A COMPUTER-BASED BARTHESIAN READER'S ASSISTANT

 XEBRA RIDES ROLAND BARTHES' TEXTUAL NETWORK OF SIGNIFIERS

DISSERTATION 

Presented in Partial Fulfillment of the Requirements for

the Degree Doctor of Philosophy in the Graduate

School of The Ohio State University

By

Bradley C. Watson, B.A., M.L.S., M.A., M.C.S.

* * * * *

The Ohio State University

1994

Permissions

Excerpts from S/Z by Roland Barthes and translated by Richard Miller. Translation copyright(C) 1974 by Farrar, Straus & Giroux, Inc. Reprinted by permisssion of Hill and Wang, a division of Farrar, Straus & Giroux, Inc.

To Alice

Alice A. Watson
Born: April 14, 1955
Died: November 3, 1996

What can I say?
That she was/is the finest human being I have ever had the honor to know?
That I loved her with all my heart and my soul?
That God was good to me to allow me to be a part of her life?
That a finer wife no man has ever had?

It's all true.

I miss her deeply.

That is truer still.

ACKNOWLEDGEMENTS

Acknowledging the contribution of the following individuals is admittedly insufficient reward for their support, but certainly it is one small part of the debt I owe them all. My thanks go first to my committee: Walter A. Davis (Mac), my offical advisor and unofficial encourager who wanted to see what computers could do to literary theory; Martin Dillon, my computer/literature interface expert who already had more than a notion in that matter; and Jessica Prinz, who saw the possibilities of the project early and stayed late to see the final outcomes. Indeed, all three stayed the course with me when I and this project needed them most. Next, many thanks must go to my fellow adventurer in literature, James A. Shanks, who dedicated considerable time to the reading of this paper, red ink pen capably in hand and used more than I care to admit. And finally, to Alice Watson, my wife and best friend, who made for me the time and space to work on this project, I give all my thanks and love.

VITA

February 14, 1950 . . . . . . . . Born - Dayton, OH.

1972 . . . . . . . . . . . . . . . . . . .B.A., University of Dayton, Dayton, OH.

1974 . . . . . . . . . . . . . . . . . . .M.L.S., George Peabody College For Teachers, Nashville, TN.

1975-1976 . . . . . . . . . . . . . .Assistant Reference Librarian, University of Dayton, Dayton, OH.

1977-1978 . . . . . . . . . . . . . . Motel Manager, L-K Restaurants and Inns, Inc., Marion, OH.

1979-1983 . . . . . . . . . . . . . . Programmer, Air Force Logistics Command, Wright-Patterson Air Force Base, Dayton, OH.

1980 . . . . . . . . . . . . . . . . . . .M.A., Wright State University, Dayton, OH.

1983 . . . . . . . . . . . . . . . . . . .M.C.S., University of Dayton, Dayton, OH.

1983-1984 . . . . . . . . . . . . . . Programmer/Analyst, Mead Data Central, Dayton, OH.

1984-1986 . . . . . . . . . . . . . . Senior Principle Programmer Analyst, NCR, Inc., Dayton, OH.

1986-1996 . . . . . . . . . . . . . . Research Scientist, OCLC Online Computer Library Center, Inc., Dublin, OH.

1996-Present . . . . . . . . . . . . .Consulting Systems Analyst, OCLC Online Computer Library Center, Inc., Dublin, OH.

PUBLICATIONS

 Watson, Bradley C., Terry Noreault, and Howard Turtle. "Designing a CD ROM Information Structure." Chap. in The CD ROM Handbook,. ed. Chris Sherman. New York: McGraw-Hill Book Company Inc. 1988.

________, and Jonathan Fausey. "Relative Performance of Three CD-ROM Network Access Products," OCLC Micro 5(4) (August, 1989): 20-21.

________. "Facsimile: As You Like It (Or: A 'Graphic' View of Telecommunications and the Library," Library Hi Tech News (November, 1989): 7-11.

________. "Sidebar: For Those Who Want to Start Small (An Overview of the LANtastic LAN)," Library Hi Tech News (April, 1990): 15.

________, and Jonathan Fausey. "Relative Performance of Two More CD-ROM Network Access Products," OCLC Micro 6(3) (June, 1990): 14-15. (With Jonathan Fausey).

 ________, and Robert Davis. "ODA and SGML: An Assessment of Co-Existence Possibilities," Computer Standards & Interfaces 11 (1990/91): 169-176.

________. "ODA? SGML? SGML? ODA?," in EDD '94 Recipes For Success: Bellcore/BCC Conference on Electronic Document Delivery Final Program & Conference Proceedings Held in Somerset, New Jersey, 11-12 May 1994, by Bellcore/BCC. Murray Hill, New Jersey: Bellcore, 1994, 96-102.

FIELDS OF STUDY

Major Field: English

Studies in Literary Theory, Walter A. Davis
20th Century British and American Literature, Jessica Prinz
Samuel Johnson, James Battersby

 TABLE OF CONTENTS

ACKNOWLEDGEMENTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .ii

VITA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .iii

LIST OF TABLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .x

LIST OF FIGURES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xii

INTRODUCTION An Hypotheses: Formulation and Inquiry . . . . . . .1

The Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
The Inquiry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
The Plan Of The Dissertation . . . . . . . . . . . . . . . . . . . . .7

Barthes' S/Z: A Reading System of Worth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .8
Barthes' Reading System: A Further Look . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .10
Barthes' Reading System: Against . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
Barthes' Reading System: For, Even So . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .29
Barthes' Reading System: An Enabling Strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
Barthes' Reading System: An Evolution Plan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .41
Barthes' S/Z: What Does It Do? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .46
The Project: Chapter by Chapter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .50

The Barthesian System: Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
The Barthesian System Processes: Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
The Primal Barthesian Act: Lexia Making . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
The Barthesian Labeling Language:

Signifier Identifying and Labeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
On Barthes' Voices: Truth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
On Barthes' Voices: The Person . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
On Barthes' Voices: Science . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
On Barthes' Voices: Symbol . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
On Barthes' Voices: The Empirics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
On Barthes' Voices: A Statistical Picture . . . . . . . . . . . . . . . . . . . . . . . . . . 80
On Barthes' Voices: A Summation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
The Barthesian Grammar: The Codes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
The Barthesian Labeling Language: An Example Application . . . . . . . . . . . .94
The Barthesian Labeling Language: A Closing Summary . . . . . . . . . . . . . . 100

The First, Middle, & Final Barthesian Act: Rereading . . . . . . . . . . . . . . . . . . . . . .102

The Barthesian System: Its Problematic Attributes . . . . . . . . . . . . . . . . . . . . . . . . 104

The Barthesian System: Its Problematic Theory . . . . . . . . . . . . . . . . . . . . 105
Barthes In Theory: Precision, Consistency, Completeness . . . . . . . . . . . . 105
The Barthesian System: Sufficient To Its Purpose? . . . . . . . . . . . . . . . . . .109
Barthes In Theory: Wholes and Parts . . . . . . . . . . . . . . .. . . . . . . . . . . . . 111
The Barthesian System: Its Problematic Practices . . . . . . . . . . . . . . . . . . .113
Barthesian System Practice Problematic: Barthes' View . . . . . . . . . . . . . . 114
Barthesian System Practice Problematic: The Effort . . . . . . . . . . . . . . . . . 115
Barthesian System Practice Problematic: On Completeness . . . . . . . . . . . .119
Barthesian System Practice Problematic: On Precision . . . . . . . . . . . . . . . 122
Barthesian System Practice Problematic: On Consistency . . . . . . . . . . . . . 124

The Barthesian System: Addressing Its Problems . . . . . . . . . . . . . . . . . . . . . . . . . 125

Addressing Barthesian System Problematics: Its Theory . . . . . . . . . . . . . . .125
Addressing Barthesian System Problematics: Its Practice . . . . . . . . . . . . . . 128

Practice Problematics: Level of Effort . . . . . . . . . . . . . . . . . . . . . . .128
Practice Problematics: Difficulty of Effort . . . . . . . . . . . . . . . . . . . . 133

Addressing Barthesian System Problematics: In Sum . . . . . . . . . . . . . . . . . 135

Xebra In Context: Related Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .136

Computer Criticism Tools: A Typology . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
Computer Criticism Tools: A Low And The High . . . . . . . . . . . . . . . . . . . .139
Computer Criticism Tools: The Mid-Level . . . . . . . . . . . . . . . . . . . . . . . . .140
Xebra and Computer Criticism Tools: In Summary . . . . . . . . . . . . . . . . . . .151

Xebra: A Functional Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152

The Xebra Functions: Defined . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
The Xebra Functions: Illustrated . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .159

Xebra: The Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .172
Xebra: The Proving Of Its Utility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .180

The Xebra Evaluation: Choosing The Text . . . . . . . . . . . . . . . . . . . . . . . . . 181
The Xebra Evaluation: The Evaluation Criteria . . . . . . . . . . . . . . . . . . . . . . 183
The Xebra Evaluation: The Reading Done . . . . . . . . . . . . . . . . . . . . . . . . . .194

The Barthesian/Xebran Reading: Phase One . . . . . . . . . . . . . . . . . . 194
The Barthesian/Xebran Reading: Phase Two . . . . . . . . . . . . . . . . . . 197
The Barthesian/Xebran Reading: The Statistical View . . . . . . . . . . . .210

Evaluating Xebra: Criterion One . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .222
Evaluating Xebra: Criterion Two . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .241
Evaluating Xebra: Criterion Three . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .247
Evaluating Xebra: Criterion Four . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
Evaluating Xebra: Criterion Five . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .270
Evaluating Xebra: Criterion Six . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .293
Evaluating Xebra: Criterion Seven . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 302
Evaluating Xebra: Criterion Eight . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 312
Evaluating the Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 317 

Reflections on Time, On Xebra, On Whys and Wherefores . . . . . . . . . . . . . . . . . . . .325
On Xebra: The Near Present Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .327
On Xebra: The More Distant Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .332
On Xebra: Towards The Always Already Rising Sun . . . . . . . . . . . . . . . . . . . . . . . . 340

BIBLIOGRAPHY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .343

LIST OF TABLES

LIST OF FIGURES

INTRODUCTION

An Hypothesis: Formulation and Inquiry

I must begin with a good body of facts and
not from a principle (in which I always suspect
some fallacy) and then as much deduction as you please.

Charles Darwin
Letter to J. Fiske
December 8, 1874¹

  The Formulation

The processes of determination and use of the "facts of a text" have a long history within the field of literary theory and criticism. There have been both theoretical and empirical approaches, with each having its own set of problems to be overcome in order to produce a model that critics can apply to actual texts. Roland Barthes' reading system, as discussed and demonstrated in his text, S/Z, is a largely empirical system designed to determine, as such, the facts, "signifieds" or "semantic substance"², in Barthesian terms, of a text as opposed to how they might be shown to work in terms of an interpretation of the text.³ The system assumes as its basic principle that the text itself is the proper object of criticism.⁴ This principle has been stated in many different ways over the centuries, with Roland Barthes' system being one of many that have attempted to ground themselves in their own version of it.

¹Familiar Medical Quotations, s.v. "Scientific Method".

²Roland Barthes, S/Z. trans. Richard Miller, (New York: Hill and Wang, 1974), 14.

³Ibid., 14.

⁴Ibid., 12.

If one accepts the principle, the problems that are left are based in the specifics of each such attempt to go back to the text. This dissertation bases itself in the hypothesis that Barthes' system, while constrained by problems, both of theory and practice, is ultimately capable of yielding sufficient value as to warrant a strong effort at overcoming those problems.

Thus, this dissertation assumes that Barthes' project is a reasonable project, as such, then it addresses certain specific problems that are inherent in his approach and attempts to bring them to a resolution which preserves the spirit, and much of the letter, of its rules, models, and paradigms, as parts, as well as of the system itself, as a whole. The value gained from this activity is reasonable access to the power of Barthes' system, a power that consists of unearthing facts of a text in a systematic, comprehensive, precise manner that allows for their direct use in performing acts of interpretation on texts. While these attributes constitute the controlling goals behind any systematic reading method, it is the basic tenet of this dissertation that a Barthesian reading yields up this knowledge so completely, precisely, and consistently, and in such a useable form, that ignoring its potential value to practicing critics is an unacceptable loss.

The specific problems of a Barthesian reading are based in points of theory and practice. The fundamental arguments against the system, from a theory stance, stem from three canon: theory of structure, theory of meaning, and theory of tools. The basic arguments against the system, from a practice stance, stem from matters of size and complexity.

Regarding the theory canon, Barthes, on the one hand, deliberately violates the grammatical structure of a text, which is to be, according to Barthes, "interrupted without any regard for its natural divisions (syntactical, rhetorical, anecdotic),"⁵ while on the other, he willingly accepts the certainly greatly expansive, if not truly infinite, cascading of meaning inherent in connotation. For some theorists, both acts would be unacceptable; for others, just one. But either way, Roland Barthes' system becomes anathema to them.

⁵Ibid., 15.

On a more general plane, there is the problem of what tools can or do accomplish for the user regarding specific ends. That is, given a fishnet of two inch diameter mesh, a user is likely to conclude that all fish are two inches or more in diameter, as Sir Eddington, astronomer and mathematician, stated. Such a comparison can be shown to mean that when applying Barthes system, it is only possible, at the very best, to find those entities that Barthes designed his "fishnet" to cull from the sea of words that make up any given text.

For theorists with an empirical mind set, none of these objections are necessarily a problem as long as the results of the system are useful. Usefulness then becomes the touchstone of any evaluation of the Barthesian system, that is, when given a particular definition of "useful."

As for the problems related to practice, they translate, finally, into the degree of effort on the part of the critic, which, in this instance, is significant. The performance of a Barthesian reading on even a small piece of text, such as one hundred words, can easily result in ten pages of closely written analysis. Multiply this by the three, or even four, times larger text of an average novel, and the effort quickly stretches out into something that most critics are not willing to undertake, simply from the standpoint of effort necessary to complete a given project.

Given these sets of problems, both of theory and practice, and maintaining the assertion that a Barthesian reading is a productive means to a desirable end (that is, the unearthing of the facts of texts), it is reasonable to attempt a resolution of the problems. Taking the empirical approach, one would first manage the problems of practice on the premise that, having resolved them, the problems of theory become more accessible to study. In order to overcome the practical problems, this dissertation turns to a tool which has proven quite useful in a wide variety of fields whenever a problem involves great manual effort: the computer. Thus, an inquiry into the feasibility of implementing Roland Barthes' system on a computer becomes the objective. The next section discusses the nature and substance of this inquiry.

The Inquiry

This dissertation takes as its means to an end the demonstration of a computer-based tool called Xebra (eXperimental Eclectic Barthesian Readers' Assistant), which encapsulates a solution set for the problems of practice in the Barthesian system so that literary criticism practitioners can apply it to the texts of their choice, while theorists may use it as a means for studying the benefits of deeply detailed criticism.

This approach is taken because computers exist to do work which is extensive, intensive, and repetitive; applying Barthes' system is precisely and exactly that species of labor. By relieving the labor burden, which constitutes the bulk of the objections to the system from the view of practice, it becomes possible to focus on matters of theory; that is, in determining the usefulness of such a process.

To that end, Xebra and a test of its and the Barthesian method's utility is described in this dissertation. William Faulkner's "The Bear" serves as the test input text, because of its relatively large size, rich complexity, and strong critical history. The size and complexity of "The Bear" stretch the system to the limits, both theoretical and practical, while the critical history serves as a basis for judging the test results.

Specific test criteria for this demonstration were drawn from the text, The Act of Interpretation, by Walter Davis, in which eight sets of "facts of the text" concerning Faulkner's "The Bear" are defined which, according to Davis, must be known items for any act of interpretation regarding "The Bear."⁶ By comparing the results of performing a Barthesian reading to Davis' "facts of the text," it is possible to measure how well Xebra, as an implementation of Barthes' method, achieves the desired results.

While the demonstration of the capability of Xebra to successfully overcome the practical problems inherent in the Barthesian system forms the core of this dissertation, attention to and focus on the theoretical implications of both Barthes' and the dissertation's projects, especially where the two are not one, is interwoven with this core.

⁶ Walter A. Davis, The Act of Interpretation: A Critique of Literary Reason, (Chicago: University of Chicago Press, 1978), 9.

The Plan Of The Dissertation

The dissertation is divided into five chapters. CHAPTER I consists of a detailed statement of the basic hypothesis of the dissertation, with supporting description, discussion, and argumentation. CHAPTER II is an indepth description and evaluation of Barthes' method and the proposal of a set of changes to overcome known or perceived problems with the method. CHAPTER III is a description of Xebra, other computer-based efforts in literary criticism, and Xebra's relationship to them, and how Xebra can and was evaluated as to its effectiveness as a Barthesian reader's assistant. CHAPTER IV is presents the performance of the evaluation of Xebra using Faulkner's "The Bear" as the test input text. CHAPTER V is a discussion of future directions for pushing Xebra's reach further into the world of literary criticism, defining more precisely what can be done in practice by using computers to extend critics' ability to analyze texts from a variety of views.

CHAPTER I

Barthes' Textual Network of Signifiers

 Barthes' S/Z: A Reading System of Worth

In his text, S/Z, Roland Barthes demonstrated an empiric, systematic methodology for determining "the facts of a text" for prose fiction based on connotative meanings of words and phrases, or "signifiers."¹ A precise, consistent, complete, and useable systematic approach to critical reading has been sought after with great energy for most of the 20th century, if not since the time of Plato and Aristotle. Barthes' system, however, is not the "Holy Grail of Literary Theory." First, its focus is only on that part of the process of literary criticism that precedes the actual act of interpretation, that is, the making of an inventory of the "features" of the text.² Second, it is not without blemish, both in theory and practice. Even so, it is the grounding tenet of this dissertation that Barthes' system yields rewards worthy of pursuit, especially when his manual model is effectively restated as a computer-based model.

A reading of S/Z, with its detailed account of a Barthesian reading of Balzac's Sarrasine, leaves little doubt that "the facts of the text" are, indeed, uncovered by the approach. Given the bountiful harvest of facts that Barthes presents in his reading of Sarrasine, a critic performing an act of interpretation involving the text could certainly advance confidently with the requisite supporting evidence at hand, thus avoiding being one of those "critics [who] often proceed upon insufficient or precariously selective attention to such details;"³ such avoidance being a reasonable, indeed, necessary goal of all critics.

That valuation of the Barthesian reading system having been made, it is important to note the countervailing view that Barthes did not altogether avoid problematic methods, assumptions, and assertions in the formulation of his system. As an empirical project, some features of the Barthesian system are open to theoretical objections, while others are, in fact, problems of an experiential nature.

Given these attributes of the Barthesian system, its rewards and its difficulties, this dissertation focuses on demonstrating a tool that overcomes the empirical problems, while concurrently exploring the benefits of using the Barthesian method of reading. The non-empirical objections to the approach will largely be only noted. Where Barthes himself presents a countering argument to the theoretical view of his project, Barthes' own case will be presented, without amplification, and with as little explanatory material as seems appropriate.

The following section explores the Barthesian reading system more deeply with the goal of underscoring the value of the system. It begins with a summary of Barthes' argument for the system, then proceeds with a further deepening of the point that herein lies value for the practicing literary critic, as well as the theoretician.

¹Roland Barthes, S/Z. trans. Richard Miller, (New York: Hill and Wang, 1974), 14-15.

²Walter A. Davis, The Act of Interpretation: A Critique of Literary Reason, (Chicago: University of Chicago Press, 1978), 9.

³Ibid., 9.

Barthes' Reading System: A Further Look

Barthes begins S/Z by addressing the need for a typology of texts. He notes early attempts at defining such as a way into introducing his own typology. Predicating a valuation of texts by using a definition of what would constitute an ideal text, Barthes' typology ranks texts in terms of how far they differ from this ideal, with all texts falling into one of two general categories: "the writerly" and the "readerly."⁴ The former, if it existed, would match his definition of the ideal text, while the latter, which Barthes also calls "the classic text,"⁵ contains all other texts. Thus, the classic text is "The Other" of the ideal, "its countervalue, its negative, reactive value."⁶

The Barthesian ideal text, the writerly text, is a "galaxy of signifiers, not a structure of signifieds; it has no beginning; it is reversible;."⁷ This text is "what can be written (rewritten) today," not just by the author, per se, but by the reader as well, "because the goal of literary work (of literature as work) is to make the reader no longer the consumer, but a producer of the text."⁸ Having access to the signifiers, a reader of a writerly text can make meanings, signifieds, that the first author of the text has no control over. Such a reader can enter the galaxy (text) at any star (signifier) and track towards any other star (signifier) as it pleases this text producing reader.

The readerly, or classic, text is, on the other hand, "what can be read, but not written,"⁹ because a reader of a classic text has "no more than the poor freedom either to accept or reject the text."¹⁰ Thus, such a text is a less than ideal text, differing from the ideal, the limit-text, to the degree that it bounds meaning by a structure of signifieds that has a beginning, a middle, and an end, and it is incompletely reversible (that is, not open to rewriting of meaning by the reader). The reader of a classic text is constrained to move through the text in a (largely) pre-ordained order, encountering, consuming signifieds (largely) as the author intended. But not necessarily totally, for even the readerly text has some portion of the ideal or writerly text within its interior.

Barthes calls this quality of a classic text its "plural,"¹¹ which can either be "more or less,"¹² that is nearer or farther from the ideal. This plural of the classic text, its degree of reversibility, of accessibility of the signifiers to the reader's will, is, for Barthes, the basis for making an evaluation of a classic text. Thus, the act of interpretation of a text is "to appreciate what plural constitutes it."¹³ This type of text evaluation, by definition, can only be applied to readerly texts, and not all of them. A writerly text, according to Barthes, "demolishes any criticism which, once produced, would mix with it.,"¹⁴ for in its infinite plurality it has no exteriority which could be inhabited by an evaluation.

⁴Barthes, 3-4.

⁵Ibid., 4.

⁶Ibid., 4.

⁷Ibid., 5.

⁸Ibid., 5.

⁹Ibid., 5.

¹⁰Ibid., 5.

Barthes calls this quality of a classic text its "plural,"¹¹ which can either be "more or less,"¹² that is nearer or farther from the ideal. This plural of the classic text, its degree of reversibility, of accessibility of the signifiers to the reader's will, is, for Barthes, the basis for making an evaluation of a classic text. Thus, the act of interpretation of a text is "to appreciate what plural constitutes it."¹³ This type of text evaluation, by definition, can only be applied to readerly texts, and not all of them. A writerly text, according to Barthes, "demolishes any criticism which, once produced, would mix with it.,"¹⁴ for in its infinite plurality it has no exteriority which could be inhabited by an evaluation.

On the other end of the scale, since a readerly text supports an act of interpretation only to the extent that there is a limited plural to its meanings, a readerly text that is truly univocal, that is, having no plural, would, like the writerly, escape the critic's grasp. By definition, there would be but one meaning for such a text, and thus no interpretation necessary or possible.¹⁵

The question arises as to how one is to discover the plural of any given text so that one can appreciate it. The answer Barthes provides is connotation. More specifically, it is the totality of the connotations associated with the individual signifiers. As Barthes states, "connotation is the way into the polysemy of the classic text, to that limited plural on which the classic text is based."¹⁶

¹¹Ibid., 5.

¹²Ibid., 5.

¹³Ibid., 5.

¹⁴Ibid., 5.

¹⁵Ibid., 6.

¹⁶Ibid., 8.

It is important to note what Barthes means by the term "connotation" and the term "denotation." Barthes defines what connotation several ways, beginning by referring to Hjelmslev's definition of connotation and denotation as:

a secondary meaning, whose signifier is itself constituted by a sign or system of primary signification, which is denotation.¹⁷

He then goes on to relate two prevailing views of the system of denotation. One, that of the philogists and linguists, is to declare "every text to be univocal, possessing a true, canonical meaning,"¹⁸ and thus "banish the simultaneous, secondary meanings to the void of critical lucubrations."¹⁹ In the other view, according to Barthes, semiologists

contest the hierarchy of denotated and connotated; language they say, the raw material of denotation, with its dictionary and its syntax, is a system like any other; there is no reason to make this system the privileged one, to make it the locus and the norm of a primary, original meaning, the scale of all associated meanings²⁰

Barthes disputes both views in his quest to establish not just a method for reading texts, but a standard by which the method measures the value of the texts read through it.

¹⁷Ibid., 7.

¹⁸Ibid., 7.

¹⁹Ibid., 7.

²⁰Ibid., 7.

First, regarding the connotations that the philogists would ban, he defines, in a positive sense, connotation in terms of its multivalency regarding meaning by relating sites across the text to each other and to sites in other texts. In defining connotation, he says,

it is a determination, a relation, an anaphora, a feature which has the power to relate itself to anterior, ulterior, or exterior mentions, to other sites of the text (or of another text)²¹

Thus, a connotative meaning is a connective, a bridge, from one site in a text to others, either in the text, or in other texts. This allows a reader to connect patterns of meaning in the journey that is a reading of the text.

Further, he says:

Topically, connotations, are meanings which are neither in the dictionary nor the grammar of the language in which the text is written (this is, of course, a shaky definition: the dictionary can be expanded, the grammar can be modified).²²

In other words, the connotations Barthes is speaking about are the secondary meanings that words and phrases take on due to their use as signifiers by writers, and their interpretation as such by readers. They are not meanings that are prescribed in some way by some official body.

As a native French-speaker, Barthes was probably particularly sensitive to this distinction, since French, as a language, is officially a legal entity controlled by the French government, and thus its grammar and its meanings are, at least theoretically, totally prescriptive and determined, thus determinant, thoroughly and completely. Clearly, Barthes recognizes the illusionary aspect of the French government's project. Balzac, through the creation of new connotations in Sarrasine, obviously expanded the French dictionary and grammar, without seeking legislative approval, neither before nor after his additions were existent. Language is a living entity that grows as it is used, and Barthes' methodology does not just recognize this fact, it makes use of it.

Then finally, Barthes says:

Topologically, connotation makes possible a (limited) dissemination of meanings, spread like gold dust on the apparent surface of the text (meaning is golden).²³

²¹Ibid., 8.

²²Ibid., 8.

²³Ibid., 8.

Thus, for Barthes, the ability of connotation to supply multiple meanings to signifiers is the means by which texts achieve their varying degrees of plurality of meaning, which is the counter, the means for valuation, of a given text.

As Barthes argues:

if there are readerly texts, committed to the closure system of the West, produced according to the goals of this system, devoted to the law of the Signified, they must have a particular system of meaning, and this meaning is based on connotation. Hence, to deny connotations altogether is to abolish the differential value of the texts, to refuse to define the specific apparatus (both poetic and critical) for the readerly texts--it is to make the limited text equal to the limit-text, to deprive oneself of a typological instrument.²⁴

Unlike the philogists, Barthes views text as inherently multi-vocal, and indeed, sees its value in that very multi-vocal nature. Through the multiple meanings of the signifiers, one can take a variety of paths through a text, connecting a variety of sites into a pattern of meaning, with the patterns changing, indeed, the connected network of sites changing as one changes which connotations to follow in blazing the privileged path. The boredom of univocality is not Roland Barthes'; rather it is left to those philogists who must have it for their own stability.

As for the semiologists who find the splitting apart of meanings into two types, denotative and connotative, to be less than an ethical act, Barthes has yet another view to put forth of the two types. For him, the counter-balancing of the systems of denotation and connotation, where denotation is held to be the primary meaning, does two things that are important for the classic or readerly text, which is his target reading material.

First, it "enables the text to operate like a game, each system referring to the other according to the requirements of a certain illusion."²⁵ Second, "this game has the advantage of affording the classic text a certain innocence."²⁶ Thus, the systems allow a text to enter into a game of illusion, of mysteries, where the main illusion is based on the innocent belief that there is a truth to be found in the language of the text, that, indeed, language is capable of expressing truth. Indeed, it is important that,

a sentence, whatever meaning it releases, subsequent to its utterance, it would seem, appear[s] to be telling us something simple, literal, primitive: something true, in relation to which all the rest (which comes afterwards, on top) is literature.²⁷

²⁴Ibid., 7-8.

²⁵Ibid., 9.

²⁶Ibid., 9.

²⁷Ibid., 9.

Barthes, then, agrees with the semiologists, with regard to denotation, in that he believes

denotation is not the first meaning, but pretends to be so; under this illusion, it is ultimately no more than the last of the connotations to who finds value in a system²⁸

 However, in so agreeing, he does not, like the semiologist, want to throw denotation out. Barthes' argument for keeping denotation as "the old deity, watchful, cunning, theatrical,"²⁹ centering, as it does, on its functioning structurally with connotation in such a way as to allow a text to engage in a game of illusion, and ideologically, in its allowing the game to rely on the belief in the innocence of language, is an empirical argument relying on the experiential value of denotation to a reader or to a critic. Thus, for Barthes, denotation, as well as connotation, is necessary for a proper valuation and appreciation of the classic, readerly text of Western production.

But having established the what and why of connotation, in what manner does one pay attention to the connotations, as Barthes says we must? That is a central question that leads into the heart of the Barthesian method. By definition, a systematic approach to reading must be objective and the results reproducible. That this is true of a Barthesian reading is a point to be demonstrated and is directly related to the problem of paying appropriate attention to the connotations.

In terms of application, Barthes' system approaches a text analytically, a means of fact discovery whose roots in literary theory go back at least to Aristotle. Both Barthes and Aristotle found good use in the analytic technique of dividing the large in order to conquer the small, which then leads, theoretically, to resultantly triumph over the large. In the computer science field, E. Dijkstra, a leading authority on programming models, has stated why he uses the divide and conquer method in this manner:

Some people might think the dissection technique just sketched a rather indirect and tortuous way of reaching one's goals. My own feelings are perhaps best described by saying that I am perfectly aware that there is no Royal Road to Mathematics, in other words, that I have only a very small head and must live with it. I, therefore, see the dissection technique as one of the rather basic patterns of human understanding and think it worthwhile to try to create circumstances in which it can be most fruitfully applied.³⁰

Whether or not Roland Barthes also thought he had "a very small head" is not known, but certainly he deliberately set out to conquer the complexity of texts by dissection. He created a method (circumstance) that attempts to "fruitfully" apply dissection to the problem of text understanding. Thus, in Barthes' system, the critic divides a text into small, contiguous pieces, which he refers to as lexias, with the atomic semantic units, signifiers, acting as the guiding principle for determining where to cut. As Barthes states: "all we require is that each [lexia] should have at most three or four meanings to be enumerated."³¹

Having cut the text into lexias, the critic then proceeds to "star the text, separating, in the manner of a minor earthquake, the blocks of signification."³² Each such "star" marks in the lexia a signifier which the critic must then label in terms of its connotation(s). For this operation, Barthes created a very specific coding or labeling grammar that categorizes signifiers into five major types. This language will be examined closely in Chapter II. For now, it is sufficient to note that the signifier is to be split out, marked, and then labeled as to type and connotative meaning(s). The labeling can also include, and often does in Barthes' practice as represented in S/Z, free text commentary beyond the "codes" Barthes invented.

²⁸Ibid., 9.

²⁹Ibid., 9.

³⁰E. Dijkstra, "Programming Considered as a Human Activity." in Classics in Software Engineering, ed. Edward Nash Yourdon, (New York: Yourdon Press, 1979), 5-6.

³¹Barthes, 13-14.

³²Barthes, 13.

In S/Z, Barthes went no further than the lexia making, starring, and coding, with commentary. However, he does state that the goal of this activity is to:

propose the semantic substance (divided but not distributed) of several kinds of criticism (psychological, psychoanalytical, thematic, historical, structural); it will then be up to each kind of criticism (if it should so desire) to come into play, to make its own voice heard, which is the hearing of one of the voices of the text. What we seek is to sketch the stereographic space of writing (which will here be a classic, readerly writing).³³

Thus, the act of interpretation, while not entered into by Barthes in S/Z, is provided for, and is, in fact, the natural next step, given the "semantic substance"; that is, the facts of the text that his method reveals. While Barthes only enumerates five specific types of criticism, it is not unreasonable to assume that, by implication, all forms of criticism that need a "semantic substance" as input to their process can be supported by a Barthesian reading.

For example, in applying a dialectic materialism-based interpretive model to a text, semantic structures involving economic values and actions are of interest. If Barthes' system is applied to a text, the atomic semantic units of those types could be identified and labeled and thus made available to the critic. Similarly, if the critical model to be applied is feminist-based, the atomic semantic units involving gender-based values and behaviors could be labeled and made available to the interpreter as well.

The key Barthesian operation, then, is identifying the particular signifiers present in a text. Explaining the systematizing of this operation via an extended example absorbs, quite understandably, the greatest amount of Barthes' energy throughout S/Z. How to actually do the operations, what problems to avoid, what pitfalls exist, what actions and habits are most likely to lead to success or failure, are largely ignored. In these matters, it will be the task of this dissertation to make explicit, when possible, that which Barthes has not.

The whole of Barthes' system, then, depends upon the hypothesis (which is never explicitly stated or simply labeled as such in S/Z, and is just assumed as true), that his signifiers can be found, labeled in terms of their connotations, and then placed, by critics, within larger structures of meaning constructed by variously connecting the multi-valent signifiers across the text, thus forming textual networks of signifiers, of meaning, within the text--its black clouds, star clusters, and hydrogen masses. Stated differently, these signifiers, when found and labeled appropriately, are assumed to form, or more accurately, to represent, a set of facts of the text, which can then be manipulated with those theoretical tools a literary critic might use in order to produce a reading.³⁴

³³Ibid., 14-15.

³⁴Ibid., 13-15.

It is the tenet of this dissertation that, if, in fact, Barthes' hypothesis is true, then Barthes has given the world of literary criticism an empirical tool of great value. This prediction of the value of Barthes' achievement is arrived at through the application of some basic axioms, first principles, indeed, of literary theory and criticism.

First, it is an axiom of literary criticism and theory that one cannot adequately produce, construct, create, or make an interpretation of a text without knowing, observing those "features" of a text which constitute its facts. "Facts" here refer to those entities of a text that are "self-given to attentive reading and need only be pointed out for their existence and importance to be established."³⁵

Further, these facts "are not really open to question, nor do they pose any major epistemological problems,"³⁶ or, put differently by Barthes, himself, "the meanings I find are not established by 'me' or others, but by their systematic mark."³⁷ That is, these facts, or meanings that Barthes (and by extension, all criticdom) is determined to uncover are in the text, not somewhere else, and are their own sign within a system of signs that constitutes the text.

It is another axiom that making these observations of fact is "hard work",³⁸ requiring "careful reading and patient attention to detail."³⁹ Indeed, Barthes says "it is a form of work."⁴⁰ Thus, he constructed a process that, when applied to a text, systematically "propose the semantic substance (divided but not distributed) of several kinds of criticism,"⁴¹ for, "there is no other proof of a reading than the quality and endurance of its systematics;."⁴²

Given that the facts of a text are a "necessary prelude"⁴³ to the act of interpretation, combined with the known level of difficulty in discovering them, it is a short step to this assertion: a system which, in a fundamental sense, guarantees access to the facts of a text in a form that supports the act of interpretation directly, is a system worth learning and applying. As stated above, it is the basic view of this dissertation that the system which Roland Barthes created can be shown to be such a system.

But there are problems with Barthes' system; problems that need resolution. The following two sections explore the topology of these problems, first from the negative, then from the positive view. These are followed by a third section that proposes a tentative resolution set.

³⁵Davis, 9.

³⁶Ibid., 9.

³⁷Barthes, 11.

³⁸Davis, 9.

³⁹Ibid., 9.

⁴⁰Barthes, 10.

⁴¹Ibid., 14

⁴²Ibid., 14..

⁴³Davis., 9.

Barthes' Reading System: Against

As an empirical approach, the Barthesian reading system inevitably lays itself open to theoretical, as well as empirical, claims of fault. That is, there are problems of theory and of practice in the system that need, at the least, to be acknowledged, if not completely resolved.

Taking the theoretical first, the Barthesian reading methodology appears to violate at least two theories of text, one concerning structure and one concerning meaning, as well as a theory of tools. The claims of fault advanced under the two theories of text are based upon assertions regarding the relationship of wholes to parts, while the claim of fault advanced under the theory of tools is based upon assertions regarding the properties of filters.

For structural theory, the question is: do parts determine wholes or do wholes determine parts? This is a fundamental question of structure that has never been answered in a way that finally and forever banishes all doubt as to its truth. That being the case, defending a practice such as Barthes' from claims of violating the characteristics of wholes is necessary. For, if one asserts that wholes determine parts, then the enterprise of cutting up prose fiction works into strings of text known as lexias, each of which is asserted to be itself composed of a finite, definite set of signifiers or meanings, comes into question. Does not the very process of text division, of lexia creation, coupled with the labeling of its signifiers deny the very existence of the whole of which each lexia is ostensibly a part? Barthes says as much when he states:

the work of the commentary, once it is separated from any ideology of totality [emphasis added], consists precisely in manhandling the text, interrupting it. What is thereby denied is not the quality of the text (here incomparable) but its "naturalness."⁴⁴

More directly, Barthes, when discussing the plurality of possible interpretations of a text, says

The interpretation demanded by a specific text, in its plurality, is in no way liberal: it is not a question of conceding some meanings, of magnanimously acknowledging that each one has its share of truth; it is a question, against all in-difference [sic], of asserting the very existence of plurality, which is not that of the true, the probable, or even the possible. This necessary assertion is difficult, however, for as nothing exists outside the text, there is never a whole of the text (which by reversion form an internal order, a reconciliation of complementary parts, under the paternal eye of the representative Model):⁴⁵

⁴⁴Barthes, 15.

⁴⁵Ibid., 6.

Thus, Barthes simply denies the notion of a totality, of a whole of the text. At the least, then, there appears to be uncertainty as to the theoretical soundness of Barthes' approach from a structure theory point of view.

For semantic theory, a basic question is: are there such entities as connotative meanings? From this question there follows: if there exist connotative meanings, is there any limit to their number? From that: if not, is there, finally, any meaning at all? As in the wholes versus parts problem for structural theory, explorers of semantic theory have never arrived at an agreed upon position regarding either the base question or its corollaries. Barthes has his own responses to each of these three questions, which will be explored in the next section, but again, the theoretical soundness of Barthes' enterprise is in question.

Finally, for tool theory, the question is: what does the application of a filter to a stream of information actually accomplish? Sir Eddington, with his positing of a fish net of a particular size mesh returning to its user only those fish that match or exceed that size, thus forcing the fisherman to conclude that all fish are of that size or larger, has forever linked an answer to the question in such a way as to make any other answer appear as utter nonsense.

The popularization of this principle is the statement that one can only see that which one's glasses allow. Thus, if one wearing rose colored glasses, one could only conclude that the world was colored rose. For the Barthesian enterprise, this principle appears to mean that only those signifiers that fit the mesh of the signifier net thrown into the sea of the text by the literary critic will be found, while all others, if, in truth, there are others, will be lost forever. Again, at the least, the theoretical soundness of the Barthesian enterprise is in question.

Assuming that these theoretical claims of fault can be dealt with, there remain problems that empiricists must raise in their own name. These problems arise when a Barthesian reading is attempted in the empiricist's world of experience, as opposed to the theorist's world of mind. Problems of effort required, given the size and complexity of texts, as well as problems of application involving difficulties of precision, completeness, and consistency, which again are related to size and complexity of real texts, are two basic areas that need to be addressed.

A Barthesian reading of even a small text requires significant expenditure of effort on the part of the critic. Cutting up the text, then exploring the multitude of connotations involved, since even one signifier is likely to carry with it as many as three or four, is not a simple task, especially if the text under study is both physically large and structurally and semantically complex. The prospect of having to make such an expenditure of effort is, itself, a barrier to the application of Barthes' system.

However, even more importantly, the degree of effort involved is also an indication of the difficulty of actually applying the system usefully. It is not simple or easy to attain the degree of precision, completeness, and consistency of labeling that one needs in order to be certain of having accomplished Barthes' goal

of working back along the threads of meanings, of abandoning no site of the signifier without endeavoring to ascertain the code or codes of which this site is perhaps the starting point (or the goal);⁴⁶

To be successful, it would appear to be paramount that this operation be capable of precise, consistent, and complete application by its user with results that are equally precise, consistent, and complete. Barthes, however, does not say anything about how to go about assuring precision, consistency, and completeness. He does not even state that it is necessary to do so. The closest he comes is to say that a reading of a text using his approach is done under the assumption that the text has already been "read," that the first reading, the naive reading, is only one of many possible readings and that the work of the critic is to get at these multiple readings.⁴⁷ It is a by-product of reading the text multiple times that the very process Barthes applies to the text is made possible. One cannot label, as required by his system, something in the first paragraph if one is not aware of what the last paragraph contains, as well as those in between.

⁴⁶Ibid, 12.

⁴⁷Ibid., 15-16.

At this point, it is possible to conclude that while Barthes may have made strong progress toward the end of defining a workable systematic approach to reading texts, he did not finally reach it. Even putting aside the theorists' objections, the empirical ones finally defeat him. The proof (empirical) is in the fact that other than Barthes' own extended example of a Barthesian reading in S/Z, in which he dissects Sarrasine, no others exist. A system that is not used, is a system that has failed.

Perhaps the goal is unreachable using the path that Barthes took. Perhaps the goal is, in actuality, unreachable by any means. However, before making such judgements, there is more to be said, more to be learned.

In order to deepen the case for the need to pursue such an effort regarding the Barthesian system, as well as to underscore the potential value thereof, the following section addresses, from a theoretical stance, the positive attributes of the system, as well as countering, where possible, the theoretical objections as advanced above.

Barthes' Reading System: For, Even So

It is an empirical truism that no two (or more) competing theoretical stances concerning a given aspect of reality ever completely stand up when taken into the world and tested in experience. And so it is with the wholes versus parts problem concerning structures. Practice in many disciplines has shown that neither is consistently true or false. Whether one is a computer scientist attempting to design a program, or a literary critic attempting to produce an interpretation of a text, the fact remains that both the whole of the program/text and the parts of which it is composed are, paradoxically, first principles that must be satisfied simultaneously.

In practice, the computer scientist will design the structure of the whole (top down, in computer jargon) while concurrently designing its parts (bottom up, in computer jargon) whenever necessary for determining some aspect of the structure of the whole. Similarly, the literary critic will seek to understand basic facts of the text, while concurrently fitting them into a posited structure of the whole. If the part does not fit the structure of the whole, as the critic has supposed it to be, then either the posited structure is wrong, or the part is insufficiently understood, or both. In any case, the parts determine wholes, and the wholes determine parts.

At first glance, Barthes appears to have dealt with this problem of competing theories, not by giving both their due, but by denial. First, he denied the notion of a ranked plurality of meanings for a text in terms of a whole of the text. Further, he insisted that all the meanings of the text, the plurals of the text, just were, that they existed as individual wholes which did not in some sense partake in a dividing up of the "whole of the truth" of the text.⁴⁸

Thus, the theoretical objections to Barthes' violation of the wholes determining parts principle is not refuted, but simply denied a basis in reality. Wholes of texts simply do not exist; to assert that they do is to be trapped in (false) ideology.⁴⁹ He then proceeds to perform a bottom up operation on the text of Sarrasine that yields parts (signifiers) that are then to be used to aid the various criticisms in their task of producing a specific interpretation of a text. The proof of his method (he implicitly insists by giving over the whole of S/Z to the reading of Sarrasine), is to be sought in experience, not in a theory of structure, of truth of texts.

A more conservative approach would be to simply accept the proof of experience, without attempting to reconcile practice with theory. However, even empiricists have the urge to retreat occasionally into the world of mind, into the realm of theoretic thought. Thus, when Barthes seeks to assert that "there is never a whole of the text,"⁵⁰ he appears to attempt to do battle with the theorists on their own ground, always a difficult operation, but he appears to do so in the simplest form: denial.

⁴⁸Ibid, 6.

⁴⁹Ibid, 6.

⁵⁰Ibid, 6.

Barthes is not so simple. When he denies the totality of the text, what he really is attempting to address is the "infinity" of the text, and more specifically, the ideal text of the writerly.⁵¹ However, when he addresses the readerly text, he speaks of their meanings in terms of metonymic, existential attributes in a manner that is meant to illustrate the quasi-infinite capability of even a readerly text to mean or to signify. On the metonymity and existentiality of such texts, Barthes states:

To read is to find meanings, and to find meanings is to name them; but these named meanings are swept toward other names; names call to each other, reassemble, and their grouping calls for further naming: I name, I unname, I rename: so the text passes: it is a nomination in the course of becoming, a tireless approximation, a metonymic labor.⁵²

Further, as Barthes goes on to say, reading such a text

consists in coupling systems [of meaning], not according to their finite quantity, but according to their plurality (which is a being, not a discounting): I pass, I intersect, I articulate, I release, I do not count.⁵³

That is, that to the degree a classic text possesses a plural, it participates in that aspect of the ideal text that enjoys the lack of exteriority, of totality. Barthes asks, "what is the sum of the text?"⁵⁴ then answers by not answering, implying there is no "sum of the text,"⁵⁵ as such, but there is a "tireless approximation"⁵⁶ on the reader's part to reach such a sum which invariably fails.

⁵¹Ibid, 6.

⁵²Ibid, 11.

⁵³Ibid, 11.

⁵⁴Ibid, 11.

⁵⁵Ibid, 11.

⁵⁶Ibid, 11.

The reason the sum is never reached for the ideal text is that it is, by definition, infinite. For the readerly text, the reason lies in two mechanisms: the limited plural of which it is composed, and the human facility for forgetting. Indeed, Barthes is very positive about this correlation between the plural of the reader text and forgetting. For Barthes:

Forgetting meanings is not a matter for excuses, an unfortunate defect in performance; it is an affirmative value, a way of asserting the irresponsibility of the text, the pluralism of systems (if I closed their list, I would inevitably reconstitute a singular theological meaning): it is precisely because I forget that I read.⁵⁷

Barthes' answer then to the problem of wholes and parts is not completely a denial of wholes, but a skirting of the issue: the infinite obviously has no whole we can grasp in our finiteness, while finite texts are an analogue of the infinite texts because humans forget, and thus cannot even grasp the whole of a finite object. If there is a "sum of the text,"⁵⁸ we cannot know it. So we are left knowing its parts, which are related to any such whole metonymically. For Barthes, not only is this true, it is good.

Barthes' definition of reading, as a ceaseless approximation of the whole of a text through the finding and naming of individual meanings, deals not only with structural theory objections, but with those of semanticists who are concerned about his use of connotation as his basis for finding those meanings. Meaning, in the sense of what Barthes refers to as the "singular theological meaning,"⁵⁹ is not what one wants, according to Barthes. Instead, one wants to participate in the ongoing process of the becoming of the text in terms of meaning. That is, instead of denying that his method might violate some theoretician's view of what constitutes texts and their meaning(s), he advances his own theory upon which he proceeds to build a system of reading. He then invites the reader of S/Z to determine whether the practice and theory are reasonable by studying, that is, by experiencing, his extended example, the Barthesian reading of Sarrasine.

⁵⁷Ibid, 11.

⁵⁸Ibid, 11.

⁵⁹Ibid, 11.

It is this experiencing of the reading of Sarrasine through the mind of Barthes that must finally convince (or not convince, as the case may be) a scholar of the efficacy of the Barthesian approach. Empirical systems demand empirical valuations; to do otherwise is to deny them of their very basis, of their identity as systems in the world of experience. Certainly, this dissertation takes the stance that, indeed, there is, in the Barthesian approach, a good worthy of the pursuit.

That is not to say, however, that there are not problems, problems which are recognized as such by empiricists, and are, as it should be, resolvable through empirical methods. These problems are, as noted earlier, centered in the practice of the system and derive largely from the size and complexity of the task of applying the system to a text. There is a growth curve in effort as one moves from very small texts toward the ever larger that is, if not exponential, certainly multiplicative. As one moves along this curve, two things happen. First, the critic, realistically, gets tired. Second, the critic loses the ability to hold effectively to the task at hand. These are problems that need addressing if the system is to be used.

As noted earlier, the system needs to be used, for a practicing critic must know the text which is the object of the critical act; that is, at the very least, the facts of the text must be in the critic's mind. Without these facts the act of interpretation, which Barthes calls "the hearing of one of the voices of the text,"⁶⁰ cannot go forward. The Barthesian project has yielded a system, which, when used, is an empowerment for those who would practice criticism, arising from its capability for uncovering those facts of a text that are essential to the critical project.

⁶⁰Ibid, 15.

Paying detailed attention to a text is always difficult. However, the discipline and structure of Barthes' system provides a means for focusing on and addressing the text, making such problems as completeness, precision, and consistency tractable, if not simple. Further, Barthes' system has the additional benefit of providing a definition of the quarry--the facts of the text, or, as he names them, the signifiers of the text.

While it is true that this definition, based as it is on the connotative meanings of signifiers, is not a stable, absolutely determinant definition that can be applied without fear of error, of lack of completeness, precision, and consistency, Barthes would undoubtably argue that this aspect of the definition is precisely its strength. He would so argue, because, for him, a text that is accessible to interpretation is inaccessible to a determination of a "singular theological meaning"⁶¹ due to its limited plural. Thus, on the premise that it takes a thief to catch a thief, it takes an unstable tool to catch an unstable object. Or, more precisely, it is, in fact, the instability of the signifiers through connotative meanings that imbues the classic text with its own instability, its limited plural, and, therefore, makes possible the act of interpretation.

⁶¹Ibid, 11.

Up to this point, the focus has been on what a Barthesian reading of a single text might mean to the practice of literary criticism. Changing that focus toward literary theory, there are, at least, two interesting problems of theory that Barthes' method appears to promise a reasonable environment for study. First, using the Barthesian approach, it appears possible to study the question of the value of a deeply detailed critical reading. Second, the Barthesian approach appears to offer a way to study the interconnection of texts in terms of meaning.

The ability to study the first question follows naturally from what the Barthesian methodology is: a systematic approach to detailed reading of a text. However, the second question is not so obvious. Barthes created an entire language, with a grammar and a vocabulary, which he uses to encode his signifiers in terms of their signifying in a given text through the connotative meanings of each. This language reaches out into the whole world of experience, implicitly and explicitly, and so is not necessarily constrained to a single text.

For example, Barthes states that "The Kidnapping refers to every kidnapping ever written."⁶² Thus, in a Barthesian reading of The Kidnapping, all of the connotations of the phrase would be encoded, which would be valid not only for that particular text, but for all other texts in which kidnappings are a part of the material substance. This fact leads to the observation that the connotative meanings that a Barthesian reading encodes might serve as stepping stones to connecting not only the patterns of meaning of a single text, but to relate that text to other texts, forming a universe of texts composed of "galaxies of signifiers".⁶³ It is, at least, an open question that deserves study.

But all of the above does not indicate that no problems exist with the Barthesian methodology, as such, and as noted above, they do have to be addressed and resolved. The following section, then, outlines the approach this dissertation will take toward accomplishing that end.

⁶²Ibid, 20.

⁶³Ibid, 5.

Barthes' Reading System: An Enabling Strategy

This dissertation proceeds on the hypothesis that the Barthesian reading system has an inherent value that merits pursuit. Because Barthes' system, as he outlined it in S/Z, has been largely ignored in practice, empirical evidence is scarce for making any final judgements about it as such, or for showing how it contributes to the larger issue of whether the ends it is professed to serve are, indeed, feasible ends. Only in the application and use of a system are its frailties and virtues finally brought forth.

As argued in the preceding section, there is evidence for the usefulness of the system as it stands, but the evidence exists, unfortunately, only in S/Z. Even more to the point, there is strong evidence that the system, as it stands, is not useable; that it is, in fact, not being used. Thus, a way is needed to improve the system while also gathering evidence that the system, as improved, is both useable and worth using. Until these two empirical measures of value are satisfied, the Barthesian system will remain an artifact residing only in Roland Barthes' S/Z.

In an attempt to avoid such a fate, this dissertation takes a two-step approach. First, the Barthesian system is evolved through a process of clarification, generalization, and simplification, using a computer-based model as the underlying operational mode, replacing the manual-based model of the original, to achieve needed gains in usefulness and useability. Second, the evolved system is evaluated for enhanced efficacy and accessibility.

Using a computer-based model for the Barthesian methodology is a consequential change from the original model of doing everything manually. The sheer expenditure of effort required in applying Barthes' system, especially on larger texts, is enough to discourage its manual application. This becomes obvious when the margin for execution error, which gets larger as the size and complexity of the text under study increases, is considered, mandating a labor saving device be used. By computerizing the methodology, both the actual effort and the error potential should be reduced significantly, leading to a better tool that can be used effectively, both for critical practice and for literary theory.

In support of this position, it is important to note at this point that Barthes' system has been singled out in the past as a candidate system for computer implementation using current state-of-the-art techniques. This was done for one particular reason: the infeasibility of manual effort. John B. Smith, a pioneering and continuing researcher in the field of computer-based literary criticism, in his essay "Computer Criticism," suggested that systems such as Roland Barthes' would necessarily have to wait for computer implementation to take full advantage of any usefulness to be gained from it. His argument for this, just as the argument made here, relied on the intensity and complexity of the manual effort required, as well as the discipline that the computer mandates in terms of strictly defining one's methodology.⁶⁴

⁶⁴John B. Smith, "Computer Criticism." Style 12 (1978): 337.

The approach taken in this dissertation to reach a better Barthesian tool by evolving the Barthesian reading system from its current state via a change in its underlying mode of operation, is empirical. It is not an approach designed to satisfy the theoretical objections to the basic attributes of Roland Barthes' system. Those objections are assumed by this dissertation to be irrelevant to Barthes' project and to the dissertation project.

Given that the Barthesian system is an empirical system, the measure to be applied to determine the degree of success of the system is likewise purely empirical: does the application of the tool, as it is evolved on the computer, give back to the user good value in terms of aiding in the discovery of the facts of the text? A second measure will be the value the tool might have for literary theoreticians.

Barthes' Reading System: An Evolution Plan

Roland Barthes created a system that generally fits the definition of processes for which a computer-based model is appropriate. That is, processes that are repetitive, intensive, and employ computable algorithms usually benefit from computer implementation. Barthes' system possesses all three attributes, in varying degrees. It is precisely that degree of variance that determines how successfully a given process can be implemented via computer.

For example, the Barthesian system requires that the text under study be divided up into units termed lexia. These units are rather vaguely defined by Barthes from the point of view of computable determinacy:

The tutor signifier [that is, the text] will be cut up into a series of brief, contiguous fragments, which we shall call lexias, since they are units of reading. This cutting up, admittedly, will be arbitrary in the extreme; . . . The lexia will include sometimes a few words, sometimes several sentences; it will be a matter of convenience: it will suffice that the lexia be the best possible space in which we can observer meanings; its dimension, empirically determined, estimated, will depend on the density of connotations, variable according to the moments of the text: all we require is that each lexia should have at most three or four meanings to be enumerated.⁶⁵

From this definition it is clear that a computer program, in order to cut a text into lexias, would have to be able to recognize a "meaning," a "connotation," a "signifier." At present, natural language processing algorithms are simply insufficient for that task, directly. But this is not necessarily an insurmountable obstacle.

⁶⁵Barthes, 13-14.

There are two avenues open: invent new algorithms, or use existing ones to approximate Barthesian operations. Instead of attempting to discover new processing algorithms, which would be difficult, if not, in fact, impossible, the approach used was to adapt current computing capabilities to approximate, not just the lexia cutting, but as much of Barthes' system as feasible, with the goal of obtaining a workable, useful version of the Barthesian reading methodology, that is simpler, clearer, and, if possible, more general than Barthes' original.

Simpler, in this context, means that the complexity of the text does not necessarily translate directly into more difficulty when applying the system. With the computer tracking the clerical aspects of the operation, the user should find that effort is largely confined to the intellectual realm of determining what to label, not in managing the labels and their relationships within the text. Thus, the task should be simpler using Xebra than using a purely manual approach.

Clearer means that, wherever Barthes' directions as to how to apply the system appear ambiguous, an attempt has been made within the Xebra system to remove, or at least lessen, the ambiguity. This is not always possible, given that not only are the objects of study (prose texts), ambiguous, but Barthes chose a tool (connotation), which is, by definition, grounded in ambiguity. But where gaps appear in the original operational modes or in the grammar of the labeling scheme as defined in S/Z, Xebra has been designed with clarifications to the original.

More general means that an attempt has been made to allow users of Xebra to redefine the Barthesian system as deemed appropriate or desirable. For example, while Barthes may not have explicitly looked to link texts via the labels, a computer-based system can support such linking quite readily. Or, as another example, such a system need not be constrained to just the Barthesian label language, neither its vocabulary nor its grammar, so that other critical reading methods depending on labeling items in a text can be supported directly. Indeed, a user could make changes to the Barthesian language, as necessary or desirable. More general also means that there is support beyond the strict line that Barthes draws between what he was doing and what critical interpretations do. That is, a more general system would support moving from Barthes' stated end goal "to sketch the stereographic space of writing"⁶⁶ to the formal work of the practicing literary critic: that is, the performance of an act of interpretation.

⁶⁶Ibid., 15.

The computer-based system developed for this dissertation is named Xebra, for eXperimental Eclectic Barthesian Reader's Assistant. Xebra has two major components: one to support lexia creation and another to support signifier label creation and manipulation. The lexia creation component consists of three parts, each representing three different approaches to cutting a text into lexias: punctuation mark parsing, noun phrase parsing, and, as a direct analogue of the original system, manual parsing. A user could use any of the methods on a given text, including all three. The second component is essentially a database system for storing and manipulating the lexias and their associated labels. It allows a user to discover the patterns of meanings within the text, as represented by the labels, through a sophisticated query mechanism that supports sorting and grouping of labels by various attributes, thus moving from the strict Barthesian reading into the critical act of interpretation. A fuller description of Xebra is given in CHAPTER III.

Given Xebra, the next step towards evolving the Barthesian reading system involves empirically demonstrating the value gained from the effort to improve the system. To accomplish this, the performance of an actual Barthesian reading with Xebra, using a strong text from the literary canon as the proving ground, would fit best with the empirical nature of both the system itself and the direction of this dissertation.

Having determined that this is the appropriate direction to take, the next step is to choose the text. The text should be relatively complex and physically large with a pedigree of critical interpretation that can be used as a standard for measuring the effectiveness of the system. Once the text is chosen, the next steps are to then perform the reading, followed by an evaluation of the results in terms of the critical test standards.

The text chosen for this demonstration of the Barthesian system, as embodied in Xebra, was "The Bear" by William Faulkner. This text was chosen for several reasons, but the most important were that it has the necessary size and complexity, coupled with a critical history that is very accessible. Indeed, Walter Davis, in his text The Act of Interpretation, outlined eight areas of factual knowledge that any would-be critic of "The Bear" must have at hand, as well as critical examples of such facts in each area.⁶⁷ From these, a standard set of criteria were developed that could be objectively evaluated against the results from a Barthesian reading of "The Bear."

⁶⁷Davis, 9.

Given an evolved Barthesian reading system, a text for study, and objective criteria for measuring the study's results, the project of this dissertation comes to the performance of the test reading with the goal of demonstrating the effectiveness of the system. However, one final area of concern is the role S/Z plays in this project of constructing and testing a computer-based version of the Barthesian reading system.

Barthes' S/Z: What Does It Do?

S/Z is many things, two of which are crucial to the present enterprise. The first is Roland Barthes' own explanation of his system by description and example. The second is his recording of a reading, a commentary of Sarrasine, produced in accordance with that system. The former, the explanation of his system, served directly, and largely unproblematically, as input into a major portion of the requirements for building Xebra, the computer-based version of Barthes' system. The latter served in that capacity as well, but it did so somewhat problematically. The difficulty was inherent in a persistent impression that, in S/Z, there is embedded a reading or commentary masquerading as a criticism of the text, the product of an act of interpretation.

One can easily leave S/Z with the belief that Barthes has performed a significant explication of the text, Sarrasine; that he has, in fact, given the reader an authoritative exposition of the meaning of Sarrasine. Thus, one is tempted to argue that Barthes does, indeed, set forth a criticism of the text, a criticism that physically exists in his notes on each lexia and its labels, this despite his statement in S/Z that "we shall not set forth the criticism of a text, or a criticism of this text,"⁶⁸ and his section in S/Z entitled "XXXIX. THIS IS NOT AN 'EXPLICATION DE TEXTE'"⁶⁹. If one subscribes to this position, then it makes the stated Barthesian enterprise of merely pointing to the facts, the signifieds, as being problematic.

However, such an argument would need to be based on a definition of criticism of a text that is not consistent with Barthes' own. To him, such a criticism necessarily seeks to articulate "a metameaning which would be the ultimate construction"⁷⁰ one could make out of the atomic signifiers that he seeks merely to enumerate, label, and comment upon. Instead, his labels and accompanying commentary aim simply "to sketch the stereoscopic space of writing,"⁷¹ or, in other words, to label the signifiers so that each of the various kinds of critical models can "come into play, to make its voice heard,"⁷² thus, deliberately avoiding the articulation of an ultimate metameaning of the strands of signifieds.

⁶⁸Barthes, 14.

⁶⁹Ibid., 90.

⁷⁰Ibid., 14.

⁷¹Ibid., 15.

⁷²Ibid., 14-15.

Indeed, in Sarrasine, it is Balzac who has distributed the signifiers through the lexias, not Barthes. So any sketch of the stereoscopic space of the text will inevitably repeat and repeat and repeat itself, and it is the repetition of label and commentary that gives the impression of a value judgement being made by the commentator, Barthes, when in fact it was Balzac who placed a value on those signifiers thus repeatedly used through his distribution of the those signifiers. Barthes is merely reporting the fact of their existence.

Finally, after further analysis, one discovers that the real heart of the reason why one believes that one has, indeed, read an "'explication de texte'"⁷³ is the sections of S/Z which are set off by roman numerals. In these Barthes often gives an in-depth discussion of particular aspects of the text that some variable number of just previously rendered, labeled lexias allows the reader to understand through the labels.

For instance, section "LXIII. Psychological Proof" of S/Z, contains Barthes' discussion of how the proceeding fifteen lexias, numbers 332 through 346, demonstrate that "the Sarrasinean snare shifts from proof to proof"⁷⁴ and that these proofs (all false) are largely psychological and self-inflicted, so that Sarrasine ensnares himself into believing that the castrato is a woman.⁷⁵ Clearly sections such as this are not part of Barthes' reading, per se. They contain judgements and interpretations of the text; they are a part of an "'explication de texte',"⁷⁶ though they do not, when taken together, necessarily make a complete one. They are, in fact, examples of what a critic might do with the data that the reading is identifying, labeling, and preserving for the critic's use.

⁷³Ibid., 90.

⁷⁴Ibid., 147.

⁷⁵Ibid., 147-148.

⁷⁶Ibid., 90.

Other such sections (all of which are marked off by roman numerals, thus separating them from the ongoing, continued reading, instead of being explicit discussions of Sarrasine), are discussions of a theoretical nature, either about his methodology (for example, "XI. The Five Codes"⁷⁷), or about more general topics, such as literature (for example "LXXXIV Literature Replete"⁷⁸), often using his reading of Sarrasine and his methodology as a spring board.

Thus, S/Z contains, is made up of, is irretrievably woven with, the strands of at least three projects: an ongoing reading of Sarrasine using the Barthesian method, an explication of that method, and an argument for the Barthesian model of literature. To construct Xebra, this web, this tapestry, this galaxy of signifiers, needed to be traced, disentangled, and unwoven, in order to finally reach those aspects that were pertinent to the project of constructing and testing a computer-based Barthesian reading system to be known as Xebra.

A Barthesian reading is one process, a Barthesian criticism another, a Barthesian argument still another, and it is the task, the "labor of language"⁷⁹ of the reader of S/Z, to keep these processes separate since Barthes saw no need to do so himself. Thus, once separated out, it is clear that there is, indeed, a difference between enumerating and labeling the signifieds or facts of the text and consequently drawing conclusions or making value judgements concerning those signifieds, those facts. The Barthesian methodology, and by extension, Xebra and this dissertation, is concerned with the latter, not the former.

⁷⁷Ibid., 18-20.

⁷⁸Ibid., 200-201.

 ⁷⁹Ibid., 11.

The Project: Chapter by Chapter

There are five chapters in this dissertation. CHAPTER I presents the grounding hypothesis of the project regarding the viability of the Roland Barthes' reading system, with supporting material. It also includes an overview of what is involved in the testing of the hypothesis. CHAPTER II begins with an overview of the Barthesian system as it was presented in Barthes' S/Z, followed by a detailed description of that system, attribute by attribute. It ends with a detailed presentation of the problems presented by the system for anyone interested in applying it, both from a theory and a practice perspective. CHAPTER III presents Xebra, first in the context of other computer-based critical reading systems, second, in terms of its functionality, and finally, in terms of how its functionality can be and was evaluated for effectiveness. CHAPTER IV presents the evaluation itself of doing a Barthesian/Xebran reading of William Faulkner's "The Bear." Chapter V presents a discussion of where the Barthesian system, and Xebra, might fit in the future fabric of literary practice and theory. In order to do this, a review of the results of the test described in Chapter IV is given, along with suggested avenues for improvement. This is followed by a series of proposals for how best the Barthesian reading system, as embodied in Xebra, as well Xebra as a tool of more general capability, might be put to work.