meta-talk functions in two planes, referential, informational plane when it serves as an organizational bracket, and on an expressive, symbolic plane when it serves as an evaluative bracket" (231), thus anticipating Crismore's distinction between informational and attitudinal metadiscourse. And like Crismore, Schiffrin includes verbs that name speech acts in her typology, noting that "one group of meta—linguistic verbs names acts of speech: verbs of saying, such as say, tell, ask, assert. Other verbs indicate that something will be done to a piece of talk: clarify and define, for example. And still others name speech events, for example, gpggg and 19527 (204). Other titles could be added to this list of works that include concepts similar to metadiscourse; likely candidates for inclusion are the works by van Dijk and Halliday mentioned earlier in this chapter. Although these works suggest the range of fields that are concerned with discourse about discoursing, the studies themselves do not use the term ”metadiscourse,” so they fall outside of the etymological considerations of my own work. While I will not consider these tangential works in any detail in the chapters that follow, I include them here to indicate the numerous rubrics under which researchers have classified 63 text structures that do not convey information of primary importance. In the next chapter, I will analyze in detail the studies that use the term ”metadiscourse." Through my analysis, I will attempt to demonstrate that the existing definitions and typologies of metadiscourse are inadequate, thus establishing the need for my own recommendations concerning a theory of metadiscourse. CHAPTER TWO A CRITICAL ANALYSIS OF EXISTING METADISCOURSE MODELS The thesis that I will support in this chapter is that the metadiscourse models I summarized in Chapter One all possess flaws that undermine their utility in text analysis. Concomitant to that thesis is my belief that the models do contain the seeds of a useful analysis, but that significant adjustments in the precision of the definitions of key terms and refinement of the constituents grouped under the rubric of metadiscourse are required. Before I begin a detailed analysis of each of the major models, I will introduce the basic critical criteria to which I will refer throughout my discussion of the models. Criteria for Analyzing Metadiscourse Models In any attempt to construct a model of systematic phenomena, a primary consideration is the precision with which the basic terms of analysis are defined. In the case of metadiscourse models, numerous possibilities exist for framing the basic analysis--metadiscourse may be considered as a syntactic, semantic, or pragmatic phenomenon, and most existing models define key terms with references to all three of these domains of study. However, the fundamental question in considering the 64 65 utility of definitions is not whether the domain used for specification of the key terms is appropriate; important questions concerning whether the terms of analysis should be syntactic or pragmatic, intensional or extensional, and writer-based or reader-based are all rendered moot if the definitions are vague or ambiguous. For this reason, the first criterion in evaluating metadiscourse models is whether the definitions of metadiscourse and its sub— categories are sufficiently precise to enable theorists to agree concerning the specific entities that the definitions describe. Another basic consideration in model construction is whether any purpose is served by including each of the sub-categories of constituents in the model. In positing sub-categories of constituents, theorists sometimes include a sub-category that has little in common with the other sub-categories in the model. By grouping a disparate sub-category along with other sub-categories, theorists reduce the number of generalizations that can be offered concerning the phenomena being modeled. If few statements can be made about the phenomena as a collective entity, then the descriptive analysis is of little value. To remedy this problem, it sometimes is necessary to exclude a sub-category from a model, relegating it instead to a different domain of study. In analyzing models of metadiscourse, my second criterion is whether each of the sub—categories in the models possesses properties sufficiently similar to the properties of the other sub—categories so that useful 66 generalizations can be offered concerning metadiscourse as a collective entity. Closely related to the second criterion is the question of whether any purpose is served by grouping constituents under a sub-category. In allocating constituents to sub- categories, theorists sometimes attempt to subsume numerous forms under a common heading; however, by grouping disparate constituents together in a sub-category, theorists reduce the number of generalizations that can be offered concerning that sub—category. Because this reduces the value of the descriptive analysis, theorists should limit each of their sub—categories to include only those constituents that have common properties or that are subject to the same rules concerning use. In analyzing metadiscourse models, my third criterion is whether each sub—category in the models is limited sufficiently so that theorists can advance generalizations that are true for all of the constituents in the sub-category. A fourth consideration in evaluating metadiscourse models is the extent to which the terms of analysis are useful in revealing the important properties of the phenomenon under consideration. As I noted above, existing models of metadiscourse include syntactic, semantic, and pragmatic terms; however, the terms drawn from these three domains may not be equally useful in identifying the salient characteristics of metadiscourse. In a paper I co-authored with Frank Parker, we noted the importance of selecting appropriate basic terms for use in a theoretical model, 67 and we suggested that just as a syntactic model that relies on phonological terms would be unrevealing, so too would it be unrevealing to construct a text model that relies on syntactic terms; in both cases, the appropriate terms of analysis for the phenomena under consideration cannot be provided by a model intended for use at a lower level of analysis (AfGLtC 7—8). Williams advances essentially the same point, noting that "if sentences are shaped by the discourses they appear in, quantitative analyses cannot capture that shape, particularly when only those features relevant to the structure of sentences constitute the terms of the theory” (”N-LLatToS" 27). For metadiscourse models, the appropriate terms of analysis remain in dispute, as is evidenced by the conflicting models currently proposed. As I analyze those models, my fourth criterion will be the extent to which the key terms in each model seem appropriate to reveal important aspects of metadiscourse structure. Using the four criteria presented above, I now will consider the relative merits of the metadiscourse models described in Chapter One. Through my analysis, I will attempt to show that all of the competing models violate at least one of the criteria, and I also will note the features in the existing models that should be preserved in a more satisfactory model of metadiscourse structure. Because my own model will place metadiscourse within the context of speech act theory, I will pay particular attention to those aspects of the existing models that suggest the utility of a speech act model of metadiscourse. 68 An Analysis of Zellig S. Harris' Model In considering Zellig S. Harris' comments concerning metadiscourse, it is important to remember that his most extensive discussion of the topic is in an_essay intended to serve a specific, narrow purpose: Harris is interested in developing a system for extracting important information from scientific articles in order to construct abstracts of those articles. ,For this reason, Harris' first definition of metadiscourse is specified in the domain of scientific articles only, and his commentstrovide little direction to those who would attempt to construct a definition of metadiscourse that is useful in a broader array of text studies. Noting that metadiscourse kernels "contain words entirely different from those of the main kernels" (”LTfIR” 465), Harris provides little elaboration of the forms of metadiscourse, saying only that ”metadiscourse kernels . . . talk about the main material (e.g., dis— cussing the problems of the investigators)” ('LTfIR” 464— 465). Because Harris does not specify the differences between words in metadiscourse kernels and those in other kernels, his definition violates my first criterion-—it is not specific enough to ensure that researchers will agree concerning which entities are metadiscursive. And his observation that some metadiscourse kernels talk about the investigators' problems is germane only to research reports. Harris provides only one specific example of a metadiscourse kernel in his essay, the introductory clause ”We have found that" (”LTfIR" 466), and he 69 classifies it in the 0 section of his typology, which also contains connectors such as ”because" and "if" ('LTfIR” 465). Because the introductory metadiscourse clause is different in both form and content from single—word con— nectors, this classification violates my third criterion—- it groups disparate constituents under the same heading. Although the narrowness of Harris' intention and his lack of specificity in identifying metadiscourse forms minimize the utility of his study, his pioneering work does identify introductory clauses as a form of metadiscourse. This classification anticipates other researchers' listings of illocution markers as a type of metadiscourse, and it provides an antecedent for my own discussion of the role that illocution markers serve in a model of metadiscourse. This is the only aspect of Harris' first study that I will preserve in my own model. Harris' second study that makes reference to meta— discourse places the term within the context of a "meta- language" that includes all the grammar rules of a language and also those words that refer to linguistic structures; metadiscourse is a subset of metalanguage, and "metadiscourse sentences can be inserted into the discourse about which they speak” (”AOiLS' 608). As an example, Harris cites the metadiscourse clause "This sentence is” (”AOiLS' 608), and he notes that ”if some discourse occurs only in particular circumstances, we can conjoin this statement about the discourse to the discourse itself" ("AOiLS" 608). Although Harris' 70 definition anticipates later definitions of metadiscourse as ”discourse about discoursing,“ the specific example he provides suggests that he would limit the term to self-referential sentences that contain clauses referring to their own linguistic structure. While this is a fairly specific explanation of the structure and function of metadiscourse, the role that Harris provides for the term is extremely limited. By limiting the term to a set of idiosyncratic structures, he limits the utility of metadiscourse as a descriptive term in a model of text structure. Harris' second definition of metadiscourse is of little practical importance to the researchers who have recently adopted the term, and it is inconsequential for my own model. An Analysis of Joseph M. Williams' Comments on Metadiscourse Among the theorists who have advanced metadiscourse models, Joseph M. Williams ranks as the most influential; his works have been cited in every recent study of metadiscourse. However, the most significant aspect of Williams' work is not the aspect that attracts the most attention: while most metadiscourse studies mention Williams' definition and categories, few acknowledge the guidelines for text studies that he provides in ”Non- Linguistic Linguistics and the Teaching of Style.” This oversight is unfortunate, because Williams' comments in that paper are useful directives to all who plan to study metadiscourse. 71 In suggesting that discourse analysis cannot begin with study of the syntactic roles of text constituents ('N-LLatToS' 35), Williams recognizes the need for a theoretical vocabulary that reflects the higher-order structure of a text. In the multi-level analysis of text structure that he offers, Williams tries to capture the rhetorical structure of a text by introducing a vocabulary that indicates the functions that text elements serve, identifying those functions with terms like ”topic” and ”comment.“ Although Williams does use syntactic terms in his description.of text structure, these terms constitute a level of analysis that complements the higher-order functional analysis~~the syntactic terms are neither the sole nor the primary means of describing a text. Williams' work shifts the emphasis in discourse analysis from study of grammatical structure to study of rhetorical/functional structure, and it is this shift in emphasis that constitutes his greatest contribution to the field of text linguistics. Produced at a time when most linguists were preoccupied by syntactic models, Williams' model marks a radical change in the nature of text studies. Within this context of rhetorical analysis, Williams provides a functional definition of metadiscourse, using the term to denote "all the elements in a sentence that } refer to the process of discoursing, as opposed to the specific reference of the discourse” (”N—LLatToS' 33). He expands upon this definition in ”Literary Style: The Personal Voice," noting that metadiscourse includes "words 72 and phrases and clauses—~even sentences--that refer . . . to the speech event that the discourse and its reader create” (”LS:TPV” 195). While this composite definition establishes metadiscourse as an entity within a functional theory of discourse structure, the comments that Williams offers in the two articles mentioned above do not constitute a satisfactory theory of metadiscourse. One problem is that Williams' taxonomy of metadiscourse types includes ”shorter discourse signals such as those that indicate cause: there- fore, consequently, as a result, pp, for, and so on" ("LS: TPV' 197); however, Williams does not establish how causal connectors relate to other metadiscursive references to the process of discoursing. In fact, connectors that establish causal links between primary discourse elements have been studied in detail by Halliday and Hasan (Cohesion 256—61) as part of a larger account of text connectives, and some causal connectors possess qualities quite different from those of Williams' other categories. Williams himself notes that causal connectors usually appear as short phrases, and I would add that cause signals usually cannot appear as introductory clauses that contain pronominal references to the discourse participants. By including causal connectors as a sub-category of metadiscourse, Williams reduces the number of meaningful generalizations that can be offered concerning metadiscourse as a collective entity, thus violating the second criterion of good design. Instead of being grouped as a sub-category of metadiscourse, causal connectors could be relegated to another component 73 of a comprehensive text model. In my own model of meta- discourse, I will consider causal connectors only insofar as they correlate with metadiscourse clauses that refer to the participants in a discourse. Another problem in Williams' analysis is that he does not provide a detailed account of the varied syntactic forms that can serve a single function. Williams does note that metadiscourse can "comment on the probability of the pro- position expressed in a sentence” through forms such as "probably," ”seemingly,” and ”it would seem that” ("LS:TPV” 197), and that it can ”express an attitude toward an event” through forms such as ”interestingly” and ”it is odd that” ("LS:TPV” 197); however, he does not point out that most of the metadiscourse examples he cites can occur either as introductory clauses that precede a sentential complement ("It is probable that,” “It would seem that," ”I find it interesting that,“ "It is odd that") or as adverbs ('probably,” ”seemingly,” "interestingly," 'oddly'). Although metadiscourse is defined in Williams' works as a rhetorical/functional category, it nevertheless is important to note the variety of forms that can convey a particular function. In his analysis of text structures other than metadiscourse, Williams is careful to note the “canonical" or most common modes of expressing information; however, he does not comment concerning what the canonical forms of metadiscourse might be. Consequently, his model of metadiscourse is not developed completely. In my own model, I will remedy this flaw by positing a canonical 74 form of metadiscourse——the illocutionary clause. I then will formulate rules for deriving other forms of meta- discourse from the canonical form. This solution will be based in part on Williams' suggestion that metadiscourse refers "to the speech event that the discourse and its reader create" (”LS:TPV" 195), since it will define the most ordinary form of metadiscourse as a clause that reveals the writer's intention to the reader. In the discussion that appears in Slylg, Williams defines metadiscourse as "writing about writing, whatever does not refer to the subject matter being addressed” (226). This definition is somewhat vague in specifying the forms and functions of metadiscourse, since it encompasses entities that Williams would not categorize as metadiscourse; for example, any book of literary criticism is "writing about writing," and many digressions in a text will "not refer to the subject matter being addressed" as the primary topic of the text. However, Williams does augment the definition with a typology of metadiscourse that is more detailed than those included in his other works. This typology divides metadiscourse into three groups, each of which contains two members: hedges and emphatics, sequencers and topicalizers, and narrators and attributors. Although these categories are adequate for Williams' textbook discussion of concision in writing, it is not evident that the categories should be preserved in a more extensive treatment of text structure. The entities that Williams classifies as metadiscourse possess such diverse properties 75 that one may question why they are grouped under a common rubric. The relationship between hedges and emphatics is clear, since the former express doubt while the latter express certainty. .Also fairly clear is the relationship between narrators and attributors, since these differ primarily in that the former use first person subjects while the latter use third person subjects. However, few similarities exist among categories taken from different couples in Williams' typology. What can one say that will apply equally to emphatics and narrators? To hedges and sequencers? A possible answer to these questions is that an excess or deficit of any metadiscourse forms can make a text difficult to comprehend. However, this is true also of material that Williams would not classify as meta- discourse; an overabundance of extraneous information of any type will cause comprehension problems, as will the omission of important information. These properties do not distinguish metadiscourse from discourse, and they do not provide a practical justification for grouping disparate categories under a common rubric. Once again, Williams violates the second criterion for designing a useful model, since his sub—categories are not sufficiently similar to allow useful generalizations. Another problem in the Style model is that it is not a principled system—-it does not contain rules to explain the interdependencies of the categories. For example, the attributors in Williams' model can be used to convey hedges, as in the clause ”Jones doubts that.” And narrators can be 76 used to convey emphatics, as in the clause "1 am certain that.” These examples suggest that some forms of meta- discourse interact with other forms in predictable patterns, and they also suggest that some metadiscourse categories may merit a status superordinate to that given to other categories. However, Williams does not explore these possibilities. Williams' Style model of metadiscourse also contains some other problems, several of which were evident in his earlier efforts. These include an inadequate account of the varied syntactic forms that can serve a single meta- discursive functiOn, and the presence of the "sequencers" category, which itself is only marginally metadiscursive and which contains causal connectors as a constituent sub-category. However, when considering the shortcomings in Williams' Style model, it is important to remember that his discussion clearly is governed by pedagogical concerns and that he is not attempting to situate meta— discourse within a rigorous theoretical framework. Although its analysis of metadiscourse is inadequate to guide future research into text structure, Style does offer a discussion that is well-suited for classroom use-~the purpose for which it is intended. An Analysis of Avon Crismore's Metadiscourse Studies In her series of studies, Avon Crismore provides a variety of definitions of metadiscourse, and she also presents several different typologies of metadiscourse 77 sub-categories. These definitions and typologies vary in their utility, with the least useful analysis provided in her early study, The Metadiscourse Component. Borrowing from the work of Teun A. van Dijk, Crismore defines meta- discourse as a system of textual relevance cues for a reader to use in determining the importance of information presented in a text (INQ 4), and she then presents nine categories and twenty—three sub-categories of metadiscourse types that she borrows from the works of van Dijk, Williams, and Randolph Quirk and Sidney Greenbaum. The problems that mar Crismore's analysis can be traced to two principle causes: (a) with the exception of Williams, none of the theorists that Crismore cites use the term ”metadiscourse,” because they pursue goals considerably different from hers in their works, with van Dijk concerned with producing a broader theory of text structure and Quirk and Greenbaum concerned with producing a detailed grammar of the English language, and (b) Crismore makes no attempt to reconcile the overlapping categories in the studies from which she borrows. The result is a mélange of disparate and redundant categories, with functional entities such as van Dijk's theme indicators and Williams' hedges appearing alongside syntactic categories like Quirk and Greenbaum's adverbial adjuncts, conjuncts, and disjuncts, and with text con— nectives appearing in slightly altered forms in all three of the models that Crismore cites. By combining these categories under a common heading, Crismore violates my first criterion of model design, since she makes no 78 effort to give precise definitions of discrete categories. She also violates my second criterion, since the diverse properties possessed by her functional and syntactic categories preclude the possibility of making meaningful statements about metadiscourse as a collective entity-— what can one say that will be true equally for theme indicators such as "the subject is” and attitudinal disjuncts such as "certainly”? For superstructure signals such as "our premises are" and particularizers such as ”chiefly"? And Crismore violates my fourth criterion by using syntactic terms as major entities in a study of a functional phenomenon. Her suggestion that ”linguists would call most metadiscourse adverbials” (ENE 8) is unrevealing, because it ignores the fact that either primary discourse or metadiscourse can be presented in adverbial form; for example, in the sentence ”Visually, it was a powerful play” (Quirk et al., AGoCE 429), "Visually” is a viewpoint adjunct adverbial that is not metadiscursive, while in the sentence ”Secondly, they have a losing record,” the word "secondly" is an enumerative conjunct adverbial that is metadiscursive. The important factor in determining whether a sentence constituent is metadiscursive is the function that the constituent serves in the sentence, not the grammatical shape of the con- stituent. In her 1983 publication, Metadiscourse: What It Is, Crismore resolves several of the problems that plague the model in The Metadiscourse Component. Once again 79 she provides a functional definition, noting that meta- discourse is ”the author's intrusion into the discourse, either explicitly or non-explicitly, to direct the reader rather than inform" (NENll 2), and adding that the purpose of metadiscourse is to direct readers "so they will under- stand what is said and meant in the primary discourse and how to 'take' the author" (fliflll 2). She then presents a simplified metadiscourse taxonomy that includes two major categories: informational and attitudinal. As sub—categories of informational metadiscourse she introduces goals, pre-plans, post—plans, and topicalizers, and as sub- categories of attitudinal metadiscourse she introduces saliency markers, emphatics, hedges, and evaluatives (M:WII 11-15). She adds that both informational and attitudinal metadiscourse can be either explicit or implicit signals of the author's presence (NlNll 12-13). Crismore's simplified taxonomy has the advantage of being a relatively unified description of the functions that metadiscourse can serve, since all the sub-categories are defined in reference to an author's guiding presence in a text. In this regard, the model adheres more closely to my second criterion of good design than her earlier model does, since all the sub—categories share the property of signaling the author's judgements concerning primary discourse. However, the definition still contains several imprecisions; the distinction between directing and in- forming is not drawn clearly, and even the attitudinal forms that Crismore presents are informative to some degree, since 80 they inform the reader of the author's attitudes. Even more problematic is the inclusion of a category called ”informational metadiscourse” following a definition of metadiscourse as writing that directs rather than informs; greater precision is needed in the definition to resolve this contradiction. And Crismore does not provide an adequate explanation of the distinction between explicit and non-explicit (or implicit) forms of metadiscourse. She does suggest that ”the larger metadiscourse phrases and sentences indicate more explicit author intrusion into the primary discourse while shorter metadiscourse words such as luckily or clearly indicate a more subtle intrusion on the author's part” (M:WII 15), but Crismore does not establish a link between length and affective saliency, so a dis- tinction in explicitness based on length rather than on the content of the metadiscourse seems unjustified. In any event, the term ”non-explicit" is a misnomer, because even the shortest forms of metadiscourse must be at least partially explicit in order for Crismore to identify them. All of these problems constitute violations of the first criterion for evaluating metadiscourse models--Crismore's definition of metadiscourse is not sufficiently precise to enable theorists to agree concerning which entities are forms of metadiscourse. Because Crismore's definition of metadiscourse is imprecise, the information she offers concerning the types and frequency of metadiscourse in textbooks is invalid——without a precise definition of metadiscourse, other researchers cannot corroborate 81 her findings. In my own metadiscourse model, I will utilize a distinction similar to the one that Crismore draws between informational and attitudinal metadiscourse; however, I will not borrow her misleading terminology, and I will provide a more precise description of the functions that metadiscourse serves. I also will use the term ”explicit” to describe certain forms of metadiscourse, but I will not define explicitness in terms of metadiscourse length; instead, I will offer a structural description of the features that typify explicit metadiscourse, and I will compare these features to those that typify partially explicit forms of metadiscourse. In ”A Message to Authors about Metadiscourse Use in Instructional Texts,” Crismore defines metadiscourse as ”a level of discourse that adds nothing to the propositional content” (66), and she repeats the distinction between informational and attitudinal forms of metadiscourse. However, she presents somewhat different taxonomies of the two categories, noting that informational metadiscourse includes macropropositions, justification statements for macropropositions, purpose and goal statements, and state- ments naming the discourse act or organizing strategies, while attitudinal metadiscourse includes pronouns of self- reference or audience reference, mental state or process verbs, and sentential adverb constructions ('AMtA' 66). The wording of Crismore's definition is misleading, since it is clear that metadiscourse does add something Tto' 82 propositional content-—it adds the author's perspective concerning the validity or quality of the propositional content. Perhaps what Crismore means to say is that metadiscourse does not add propositional content to the text. However, even this assertion would be problematic, because Crismore does not define "propositional content," and a common definition of the term would encompass several of the sub—categories of informational metadiscourse that she identifies. As David Crystal notes, in linguistics the term "proposition" refers to "the unit of meaning which constitutes the subject-matter of a statement in the form of a simple declarative sentence" (AFDoLaP 288); how- ever, this definition would include the topic sentences that Crismore includes as macropropositional forms of metadiscourse, and it also would include purpose and goal statements. Without knowing how Crismore defines "propositional content," it is pointless to speculate concerning which entities are forms of metadiscourse—-her definition of metadiscourse violates the first criterion for evaluating metadiscourse models. A problem also exists in the distinction that Crismore draws between informational and attitudinal forms of meta- discourse, because some of her attitudinal forms overlap with the informational forms. For example, Crismore identifies pronouns of self—reference as attitudinal metadiscourse; however, these pronouns also could occur in constituents that she identifies as goal statements, as is exemplified by the sentence "I will show why Smith 83 is wrong.” Once again Crismore has fallen into the trap of defining functional entities in grammatical terms, and once again her analysis is imprecise and unrevealing, thus violating both the first and fourth criteria for good design of a model. Given the fundamental problems in Crismore's analysis of metadiscourse, it hardly is surprising that the results of the experimental study reported in ”A Message to Authors” are inconclusive. Compounding the theoretical problems are flaws in the design of the study, among which are Crismore's failure to balance the length of her sample text passages and her uncertainty concerning the appropriate plaCes to insert metadiscourse into the Ginn text. The only aspect of Crismore's report that is significant for my own study is her suggestion that ”statements naming the discourse type or act" ("AMtA” 66) are forms of metadiscourse. This phrasing anticipates my own analysis of metadiscourse within the context of speech act theory. In ”The Rhetoric of Textbooks: Metadiscourse," Crismore again reports on the textbook survey reported previously in Metadiscourse: What It Is, and she adds little to the findings discussed in the earlier work. However, she does include yet another peculiar definition of metadiscourse, noting that it is ”contentless writing about writing [that] includes comments about the discourse plans, the author's attitudes, the author's confidence in his following assertion, and the use of self-references and references to the readers—-the interpersonal part” (”TRoT:M" 280). 84 Considering the variety of information that Crismore suggests metadiscourse can convey, it is odd that she use the word ”contentless” in referring to it; indications of an author's plans and attitudes certainly constitute types of content. It appears that Crismore again is attempting to distinguish between metadiscourse and primary information, but such a distinction must be based on the types of content that the two categories convey, not on the simple presence of content. Crismore's suggestion that metadiscourse is contentless creates the impression that metadiscourse is insignificant, despite the fact that Crismore herself is aware of the importance of meta- discourse. However, the remainder of her definition is useful, because all of the functions that Crismore attributes to metadiscourse reflect the presence of a writer and a reader who use a text as a communicative medium. In her emphasis of the interpersonal functions of metadiscourse, Crismore provides the basic ideas needed for a speech act theory of metadiscourse. In The Case for a Rhetorical Perspective, Crismore offers yet another definition of metadiscourse, noting that "metadiscourse is a level of discourse that adds another proposition to the subject matter” (10). This definition constitutes a complete reversal of the definition offered in ”A Message to Authors," which states that meta- discourse "adds nothing to the propositional content" (66). Because Crismore still does not define the term ”pro- position,” this new definition is no more adequate than 85 her earlier effort; however, the change does suggest that Crismore recognizes that metadiscourse is informative in its own right and is not contentless. In expanding upon her definition, Crismore places metadiscourse within the context of speech act theory, observing that ”metadiscourse calls attention to the communicative speech act itself, seeks to engage the reader as an active human being, and signals the presence of the author" (TCfaRP 11). However, she does not explore the connection between metadiscourse and speech act theory in any detail. Instead, she again notes that metadiscourse can be either informational or attitudinal, but she does not provide a detailed discussion of the structural properties of metadiscourse, thus omitting any description of the variant metadiscourse forms that can serve a single function. Consequently, her analysis only suggests a promising direction for other metadiscourse researchers to pursue; her own work is not sufficiently well—developed to serve as an authoritative metadiscourse theory. As I noted earlier, Crismore's doctoral dissertation is a compendium of her earlier research, and it adds little to the ideas presented in the papers that I have critiqued. In constructing my own metadiscourse theory, I will expand upon Crismore's suggestion that metadiscourse identifies communicative speech acts, and I will use some of her terminology to label the forms and functions of meta- discourse. However, I will attempt to avoid the imprecision and inadequate development of ideas that diminish the 86 utility of her studies. An Analysis of William J. Vande Kopple's Studies William J. Vande Kopple's first comments concerning metadiscourse appear in his 1980 doctoral dissertation, in which he observes that metadiscourse is ”discourse about the act of discoursing, discourse which does not add pro— positional information but which signals the presence of the author” (”EEfFSP” 50), a definition that anticipates several elements of Crismore's later definitions; Like Crismore, Vande Kopple attempts to distinguish between metadiscourse and propositional content without providing a definition of the term "proposition,” so his definition of metadiscourse is insufficiently precise. And Vande Kopple also anticipates Crismore's work when he suggests that metadiscourse ”calls attention to the speech act itself, often marking stages in the development of the primary discourse, displaying the author's position on the primary discourse, or molding the reader's attitude about the primary discourse" ("EEfFSP‘ 51), thus becoming the first researcher to link metadiscourse with speech act theory. Vande Kopple's dissertation does not contain a detailed taxonomy of metadiscourse types, and it does not attempt to specify the structural varieties of metadiscourse that can serve a single function, so his work does not con— stitute a complete metadiscourse theory. However, his use of introductory clauses as examples of metadiscourse anticipates the important role that these entities will 87 play in my own model. And his tentative assertion that metadiscourse and primary information are not equally memorable suggests an empirical justification for treating metadiscourse as a discrete category of text constituents, although more evidence is needed to support this assertion. Despite the fact that metadiscourse is only a tangential topic in Vande Kopple's dissertation, his analysis, in- complete as it is, makes a substantial contribution to metadiscourse research. In ”Some Exploratory Discourse on Metadiscourse," Vande Kopple offers a more detailed study of metadiscourse. He again establishes a distinction between metadiscOurse and propositional content, noting that on ”the level of metadiscourse, we do not add propositional material but help our readers organize, classify, interpret, evaluate, and react to such material" ("SEDoM' 83); however, he still does not define ”propositional material.” Perhaps more disturbing is the fact that this later definition of meta— discourse omits any mention of speech act theory, thus ignoring a connection that could integrate metadiscourse into a broader theory of discourse. Vande Kopple does not provide a reason for this change in his definition, but a probable explanation is that several of the categories he includes in his metadiscourse taxonomy do not figure in speech act theory. In fact, Vande KOpple's seven categories and eight sub-categories of metadiscourse are so disparate that one may question why they are grouped together under a common 88 heading. Vande Kopple violates the second criterion for good design of a model because his categories are not sufficiently similar to allow meaningful generalizations concerning them, since little of what one can say concerning code glosses will apply to attitude markers, and little of what one can say concerning the five varieties of text connectives will apply to validity markers. And in some cases where the categories are similar, the distinction between them seems arbitrary; for example, there is little functional difference between an attributor such as ac— cording to Einstein" ("SEDoM" 84) and a narrator such as ”Einstein reports that," so the two forms do not merit distinct categories. Instead, they should be treated as variant structural forms of the same functional category. However, Vande Kopple does not provide a detailed account of the varied syntactic forms that can serve a single function, so his model is incapable of explaining the similarity of the above examples; instead, it offers an unrevealing analysis that violates the fourth criterion of model design. And while Vande Kopple does note that some groups of words can fulfill the functions of more than one category (”SEDoM" 85), his analysis of the inter- dependencies among his categories and sub-categories is incomplete; for example, he doesn't observe that commentary and attitude markers can be combined, as is evidenced by the example ”you may be surprised to note,” and he doesn't mention that emphatics and attributors can occupy the same passage of metadiscourse, as in the example "Smith is 89 certain that." Although most of these problems also are present in Williams' Style model and in Crismore's works, their presence in Vande Kopple's model is particularly disappointing in light of the promise shown by his first comments concerning metadiscourse. The Prospects for an Adequate Theory of Metadiscourse In identifying the flaws in the metadiscourse theories discussed above, I have not intended to suggest that it is impossible to construct an adequate theory of metadiscourse; in fact, I believe that the existing theories contain the seeds of an adequate analysis. Although the models differ in their definitions and categories, all are marred by imprecise formulations and inadequate development; however, they also suggest considerable agreement among theorists concerning some of the features that a metadiscourse model must possess. Common to most of the theories is the belief that an adequate account of metadiscourse must include categories that identify the roles that a writer and a reader play in using a text as a communicative medium. These categories must account for specific references in the text to the writer and reader, and the categories also must identify the communicative functions that passages of metadiscourse serve. In fact, a substantial corpus of research that attempts to explain many of the phenomena considered in metadiscourse theories already exists. It is not a coincidence that both Crismore and Vande Kopple mention speech act theory in their 90 writings on metadiscourse, because the research that speech act theorists have been conducting for over thirty years is guided by goals similar to those that metadiscourse theorists pursue. Perhaps the most significant omission in existing metadiscourse research is its failure to exploit thoroughly the possibilities offered by speech act theory. In the next chapter, I will provide a brief survey of important works on speech act theory. In so doing, I will establish a foundation for a speech act model of meta- discourse. CHAPTER THREE BASIC CONCEPTS IN SPEECH ACT THEORY In this chapter, I will provide an overview of several basic concepts in speech act theory. In so doing, I will not attempt to survey all of the available speech act literature; such a survey is not necessary to produce an adequate account of metadiscourse within the context of speech act theory, since the concerns of metadiscourse theorists are narrower than those of speech act theorists. Instead, I will devote most of my attention to works by the two most influential speech act theorists: J. L. Austin and John R. Searle. Readers who wish to pursue the subject in greater depth can consult the fine synoptic studies that already are available, among the most useful of which are the chapters on speech act theory in Malcolm Coulthard's An Introduction to Discourse Analysis and in Stephen C. Levinson's Pragmatics. J. L. Austin and the Origins of Speech Act Theory Speech act theory begins with the William James Lectures delivered by J. L. Austin at Harvard University in 1955; these lectures appear in the seminal book, Egg to Do Things with Words. Throughout the lectures, Austin explores the properties possessed by performative 91 92 utterances, a class of statements that differ from the common constative utterances of a language. Noting that ”for too long [it was] the assumption of philosophers that the business of a 'statement' can onlbee 'to describe' some state of affairs, or to 'state some fact', which it must do either truly or falsely” (NEEE 1), Austin observes that some utterances are not constative--they'do not make statements that are either true or false. Among the examples he cites are ”I do,” when it is uttered in a wedding ceremony, and "I bet you Sixpence it will rain tomorrow" (NEQE 5). Austin calls these utterances ”per— formatives," and he adds that "the issuing of the utterance is the performing of an action—-it is not normally thought of as just saying something” (NEEE 6—7). Performatives are utterances ”in which to ggy something is to E9 something; or in which py saying or lg saying something we are doing something” (NEEE 12). Concerning the "I do” example, he notes that ”in saying these words we are doing something-- namely, marrying, rather than reporting something, namely EEEE we are marrying” (NEEE 13). While performatives are not themselves true or false, Austin notes that the execution of a performative can be a failure, depending upon the manner and context in which the performative is uttered (NEEE 14). He then specifies four ”felicity conditions" that are necessary for the "happy" execution of a performative: (A. 1) There must exist an accepted con- ventional procedure having a certain conventional effect, that procedure to include the uttering of certain words 93 by certain persons in certain circum— stances, and further, (A. 2) the particular persons and circumstances in a given case must be appropriate for the invocation of the particular pro- cedures invoked. (B. 1) The procedure must be executed by all participants both correctly and (B. 2) completely. (HtDT 14—15) Austin refers to violations of these four conditions as ”misfires,” and he notes that such violations prevent the performance from being achieved. For example, if one of the partners in a marriage ceremony already is married, or if the person who performs the ceremony is not authorized to do so, then the marriage is not completed successfully, even if the appropriate words are spoken (HtDT 14-15). Austin then attempts to identify methods for recognizing performative utterances. He begins by con- sidering the grammatical form of performatives, noting that all of the examples he has discussed "begin with or include some highly significant and unambiguous expression such as 'I bet', 'I promise', 'I bequeath'--an expression very commonly also used in naming the act which, in making such an utterance, I am performing--for example betting, promising, bequeathing” (HtDT 32). He then notes that the classic examples of performatives use ”verbs in the first person singular present indicative active" (HtDT 56) and that this form often indicates whether an utterance is performative, since "there is an asymmetry of a systematic kind between it and other persons and tenses of the very same verb" (HtDT 63), exemplified by ”I bet," which is performative, and "I betted" and "he bets,” which are not 94 performative (HtDT 63). However, Austin notes that the use of the first person singular present indicative active is not the only form for conveying performatives. He observes that another common type of performative "has the verb in the secOnd or third person (singular or plural) and the verb in the passive voice: so person and voice anyway are not essential" (NEEE 57). As examples of this form, he cites "You are hereby authorized to pay” and "Passengers are warned to cross the track by the bridge only” (NEEE 57). Having identified two standard forms of performatives, Austin notes that still other forms are possible. He points out that "if we turn away from . . . highly formalized and explicit performative utterances, we have to recognize that mood and tense (hitherto retained as opposed to person and voice) break down as absolute criteria" (NEEE 58). He then provides examples of performatives that do not conform to either of the standard forms: Mood (whatever this may be in English as opposed to Latin) will not do [to identify performatives], for I may order you to turn right by saying, not 'I order you to turn right', but simply 'Turn right'; I may give you permission to go by saying simply 'You may go'; and instead of 'I advise (or ”re- commend') you to turn right' I may say 'I should turn to the right if I were you'. Tense will not do either, for in giving (or calling) you off-side I may say, instead of 'I give (or ”call") you off—side', simply 'You were off-side'; and similarly, instead of saying 'I find you guilty' I may just say 'You did it'. Not to mention cases where we have only a truncated sentence, as when I accept a bet by saying simply 'Done', and even cases where there is no explicit verb at all, as when I say simply 'Guilty' in finding a person guilty. . . . (HtDT 58) 95 In light of the variety of forms that performatives can assume, Austin concedes that grammatical criteria are not themselves sufficient for identifying perfOrmative utterances (NEEE 60). However, he also notes that the two standard forms possess the advantage of being relatively unambiguous expressions of performatives, while the implicit forms are easier to misconstrue: There is something which is at the moment of uttering being done by the person uttering. . . . The 'I' who is doing the action does thus come essentially into the picture. An advantage of the original first person singular present indicative active form——or likewise of the second and third and impersonal passive forms with signature appended--is that this implicit feature of the speech—situation is made eXplicit. Moreover, the verbs which seem, on grounds of vocabulary, to be specially performative verbs serve the special purpose of making explicit (which is not the same as stating or describing) what precise action it is that is being performed by the issuing of the utterance: other words which seem to have a special performative function (and indeed have it), such as 'guilty', 'off—side', &c., do so because, in so far as and when they are linked in 'origin' with these special explicit performative verbs like 'promise', 'pronounce', 'find', &c. (HtDT 60-61) In the above passage, Austin introduces a distinction between explicit performatives that use one of the two standard forms and implicit performatives that omit specific reference to either the speaker or to time act performed. He further notes that ”what we should feel tempted to say is that any utterance which is in fact a performative should be reducible, or expandible, or analysable into a form, or reproducible in a form, with a verb in the first person singular present indicative active (grammatical)" (HtDT 61—62). Although Austin notes that the analysis of 96 implicit performatives is not foolproof, since the same utterance can serve different functions in different con- texts (NEQE 62), the difficulties in analysis are themselves reflections of the potential for readers to misconstrue implicit performative utterances. (For a more detailed account of performative structure, see Austin's Philo- sophical Papers, 233—52.) Having identified two canonical forms and numerous implicit forms of performatives, Austin reconsiders the distinction between stating and doing that served as his original justification for distinguishing between con- stative and performative utterances. In so doing, he discusses a category of performatives that he labels ”expositives,” in which ”the main body of the utterance has generally or often the straightforward form ofIa 'state- ment', but there is an explicit performative verb at its head which shows how the 'statement' is to be fitted into the context of conversation" (NEQE 85). Among the examples of expositives that Austin provides are ”I argue (or urge) that there is no backside to the moon,” and “I conclude (or infer) that there is no backside to the moon" (NEQE 85). Austin notes that expositive performatives (as typified by the explicit form ”I state") pose a problem for his analysis, because "when we come to pure explicit performatives such as 'state' or 'maintain', surely the whole thing is true or false even though the uttering of it is the performing of the action of stating or maintaining" (NEEE 90). Since Austin's original distinction between constatives and 97 performatives is based on the assumption that performatives do not possess truth properties, expositive performatives seem to contradict his analysis. To resolve this problem, Austin first suggests that ”we could distinguish the per— formative opening part (I state that) which makes clear how the utterance is to be taken, that it is a statement (as distinct from a predication, 8c.), from the bit in the that-clause which is required to be true or false” (NEEE 90); however, he quickly notes that "there are many cases which, as language stands at present, we are not able to split into two parts this way, even though the utterance seems to have a sort of explicit performative in it: thus 'I liken 5 to y', 'I analyse E as y'" (NEQE 90). Austin concludes that a different type of analysis is necessary to replace his distinction between constatives and per- formatives. Noting that "for some years we have been realizing more and more clearly that the occasion of an utterance matters seriously, and that the words used are to some extent to be 'explained' by the 'context' in which they are designed to be or actually have been spoken in a linguistic interchange” (NEEE 100), Austin introduces a new distinction between the meaning and the force of an utterance. In so doing, he identifies three acts that an utterance can perform. The first of these is a locutionary act, ”which is roughly equivalent to uttering a certain sentence with a certain sense and reference, which again is roughly equivalent to 'meaning' in the 98 traditional sense" (HtDT 109). The second is an ll: locutionary act, which is the ”performance of an act in saying something as opposed to performance of an act of saying something" (NEEE 100) and which includes acts ”such as informing, ordering, warning, undertaking, &c., i.e. utterances which have a certain (conventional) force” (NEEE 109). The third is a perlocutionary act, which expresses the effect a speaker hopes to achieve upon a listener ”2y saying something, such as convincing, persuading, deterring, and even, say, surprising or misleading” (NEEE 109). Austin notes that these three acts represent ”different senses or dimensions of the 'use of a sentence' or of 'the use of language'“ (NEQE 109—110). In distinguishing among the acts, he provides the following examples: Act (A) or Locution He said to me 'Shoot her!’ meaning by 'shoot' shoot and referring by 'her' to her. Act (B) or Illocution He urged (or advised, ordered, 8c.) me to shoot her. Act (C.a) or Perlocution He persuaded me to shoot her. Act (C.b) [Perlocution] He got me to (or made me, 8c.) shoot her. (HtDT 101-102) These examples are the possible acts that can be performed by the sentence "Shoot her,” and they suggest that a single utterance can perform all three of the acts that Austin identifies. Restating the distinctions among his categories in a somewhat simplified form, Austin notes that a locutionary 99 act "has a meaning,“ an illocutionary act "has a certain force in saying something,” and a perlocutionary act "is the achieving of certain effects by saying something" (HtDT 121). Since an utterance can perform all three acts, it follows that a single utterance can have not only truth properties relating to its meaning, but also "happiness” properties relating to its success in conveying the il- locutionary force that the speaker intends. In introducing three types of acts, Austin has replaced his original dis- tinction between constatives and performatives with a richer analysis of the multiple properties that a single utterance may possess. And while he is unable to isolate grammatical components of an utterance that convey a particular act in every case, Austin retains a canonical form for illocution markers, since explicit performatives such as "I state” often are used to convey illocutionary force. As Malcolm Coulthard observes, The illocutionary act, being achieved through the uttering of certain words, is potentially under the control of the speaker; provided he uses the correct explicit performative in the appropriate circumstances he can be certain that the act will be happy--no one can prevent someone from warning or advising them, except by refusing to listen. (18-19) In attempting to identify the types of illocutionary forces that utterances can convey, Austin lists five general classes of illocutionary acts. These include verdictives, which ”consist in the delivering of a finding, official or unofficial, upon evidence or reasons as to value or fact, so far as these are distinguishable” (HtDT 153) and which in— clude examples such as "acquit," ”convict," "assess,” and 100 ”estimate" (HtDT 153); exercitives, which give ”a decision in favor of or against a certain course of action . . . a decision that something is to be so, as distinct from a judgement that it is so” (NEEE 155) and which include "appoint,” "dismiss," ”order,” "direct,” and “urge” (NEEE 155-56); commissives, which ”commit the speaker to a certain course of action” (HtDT 157) and which include "promise,” ”intend," and ”guarantee" (HtDT 157-58); behabitives, which "include the notion of reaction to other people's behavior and fortunes and of attitudes and expressions'of attitudes to someone else's past conduct or imminent conduct” (NEE! 160) and which include ”apologize," ”thank,” ”deplore,” and “congratulate” (HtDT 160); and expositives, which "are used in acts of exposition involving the expounding of views, the conducting of arguments, and the clarifying of usages and of references" (NEQE 161). Austin notes that many examples of expositives also may be analyzed as being forms of other categories, but he cites numerous examples that he takes to ”have reference to the communicational situation" (NEEE 162). These include ”state,” "describe,” "inform,” "report,” "argue," ”revise,” ”begin by," ”conclude by,” ”illustrate," and "understand" (NEQE 162—63). For researchers interested in metadiscourse theory, Austin's work is significant in several ways. First, there is a striking similarity between Austin's analysis of locutionary and illocutionary acts and the distinction that metadiscourse researchers attempt to make between primary discourse and metadiscourse. While locutionary 101 acts and primary discourse are categories that characterize utterances having a particular sense and reference, il— locutionary acts and metadiscourse are categories that characterize a speaker's communicative intentions. Most notable is the similarity between expositive illocutionary acts and metadiscourse; Austin's observations that ex— positives ”have reference to the communicational situation” (NEEE 162) and are "the clarifying of reasons, arguments, and communications" (NEEE 163) are akin to Williams' definition of metadiscourse as ”all the elements in a sentence that refer to the process of discoursing, as op— posed to the specific reference of the discourSe' ("N-LLat T08” 33) and to Crismore's observation that the purpose of metadiscourse is to direct readers ”so they will understand what is said and meant in the primary discourse and how to 'take' the author” (Ngflll 2). The similarity of these comments suggests the possibility of defining metadiscourse as a type of expositive illocutionary act—-a possibility that I will explore in my own model. Also significant is Austin's analysis of performative structure; while metadiscourse researchers have struggled to describe the variant forms of metadiscourse types, Austin posits two explicit performative patterns that are canonical forms for expressing illocutionary acts, and he analyzes implicit performatives as being reducible to the explicit forms. In so doing, Austin suggests a sensible solution to the problems in structural description that have plagued metadiscourse models--a solution that I will utilize in 102 constructing my own model of metadiscourse. John R. Searle on Speech Act Theory Perhaps the.most influential speech act theorist since Austin, John R. Searle both borrows from and extends upon Austin's work. In his first book, Speech Acts, Searle offers an analysis of speech acts that remains one of the most important works in the field. Searle begins by asserting that speech acts ”are the basic or minimal units of linguistic communication" (EN 16) and that ”a great deal can be said in the study of language without studying speech acts, but any such purely formal theory is necessarily incomplete” (EN 17). Searle then identifies four types of speech acts: (a) utterance acts, which are simple acts of uttering words and sentences, (b) propositional acts, which are acts of referring and pre- dicating, (c) illocutionary acts, which include stating, commanding, and promising, and (d) perlocutionary acts, which express the effects that illocutionary acts have upon a hearer (EN 24—25). The first two categories pro- vide a finer distinction between acts that Austin would classify simply as ”locutionary," since Searle distinguishes between simple linguistic utterances and those that make specific references. In distinguishing illocutionary acts from perlocutionary acts, however, Searle repeats Austin's usage of those two terms. In elaborating on the distinctions among utterance acts, propositional acts, and illocutionary acts, Searle 103 notes that "utterance acts consist simply in uttering strings of words" (EN 24), while propositional and il- locutionary acts "consist characteristically in uttering words in sentences in certain contexts, under certain con- ditions and with certain intentions" (EN 24—25). This dis- tinction renders utterance acts only indirectly important in the study of linguistic communication. Concerning the distinction between propositional and illocutiOnary acts, Searle notes that "a proposition is to be sharply dis— tinguished from an assertion or statement of it” (EN 29). To exemplify this point, he provides seven examples of sentences that contain the same proposition but that do not perform the same illocutionary act. These example sentences include "Sam smokes habitually," "Sam, smoke habituallyE', and "Does Sam smoke habitually?" (EN 22), of which only the first is the assertion of a proposition. Searle then pro— vides the following analysis of the distinction between propositions and illocutionary acts: Stating and asserting are acts, but propositions are not acts. A proposition is what is asserted in the act of asserting, what is stated in the act of stating. The same point in a different way: an assertion is a (very special kind of) commitment to the truth of a proposition. (EN 29) Searle's use of terminology is inconsistent, because he also states that ”the expression of a proposition is a pro- positional act” (EN 29); however, what is important is that he establishes a clear distinction between propositions and illocutionary acts. Concerning the grammatical structures of illocutionary 104 and propositional acts, Searle provides the following observations: The characteristic grammatical form of the illocutionary act is the complete sentence (it can be a one-word sentence); and the characteristic grammatical form of the pro- positional acts are parts of sentences: grammatical predicates for the act of pre- dication, and proper names, pronouns, and certain other sorts of noun phrases for reference. (SA 25) Searle also adds that "propositional acts cannot occur alone; that is, one cannot just refer and predicate without making an assertion or asking a question or performing some other illocutionary act” (EN 25). While the suggestion that the characteristic form of an illocutionary act is a complete sentence seems to con- tradict Austin's assertion that illocutionary acts can be reduced to explicit performative clauses, the point that Searle is making is that, just as propositional acts cannot occur alone, so too must illocutionary acts always contain a proposition. The complete illocutionary act is conveyed not only in the parts of a sentence that express the 11- locutionary force, but also in those parts that express the proposition that is asserted, stated, questioned, etc., in the illocutionary act. It is impossible to assert or state without asserting or stating something, and that something is the propositional content of the illocutionary act. In fact, Searle does attempt to distinguish between the grammatical components that convey a proposition and those that convey the nature of an illocutionary act, as 105 is evidenced by his division of a sentence into a pro- positional indicator and an illocutionary force indicator: From [a] semantical point of view we can distinguish two (not necessarily separate) elements in the syntactical structure of the sentence, which we might call the propositional indicator and the illocutionary force indicator. The illocutionary force indicator shows how the proposition is to be taken, or to put it another way, what illocutionary force the utterance is to have; that is, what illocutionary act the speaker is performing in the utterance of the sentence. . . . I may indicate the kind of illocutionary act I am performing-by beginning the sentence with 'I apologize', 'I warn', 'I state', etc. Often, in actual speech situations, the context will make it clear what the il- locutionary force of the utterance is, without its being necessary to invoke the appropriate explicit illocutionary force indicator.. (EN 30) In noting that the propositional indicator and the il— locutionary force indicator are not necessarily separate in the syntactic structure of a sentence, Searle retains Austin's analysis of implicit performative structure. And Searle's examples suggest that the term ”explicit il- locutionary force indicator” is synonymous with Austin's first standard form of explicit performatives. In the chapters that follow, Searle discusses the rules that constitute and regulate speech acts, and he provides a detailed analysis of the illocutionary act of promising. However, he does not provide a detailed taxonomy of il- locutionary acts; for this we must turn to a later pub- lication. In ”A Classification of lllocutionary Acts," Searle discusses several problems in Austin's taxonomy of il— locutionary acts, and he then introduces an alternative 106 taxonomy of his own. As a basis for analyzing the relative merits of the two taxonomies, Searle first-discusses twelve dimensions in which illocutionary acts may differ from one another. Among the most significant of these dimensions are ”differences in the point (or purpose) of the (type of) act” (”ACoIA" 2), which Searle labels as differences in the illocutionary point (”ACoIA” 3); "differences in expressed psychological states" ("ACoIA” 4), which characterize the "belief, desire, intention, regret or pleasure [expressed] in the performance of the speech act" ('ACoIA” 4); ”dif— ferences in the force or strength with which the il-‘ locutionary point is presented” (”ACoIA" 5), which identify the varying degrees of commitment of a speaker to il- locutionary points; and ”differences in relation to the rest of the discourse" (”ACoIA” 5), which relate some performative expressions to the rest of the discourse in particular ways. This last dimension is of particular interest to metadiscourse theorists, because it isolates as a discrete class of illocutionary acts those forms that serve to relate utterances to other utterances in an ongoing discourse and to the context in which the discourse takes place. Searle discusses several examples of expressions that serve these functions: Consider, for example 'I reply', 'I deduce', 'I conclude', and 'I object'. These ex- pressions serve to relate utterances to other utterances and to the surrounding context. The features they mark seem mostly to involve ut- terances within the class of statements. In addition to simply stating a proposition, one may state it by way of objecting to what some— one else has said, by way of replying to an 107 earlier point, by way of deducing it from certain evidentiary premises, etc. 'How- ever', 'moreover' and 'therefore' also perform these discourse—relating functions. (”ACoIA' 5) In suggesting that single—word connectives serve discourse— relating functions, Searle implies that these words con- stitute a type of illocutionary force indicators; however, he does not comment concerning whether these forms are expandable into fully explicit performative clauses. Nevertheless, Searle's discussion of discourse—relating illocutionary acts establishes a precedent for considering these acts as a discrete category, and these acts will be central to the model of metadiscourse that I will offer. In criticizing Austin's taxonomy of illocutionary acts, the most important objection that Searle offers is that “there is no clear or consistent principle or set of principles on the basis of which the taxonomy is con— structed" ('ACoIA' 8). Searle notes that only commissives are defined unambiguously with illocutionary point serving as the basis of the definition, while expositives ”seem to be defined in terms of discourse relations” (”ACoIA" 8), and exercitives ”in terms of the exercise of authority“ ("ACoIA' 8). He considers behabitives the least adequately defined of Austin's categories, and he notes that this group "seems to involve notions of what is good or bad for the speaker and hearer . . . as well as expressions of at- titudes“ ("ACoIA" 8). Because Austin'c classification system uses no clear principle, several problems arise. Searle observes that "there is a great deal of overlap 108 from one category to another and a great deal of hetero- geneity within some of the categories" (”ACoIA" 8). As examples of the first problem, he points out that ”affirm,” ”deny,“ "identify,” and several other verbs that Austin classifies as expositives could also be considered as ver- dictives, and that ”the few cases which are clearly not verdictives are cases where the meaning of the verb has purely to do with discourse relations . . . or where there is no question of evidence or reasons“ ("ACoIA" 9). As examples of the second problem, Searle observes that "Austin lists 'dare', 'defy', and 'challenge', alongside 'thank', 'apologize', 'deplore' and 'welcome' as be? habitives” ("ACoIA" 9), despite the fact that these verbs seem to constitute two groups with distinct properties. As an alternative to Austin's taxonomy, Searle pro- poses five categories of illocutionary acts: (1) repre- sentatives, whose point is "to commit the speaker (in varying degrees) to something's being the case, to the truth of the expressed proposition" ("ACoIA" 10) and which includes ”most of Austin's expositives and many of his ver— dictives as well” ("ACoIA" 11), (2) directives, which are "attempts . . . by the speaker to get the bearer to do something” (”ACoIA” 11) and which includes most of Austin's behabitives and many of his exercitives, (3) commissives, which ”are those illocutionary acts whose point is to com- mit the speaker (again in varying degrees) to some future course of action" (”ACoIA" 11), (4) expressives, whose point is to express a "psychological state . . . about a 109 state of affairs specified in the propositional content” (“ACoIA” 12), including acts such as thanking, apologizing, and congratulating, and (5) declarations,which bring about ”the correspondence between the propositional content and reality” ('ACoIA" 13), including such acts as appointing, firing, and marrying. Searle concludes by noting that his taxonomy is a classification of illocutionary §E£§ and not of illocutionary verbs, since the same verb may be used to convey more than one type of act. For example, the verb ”advise" can function as either a directive or a repre- sentative, as is evidenced by the sentences "I advise you to leave” and "Passengers are hereby advised that the train will be late” (”ACoIA" 22). In the 1985 book, Foundations of Illocutionary Logic, Searle and Daniel Vanderveken revise the taxonomy that Searle introduces in ”A Classification of Illocutionary Acts.” In this later work, Searle and Vanderveken present a taxonomy of illocutionary points as opposed to a taxonomy of illocutionary EEEEv since all of Searle's earlier "act" categories are based on the point or purpose that the acts serve. Beyond this change in terminology, Searle and Vander— veken retain most of Searle's earlier work, with the only other significant change being the substitution of the category ”assertive point" for representatives. In intro- ducing this new term, they note that ”with the assertive point the speaker presents a proposition as representing an actual state of affairs in the world of utterance” (FoIL 37), a definition that is consistent with Searle's 110 earlier definition of representatives. All of Searle and Vanderveken's other categories-—commissive point, directive point, declarative point, and expressive point--are defined in the same terms that are used in Searle's earlier taxonomy of illocutionary acts. Toward a Speech Act Theory of Metadiscourse Although Austin and Searle disagree concerning the appropriate categories for classifying illocutionary acts, they are in accord on a matter that is of crucial importance to metadiscourse theorists: both believe that discourse- relating structures are an identifiable subset of 11— locutionary acts. Whether these structures are classified as expositives, representatives, or assertives is an issue of relative insignificance; the important point is that these structures constitute a class of entities that serve a common function--the clarifying of communicative discourse.} In fact, it would be reasonable to use the term ”meta- discourse” as a label for this class of illocutionary acts, because the function that Austin and Searle identify for these structures is also a central element in definitions of metadiscourse as "discourse about discoursing” (Williams, Style 81), discourse that ”calls attention to the com- municative speech act itself" (Crismore, TCfaRP 11), and discourse that “calls attention to the speech act itself, often marking stages in the development of the primary discourse" (Vande Kopple, "EEfFSP" 51). It appears, then, that metadiscourse can be considered as a discrete category I / 111 of illocutionary acts within the context of speech act theory. That such an analysis is not only reasonable but also advantageous for metadiscourse theorists is evidenced by the precision that characterizes Austin and Searle's definitions and analyses of data—-precision that is con— spicuously absent from existing theories of metadiscourse. In their definitions of basic terms and their structural analyses of the variant forms for conveying illocutionary intent, Austin and Searle's works are far superior to comparable works by metadiscourse theorists. This suggests that metadiscourse theorists would benefit from using Austin and Searle's terminology and analyses insofar as these are appropriate for the goals that metadiscourse theorists pursue. In the chapter that follows, I will offer a speech act theory of metadiscourse. In so doing, I will make extensive use of the terminology and concepts discussed in this chapter. CHAPTER FOUR A SPEECH ACT THEORY OF METADISCOURSE In this chapter, I will offer a theory of metadiscourse that lodges the term within the context of speech act theory. In so doing, I first will advance a functional definition of metadiscourse in terms of the illocutionary acts that metadiscourse identifies. I then will provide a taxonomy of the primary categories of metadiscourse, and I will identify canonical forms that use explicit performative structures for each category. After identifying the canonical forms, I will discuss the partially explicit forms that are possible for each category. I then will consider some secondary categories of metadiscourse, and I will identify canonical and partially explicit forms for these categories. I will conclude by analyzing my own theory in light of the four criteria that I provided in Chapter Two. A Speech Act Definition of Metadiscourse As a basis for the theory that I will advance, I offer the following definition of metadiscourse: Metadiscourse: illocutionary force indicators that identify expositive illocutionary acts. This definition uses Searle's distinction between 112 113 illocutionary force indicators that show how a proposition is to be taken and propositional indicators that convey statements with truth properties (EN 30). It also uses Austin's category of expositive illocutionary acts that clarify communication (EEEE 163), and together these terms specify that metadiscourse consists of structures that identify the communicative functions served by pro— positions in a passage of discourse. In choosing the term "expositives” to characterize the illocutionary acts that metadiscourse identifies, I am aware of Searle's objections concerning Austin's classification of expositive verbs; how- ever, I believe that Austin's term best characterizes the communicative functions that these illocutionary acts serve, and I will eliminate the problem of overlapping categories by offering precise definitions of the various categories of expositive illocutionary acts. A Taxonomy of Primary Expositive Illocutionary Acts In identifying the various categories of expositive illocutionary acts, I first will consider those acts that I classify as primary. The characteristic that distinguishes primary acts from secondary acts is that primary acts can be expressed in canonical form with explicit performative structures that use first person subject pronouns, while secondary acts use either second or third person subject pronouns. I shall call the most basic primary expositive il— locutionary act the simple expositive act, the function of 114 which is to state. In explicit performative form, this act is performed by clauses such as "I state,” ”I note," and ”I assert,” all of which serve the same essential function. These clauses can be diagramed in the fol— lowing manner: illocutionary force indicator attributive subject illocutionary predicate lst pprson . expositive I state note assert Figure 1: Simple Expositive Performative Clauses In classifying illocutionary force indicators, I take the term "explicit” to mean that the indicator identifies both the illocutionary act that is performed and the person who performs it. Although the canonical form of a simple expositive act is a performative clause with a simple pre- sent verb, other explicit forms are possible; for example, clauses such as ”I must note" and "I should state" are explicit indicators of simple expositives, despite the fact that they contain verb phrases with modal auxiliaries, instead of simply containing main verbs in the present tense. In addition to explicit illocutionary force indicators, partially explicit indicators also are possible; these forms identify an illocutionary act without referring to the per- son who performs the act. In the case of simple expositive acts, the partially explicit forms include clauses such as 115 ”It is notable" and adverbs such as ”notably," as well as other illocutionary words, phrases, and clauses that do not identify an attributive subject. Under the heading of ”primary expositive illocutionary acts," I also would include several complex expositive acts; these forms differ from the simple expositive act in that they convey features of the expositive act that supplement the basic information that a speaker/writer is stating something. One category of complex acts is the relational expositive act, which identifies sequential and causal links among passages of propositional discourse. The canonical form of the relational expositive act is an explicit per- formative clause that also contains an adverbial to identify the relationship of propositional passages; this includes clauses such as ”I first state," "I also note," and ”I therefore assert.” The following diagram illustrates these relational clauses: illocutionary force indicator attributive subject illocutionary predicate lst person adverbial expositive I se uential state 49 note first assert also causal therefore Figure 2: Relational Expositive Performative Clauses 116 Explicit forms of relational expositive indicators that do not follow the performative formula in their verb forms include.clauses such as ”I first would state,” ”I also should note,” and "I therefore must assert,” all of which contain modal verbs. Another explicit form uses a first person possessive pronoun, as in the clause ”My first point is.” Of the partially explicit forms} one of the most common is the use of only the adverbial to introduce the propositional indicator, as is exemplified by ”First, she does not like lobster," "ngp, it rained all day," and ”Therefore, we shouldn't go fishing." These indicators contain no mention of the attributive subject, and the illocutionary act itself is identified only indirectly by the adverbials. However, the reader/listener can infer both the speaker/writer and the nature of the illocutionary act from the adverbials. In the following diagram, implied constituents of the illocutionary indicator are enclosed in parentheses: illocutionary force indicator attributive subject illocutionary predicate lst person adverbial expositive (I) (state) sequential first also causal therefore Figure 3: Partially Explicit Relational Adverbs 117 A second type of complex expositive act is the evaluative expositive act; this category includes those illocutionary acts that indicate the speaker/writer's assessment of propositional material. In addition to performing acts of stating, evaluative expositive acts express the speaker/writer's attitudes concerning the validity of propositions, and they also may express other judgmental reactions to propositions. In fully expanded form, validity is indicated by embedded explicit per— formatives such as "I state that I believe” and ”I state that I doubt”; however, these often are reduced to simple explicit performatives such as "I believe" and "I doubt.“ The fully expanded form for indicating other reactions is exemplified by embedded explicit performatives such as "I state that I like" and "I state that I dislike"; how- ever, these often are reduced to ”I like” and "I dislike.” The following diagram illustrates the reduced forms of evaluative expositives: illocutionary force indicator attributive subject illocutionary predicate lst person evaluative expositive I validity believe doubt reaction like dislike Figure 4: Evaluative Expositive Performative Clauses 118 Evaluative expositive acts also can be indicated by explicit forms that do not use the full performative for— mula; for indicating validity, examples of these forms include near-performatives such as ”I am convinced" and "I have no doubts,” as well as possessive pronominal forms such as “My belief is,” while examples that indicate other reactions include "I am surprised," "I am amused," and ”I am concerned.” However, the range of evaluative expositive acts is better suggested by the partially explicit forms; examples of these for indicating validity include clauses such as "It is certain,” "It is possible,” and “It is doubtful,” phrases such as ”in fact" and ”without a doubt," and adverbs such as "certainly,” ”possibly," and "arguably," as well as modal verbs such as ”must," ”may," and ”might." Examples of partially explicit forms that indicate other I reactions include clauses such as "It is surprising" and "It is confusing” and adverbs such as "amusingly" and ”disturbingly.” None of these examples indicates the person who is performing the evaluative act, and the examples also do not indicate that the evaluative act is contained in an act of stating. In the cases of the adverbs and modal verbs, the illocutionary indicator often is embedded within a propositional indicator, as in the examples "They will certainly win" and "John may win the match"; however, it is possible to construct explicit embedded clauses that correspond to these forms, as is exemplified by "I state that I am certain that they will win" and "I state that it may be possible that John wins the match.” 119 As is the case for simple expositive acts, evaluative expositives also can appear in relational forms, as in the examples ”I also dislike” and "I therefore believe.” The numerical sequencers in evaluative indicators often take the forms exemplified by ”I first state that I doubt” and ”First, I believe.” The partially explicit forms of the relational evaluative act include examples such as "It therefore is doubtful," "It also is possible,” and ”First, it is likely.“ Another type of complex expositive is the commissive expositive act, which indicates that the speaker/writer is committed to performing a specific expositive act concerning specific propositional material in the discourse or text that follows the commissive act. For example, a speaker may indicate that she will state the causes of the Civil War. In fully expanded form, this commissive act has the structure ”I state that I will state"; however, the canonical form of the commissive act is that of a near- performative with a verb possessing a future aspect, such as ”I will state” or ”I will note.” It also is a characteristic of this form to introduce a simple indication of reference, instead of introducing a full propositional indicator. For example, ”the causes of the Civil War' merely indicates a specific group of referents; it does not advance a proposition that possesses truth properties. The following diagram illustrates the canonical structure of commissive expositive acts: 120 illocutionary force indicator attributive subject illocutionary predicate lst person commissive expositive I will state will note Figure 5: Canonical Commissive Expositive Clauses Also possible are relational commissive expositive EEEEI which not only commit the speaker/writer to performing a specific expositive act concerning specific propositional material, but which also link the act and material to other passages of the discourse or text. Canonical examples of relational commissive expositive acts include ”I first will state,” “I next will note," and ”I therefore will consider," all of which are illustrated by the following diagram: illocutionary force indicator attributive subject illocutionary predicate lst person adverbial commissive expositive I . will state sequent1al will note . will consider f1rst also causal therefore Figure 6: Canonical Relational Commissive Clauses Another explicit form for indicating a relational com- missive act is the possessive pronominal clause, exemplified by ”My first subject will be" and "My next topic is,“ both 121 of which indicate content that will follow, despite the present aspect of the verb in the latter example. Partially explicit forms include agentless passives such as "The next topic to be considered is” and clauses such as ”The next subject will be,” as_well as use of only the adverbial, as exemplified by ”First, the causes of this growing problem.” Because the illocutionary force indicator and the indication of reference in the last example do not combine to form a complete sentence, the structure is somewhat more common in speaking than in writing. Yet another type of primary complex expositive is the reiterative expositive act, which restates an expositive act and its corresponding propositional or referential material. The canonical form of the reiterative act is a near—performative clause with a verb possessing a past aspect, as in the examples ”I stated that the Red Sox won the pennant in 1975” and "I have noted the causes of the Civil War.” The following diagram illustrates the canonical form of the reiterative expositive: illocutionary force indicator attributive subject illocutionary predicate lst person reiterative expositive I stated have noted Figure 7: Canonical Reiterative Expositive Clauses As is the case for the other primary expositive acts, the reiterative expositive can appear in relational forms, 122 as in examples such as "I first stated,” "I also have noted," and ”I therefore asserted." Partially explicit relational reiteratives include forms such as "Having first considered” and "Also having noted.” The primary expositive illocutionary act categories constitute a taxonomy of the metadiscursive acts of stating that can be performed directly by a speaker/writer. The specific examples listed for each type of act are merely representative rather than exhaustive; numerous other indicators that identify each act could be cited. However, the definitions that I have provided for the acts are sufficient to enable theorists to identify other instances of each act. In addition to expositive acts that are performed directly by a Speaker/writer, it also is possible for the speaker/writer to perform indirect acts of stating that are attributed to other sources. These acts form the taxonomy of secondary expositive illocutionary acts that I will discuss below. A Taxonomy of Secondary Expositive Illocutionary Acts The main characteristic that distinguishes secondary expositive illocutionary acts from primary acts is that the secondary acts do not use or imply first person subjects; instead, they attribute the act of stating to someone other than the speaker/writer of the discourse or text. These secondary acts divide into two broad classes: those that use third person attributive subjects and those that attribute the illocutionary act to the listener/reader by 123 using second person subjects. Third person illocutionary acts are mentioned briefly by Jerrold M. Sadock in Toward a Lingpistic Theory of Speech NEEE; he uses the term “covert illocutionary acts" to label them, and he provides examples that include “Officer O'Brien warned us that there were several bridges out" and "My wife told me that the dog was barking" (44). Unlike first per- son speech acts, third person acts do possess truth pro— perties; that is, in addition to performing an illocutionary act, the illocutionary indicator is itself part of a pro— position that may be true or false. This property sug- gests that third person expositive acts may be difficult to distinguish from the propositional material that con— stitutes the primary information of a discourse; however, this problem is itself a distinguishing characteristic of third person acts. In third person form, simple expositive acts are typified by illocutionary indicators such as ”She states,” ”He notes," and “Smith asserts," which are illustrated by the following diagram: illocutionary force indicator attributive subject illocutionary predicate 3rd person expositive She states He notes Smith asserts Figure 8: Third Person Simple Expositive Clauses Because third person expositives refer to statements made 124 previously by a person to whom the speaker/writer refers, the third person act is as likely to use past tense verbs as present tense verbs; typical third person expositives include examples such as ”She stated," ”He noted,” and ”Smith has observed." Although these forms appear similar to the reiterative expositives discussed earlier, third person simple expositives do not refer to illocutionary acts and propositions made previously lg the discourse or text and py the speaker/writer; instead, they indicate the first appearance in the discourse or text of acts and pro- positions taken from a previous communicative instance. Unlike first person expositive acts, third person expositives generally do not appear in partially explicit forms. Third person expositives are by definition at— tributed to a source other than the speaker/writer, so the use of a form that does not identify an attributive subject usually is inappropriate for conveying these acts; in most cases, a listener/reader will infer the existence of a first person subject if no attributive subject is specified. However, third person acts occasionally do appear in agentless passive forms such as "It is stated" and “It has been noted”; these forms can cause inter— pretative problems for the listener/reader unless the attributive subject is recoverable from extra-textual aspects of the communicative situation. As is the case for first person expositive acts, third person expositives can appear as complex acts that convey features supplementing the basic information that 125 someone is stating something. One complex type of ex~ positive act is the third person relational expositive, which includes forms such as “She first states," ”He also notes,” and "Jones therefore asserts,” as well as past tense forms such as "She first stated,” "He next observed," and ”Smith consequently noted.” Like the third person simple expositive indicators, the third person relational expositives usually do not appear in partially explicit forms. The third person evaluative expositive act is a category that contains these acts that reveal someone_ other than the speaker/writer's assessment of propositional material. As is the case for first person evaluatives, the third person forms can indicate either attitudes concerning the validity of propositions or expressions of other judg— mental reactions to propositions. In fully expanded form, validity is indicated by embedded explicit clauses such as ”I state that she states that she believes” and ”I state that he states that he doubts”; however, these often are reduced to canonical forms such as ”She believes” and ”He doubts." The fully expanded form for indicating other reactions is exemplified by embedded clauses such as ”I state that Smith states that he likes" and ”I state that Jones states that she dislikes"; however, these often are reduced to ”Smith likes” and "Jones dislikes.” The fol— lowing diagram illustrates the reduced forms of explicit third person evaluative expositives: 126 illocutionary force indicator attributive subject illocutionary predicate 3rd person evaluative expositive She validity He - Smith believes Jones deubts reaction likes diSlikes Figure 9: Third Person Evaluative Clauses Like third person simple expositives, the third person evaluatives usually do not appear in partially explicit forms. In most cases, adverbs and medals that are embedded within a propositional indicator are taken to be expressions of the speaker/writer's attitudes. However, agentless passives such as "It is believed” and ”It is considered unfortunate” occasionally can be interpreted as implying third person subjects. Third person evaluatives can appear in relational forms, as in the examples "She also believes" and ”He therefore dislikes.” As is the case for other third person acts, partially explicit forms of third person relational evaluatives are relatively uncommon. The next category, third person commissive expositives, poses a unique problem, since a cursory analysis would sug— gest that it is impossible for a speaker/writer to commit another person to performing a specific expositive act concerning specific content that will appear following 127 the commissive indicator. However, the.frequent occurrence of forms such as "She will state," ”He will note," and ”Smith will consider” suggests that thirdeerson commissives are possible. In analyzing these acts, it is helpful to consider that the eXpanded forms corresponding to the examples above are the embedded clause structures "I state that she will state," ”I state that he will note," and "I state that Smith will consider.” Despite the future aspect of the reduced indicators, they identify illocutionary acts that were performed prior to the discourse or text in which they are being restated by the speaker/writer. In fact, these forms do not commit a third person subject to per— forming a particular illocutionary act; instead, they commit a speaker/writer to including in the discourse or text an act that already has been performed elsewhere. Since the speaker/writer is the real architect of the discourse or text, the onus is on him or her to include the appropriate material following the commissive indicator. The following diagram illustrates the reduced forms of the third person commissives mentioned above: illocutionary force indicator attributive subject illocutionary predicate 3rd person commissive expositive She will state He will note Smith will consider Figure 10: Third Person Commissive Clauses 128 As is the case for other third person expositive acts, third person commissives usually do not appear in partially explicit forms unless the attributive subject is recoverable either from prior textual references or from extra—textual aspects of the communicative situation. Third person relational commissives express the entire range of sequential and causal relations that are indicated by their first person counterparts. Examples of third per- son relational commissive forms include "She first will state," "He also will note,” and ”Smith therefore will consider,“ all of which are illustrated by the following diagram: illocutionary force indicator attributive subject illocutionary predicate 3rd person adverbial commissive expositive She se uential will state He q will note Smith . will consider f1rst also causal therefore Figure 11: Third Person Relational Commissive Clauses As is the rule for the other third person acts, the third person relational commissives normally do not appear in partially explicit forms. Also possible are third person reiterative expositives, which restate expositive acts and their corresponding 129 propositional or referential information. In a general sense of the word, all third person expositive acts are reiterative, because they all identify illocutionary acts that first were performed prior to the discourse or text in which they are restated. However, the narrower use of the word, the use which I favor for this model, would limit it to those expositive acts that are restatements of acts and their associated propositional or referential content that already have appeared lg the discourse or text under consideration.I The canonical forms of third person reiteratives include "She stated," ”He noted,” and.'Smith has observed,” all of which contain verbs with a past aspect. Although I already have identified these clauses as examples of third person simple expositives, their proper classification can be determined by the absence or presence of an earlier occurrence of the same act and propositional or referential information in the discourse or text; such an occurrence indicates that the clause is a simple ex— positive, while the absence of such an occurrence indicates that the clause is reiterative. The following diagram illustrates the canonical forms of third person reiteratives: illocutionary force indicator attributive subject illocutionary predicate 3rd person reiterative expositive She stated He noted Smith has observed Figure 12: Third Person Reiterative Clauses 130 Third person reiteratives also can appear in relational forms, examples of which include ”She first stated," ”He next observed,” and "Smith consequently noted.5 As is the case for other third person reiteratives, the third person relational reiterative forms can be distinguished from their simple expositive counterparts by searchinngor prior occurrences in the discourse or text of the same acts and propositional or referential content. The second broad class of secondary expositive il— locutionary acts constitutes another means by which a speaker/writer can perform acts of stating that areI attributed to another source. With this class of acts, the speaker/writer attributes statements to a listener/ reader through the use of second person attributive sub- jects. Second person expositive acts are similar to first person and third person expositives in many regards, but additional rules apply to some of the second person categories. In second person form, simple expositive acts may be identified by illocutionary indicators such as ”You note” and ”You assert"; however, the second person forms often use modal verbs to weaken the force with which an act is predicated of a listener/hearer, resulting in indicators such as "You may note" and ”You may assert.” I take these weaker forms to be the canonical type of second person simple eXpositives. The following diagram illustrates the modal forms of the above examples: 131 illocutionary force indicator attributive subject illocutionary predicate 2nd person I expositive You may note may assert Figure 13: Second Person Simple Expositive Clauses In partially explicit form, second person simple ex— positives often appear as imperative clauses such as "Note that” and ”Consider that." Although these clauses do not explicitly identify an attributive subject, the imperative form allows the listener/reader to infer that ”you” is the intended subject of the above illocutionary indicators. The following diagram of the partially explicit second person expositives cited above includes ”you" in parentheses to indicate that it is an implied constituent of the in— dicators: illocutionary force indicator attributive subject illocutionary predicate 2nd person expositive (You) Note Consider Figure 14: Imperative Second Person Expositives As is the case for first person and third person acts, second person expositives can appear as complex acts that convey features supplementing the basic act of stating. One of these complex types is the second person relational expositive act, which includes forms such as ”You first may 132 note,” "You also may consider," and ”you therefore may assert.” These forms are illustrated below: illocutionary force indicator attributive subject illocutionary predicate 2nd person adverbial expositive You sequential may note may consider first may assert also causal therefore Figure 15: Second Person Relational Expositives In partially explicit form, second person relational ex- positives include examples such as "First note” and “Also consider.” Causal indicators and strong expositives such as "assert" normally do not appear in partially explicit re— lational forms of second person expositives. Another type of complex expositive act is the second person evaluative expositive, which predicates the listener/ hearer's assessment of propositional material. As is the case for other evaluative acts, the second person forms can indicate either attitudes concerning the validity of propositions or expressions of other judgmental reactions to propositions. In expanded form, validity is indicated by embedded clauses such as ”You may state that you believe” or “You may state that you doubt”; however, these often are reduced to canonical forms such as "You may believe" 133 and "You may doubt.” The expanded form for indicating other reactions is exemplified by embedded clauses such as “You may state that you like" and ”You may state that you dislike”; however, these often are reduced to ”You may like" and "You may dislike.” The reduced examples cited above are illustrated by the following diagram: illocutionary force indicator attributive subject illocutionary predicate 2nd person evaluative expositive You' validity may believe may doubt‘ reaction may like may dislike Figure 16: Second Person Evaluative Clauses Unlike the second person simple expositives, the second person evaluatives usually do not appear in partially explicit imperative forms. While it occasionally may be appropriate to indicate an evaluative act with an imperative clause such as "Believe that," in most cases imperatives such as ”Doubt that," ”Like that,” and ”Dislike that” will appear inappropriate. Second person evaluatives can appear in relational forms, as in the examples ”You first may believe," ”You also may doubt," and ”You therefore may like.” As is the case for other second person evaluatives, partially explicit second person evaluatives generally are inappropriate. 134 Second person commissives are similar to their third person counterparts in that they really do not commit anyone other than the speaker/writer toperforming a particular illocutionary act involving specific content. Consider the example "You will note the causes of the Civil War. These include . . ." This commissive act does not mandate that the listener/reader will identify the causes of the Civil War; instead, it calls attention to the statement of causes that the speaker/writer will provide. Canonical second person commissives use modal verbs, as in the examples ”You will note” and ”You must consider." These examples are illustrated by the fol- lowing diagram: illocutionary force indicator attributive subject illocutionary predicate 2nd person commissive expositive You will note must consider Figure 17: Second Person Commissive Clauses As is the case for first person and third person commissives, the second person indicators introduce a simple indication of reference instead of introducing a full propositional indicator. This distinction is useful for distinguishing partially explicit imperative second person commissives from their simple expositive counter- parts, as is the fact that imperative second person simple expositives often are followed by a "that" to indicate 135 that a sentential complement follows the illocutionary force indicator, while imperative second person commissives are not. For example, the sentence "Consider that the Red Sox have lost fifty games" contains a partially explicit second person simple expositive indicator followed by a full propositional indicator, while the sentence ”Consider the causes of the Civil War" contains a partially explicit second person commissive indicator followed by referential content. In relational forms, explicit second person commissives are indicated by clauses such as ”You first will note," ”You also must consider," and "You therefore will note." Partially explicit forms include "First note,” "Next con- sider," and "Therefore note.” Second person reiteratives perform illocutionary acts that are restatements of acts and their associated pro— positional or referential content that already have appeared in the discourse or text under consideration. In performing second person reiterative acts, it is not necessary for the act and content that are repeated to have been attributed originally to the listener/reader; it is possible to have a second person reiteration of an act originally attributed to a first person or a third person subject. For example, an act that originally appears in a form such as ”I note that the Red Sox have lost fifty games" or ”Smith notes that the Red Sox have lost fifty games" can be reiterated as ”You noted that the Red Sox have lost fifty games.“ Canonical second person reiteratives include clauses such 136 as ”You noted" and "You have observed," both of which are illustrated by the following diagram: illocutionary force indicator attributive subject illocutionary predicate 2nd person reiterative expositive You noted have observed Figure 18: Second Person Reiterative Clauses In relational forms, explicit second person reiteratives are indicated by clauses such as ”You first noted,” ”You also observed,; and "You therefore have noted."' While partially explicit forms of nonrelational second person reiteratives are uncommon, partially explicit imperative relational forms such as "Note again" sometimes do appear in discourses or texts. As is the case for the examples of primary expositive illocutionary acts that I have cited, the examples of secondary acts are representative rather than exhaustive. However, the examples and definitions that I have provided are sufficient to enable theorists to identify other in— stances of secondary expositive illocutionary acts. An Evaluation of the Proposed Metadiscourse Model Just as I used the four criteria that I introduced in Chapter Two to evaluate the existing models of meta- discourse, it also is possible to use those criteria to evaluate my own model. Such an evaluation suggests that 137 several advantages result from considering metadiscourse within the context of speech act theory. The first criterion requires that definitions of key terms in a theoretical model be sufficiently precise to enable theorists to agree concerning the specific entities that the terms describe. By defining metadiscourse as those illocutionary force indicators that identify ex— positive illocutionary acts, I am able to utilize two specific distinctions drawn by speech act theorists: the distinction between structures that identify illocutionary acts that possess "happiness” properties and structures that identify propositional material containing truth properties, and the distinction between illocutionary acts that identify the communicative functions served by propositional or referential material and illocutionary acts that serve other functions. Although these two distinctions are only loosely applicable to several categories of secondary ex- positive illocutionary acts, the definition is sufficiently precise to establish metadiscourse as a functional category containing discrete entities. By further specifying that all metadiscursive entities can be analyzed as canonical performative or near—performatives structures and their partially explicit variants, I establish unambiguous guidelines for identifying instances of metadiscourse. In this regard, my speech act model is far superior to other attempts to define metadiscourse. The second criterion requires that each of the sub— categories in a model must possess properties sufficiently 138 similar to the properties of the other sub—categories so that useful generalizations can be offered concerning metadiscourse as a collective entity. Although the primary and secondary categories that I have described differ some- what in their specific functions and forms, all of them are united by a common general function and form. 'While the function of a metadiscursive act can be categorized as simple expositive, evaluative, commissive, or reiterative, with relational forms possible for each of these sub- categories, all of these sub—categories serve a common communicative function——the clarifying of a speaker/writer's intentions in producing passages of a discourse or text. And while the person of an illocutionary subject or the specific form of a verb may vary from one sub—category to another, the constituents of every sub—category can occur in common forms that I have termed ”canonical"—-the per- formative and near-performative clauses. That these two generalizations concerning the function and form of the sub-categories are possible suggests that my metadiscourse model conforms to the second criterion for good design. The third criterion addresses the question of whether any purpose is served by grouping constituents under a sub-category heading, and it specifies that theorists must be able to advance generalizations that are true for all of the constituents in a sub—category. My model divides metadiscourse along several lines: it separates primary acts using first person attributive subjects from secondary acts using third person and second person 139 subjects, and it separates expositive acts into the various simple and complex sub-categories that I have noted above. The division of metadiscourse into sub-categories based upon the person of the attributive subject is necessary in order to characterize the significant differences that exist among forms that use different subjects. As.I observed earlier, the fact that all forms of metadiscourse share a general function and certain formal properties does not mean that the forms are similar in all regards. First per- son acts allow single adverbs as partially explicit il- locutionary force indicators, while third person and second person acts do not. Furthermore, second and third person illocutionary force indicators can serve as part of a larger proposition that possesses truth properties, while first person indicators cannot. And first person and third person forms can use certain illocutionary predicates that are not appropriate with second person subjects. These differing characteristics constitute generalizations that are true for certain sub-categories of metadiscourse but not true for metadiscourse as a collective entity, thus justifying the division of metadiscourse into primary and secondary sub-categories. Concerning the distinctions among the simple and complex sub—categories, the definition of each complex sub—category specifies the properties that distinguish it from other simple and complex sub—categories; for example, evaluative acts contain an expression of at- titude that other acts do not, and commissive acts dictate the content that follows them in a manner that other acts 140 do not. These distinguishing properties justify the division of metadiscourse into several sub-categories that identify different types of expositive acts. The fourth criterion requires that the terms of analysis used in a model be appropriate to reveal im- portant properties of the phenomenon under consideration. In constructing my model, I have been mindful of Searle's assertion that speech acts ”are the basic or minimal units of linguistic communication" (EN 16), an assertion that implies the necessity of using functional terms in order to characterize the communicative purposes that speech acts serve. Because my model treats metadiscourse as One com- ponent of a larger theory of speech acts, my definition of metadiscourse and taxonomy of expositive sub—categories use many of the terms of analysis that Austin and Searle provide in their studies. By so doing, I avail myself of an existing vocabulary that reveals the precise functions that metadiscourse can serve. However, I also recognize that a useful account of metadiscourse cannot be specified in functional terms alone; it also is necessary to identify the various syntactic forms used to convey expositive functions. To accomplish this, I again borrow terminology from Austin and Searle, and I offer a detailed account of the explicit and partially explicit structures that can express each expositive act. By offering detailed analyses of not only the functions but also the forms of meta- discourse, my model conforms to the fourth criterion for good design. 141 Because my speech act model of metadiscourse adheres to all four of the design criteria I provided in Chapter Two, it stands as-a precise and useful alternative to other existing metadiscourse models, none of which con- form to all of the criteria. In offering my model, it is my hope that other researchers will find it valuable in conducting experimental studies of metadiscourse, studies which until now have been hampered by inadequate analyses of the phenomenon being considered. In the final pages of this dissertation, I will con— sider several of the research questions that my model could be used to explore. CONCLUSION In this dissertation, I have attempted to provide a clear and precise analysis of metadiscourse within the con- text of speech act theory. By considering metadiscourse as a carefully delimited component of a theory that has a long history of productive scholarship, I believe that we may resolve the basic problems of defining and classifying metadiscursive structures, thus allowing us to proceed with the important work of determining the role that meta- discourse plays in text comprehension. However, the terms of analysis that we employ in our models of metadiscourse have a significance that extends beyond their ability to isolate discrete structures. In fact, the terms themselves dictate to a considerable extent the research questions that we are able to consider. As Williams has noted, "If our theory includes the category 'word' . . . then we count words. . . . If the system includes the category 'clause,' we count clauses" ("N-LL atToS' 26). In light of the important role that theoretical entities play in determining research practices, I will conclude by considering some of the research questions that are suggested by my model. Although several of these questions address concerns that have been raised by other metadiscourse theorists, my speech act model provides 142 143 a new perspective on these familiar questions. As I noted above, the most important questions concerning metadiscourse address the role that it plays in a reader's processing of a text. The primary question is whether metadiscourse use has an effect on a reader's ability to comprehend and/or remember propositional material. While Williams suggests that overuse of meta— discourse can impair the readability of a text (Egylg 81). Crismore believes that judicious use of metadiscourse actually can enhance a text's readability and memorability (TCfaRP 9). These two positions are compatible and both seem reasonable, but little experimental evidence is available to support either of them. However, the meta— discourse model that I have provided should facilitate detailed study of these questions. Using my system for classifying metadiscourse types, researchers could consider whether any of the sub—categories cause a reader to process a text differently than a text with the same propositional content but no metadiscourse would be processed. Other questions to consider include the following: (1) whether primary forms of metadiscourse using first person subjects have a different effect on text processing than do secondary forms using third person or second person subjects, (2) whether relational and nonrelational forms have different effects, (3) whether a sub-category such as evaluatives affects processing differently than does a sub-category such as simple expositives, and (4) whether explicit forms of metadiscourse have a different effect on the processing 144 of propositional content than their partially explicit counterparts do. In addition to considering the effects of metadiscourse on the processing of propositional material, researchers also should consider whether the sub-categories of meta- discourse themselves differ in readability and memorability. Vande KOpple has observed that readers tend to recall less metadiscourse than primary discourse ('EEfFSP' 52), but he does not provide a detailed analysis of the relative readability and memorability of the various metadiscourse sub-categories. Using my taxonomy, researchers could consider the following questions: (1) whether primary types of metadiscourse are themselves more easily read and remembered than secondary types, (2) whether relational types are more or less readable and memorable than non- relational types, (3) whether a sub-category such as evaluatives is more or less readable and memorable than other sub-categories, and (4) whether explicit forms of metadiscourse are more or less readable and memorable than partially explicit forms. While a precise definition and a systematic taxonomy of metadiscourse do not assure that productive experimental research will follow, it is impossible to conceive of productive studies without the prior existence of such a definition and taxonomy. If my speech act theory of meta- discourse can provide the basic terminology needed to guide future research, then it will constitute a modest but significant contribution to the field of text analysis. BI BLIOGRAPHY BIBLIOGRAPHY Austin, J. L. How to Do Things with Words. Ed. J. O. Urmson and Marina Sbisa. 2nd ed. Cambridge: Harvard UP, 1981. -—-. Philosophical Papers. Ed. J. O. Urmson and G. J. Warnock. 2nd ed. Oxford: Clarendon, 1970. Beauvais, Paul J., and Frank Parker. Analogizing from Generative Linguistics to Composition. ERIC, 1984. ED 236 672. Chomsky, Noam. Syntactic Structures. The Hague: Mouton, 1957. Coulthard, Malcolm. An Introduction to Discourse Analysis. London: Longman, 1977. Crismore, Avon. The Case for a Rhetorical Perspective on Learning from Texts: Exploring Metadiscourse. Center for the Study of Reading, U of Illinois, 1985. Technical Report No. 342. --—. "A Message to Authors about Metadiscourse Use in Instructional Texts." Changing Perspectives on Research in Reading/Language Processing and Instruction. Ed. Jerome A. Niles. Thirty-third Yearbook of the National Reading Conference, 1984. 66—74. ---. ”Metadiscourse as Rhetorical Act in Social Studies Texts: Its Effect on Student Performance and Attitudes." D155. 0 of Illinois, 1984. -——. The Metadiscourse Component: Understanding Writing about Reading Directives. ERIC, 1982. ED 217 374. -——. Metadiscourse: What It Is and How It Is Used in School and Non-School Social Science Texts. Center for the Study of Reading, U of Illinois, 1983. Technical Report No. 273. —--. "The Rhetoric of Textbooks: Metadiscourse.” Journal of Curriculum Studies 16 (1984): 279-96. 145 146 ---. Suggestions for Improving a Poorly-Written Sixth Grade Social Studies Chapter. ERIC, 1982. ED 228 113. ---. ”Summary of 'Metadiscourse as RhetOrical Act in Social Studies Texts: Its Effect on Student Performance and Attitudes." Unpublished paper, 1985. Crystal, David. A First Dictionary of Linguistics and Phonetics. Boulder: Westview, 1980. van Dijk, Teun A. "Relevance Assignment in_Discourse Comprehension.” Discourse Processes 2 (1979): Donaldson, Fred. "Metacommunication in Rough and Tumble Play.”' Reading Improvement 13 (1976): 235-39. Grice, H. P. ”Logic and Conversation.” Syntax and Semantics, Vol. III: Speech Acts. Ed. Peter Cole and Jerry L. Morgan. New York, Academic Press, 1975. Halliday, M. A. K. Explorations in the Functions of Language. New York: Elsevier North-Holland, 1973. Halliday, M. A. K., and Ruqaiya Hasan. Cohesion in English. English Language Series 9. London: Longman, 1976. Harris, Zellig S. ”Algebraic Operations in Linguistic Structure.” Papers in Structural and Transformational Linguistics. Formal Linguistics Series 1. Dordrecht: D. Reidel, 1966/70. 603-11. ---. "Linguistic Transformations for Information Retrieval.“ Papers in Structural and Transformational Linguistics. Formal Linguistics Series 1. Dordrecht: 1959/70. 458-71. Lautamatti, Liisa. "Observations on the Development of the Topic in Simplified Discourse." Text-Linguistics, Cognitive Learning, and Language Teaching. Ed. Viljo Kohonen and Nils E. Enkvist. Turku: U of Turku, 1978. 71-104. Levinson, Stephen C. Pragmatics. Cambridge Textbooks in Linguistics Series. Cambridge: Cambridge UP, 1984. Meyer, Bonnie J. F. The Organization of Prose and Its Effects in Memory. Amsterdam: North-Holland, 1975. ---. “What Is Remembered from Prose: A Function of Passage Structure." Discourse Production and Comprehension. Ed. Roy 0. Freedle. Advances in Discourse Processes 147 Series 1. Norwood: Ablex, 1982. 307-36. Meyer, Bonnie J. F., David M. Brandt, and George J. Bluth. ”Use of Top-Level Structure in Text: Key for Reading Comprehension of Ninth Grade Students.” Reading Research Quarterly 16 (1980): 72-103. Quirk, Randolph, and Sidney Greenbaum. A COncise Grammar of Contemporary English. New York: Harcourt, 1975. Quirk, Randolph, Sidney Greenbaum, and Jan Svartvik. N Grammar of Contemporary English. London: Longman, 1980. Ragan, Sandra L., and Robert Hopper. ”Alignment Talk in the Job Interview.” Journal of Applied Communication 9 (1981):-85-103. Sadock, Jerrold M. Toward a Linguistic Theory of Speech Acts. New York: Academic Press, 1974. Schiffrin, Deborah. ”Meta-Talk: Organizational and Evaluative Brackets in Discourse.” Sociological Inquiry: Language and Social Interaction 50 (1980): 199-236. Searle, John R. "A Classification of Illocutionary Acts.“ Language in Society 5 (1976): 1-23. ---. Speech Acts: An Essay in the Philosophy of Language. Cambridge: Cambridge UP, 1984. Searle, John R., and Daniel Vanderveken. Foundations of Illocutionary Logic. Cambridge: Cambridge UP, 1985. Vande Kopple, William J. "Experimental Evidence for Functional Sentence Perspective.” Diss. U of Chicago, 1980. ---. "Some Exploratory Discourse on Metadiscourse.” College Composition and Communication 36 (1985): 82-93. Williams, Joseph M. "Literary Style: The Personal Voice." Style and Variables in English. Ed. Timothy Shopen and Joseph M. Williams. Cambridge: Winthrop, 1981. 117-216. ---. "Non-Linguistic Linguistics and the Teaching of Style.” Linguistics, Stylistics, and the Teaching of Composition. Ed. Donald McQuade. Akron: U of Akron, 1979. 24-40. ---. Style: Ten Lessons in Clarity and Grace. 2nd ed. 148 Glenview: Scott, Foresman, 1985.