CHAPTER II 

The Barthesian Reading Project: Its What and How

The Barthesian System: Introduction

Roland Barthes' purpose in S/Z is to empirically demonstrate a system for reading texts prior to, and in support of, the act of interpretation. The goal of this chapter is to focus in detail on the what and the how of a Barthesian reading. This will be accomplished by first, abstracting a set of processes and rules from the context Barthes uses in S/Z, where he largely describes by demonstration, then second, by examining the abstracted set of processes and rules. However, before beginning that discussion, a review of the theoretical basis for the Barthesian system, its why, is presented to provide context for the system as a whole.

As discussed in CHAPTER I, Barthes was seeking in S/Z to put forth a reading system that was consistent with his view of text evaluation. Thus, his system is based upon a theory of a typology of texts which divides texts into two categories, the readerly (classic) and the writerly. These categories differ from each other in one significant way: the degree of plurality of possible meanings (referred to as its plural) any given text can be found to possess. In principle, writerly texts have an infinite plural, while readerly texts have a finite plural.¹ It is this difference in plural that forms the basis for Barthes' theory of how to evaluate a text, which in turn supports his reading system.

Barthes admits two points concerning the writerly text. For one, "we would have a hard time finding it in a bookstore."² For another, the writerly text's

model being a productive (and no longer representative) one, it demolishes any criticism which, once produced, would mix with it.³

So the writerly cannot be read because it does not exist, and if it did, it cannot be interpreted critically because it destroys any such attempt because it produces, it does not represent. Given this, the writerly model has but one use in Barthes' system: as the standard against which one can measure the readerly or classic text. That is, the value of a classic text is proportional to the degree that it approaches the writerly text in terms of the plurality of meanings it is possible to produce from it. As Barthes explains, this is true:

Because the goal of literary work (of literature as work) is to make the reader no longer a consumer, but a producer of the text.⁴

¹Roland Barthes, S/Z. trans. Richard Miller, (New York: Hill and Wang, 1974), 4.

²Ibid., 5.

³Ibid.

⁴Ibid., 4.

Having posited that the value of a text is dependent upon the meanings one can produce from reading it, Barthes' project becomes one of discovering and examining individual signifiers that are the productive building blocks of any meaning a reader of a text can produce. Because signifiers have multiple meanings, according to the system of denotative and connotative meaning that is a primary basis for semantic theory in Western thought, their signifieds take on a variety of semantic attributes.⁵ Thus, the act of interpretation is largely a matter of tracing the results of accepting any given set of connotative meanings for the text's signifiers in relation to their signifieds. Such a set would be determined largely by the critical model applied. For example, in a Marxist model, economic connotations would be favored over any non-economic, such as psychological or archetypal.

Before entering more deeply into the following discussions on Barthes' project, it is important to note his definition of denotative meaning in relation to connotative meaning. According to Barthes,

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.⁶

Clearly, for Barthes, "denotative" is a name given to one of the meanings of a signifier, thus placing this one meaning, quite arbitrarily, in the position of "first meaning" in order to set up the play of meanings.

This play is a verbal analog to the children's game, "King of the Hill," where each child seeks to attain and then keep the top, or King, position on a hill, with all others being held as subordinant to the King-position holder. In the verbal analog, the various meanings of signifiers attempt to drown out each other, becoming in the mind of the reader the one meaning, the denotative meaning. Thus, for the Marxist, a connotative meaning of economic significance drowns out the connotative meaning of feminist significance, except of course where the two are one, which would be the case for a Marxist-Feminist critic.

Given that Barthes' base project is to discover, to find the signifiers with their sets of denotative/connotative meanings, he needed and produced a systematic approach to the process. It is precisely this systematic process, here called a Barthesian reading, that the rest of CHAPTER II explores in detail as to its what and how.

⁵Ibid., 8.

⁶Ibid., 9. 

 The Barthesian System Processes: Introduction

As described in S/Z, a Barthesian reading entails several distinct processes, interdependent and intertwined in space and time, which constitute the what, with rules connected to each, which constitute the how, of the Barthesian system. The processes are:

1) Text Restructuring

2) Text Rereading

3) Signifier Identifying

4) Signifier Labeling

Each of these processes are to be explored here as separate entities, as well as parts in a whole.

Viewed linearly, the first process in a Barthesian reading is a restructuring of the text that is an establishing of "a series of brief, contiguous fragments, which we shall call lexias, since they are units of reading."⁷ At one level, as a sub-process of the restructuring, there has to be a rereading of the text, in fact, an iterative rereading, in order to establish the lexias. Another sub-process of the restructuring is the identifying of the signifiers of the text. At a level equivalent to the restructuring, there is a text rereading process that is done in the context of the restructured text, of the lexias. The reading of these lexias consists of two processes: first, an identifying, again, of the signifiers contained in the lexias, and second, a labeling of these signifiers using a language developed by Barthes, complete with a grammar and a vocabulary.

The Barthesian reading process is a complex set of operations, with important interactions among the various subprocesses, which are iterated through until the Barthesian reader is finished. It is important to distinguish between the reader and the reading. The reading is only done when all possible connotative meanings have been noted, explored, and labeled. A reader might well declare the project done much sooner than that. As Barthes says:

With regard to the plural text, forgetting a meaning cannot therefore be seen as a fault. Forgetting in relation to what? What is the sum of the text? Meanings can indeed be forgotten, but only if we have chosen to bring to bear upon the text a singular scrutiny. Yet reading does not consist in stopping the chain of systems, in establishing a truth, a legality of the text, and consequently in leading its reader into "errors"; it consists in coupling these systems, not according to their finite quantity, but according to their plurality . . .: it is precisely because I forget that I read.⁸

Thus, there are always more meanings, more signifiers to be found, found and forgotten and read again.

The goal of a Barthesian (re)reading is "to establish a plural"⁹ of the text in terms of its ability to represent multiple meanings, as discussed in CHAPTER I. Having established this plural, this multivalency, and represented it via the Barthes' labeling language, the critic, armed with a critical model, may then proceed to the act of interpretation, using the signifiers unearthed in the Barthesian reading as the act's basis.¹⁰ This act, however, is not part of a Barthesian reading, but rather is subsequent, and therefore dependent upon, such a reading.

Of the four primary processes of a Barthesian reading, establishing the lexias of a text is a distinctly Barthesian act, and appropriately enough, the act that starts off any application of Roland Barthes' system. The next section explores this act of lexia making.

⁷Ibid., 13.

⁸Ibid., 11. 

⁹Ibid., 15.

¹⁰Ibid., 14.

The Primal Barthesian Act: Lexia Making

Rereading, as such, is not a Barthesian approach only--any deeply detailed reading for the facts requires that one reread, many times. Further, all systematic reading approaches, which have the goal of determining the facts of the text, must in some way focus on small portions of the text, usually at the sentence and/or paragraph level. However, what Barthes did was to not only define a particular space within which the rereading takes place, but one that is independent of the original syntactic and discourse-level structures of the text as a whole. That is, a lexia cuts across sentence, paragraph, section, and chapter boundaries at the will of the Barthesian reader.

According to Barthes,

this cutting up, admittedly, will be arbitrary in the extreme; it will imply no methodological responsibility, since it will bear upon the signifier, whereas the proposed analysis bears solely on the signified. 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 observe 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 [signifiers] to be enumerated.¹¹

By examining Barthes' practice in S/Z, one discovers that lexias are indeed small segments of text, usually less than one hundred words, though not always, that are derived sequentially from the main text without regard to syntactical structures, beginning with the title and proceeding to the last word of the text. Lexias do not overlap in terms of text content. In other words, if a portion of text is assigned to one lexia, it does not appear in any other. The size, and thus the composition, of these segments is determined by semantic structures, which Barthes refers to as signifiers, contained in the segment. Barthes states that lexias must have at least one such structure, but no more than three or four.¹²

¹¹Ibid., 13-14.

¹²Ibid.

Lexia-deriving is similar to cutting a piece of string, with a complex pattern formed by colored beads on it, into smaller pieces, starting at one end and proceeding toward the other, being careful not to cut beads, and only worrying about the bead patterns (small semantic structures) within the smaller unit, ignoring, at the time, the larger patterns of which they might be a part.

Clearly, Barthes' goal in defining a lexia, or reading unit, is to help focus all attention necessary on the smallest contributor to the patterns of meaning in a text as a whole. By requiring no more than three or four such contributors or signifiers, the focusing can be very strong, indeed. Further, by ignoring the larger grammatical structures of which the signifiers are embedded, Barthes' intention is to isolate them from all possible semantic contamination by larger structures. Barthes wanted, in both the lexia-deriving and the signifier labeling operations, to direct a reader's attention not just primarily, but solely on each signifier as it appears in the text, with any further global pattern discovery left to a later stage.

To illustrate, by example, what a lexia might look like, and to serve as a basis for continuing discussion concerning Barthes' method, two lexias, numbers 64 and 123, from Sarrasine (as Barthes read it in S/Z) are give here. In each case, the lexia is given in italics, as in the original, which is then followed by Barthes' signifier labels, using his "starred text"¹³ notation, for the signifiers in the lexia. Both lexias contain three signifiers, representing amongst them, all five types of signifiers as defined in Barthes' labeling language. The examples are:

(64) The old man did not want to leave this lovely creature, to whom he had attached himself with that silent and seemingly baseless stubbornness to which the extremely old are prone, and which makes them appear childish. * REF. Psychology of the old. ** SEM. Childishness. *** The castrato is attached to the young woman, opposite to opposite, the reverse of the coin by its obverse (SYM. Marriage of the castrato).¹⁴

(123) Young Marianina came in, and her innocent expression made her even more alluring than did her grace and her lovely dress; she was walking slowly and escorting with maternal care, with filial solicitude, the costumed specter who had made us flee from the music room, * ACT. "To enter": 2: the entrance itself. ** HER. Enigma 3: proposal and formulation (enigmatic as they are the relations between Marianina and the Old Man strengthen the enigma of the Lanty family: where do they come from? who are they?). *** SEM. Childishness.¹⁵

Consisting of text cut sequentially from the original, being relative short in length, and containing as they do only three signifiers each, both lexias appear to fall well within Barthes' criteria for lexias. The only questionable aspect would be that they both could be construed as being cut so as to have "regard for its natural divisions (syntactic, rhetorical, anecdotic)."¹⁶ The response to this criticism would be that Barthes' primary directive was that, regarding the definition of a lexia, "all we require is that each lexia should have at most three or four meanings to be enumerated."¹⁷ As will be shown in CHAPTER V, even this directive is not truly necessary to the proper functioning of a lexia as a unit of reading which allows it to "be the best possible space in which we can observe meanings."¹⁸

It is important to note that these two lexias contain between them examples of all five types of signifiers that form the basis for Barthes' labeling language: semes (SEM), references (REF), hermeneutics (HER), actions (ACT), and symbols (SYM).¹⁹ The following section discusses this language, using, as much as is feasible, these two lexias and their accompanying labels as a reference point.

¹³Ibid., 13.

¹⁴Ibid., 52-53.

¹⁵Ibid., 78.

¹⁶Ibid., 15.

¹⁷Ibid., 13-14.

¹⁸Ibid., 13.

¹⁹Ibid., 16-20.

The Barthesian Labeling Language:

Signifier Identifying and Labeling

As seen in the preceding lexias, Barthes' reading of Sarrasine applies five codes to label each of his five types of signifiers that can occur in any give lexia. These codes make up the primary vocabulary of his labeling language. The definition for each code is:

1. ACT--Action Code, labels signifiers that are fragments of action sequences.²⁰

2. HER--Hermeneutic Code, labels signifiers that are fragments of any enigma raised in the text.²¹

3. REF--Reference Code, labels signifiers pointing to nameable sets of knowledge outside of the text.²²

4. SEM--Semiotic Code, labels signifiers whose use is connotative in the ordinary sense of the term.²³

5. SYM--Symbol Code, labels signifiers that have a symbolic significance in the text.²⁴

According to Barthes, signifiers relate in one of five ways to their signifieds. Each of these ways is carried out by one of these five kinds of signifiers. To describe these five kinds of relating, Barthes uses the analogy of a text being "The Weaving of Voices,"²⁵ of which there are five types, with each voice corresponding to one of his signifier codes. Barthes' voices are the traces of or references to a text containing all of that which is already written, or as Barthes explains, they refer "to the Book (of culture, of life, of life as culture)."²⁶

Barthes names his voices:

1. Voice of Truth, traced by the Code: Hermeneutics

2. Voice of the Person, traced by the Code: Semes

3. Voice of Science, traced by the Code: References

4. Voice of Symbol, traced by the Code: Symbols

5. Voice of Empirics, traced by the Code: Actions.²⁷

²⁰Ibid., 19.

²¹Ibid., 19.

²²Ibid., 20.

²³Ibid., 17.

²⁴Ibid., 19.

²⁵Ibid., 20.

²⁶Ibid., 21.

²⁷Ibid., 20-21

An example Barthes uses to further his analogy is the title, The Kidnapping. He says that it immediately refers to, speaks about, all kidnappings, that it is a signifier of the REF type, spoken therefore by the Voice of Science. Thus, upon seeing the title The Kidnapping, a reader should delve into the Book of the Already-Written in order to find all that is known or spoken about kidnappings and then should bring this knowledge to the reading of the text.

In precisely the same manner, the other four codes are shorthand for four distinct voices that join with the voice of science to compose, to represent, to be, all that is written. That is, together they speak, they are, "the Book (of culture, of life, of life as culture) [emphasis added]" in a very complex weaving of traces of speech into a network that not only is that Book, but is the source of all other texts.

These voices aid in creating texts, that is, give meaning to them, by accompanying each utterance or signifier, in the text from off-stage, from that which is not the text, per se, but exists in, speaks from, the already-written texts that together form The Book. They lend to the signifiers as off-stage speakers, the connotative meanings Barthes is intent upon sifting out, pointing out, and labeling in his systematic approach to text reading.

The following five sub-sections address each voice or code in turn, exploring each voice's why, what, and how in terms of the Barthesian project. Following these is a detailing of the grammar, the syntax, of Barthes' labeling language, which uses his five Codes as its primary vocabulary.

²⁸Ibid., 20.

²⁹Ibid., 21.

³⁰Ibid., 21.

³¹Ibid., 20-21.

On Barthes' Voices: Truth

The Voice of Truth speaks through a text in conjunction with those signifiers where a mystery is played out from the announcement of its existence through various tactics for delaying revealment of true answers up to final disclosure. A text always contains one or more such mystery tracks, for such tracks are the forces that drive, the objects that attract, the reader through to the end of the text, and thus are the reasons that the text has readers. As Barthes says regarding one instantiation of the hermeneutic track concerning the Lanty fortune in the story Sarrasine:

had the discourse not moved the two speakers off to a secluded sofa, we would have quickly learned the answer to the enigma, the source of the Lanty fortune (however, then there would have been no story to tell).³²

 Barthes found six such mysteries or enigmas, that a reader must decipher in Sarrasine. Besides the source of the Lanty fortune, there are five other questions: 1), To whom or to what does the title "Sarrasine" refer?, 2), What is the secret of the old man, the grandfather?, 3), Who are the Lantys?, 4), Who is the Adonis in the portrait?, and, 5), what gender is La Zambinella?

The reader seeks Truth regarding these six mysteries by paying attention to Truth's voice print (the signifiers related to their unraveling), while the text attempts to hide those Truths.³³ The hiding of the Truth, as well as the revealing, is done through the exploitation of the denotative and connotative meanings of the signifiers that refer to the signifieds of the mysteries.

Thus, denotatively, by the rules of Italian, the name La Zambinella must refer to a woman. In fact, however, it refers to a castrato, a male who has been surgically held at the pre-puberty stage in order to keep his voice capable of the ranges normally reserved in adulthood to women. The clues to the deception depend on the reader determining that connotatively, any individual who carries a female name and is an opera singer in 18th century Italy, is, in truth, such a castrato. Such a stage name as La Zambinella was an affectation of the opera of that time and space.

³²Ibid., 31.

³³Ibid., 75-76.

On Barthes' Voices: The Person

The Voice of the Person speaks, unlike the other four voices, consciously and deliberately in the connotative mode. It takes direct advantage of this scheme of Western thought wherein there is a denotative meaning versus a connotative meaning for each signifier. The denotative is, theoretically, the authoritative or legal meaning, while the connotative is a shadow meaning, that, while there, is not the real or legal meaning. Connotative meaning thus allows a signifier to have multiple meanings.³⁴

When Balzac, deliberately it must be assumed, chose the name "Sarrasine" for his main character, a male, in a story about castration, the connotative meaning of the name is important for several reasons. One, in the French language it points to "femininity" with its final "e."³⁵ Two, it also contains the "Z" sound in its center, which in French is the sign of mutilation, thus connotatively pointing to the castration problem that Sarrasine will be faced with in the story.³⁶ Third, the "Z" sound is not in the name graphically, as the character "Z," which it should be according to French onomastics, but is, instead, present connotatively in the "S" sound.³⁷ This connotative aspect of the presence/non-presence of the "Z" underscores that gender determination is going to be a problem in the story.

³⁴Ibid., 7-9.

³⁵Ibid., 17.

³⁶Ibid., 106-107.

³⁷Ibid., 106-107.

Clearly, Barthes himself believed that the "S" and "Z" of the title were central to the whole of Balzac's story. First, he featured them in the title of his own work, joined and separated simultaneously by the "/" grapheme: S/Z. Then, in the text itself (one whole section, recursively titled "S/Z"), he addresses directly the meaning of the use of "S/Z" in Balzac's Sarrasine, particularly in the title.³⁸ The Voice of the Person, then, is an important voice in and of itself, and in fact, can be the Voice off which an entire text plays, with the other four singing harmony, or being background voices, as it were.

Barthes says that semes, the voice print, the trace of the Voice of the Person, are used to cite in the same sense as "the citar is the stamp of the heel, the torero's arched stance which summons the bull to the banderilleros."³⁹ That is, for example, with the Balzacian seme of Wealth, Balzac "cites the signified (wealth) to make it come forth, while avoiding it in the discourse."⁴⁰ As Barthes says,

The Party, the Faubourg, the Mansion are anodyne data, seemingly lost in the natural flow of the discourse: in fact they are touches designed to bring out the image of Wealth in the tapestry of the daydream.⁴¹

 In this manner, Balzac does not have to say: "Hey, everyone, the Lanty's are wealthy." Instead, through semes, signifiers that connotatively point to that which he wants known, the signified, he brings out an impression of wealth, much as a painter, through judicious use of mere particles of paint brings out the impression of a tree, or of a person. Indeed, Barthes says of the use of semes as citars:

the narrative technique is impressionistic: it breaks up the signifier into particles of verbal matter which make sense only by coalescing: it plays with the distribution of a discontinuity (thus creating a character's "personality"); the greater the syntagmatic distance between the two data, the more skillful the narrative.⁴²

Barthes also says that

the (ideological) goal of this technique is to naturalize the meaning and thus to give credence to the reality of the story: for (in the West) meaning (system), we are told, is antipathetic to nature and reality.⁴³

Thus, at least on one level, the "Person" Barthes is referring to in his phrase "Voice of the Person," could be thought to be the author of text, such as Balzac in his Sarrasine, using language to paint impressionistic pictures by employing the secondary meanings or connotations of words, deliberately and knowingly.

³⁸Ibid., 106-107.

³⁹Ibid., 22.

⁴⁰Ibid., 22.

⁴¹Ibid., 22.

⁴²Ibid., 22.

⁴³Ibid., 23.

On Barthes' Voices: Science

The Voice of Science speaks in a text whenever there is a reference made to bodies of knowledge, to cultural keepers of truth, of behavioral norms, to authoritative sources, including such texts as actual science or history or critical texts, outside of the text itself. For instance, Balzac at one point relies on the reader having the common human experience of "fairy tales." At another, he expects the reader to possess a knowledge of Literature, more specifically, of Lord Byron.⁴⁴ Connotative references to the Common Book of Human Knowledge are strewn throughout Sarrasine, and, indeed, it could be argued that one could not fully appreciate Balzac's work without access to that Book.

References, the voice prints of science, are used for many different purposes or functions, in a text. However, there is one that is certainly not intended, and is specifically tied to references to cultural codes, codes that quickly become, in essence, stereotypical, and thus ideological. Balzac relies on such cultural stereotypes as: Parisians as cynics, art as a constraint, youth as an effervescence. As Barthes says:

If we collect all such knowledge [cultural codes], all such vulgarisms, we create a monster, and this monster is ideology. As a fragment of ideology, the cultural code inverts its class origin (scholastic and social) into a natural reference, into a proverbial statement. . . . the cultural proverb vexes, provokes an intolerant reading; the Balzacian text is clotted with it: because of its cultural codes, it stales, it rots, excludes itself from writing (which is always a contemporary task): it is the quintessence, the residual condensate of what cannot be rewritten.⁴⁵

Thus, the cultural codes referenced in Sarrasine can, in fact, keep a reader from being able to read, or in Barthesian terms, rewrite, the novel.

⁴⁴Ibid., 205-206.

⁴⁵Ibid., 98-99.

However, if the Balzacian text is, indeed, capable of vexing a reader by using cultural codes, it nevertheless makes eloquent use of historical codes, making references to actual historical figures. Barthes uses a quote from Proust on Balzac, wherein Proust complains of Balzac's use of historical figures in a minor way as a way of introducing his own view of this narrative technique:

It is precisely this minor importance which gives the historical character its exact weight of reality: this minor is the measure of authenticity: Diderot, Mmd de Pompadour, later Sophie Arnould, Rousseau, d'Holbach are introduced into the fiction [Sarrasine] laterally, obliquely, in passing, painted on the scenery, not represented on the stage; . . . [thus] they give the novel the glow of reality, not the glory: they are superlative effects of the real.⁴⁶

Interestingly enough, it is the code of the history of music, and more specifically, of Italian Opera of the Eighteenth Century, upon which the entire hermeneutic chase of the reader (and of the young lady, Mme de Rochefide, who is the object of the narrator's passion and whose desire to know the answer is the reputed reason for the telling of the tale) through the story of Sarrasine and La Zambinella is based. Balzac depended upon his readers, along with his character, not knowing that castratos were used to sing female parts in operas.

In general, references to knowledge codes outside the text are used to manipulate the reader's response to the text. One goal is to get the reader to accept an illusion of reality for the text as a whole. Another is to help the author communicate to the reader through a common knowledge base on a variety of subjects, to set the scene by reference to the familiar. For instance, Balzac compares Marianina to the sultan's daughter in the story of the Magic Lamp to secure the reader's understanding of her beauty.⁴⁷ Finally, the codes can be consciously used by the author by playing them off against themselves and each other in order to introduce irony or parody, or, as in the case of the Italian Opera Code, mystery.

⁴⁶Ibid., 101-102.

⁴⁷Ibid., 32-33.

On Barthes' Voices: Symbol

The Voice of Symbol speaks in the text whenever a signifier is used in terms of a structure, either created in the text, or pre-existent in the already-written (The Book) and is used to enter into a symbolic field, such as the human body. In other words, such a signifier does not stand alone, but is related to other signifiers in the text that together form a nameable, visible, tangible object in the text. This object is a key used to open exploration of a certain kind into a given world (field), replete with culturally known, or shared, meaning on a symbolic level.⁴⁸

In Sarrasine, the symbolic field used is that of the human body. The body contains (is stamped with, is replete with), from the already-written, many known symbolic values that Balzac uses to make his own points. One such is the value that states that violation of the natural separation of gender via castration leads to disasters of several kinds (devaluation of the meaning of life, threats to the survival of the race, faulty economic valuations, etc). Belief in, or at least knowledge of, this value is vital to a reader gaining an understanding of the text.

Balzac, according to Barthes, uses three attributes of the body that are replete in symbolic value: its position as the basis for meaning itself; its position as the basis for reproduction (or sexual creation); and its position as the basis for economy. Three structures give entrance into this symbolic field: the rhetorical figure of Antithesis explores the body in terms of the creation and sustenance of meaning for individuals and the race; the cultural figure of the Castrato explores the body in terms of reproduction as a need of individuals and the race; and the cultural figure of Wealth explores the economic role of the body for individuals and the race.⁴⁹

⁴⁸Ibid., 214-215.

⁴⁹Ibid., 214-215.

Antithesis is used heavily in Sarrasine, with the human body as subject. Barthes says that this rhetorical figure's

apparent function is to consecrate (and domesticate) by a name, by a metalinguistic object, the division between opposites and the very irreducibility of this division. . . . the antithesis is the battle between two plenitudes set ritually face to face like two fully armed warriors; the Antithesis is the figure of the given opposition, eternal, eternally recurrent: the figure of the inexpiable.⁵⁰

But the wall of the antithesis can be transgressed according to Barthes, in a narrative text, and it is in Sarrasine. Ironically, it is in the breaching of the antithetical wall that meaning is produced. Barthes informs us that

it is fatal, the text [Sarrasine] says, to remove the dividing line, the paradigmatic slash mark which permits meaning to function (the wall of the Antithesis),⁵¹

which implies that one "meaning" of the text is that we dare not transgress the antithetical wall in the real world. The story (Sarrasine) shows the effects of doing so by transgressing in the symbolic field of the human body.

Early in Sarrasine, the body of the narrator, sitting between the outside and the inside, between the wintry garden outside and the warm, even hot, party inside, acts as the mediation point between the opposing abstract plenitudes symbolically represented by the outside and inside of the mansion where the party is being held. Barthes says that

mediation upsets the rhetorical--or paradigmatic--harmony of the Antithesis (AB/A/B/AB) and this difficulty arises not out of a lack but out of an excess: there is one element in excess, and this untoward supplement is the body (of the narrator).⁵²

The narrator transgresses, breaching the wall of the antithesis by describing the differences in the world outside the mansion and the world inside: death and life; cold and heat; outside and inside. The scene thus set,

by way of this excess [the narrator] which enters the discourse after rhetoric has properly saturated it that something can be told and the narrative begin,⁵³

⁵⁰Ibid., 27.

⁵¹Ibid., 215.

⁵²Ibid., 28.

⁵³Ibid., 28.

Balzac is signaling or signifying, that the story is centered on the different symbolic values that the narrator is mediating, and, in fact, that this centering is really on the results of the breaching of these antithetical values.

Later, two other bodies, that of the young lady the narrator desires, Mme de Rochefide, and the castrato, whose story she desires, are brought together, in fact, they touch. Thus, there is a transgression of the wall, the antithesis that the two represent in their bodies, when they meet and she touches him. Symbolically, the young lady (her body) means life, heat, and insideness, while the castrato means death, cold, and outsideness. Upon touching,

the physical contact between these two . . . produces a catastrophe: there is an explosive shock, a paradigmatic conflagration, a headlong flight of the two bodies brought together in so unseemly a manner: each of the partners forms the locus of an over physiological revolution: . . . the depths are emptied, as in vomiting.⁵⁴

According to Barthes,

this is what happens when the arcana of meaning are subverted, when the sacred separation of the paradigmatic poles is abolished, when one removes the separating barrier, the basis of all "pertinence."⁵⁵

That is, meaning is lost. Mme de Rochefide's response to the narrator's tale, in which Sarrasine dies due to the culpability of the castrato as a transgressor of the antithesis between maleness and femaleness, is to suspend herself in life, to remain "pensive" while she awaits death. For her, the destruction of the antithesis barrier between the two meaning systems, maleness and femaleness, leaves no meaning to life.⁵⁶ It is left to the reader of Sarrasine, to accept or reject her response.

⁵⁴Ibid., 65.

⁵⁵Ibid., 65.

⁵⁶Ibid., 216-217.

Antithesis is but one of the symbolic narrative methods Balzac uses, the other two methods focusing on the body in terms of its reproductive and economic symbolic meanings. The young lady, Mme de Rochefide, is the embodiment of the capability of a future for the human race--she is capable of reproduction, and she is sufficiently attractive to ensure that she can and will reproduce. The old man, the castrato, represents the null future where bodies are not used for reproduction, for love, but instead for economic gain--he is a rich man, but he has no offspring--there is no future for him through children.

Related to the latter is the economic contract between the narrator and the young lady--in return for his story concerning the body which served as the model for the figure of Adonis in the painting, she is to yield herself to the narrator. She, of course, breaches the contract, something that is only possible because of the removal of the barrier between antithetical plenitudes--one where it is never ethical to breach contracts, and another where it can be.

The Voice of Symbol allows the Balzacian text to explore the loss of meaning, of value in the world, through the depiction of an event (the death of Sarrasine), thus, of all humankind, that occurs through the mutilation of the human body, and by extension, the mutilation of all that the human body represents to humanity.

On Barthes' Voices: The Empirics

The Voice of the Empirics speaks in a text whenever there is action. The Empiric voice within a text speaks the logic of cause and effect, that is, a given action will have a given effect. This is a logic that is well known and fleshed out with a multitude of examples in Western thought and writings and so quite accessible as a source to Western authors and readers.⁵⁷ Like the Voice of Symbol, the Empiric voice is spread across the text in fragments that are clearly related, that form sequences. To fully comprehend them, one must bring together the signifiers of each part of each sequence of actions.

In Sarrasine, Barthes found forty-eight separate, unique action sequences spread out over the lexias, to each of which he gave a unique name, such as: "To be deep in," "To laugh," or "Dressing." Each element of a sequence, like "To be deep in," was given a qualifier (not necessarily unique within the sequence), as well, such as "To be absorbed" or "To come back again." In other words, the sequence "To be deep in" could occur over several lexias, with several qualifiers being used to differentiate the elements of the sequence in terms of their role in the sequence.

In the first lexia, the element might depict a subject being absorbed, while in the next lexia, the subject might be coming back from being absorbed. In both cases, the action was part of a sequence named "To be deep in." By looking at the lexias which are part of an action sequence, one can see the logic of cause and effect being worked out for an action in terms of the text's discourse, and in comparison and contrast to similar sequences in other texts.

⁵⁷Ibid., 18.

In S/Z, Barthes discusses the concept of action sequences in terms of their folding and unfolding through the text. A reader follows the unfolding of a sequence through the text to its end, but in order to name that sequence, the reader must fold it back up. For example, in the following text, Barthes explains what is meant by this:

What is a series of actions? the unfolding of a name. To enter? I can unfold it into "to appear" and "to penetrate." To leave? I can unfold it into "to want to," "to stop," "to leave again." To give?: "to incite," "to return," "to accept." Inversely, to establish the sequence is to find the name:⁵⁸

Thus, given the sequences "to appear" and "to penetrate," one needs to find a name that acts as an umbrella for the constituents. In this case, the constituents are part of a sequence where the individual actions have to do with "entering," thus the name "to enter" for the sequence.

Barthes maintained that these sequences were in an order that is dictated by culture and language:

The unfolding of the sequence, or, inversely, its folding, occurs by and through the authority of the great models, cultural (to give thanks for a gift), organic (to disturb the course of an action), or phenomenal (the sound precedes the phenomenon), etc. The proairetic sequence is indeed a series, i.e., "a multiplicity possessing a rule of order" (Leibnitz), but the rule of order here is cultural (habit, in short) and linguistic (the possibility of the word, the word pregnant with its possibilities).⁵⁸

⁵⁸Ibid., 82.

⁵⁹Ibid., 82.

In other words, in a sequence with the name "Farewell," the contingent actions of "to say farewell," "to confide," and "to embrace" are all actions that are culturally dictated to be part of a series of actions that results in one or more people parting company from each other. The author does not invent them, but merely uses them as the story being written requires their presence. Similarly, in the world, our world, during thunderstorms one sees the lightning then hears the thunder;thus, one would expect an action series named "Thunderstorm" to have those actions, in that order.

The task, then, of a Barthesian reader regarding the Voice of Empirics is a two-part operation. First, one must recognize the individual actions of the series within each lexia. Second, one must find or choose the appropriate name for each series. Having done this, the Empiric Voice print will thus be established and noted for future reference.

On Barthes' Voices: A Statistical Picture

Below is a table that shows how often Barthes' Voices occur in Sarrasine, the number of related sequences, and the average, smallest, and largest number of labels in a sequence. For instance, there are 135 HER labels used for six unique mysteries presented in Sarrasine, by Barthes' count, for an average of 22.5 references per mystery, with the smallest number being two and the largest being 84.

TABLE 1: CODE OCCURRENCE DATA FOR Sarrasine

This table shows that within a single, given text, such as Sarrasine, Barthes' signifiers are boundable, that is, they are not infinite in number--they are determinable. Thus, the statistics in the above table constitute an empirical basis for believing Barthes' system is applicable to a text.

However, when one considers all texts existent in the world, together, as a single unit of study, the number of unique labels and the frequency with which they occur is not so readily bounded. This is important because of Barthes' definition of connotative meaning, of which all five codes are a label for a type. His definition implies (actually, parenthetically, it states) that connotative meaning is a relation between two or more sites across texts:

Definitionally, it [connotation] 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 [emphasis added])⁶⁰

⁶⁰Ibid., 8.

Thus, one could conclude that Barthes had reason to believe that his system could be used as a tool for analyzing texts in terms of their use of his five kinds of connotative meaning. Presumably, this could be done by comparing the sets of labels for two or more texts, searching for patterns between or among them that reveal interesting critical insights in terms of how the texts differ or are the same.

For instance, by studying the use of a particular seme, such as "Wealth" or "Femininity," in two or more texts, one might be able to come to some interesting conclusions about how these semes are used by male authors versus female authors. This is, of course, just one possible use of such a cross-text analysis based on a set of Barthesian readings.

However, while the statistics in the table above support the single-text analysis done using a Barthesian reading, more study would be needed to see if the cross-text analysis makes any empirical sense. The problem of a possibly infinite number of semes, action sequences, hermeneutics, symbols, and references needs to be explored.

On Barthes' Voices: A Summation

For Barthes a text is not simply what it says, as, for instance, strict constructionist lawyers, univocalists to their core, would have it. Instead, a text is the reverberations, echoes, pointers, and references to and of other places in the text and in other texts. Secondary meanings, the connotations of words and phrases, are the key to these other "sayings," of Barthes' five Voice prints in the text: Empirics, Truth, Science, Person, Symbol. Barthes' codes (ACT, HER, REF, SEM, SYM) are labels that name these sayings, these utterances, these Voices, of meanings of the text. Or, as he says: "the convergence of the voices (of the codes) becomes writing, a stereographic space where the five codes, the five voices, intersect."⁶¹

The voices, or codes, "whose origin is 'lost' is vast perspective of the already-written,"⁶² are

a perspective of quotations, . . ., are themselves, always, ventures outside the text, the mark, the sign of a virtual digression toward the remainder of a catalog (The Kidnapping refers to every kidnapping written); they are so many fragments of something that has always been already read, seen, done, experienced; the code is the wake of that already. Referring to what has been written, i.e., to the Book (of culture, of life, of life as culture), it makes the text into a prospectus of this Book.⁶³

Thus, a text is "the weaving of voices,"⁶⁴ is a compilation, a comprisement of texts that the author has heard, seen, and experienced. It becomes a newly-shaped, newly-designed, never-before-existent tapestry. One might compare this to the art of quilting, where a quilt is constructed from "fragments" of other, already-existent articles made of cloth, into a pattern, a whole, that has never been before, which, in its own time could quite likely become part of some other quilt or article of cloth. The following sub-section details the grammar of the language Roland Barthes created for noting each of the traces of those interwoven voices in a text.

⁶¹Ibid., 21.

⁶²Ibid., 21.

⁶³Ibid., 20-21.

⁶⁴Ibid., 20.

The Barthesian Grammar: The Codes

Roland Barthes created a language for labeling the traces, the signifiers of his five voices. Earlier, two lexias and their accompanying labels were given. These are repeated here, on the next page in Table 2, for ease of reference.

TABLE 2

LEXIA LABELING EXAMPLES

(64) The old man did not want to leave this lovely creature, to whom he had attached himself with that silent and seemingly baseless stubbornness to which the extremely old are prone, and which makes them appear childish.

* REF. Psychology of the old.

** SEM. Childishness.

*** The castrato is attached to the young woman, opposite to opposite, the reverse of the coin by its obverse (SYM. Marriage of the castrato).⁶⁵

(123) Young Marianina came in, and her innocent expression made her even more alluring than did her grace and her lovely dress; she was walking slowly and escorting with maternal care, with filial solicitude, the costumed specter who had made us flee from the music room,

* ACT. "To enter": 2: the entrance itself.

** HER. Enigma 3: proposal and formulation (enigmatic as they are the relations between Marianina and the Old Man strengthen the enigma of the Lanty family: where do they come from? who are they?).

*** SEM. Childishness.⁶⁶

⁶⁵Ibid., 52-53.

⁶⁶Ibid., 78.

Lexias 64 and 123, in Table 2 above, were found by Barthes to contain three signifiers, with at least one from each of the five voices. As the examples show, each of the five label types can vary in form, but there are common attributes. While Barthes made no attempt in S/Z to detail his language formally, it is clear that a formal grammar can be described. Table 3, below, presents such a grammar. Following Table 3 is an extended description of the labels in terms of the grammar presented in the table.

Table 3 uses production rules to present the syntax of Barthes' language. These production rules are given in a modified form of that normally used in linguistic and formal grammar applications, since Barthes' language is much simpler than those normally explored in those disciplines. The primary modifications are the substitution of English terms for many of the symbols normally used, in conjunction with the dropping of many of the operations/symbols that production rule syntax presentations generally use.

A production rule is a presentation of the components of a language and the relationships those components can enter into in order to form a given component of the language. A set of production rules that describe how all the components of a language are formed is sufficient to described the syntax of that language. An example of a production rule for a simple noun phrase syntax might be:

NOUN PHRASE is produced by ((ARTICLE followed by NOUN) or (one or more sequential NOUNS))

This rule says that a noun phrase can be constructed of an article (such as "the," or "a") followed by a noun, or it can be a sequence of nouns. The rule assumes that the definitions of article and noun are in the set of production rules for the language being presented, or that they are terminal symbols that do not need definition. Vocabulary terms in any given language normally constitute the set of all terminal symbols for that language.

In this case, both noun and article are not terminals, since the listing of all possible articles or nouns in the language could serve as the definition of either of them. By definition, non-terminal symbols must appear somewhere in the set of production rules on the left-hand side, since they are made up of other components, while terminal symbols can only appear on the right hand side, since they cannot break down into smaller components.

The "()" symbols are used to group parts of rules together, similar to the presentation of equations in mathematics. In this example, they are used to show two disjunctive means for forming a noun phrase, one with an article and noun, another with only nouns.

In Table 3, as in the example, terminals are in bold, either uppercase, lowercase or a combination, and contained in double quotes. Non-terminal symbols are in bold uppercase, only. Production rule operations are in italicized lowercase.

TABLE 3

BARTHESIAN LABEL LANGUAGE GRAMMAR RULES

Rule 1 shows that a Barthesian signifier label is produced by two components: CODE and DESCRIPTOR. Rule 2 shows that the CODE component can be produced by any one, but only one, of the five three-letter terminal signs given. Rule 3, in conjunction with Rules 4, 5, 6, and 8, shows that the DESCRIPTOR component is produced by terms of the labeler's choice, though in the ACT, HER, and ANTITHESIS forms, as given in Rules 4, 5, and 6, there are certain nuances as to how the terms are arranged, specified. In all cases, the labeler has the option of amplifying or further describing the signifier being labeled by adding a NOTE to the DESCRIPTOR field of the label.

The lexias in Table 2 show several ways the grammar of the labels may vary. The REF label in Lexia 64 has as its DESCRIPTIVE_TERM, a phrase: "Psychology of the old," and the DESCRIPTIVE_TERM for the SEM label in both lexias consists of one word, "Childishness," with no NOTE attached to any of the three. The SYM label in Lexia 64, however, has a NOTE before the actual label begins, while the HER label in Lexia 123 has a NOTE at the end. The position of the NOTE is clearly immaterial, as its function is the same regardless of position--clarifying the information that the DESCRIPTIVE_TERM is to convey.

The DESCRIPTOR field for ACT labels, as in Lexia 123, and described in Rule 3 and 4, contains information from more than one DESCRIPTIVE_TERM. Most label types in the language have but one such field, the ACT label being an important exception. The Lexia 123 ACT label is:

ACT. "To enter": 2: the entrance itself.

As Rules 1 and 2 require, it begins with one of the five codes, "ACT," than as Rule 3 and 4 require, this is followed by a DESCRIPTIVE_TERM. "To Enter," which is the first use of a DESCRIPTIVE_TERM in the label, is the DESCRIPTIVE_TERM for the entire action sequence of which the particular signifier being labeled is a fragment. That is, in Sarrasine, there are fragments of a single action, labeled as "To Enter," that are spread across two or more lexias.

The action fragment being signified here is the second such fragment in the sequence as one reads the text from beginning to end. This fact is recorded in the label as a number, "2," in the NUMBER field of the ACT_DESCRIPTOR.

Following the NUMBER field, there is a second DESCRIPTIVE_TERM which describes the particular fragment being signified; that is, "the entrance itself" is signified in this lexia. Thus, while the action sequence as a whole concerns itself with the action of entering, Lexia 123 contains a fragment that is an actual entrance taking place.

Finally, while this particular example has no NOTE, it is important to remember that it could have had one according to the grammar of the Barthesian language.

The DESCRIPTOR field for the HER label is, like the ACT_DESCRIPTOR field, somewhat more complex than the normal DESCRIPTOR. Below is the Lexia 123 HER label:

HER. Enigma 3: proposal and formulation (enigmatic as they are the relations between Marianina and the Old Man strengthen the enigma of the Lanty family: where do they come from? who are they?).

First, after the CODE field containing the sign, "HER", the HER_DESCRIPTOR field follows. The HER_DESCRIPTOR field is unlike any of the other DESCRIPTOR fields, containing as it does the word "Enigma" as part of the label. This is redundant, of course, since by definition, HER labels always refer to enigmas. Nevertheless, Barthes consistently used it in his examples in S/Z.

What is not redundant, at this point, is the integer in the NUMBER field that follows the word "Enigma." This integer is the number assigned to the particular Enigma being connoted in this lexia. In Sarrasine, as noted before, there are six such Enigmas that the reader is invited by the text to unravel. Thus, unlike the ACT label where a DESCRIPTIVE_TERM is used to label the whole action sequence of which the label is but a fragment, for the enigmas a number is assigned as the sign of the whole. In Barthes' handling of Sarrasine using his labeling method, Enigma 3 happens to be the question of who the Lanty family might be. As can be seen in this example, Barthes chose to carry this information in his NOTE field for HER labels.

Following the enigma number, HER labels contain one or more of the terms shown in Rule 7, the ENIGMA_TYPE rule. As rule 7 allows, and the example in Lexia 123 depicts, the possibility of more than one of these terms being applicable exists. A question can be 'proposed' and a 'formulation' be given, for instance, all in the same lexia.

In this example, there is a proposal (a raising, according to Barthes) of the question of just who the Lantys' are. Further, the question is formulated by giving descriptive details about how two of the Lantys relate to each other in public. There is no answer here, but rather clues as to how to state the question.

In general, the terms 'theme,' 'formulation,' and 'proposal' are qualifiers of the term 'question,' though all four can stand alone, or the first three can appear in a label with 'question' or with each other. When 'question' stands alone in S/Z, it usually stands in for 'proposal'. Thus, a signifier can point to the theme or subject of the enigma, and be labeled 'theme.' Or, it can state the question, and thus be labeled 'formulation.' And finally, it can simply raise a flag, saying that there is an enigma here, but that is all that can be discerned in the lexia.⁶⁷

The terms 'Promise of answer,' 'snare,' 'ambiguity,' 'suspended answer,' 'partial answer,' and 'jammed answer,' are used to signify a delaying action in terms of disclosing the answer to the enigma. According to Barthes, it is the goal of the text to hide the answer from the reader, and from certain of its characters, as well, as long as feasible. Delaying actions, taking the form of particular rhetorical devices, are usually employed by the author for this purpose.⁶⁸ The terms listed here are Barthes' names for these devices.

Finally, 'Disclosure' or 'answer' are used to label the signifier that gives the reader and/or the characters the answer to the question.⁶⁹

⁶⁷Ibid., 84-86.

⁶⁸Ibid., 75-76.

⁶⁹Ibid., 209-210.

Closing out the examples and the label types are the SYM labels. These usually consist of the terminal sign "SYM" in the CODE field, followed by a DESCRIPTOR field containing a DESCRIPTIVE_TERM, with an optional NOTE field, as in the example in Lexia 64 in Table 2. But there is one major exception to this rule.

The exception is in the case where the rhetorical figure of Antithesis is used as a symbolic structure. For Antithesis used symbolically, Barthes clearly wanted to be able to note which of the two, or both, terms of the antithesis structure was being signified in a lexia, as well as the kind of information about the Antithesis being given in the lexia by the particular signifier being labeled.

To illustrate this concept, an example of a lexia, Lexia 55, and the Antithesis label given it in Barthes' treatment of Sarrasine, complete with a NOTE field, is given below, which is followed by a discussion of the whole:

(55) By one of those tricks of nature, the half-mournful thought turning in my mind had emerged, and it appeared living before me, it had sprung like Minerva from the head of Jove, tall and strong, it was at once a hundred years old and twenty-two years old; it was alive and dead. * SYM. Antithesis: AB: mingling (the wall of the Antithesis is breached)⁷⁰

The Antithesis signified here is the one between Death (term "A") and Life (term "B"). The only way to know this is to be aware of earlier terms of the same Antithesis where Barthes made it clear that there is an Antithesis structure in the text that has these two terms as its components. The label shows by having the terminal sign of "AB," in the ANTITHESIS_FIG field as described in Rule 6, that both terms of the Antithesis figure are being signified.

In the lexia itself, the narrator manages to "mingle" the two terms, "A" and "B", to make them one. In the classic Antithesis rhetorical figure, this is a transgression of the wall that holds the two terms apart. The word "mingling" in the DESCRIPTIVE_TERM is then used to signify that in fact the Antithesis signifier in the lexia signifies a transgression of the barrier via a mingling of the two terms. The NOTE field following the DESCRIPTIVE_TERM clarifies this further.

⁷⁰Ibid., 49.

The Barthesian Labeling Language: An Example Application

In this section an example application of the Barthesian labeling language will be worked out using Lexia 259 generated from William Faulkner's "The Bear." This lexia was produced using Xebra as part of the evaluation of the combined Barthesian/Xebran system. At this point, the goal is to show how a user of the language might decide what signifiers need labeling and how those signifiers can be labeled using the Barthesian labeling language. Table 4 contains the lexia as it was generated via Xebra, followed by its associated labels as constructed for the evaluation of the system as performed for this dissertation.

TABLE 4

SIGNIFIER LABELING SAMPLE

^aIbid., 248.

Lexia 259, above, is labeled with four of the five label types, with only REF missing. The first such label is an ACT label. When reading the lexia, it is clear that Ike is in the midst of refusing to go back to town, because, as the reader knows from earlier lexias, he wants to stay in the woods in order to witness the death of Sam Fathers. This is a death that he is certain will occur, as previous lexias have revealed, and which does occur, as subsequent lexias will show. Because this action fragment is but one in a series, it is labeled in the first DESCRIPTIVE_TERM field as a member of the "To Go Back" series.

There have been other times in the text when the action "To Go Back" has been cited, ten times, to be exact, as can be seen from the NUMBER field in the ACT label. Barthes is not completely clear in his practice of the numbering of action sequences. Sometimes, as in his "To Narrate" action sequence, he numbers all "To Narrate" fragments as part of one continuous action sequence, though in the text of Sarrasine one could be justified in finding several sub-sequences that are more related to each other than to other "To Narrate" fragments. Other times, as in his three sequences, Door I, Door II, and Door III, he breaks a sequence of the same sort of action into three components.⁷¹ In the evaluation of the Barthesian/Xebran system done for this dissertation, the former practice of numbering continuously, without sub-sequences, was the chosen practice. Thus other "To Go Back" action fragments do not necessarily apply to Ike's refusal to go back to town at this point in text, which would have been a sub-sequence had the other method been used.

⁷¹Ibid., 255-257.

The specific action taking place in this fragment of the "To Go Back" action sequence is Ike's refusal to do so, despite orders to the contrary. Thus, the second DESCRIPTIVE_TERM contains a reference to that effect.

McCaslin attempts to assure Ike that his return to town is a reasonable action, that Sam will be all right, so there is no need for Ike to stay. This attempt on McCaslin's part is two signifiers at once. First, it is a HER signifier, proceeding towards the goal of catching the reader in the "snare" of believing that maybe Ike is wrong, that indeed Sam is going to live. Second it is a SEM signifier, citing the "Blindness" of all the adult participants to the true nature of Sam's sickness. Characters in "The Bear" are continuously falling prey to this weakness of being able to see through to the truth, and the SEM "Blindness" serves to mark each instance of this problem.

In the SYM label arena, two symbols are found to be actively signified in this lexia. First, the symbol "Endings" is cited, and second the symbol of Antithesis, with its A Term of Ideal Nature, represented in Ike and Sam Fathers, and its B Term in McCaslin and Boon. Endings are seen to be wrenching, demanding affairs, that require the participants to act out their roles--Ike must participate in the coming death of Sam Fathers.

Concurrently, Ike's struggle to take up his role with regard to Sam is symbolic of the whole struggle between Ideal Nature and Non-Ideal Nature. In this lexia he is attempting to stay behind to aid in the action of canceling, uplifting, and preserving of the two terms via the deaths of Old Ben, Sam Fathers and Lion: the Holy Spirit, God, and Christ vis-a-vis Ideal Nature, respectively.

It should be clear from the above discussion, that most of the labels could not have been generated without the labeler having read and analyzed all the other lexias in the text for their constituent signifiers. For instance, one needed to realize that an enigma concerning Sam's fate was underway in order to code the HER label. Further, one needed to know that McCaslin was wrong in his diagnosis of Sam's state in order to see that the SEM "Blindness" was cited in the lexia. Still more clearly, one needed to know that throughout "The Bear" there is a struggle of two antithetical terms being played out, of Ideal and Non-Ideal Nature, before the SYM of Antithesis could be attached to this lexia. Even the ACT label could only be fully formed with the knowledge of other such "To Go Back" fragments if the numbering within the sequence was to be correct. Only the SYM for "Endings" depended on no other knowledge of the text of "The Bear" itself, and even it should not necessarily be attached if this is a unique, non-recurring symbol in the text.

Applying Barthes' labeling language to a series of lexias is an exercise in continual reading and rereading and noting and renoting of recurring signifiers and signifieds. It involves the labeler in a ever-deepening process of applying terms, re-thinking terms, re-structuring terms, until finally one has exhausted the possibilities. This aspect of the Barthesian reading system, that is the types and degree of effort required will be explored in greater detail in this and later chapters.

The Barthesian Labeling Language: A Closing Summary

The Barthesian labeling language has one purpose in the Barthesian reading system: noting signifiers in the text being read using his five codes. This noting of signifiers is the primary substance of that process identified earlier as the "Signifier Labeling" process. Before any signifier can be labeled it must be identified. Yet this is a "chicken and egg" type of situation. As stated earlier, Barthes notes that, in order to "fold" an action sequence back into its "name," that is, its label, it is first necessary to "unfold" it, then work back to a common umbrella term that manages to encompass the spirit of the action as a whole.⁷² To do this requires naming each fragment, but in order to do that properly, the name of the whole is useful. Thus, the process of identifying and labeling the signifiers can be a highly recursive procedure.

⁷²Ibid., 82-83.

In terms of the language itself, it is clear from Barthes' practice in S/Z that he was not absolutely concerned with consistency within the language in terms of its grammatical structure. He often used whatever form seemed most convenient to him for any given signifier.

However, he often took care to make certain that labels that might later need to be collated together across lexias, especially the ACT and HER and SYM Antithesis labels, could be easily brought together by any subsequent user of his data stream for that text. Thus, the primary goal of producing a useable set of data points to be used by future interpreters of the text determined his labeling practice.

Those few exceptions where Barthes did not follow the general form of particular label types (and there were several such for each type) can be thought of as outliers in the statistical population represented in S/Z. Any future user of Barthes' system, whether manually or via a computer-based tool, such as Xebra, should avoid such ambiguities, since they can cause confusion in what is already a complex undertaking. It was the design goal for Xebra, in fact, to make such usage errors difficult to execute.

 The First, Middle, & Final Barthesian Act: Rereading

As noted earlier, re-reading is not specifically a Barthesian act when reading for the facts of the text. What is important, however, in the Barthesian context, is that rereading is involved in every step of a reading as Barthes prescribes it. To review, the four acts of a Barthesian reading are:

1. Text Restructuring

2. Text Rereading

3. Signifier Identifying

4. Signifier Labeling

In order to restructure a text, that is, make lexias, the Barthesian reader must reread the text, iteratively, looking for and identifying signifiers, because the space of a lexia is defined by the number of signifiers it contains, ranging from one to four. In order to label the signifiers, the reader must again reread the text, looking for clues as to what would constitute reasonable names for the signifiers. Then, in order to put all of this together, the Barthesian reader must, at least one more time, reread the text in order to produce a final laying out of the lexias and their accompanying labels.

Barthes views that act of reading a text as necessarily consonant with rereading:

rereading is here suggested from the outset, for it alone saves the text from repetition (those who fail to reread are obliged to read the same story everywhere), multiplies it in its variety and plurality: rereading draws the text out of its internal chronology ("this happens before or after that") and recaptures a mythic time (without before or after); it contests the claim which would have us believe that the first reading is a primary, naïve, phenomenal reading which we will only, afterwards, have to "explicate," to intellectualize (as if there were a beginning of reading, as if everything were not already read: there is no first reading, . . . rereading is no longer consumption, but play (that play which is the return of the different). If then, a deliberate contradiction in terms, we immediately reread the text, it is in order to obtain, . . ., not the real text, but a plural text: the same and the new.⁷³

Rereading is the first act, the middle act (repeated many times), and the final act of a Barthesian reading. Indeed, it is, in a Barthesian sense, the only act.

This act of rereading is both the strength and weakness of the Barthesian reading system. Only through good reading can fact gathering be done well: a Barthesian reading requires a great deal of reading, in a structured manner. From a fact gathering view, this rereading helps to ensure the accuracy and completeness of the rereading process. From Barthes' view, it is the necessary and sufficient act for appreciating the plural of the text. On the other hand, the requirement of performing so much rereading can be difficult to fulfill, or as Barthes says: "to read, in fact, is a labor of language."⁷⁴

The rest of CHAPTER II focuses on this and other problems of the system, having in the first half described, first in theory, then in practice, what the system is and how it is applied.

⁷³Ibid., 16.

⁷⁴Ibid., 11.

The Barthesian System: Its Problematic Attributes

The Barthesian system is problematic in two areas: in its theory and in its practice. In CHAPTER I, the discussion of the problematic aspects of the system was general and abstract. At this time, the goal is to enter more deeply into the what and how of those problems in terms of their existence and functioning within the Barthesian reading system.

However, not all the problems referred to in CHAPTER I are addressed. As stated in CHAPTER I, the theoretical problems that are based in rationalistic objections to empirical solutions, such as those found in the Barthesian system, were only noted at that time, and are not to be addressed any further, as this is an empirical study of an empirical system. The theoretical aspects of the system that are troubling from an empirical view, however, will be addressed in the following discussion.

The major premise of this dissertation, as stated in CHAPTER I, is that problems of an empirical nature within the Barthesian system can be productively dealt with by the use of a computer-aided version of the system. CHAPTER III gives a complete view of Xebra, the computer-based Barthesian tool that is the result of implementing this premise. However, at this point, the discussion is focused on the particular qualities of the problems in the Barthesian reading system that are to be addressed within the Xebran framework and not on the Xebra-based solutions.

The Barthesian System: Its Problematic Theory

There are three kinds of problems with the Barthesian system that have a problematic basis, both in theory and in practice. These are: precision, consistency, and completeness. From a theory perspective, one needs to understand how one should approach a Barthesian reading in terms of these three attributes. From a practice perspective, one needs to know whether the Barthesian system is, in fact, sufficiently precise, consistent, and complete in and of itself in order to actually do what the system is intended to aid a reader in doing. Finally, in addition to these problems, there is another separate area of concern, that, while not particularly troublesome in practice, can nevertheless trouble a theorist. This is the issue of parts and wholes: how should one view the relationship(s) of one to the other? What follows is a discussion of these issues.

Barthes In Theory: Precision, Consistency, Completeness

Barthes was silent regarding the need for precision and consistency during the production of a Barthesian reading. One can only infer his view from his practice. From it, one sees that he was concerned with both. He displayed a strong precision and consistency in his work, first, in terms of label content, particularly in the descriptive terms used, and second, in the signifiers chosen and labeled.

As for completeness, Barthes addressed this issue in the section, "V. READING, FORGETTING"⁷⁵ in S/Z, wherein he celebrates the practice of forgetting what one reads. He appears certain that forgetting and reading are inevitably, inextricably intertwined, saying ". . . it is precisely because I forget that I read."⁷⁶ It is the finding of meanings, the act of nominating the meanings, then letting them go, that comprises the act of reading for Barthes.⁷⁷

Why is this so? The answer lies in the Barthesian question, "What is the sum of the text?"⁷⁸

Certainly, for Barthes, it is not the meanings found and named in reading. This handling of a text is merely "a nomination in the course of becoming, a tireless approximation, a metonymic labor."⁷⁹ It is a labor that has no end, can have no end, for to end it would "reconstitute a singular, theological meaning,"⁸⁰ which Barthes denies as possible. Instead, there is "an irresponsibility of the text"⁸¹ constituted in a plurality of meanings. A singular, theological meaning does not admit irresponsibility, incompleteness in the chains of meaning nominated.

⁷⁵Ibid., 10-11.

⁷⁶Ibid., 11.

⁷⁷Ibid., 11.

⁷⁸Ibid., 11.

⁷⁹Ibid., 11.

⁸⁰Ibid., 11.

⁸¹Ibid., 11.

Finally, for Barthes, there is "no sum of the text,"⁸² no completed, denominated, denoted system of meaning for the text that is, at the last, capable of being summed or totaled, and thus made representative of the whole of the text. Reading is not a matter of accounting, of receivables and payables balanced and checked, one against the other, resulting, in the end, in a single number, or even multiple numbers, that somehow sum up the text. Thus, there is no responsibility of the Barthesian reader to not forget any meaning that might be identified, noted in the course of reading the text.

However, things are not really that simple. Barthes goes on to say in the next section, "VI. STEP BY STEP,"⁸³ that his goal is to study individual texts "down to the last detail."⁸⁴ To do this, he wants to proceed in a step-wise fashion, unfolding the text's meanings gradually, "abandoning no site of the signifier without endeavoring to ascertain the code or codes that of which this site is perhaps the starting point (or the goal)."⁸⁵ Detailed analysis, taken step by step, with no abandoning of signifier sites, implies strongly a necessity, not only for precision and consistency, which Barthes' practice indicates he is interested in, but completeness as well. Otherwise the goal he states in the "STEP BY STEP" section would appear to be rendered unreachable.

⁸²Ibid., 11.

⁸³Ibid., 11-13.

⁸⁴Ibid., 12.

⁸⁵Ibid., 12.

But are Barthes' two insights, that forgetting is inevitable and necessary, but so is complete analysis, incompatible? Not necessarily, if one agrees with Barthes that regression down the chain of meanings is infinite, and that the goal is "that play which is the return of the different,"⁸⁶ which is only attained by going as far down as one can through rereading, while not worrying about that bit of the text that gets away from one during the chase.

The truth is, that as one reads S/Z, the impression is strong that Barthes, in fact, missed no signifier, no codes or threads replete with meaning within the text of Sarrasine. The power and scope of his mind is evident, and can certainly lead one to believe that only a mind of his genius could possibly be successful in such an endeavor as that which he undertook with Sarrasine. The sheer amount of textual details that had to be preserved in an orderly fashion, always accessible, ready to be put to work, is certainly sufficient to tax any competent reader.

⁸⁶Ibid., 16.

But such a reaction, which assumes only genius can perform the work, would not be a true Barthesian vision. Certainly, a Barthesian reading requires a much effort. However, one should not ignore the Barthesian principle: "it is precisely because I forget that I read."⁸⁷ Everyone forgets, so everyone must read, and reread, in order to understand the meanings of the texts before them. Anyone who can read, in the ordinary sense, can perform a Barthesian reading. Only the level of the performance, as experienced and judged by others, truly differs from performance to performance and performer to performer. This difference is largely relates to how much the effort involved interfered with the production of the reading.

The goal of any tool based on the Barthesian system should be to lower the required effort in order to raise the performance level. Thus, the function of a computer-based tool like Xebra is to make the Barthesian reading act productively accessible to any who would perform it. What follows, then, is a catalog of the types of aid a computer-based Barthesian system, such as Xebra, could render a prospective user of the Barthesian methodology.

⁸⁷Ibid., 11.

The Barthesian System: Sufficient To Its Purpose?

Whether Barthes' system is indeed precise, consistent, and complete enough to be used for its intended purpose is a matter of theory regarding practice. If it is not, in any or all of these three, than it will fail when applied, at least in some, if not all, cases. The problem comes in attempting to determine the answer to this question of sufficiency. It would appear that abstract thought alone cannot aid in solving this particular problem. Quite possibly, only empirical testing will reveal the facts of the matter. However, one can determine the boundaries of the problem and possible empirical solutions by exploring the question in the abstract, theoretically.

If the syntax used in the five categories of labels (ACT, HER, REF, SEM, and SYM) is insufficiently precise for capturing all that is in the text in terms of the workings of signifiers, this would necessitate a need to expand the label syntax in terms of depth of descriptive capability. Adding more qualifier capability would be the most likely step to take in that regard.

If the syntax is internally too inconsistent, with too many exceptions, then there exists a possibility of semantic ambiguity for the labels. The solution, in this instance, would be to attempt to eliminate the inconsistencies. This would probably entail a simplification of syntax in some instances, but, in many cases, would actually involve a deepening and widening of the syntax. This could, of course, lead to its own problems, since the more complex the syntax, the more difficult it would be to apply usefully.

Finally, if the five categories are themselves insufficient for capturing all signifiers, that is, if there are signifiers that it would be inappropriate to label as any of the five, than the system's typology is incomplete. Correcting this would entail adding categories to the system. As it stands, Barthes' system appears to have been designed to handle prose texts. If one were to attempt to apply it to poetic or dramatic texts, the possibility of a need for expansion becomes quite real.

In summary, Barthes' system may be insufficiently formed for its intended purpose or possible future purposes. This implies that any system or tool that is meant to act as an aid in applying the Barthesian reading methodology must be able to handle the expansion, modification, and contraction of the syntax and vocabulary of the labeling language.

Barthes In Theory: Wholes and Parts

Roland Barthes certainly was not silent on the issue of wholes and parts from a theory perspective. Barthes opens the discussion of parts and wholes by asserting that, in terms of his ideal text, the plural, writerly text, "nothing exists outside the text, there is never a whole of the text."⁸⁸ From this assertion, he says that it follows "for the plural text, there cannot be a narrative structure, a grammar, or a logic,"⁸⁹ and that such only exist "in proportion . . . as we are dealing with incompletely plural texts."⁹⁰ All of which leads to Barthes concluding that the meaning of a text is never a singular, always a plural, except, of course, the truly univocal text.

Barthes keeps coming back to this theme of no totality of text because of the plurality meaning. For instance, in the section on forgetting meanings, discussed above, when he asks "forgetting in relation to what? What is the sum of the text?",⁹¹ he is in the midst of asserting that "for the plural text, forgetting a meaning cannot . . . be seen as a fault."⁹² In other words, since "the systems of meaning can take over this absolutely plural text; but their number is never closed, based as it is on the infinity of language,"⁹³ one less meaning or two in a catalog of meanings for a text will not be of significance. All of which, stated differently, means that in terms of parts and wholes, a text, having no whole, no totality, no sum, has no determinable number of meanings, that is, of parts.

⁸⁸Ibid., 6.

⁸⁹Ibid., 6.

⁹⁰Ibid., 6.

⁹¹Ibid., 11.

⁹²Ibid., 11.

⁹³Ibid., 6.

In practice, a Barthesian reading returns a finite, determinable number of meanings, signifiers, and thus, a boundable number of meanings of a text. This is because no text is, or has ever been, Barthes' ideal text: infinitely plural, writerly. They are all, or at least mostly all, incompletely plural texts, that is, readerly texts. If Barthes is correct that an infinite number of meanings, of parts that cannot be summed to a totality should be possible for a text, one still must conclude, in the empirical world, that such a sum of a text, indicating its totality, is the norm, rather than the exception. Instead, a writerly text would be that exception, for as Barthes, himself, has indicated, "the writerly text is not a thing, we would have a hard time finding it in the bookstore."⁹⁴

In summary, empirically, there are wholes and parts to those texts which "make up the enormous mass of our literature."⁹⁵ If, theoretically, there are not wholes that can be founded by adding up parts, at least, for the Barthesian ideal text, the writerly text, which "we would have a hard time finding . . . in the bookstore,"⁹⁶ it makes no difference. The readerly texts are the ones that we must deal with in the world of experience, and these are not infinite in their meanings or their parts, so they have a sum of parts equal to a whole, at least in some sense.

⁹⁴Ibid., 5.

⁹⁵Ibid., 4.

⁹⁶Ibid., 5.

The Barthesian System: Its Problematic Practices

As noted in CHAPTER I, there are two kinds of practice-based problems in performing a Barthesian reading: amount of effort and difficulty of effort. The amount of effort is a problem for two reasons. First, there can be so much effort involved that one might never do it. This results in lost opportunities afforded by the Barthesian system for doing interesting and useful work with texts prior to the act of interpretation. Second, the amount of effort involved directly effects the difficulty of doing a Barthesian reading. These difficulties, which constitute the second kind of practice-based problems, are related to the precision, consistency, and completeness with which the coding of signifiers can be done especially given the large manual effort involved for any but the smallest of texts.

The next several sub-sections discuss and explore each of these problematic areas in turn, after a review of the record of Barthes' own trace in S/Z on these subjects.

Barthesian System Practice Problematic: Barthes' View

Barthes' own practice in applying his system in S/Z to the Balzac story, Sarrasine, reveals two items concerning the system's problematic performance. First, Barthes was quite capable of applying the system despite the amount of effort required. Second, while he was not overtly concerned, from a theoretical view, with the difficulties of, or even the necessity of achieving precision, consistency, and completeness, his practice shows something very interesting. He apparently believed that in the arena of doing, as opposed to theorizing one must, in fact, overcome these difficulties and produce or achieve a reading that is as precise, consistent, and complete as feasible. He, himself, was most certainly precise, consistent, and complete in his working through the text of Sarrasine.

While there were occasional inconsistencies (in terms of syntax usage of his labeling language), there appear to be no lapses (in terms of making certain that each label was consistent with all labels like it in the reading). Further, Barthes' choice of descriptive terms appears to have been consistent and precise. Where the descriptive terms were not sufficient (in terms of precision in his judgement), he made liberal use of the notes facility. The very inclusion of this facility in the language is an indication that Barthes recognized the need for being precise, as well as complete, in one's practice.

In summary, while Barthes was apparently in conflict with himself on the issues of precision, consistency, and completeness, his practice appears to support the necessity of paying close attention to all three. This being true, it becomes the function of any Barthesian reader's tool to help overcome the difficulties inherent in achieving a sufficient degree of precision, consistency, and completeness.

Barthesian System Practice Problematic: The Effort

The effort required to produce a Barthesian reading is dependent upon two variables: size of text (as measured by number of words) and complexity of semantic relationships (as measured by number of interrelationships among signifiers). Size is a readily computed value for any text, while semantic complexity can only reliably be computed by applying a system that counts the number of signifiers or their meanings, and the interconnections among them all.

That is, one needs to do a Barthesian reading in order to compute a text's semantic complexity. In actual practice the size of a text in words gives a rough notion of the semantic complexity, since the two, generally speaking, are directly proportional to each other. So the measurement of the effort to perform a Barthesian reading rests, at the last, on text size. A look at Barthes' "tutor signifier," as he refers to Sarrasine,⁹⁷ and his reading of it can be revealing concerning the amount of effort that can be expended in a Barthesian reading.

⁹⁷Ibid., 13.

The English-language version of Sarrasine runs to only 14,000 words. Yet Barthes' reading of it runs close to 50,000 words, not counting the sections where he largely is concerned with the theory of his system. Of those 50,000 words, 14,000 are the words of Sarrasine, since a Barthesian reading entails embedding the text itself in the reading. This leaves approximately 36,000 words that Roland Barthes produced in his labeling effort, or roughly two-and-a-half times as many words as Sarrasine's author managed. If these numbers can be reliably extrapolated, this would mean that a novel of 100,000 words would necessitate the critic producing 250,000 words of labeling to accompany the text. This is, indisputably, a great deal of effort.

One way to compute the semantic complexity of Sarrasine would be to note the number of labels Barthes created for the 561 lexias and the number of connections across the lexias that the labels make. Taking the data in Table 1, above, one finds that Barthes generated 861 separate labels, and 268 unique labels. Thus, on average, each label connects three lexias together. This necessitates a reader having to carry, on average, a memory of three different sites in the text for each of the unique labels, and the number can get a great deal larger. The longest sequence of lexias connected by one unique label was 84. Thus, a reader would have to keep 84 different sites of the text in memory in order to fully comprehend the signifier trace represented by that sequence, which, in this case, was the answering of one the six hermeneutics raised in the text.

Unfortunately, one cannot simply extrapolate this complexity number of three sites per label for Sarrasine to a generic 100,000 word text. The number of lexias and labels, total and unique, is a function of each text. It is generally safe to assume, however, that the complexity goes up as the number of words goes up, if, for no other reason, the fact that the more words there are, the more opportunity an author has to connect, either intentionally or otherwise, one text site to another.

No matter how one computes the size or the complexity variables, it is clear that the effort involved in Barthes' reading of Sarrasine, was not trivial. Neither, as will be shown in CHAPTER V, is the effort necessary to do such a reading on William Faulkner's story, "The Bear," which is 50,000 words long. Given that the effort is non-trivial, it is clear that any would-be user of the Barthesian system is likely to think twice before attempting the task. It is not every critic that wants to produce two-and-a-half times as much text as the original just to document the text's "facts" which, only later, feed into an actual act of interpretation and a critical reading. Roland Barthes certainly did not balk at the task, but others have and will. Which is why the focus of this dissertation is on an empirically sound solution to the problems of practice involved in his system.

A computer-based tool has the clear advantage over a manual approach to the Barthesian system in terms of effort necessary to the act of using it. The great value of computers is in their ability to seemingly do effortlessly those tasks which are largely repetitious and which require a great deal of data manipulation and storage. From the numbers given above, it is clear that a Barthesian reading qualifies on all counts: 1) the necessary rereading and searching for signifiers, for descriptive terms and for labels can all be supported by a computer; 2) the lexias and their labels can be easily stored in a computer; and 3) the searching and rereading of both can be supported in an efficient, accurate, and reliable manner.

Any computer-based Barthesian tool will bring these functions to the act of a Barthesian reading that should help in lessening the impulse to not apply the system in the first place. These functions will also help alleviate the problems of application difficulty, as the following sections will discuss.

Barthesian System Practice Problematic: On Completeness

Concerning Barthesian system completeness, the issue is whether, from a practice perspective, can one rely on the production of the facts of a text via a Barthesian reading. It is a matter of sufficiency, of course. For when are there meanings enough in the reader's net?

Barthes, as previously noted, at least on a theoretical basis, was not concerned with how complete his own reading of a text might be. However, what one could be reasonably concerned with, is a bounding of the number of meanings in a particular context. That is, a reader might legitimately be interested in being assured that all the meanings of interest to a particular critical method or methods have been, indeed, discovered and appropriately labeled. It is a matter of what subsequent model of literary criticism is to be applied during the act of interpretation, for which the Barthesian reading is meant to supply "the semantic substance."⁹⁸ A strong feminist reading will depend on a "semantic substance" that is not deficient regarding gender related signifiers, for instance. So completeness cannot be ignored from an empirical point of view.

There are two issues regarding Barthesian completeness from a practice perspective. First, does Barthes' typology of signifiers cover all of the genus? Are there more than his five voices, and their five signs (ACT, HER, REF, SEM, and SYM)? Second, does his restructuring process, accomplished through rereading, assure that a prospective Barthesian reader can truly be in possession of "the best possible space in which . . . [one] can observe meanings?"⁹⁹

⁹⁸Ibid., 14.

⁹⁹Ibid., 13.

As shown previously, these are related questions, because one has to be able to identify signifiers in order to create the lexias, which is done through rereading. Further, if there are more than Barthes' five signifier types, than one needs to be able to have a definition for each new kind that allows for recognition of actual instantiations of the species.

From an empirical perspective, a computer-based system can certainly be supportive regarding the addition, deletion, and modification of Barthesian-style label formats. In point of fact, the system could even be so flexible as to allow the reader to create a coding schema totally disjointed from the Barthesian. The former capability, of changing the Barthesian schema through additions, deletions, and modifications, is necessary for addressing any incompleteness in the signifier typology. The latter capability, however, of allowing the use of a completely different schema, is obviously not necessary to any tool devoted to the Barthesian approach, desirable though it may be in the greater scheme of computer-based literary critical practice. As will be discussed in CHAPTER III, alternative coding schema do exist, and could, conceivably, be of interest.

The second aspect of the completeness issue turns, finally, on the likelihood of a given set of labeled lexia's signifiers being completely coded. Barthes, as discussed above, is not concerned about this point; in fact, he is certain that it is not possible to achieve such completeness. Given the nature of language with its infinite capacity to express meaning, Barthes is quite likely correct. However, there are ways to lessen the number of possible missed meanings, ways that are feasible manually, but these ways are enhanced considerably when used in conjunction with computer-based tools.

For instance, if a Barthesian reader used a systematic technique to generate Descriptive_Terms as signs of the signifiers, such as that developed for taking surveys where responses are given in a free-form manner, the possibility of generating all reasonable terms that are pertinent to a given lexia will be high.

This technique involves an iterative process wherein the reader of a text would come up with a set of candidate descriptive terms and then classify these into subsets, each of which can be given an umbrella descriptive term as a name, much as is done in thesaurus construction. As reading continued, any new terms that occurred to the reader would either be added to these subsets, or a new one would be created with a new umbrella name, if necessary. This allows for a systematic range of terms and subsets of terms to be generated. A computer-based reading tool, Barthesian or otherwise, could certainly support such a process, and, given the problem of completeness, should.

In summary, then, the two primary problem areas of completeness in practical terms for a Barthesian reading are addressable. The labeling language can be extended and the descriptive term generation can be done systematically for the signifiers in each lexia. While both of these are possible to do manually, using a computer to relieve much of the effort is not only possible, it would be quite useful in terms of reducing the effort in completing a reading.

Barthesian System Practice Problematic: On Precision

Precision goes hand-in-hand with the problem of completeness. Without precise descriptive terms for the signifiers, nuances of meaning will be lost, thus, incompleteness enters into the reading. A computer-based system could aid the reader in choosing precise terms. The primary means for doing this would be through the provision of interactive access to thesaurus, dictionary, and encyclopedia tools, against which the signifiers could be compared. While the use of print versions of a thesaurus, dictionary, and an encyclopedia would certainly accomplish a great deal, computer-based versions allow for more efficient, more complete searching of the data for matches to the original signifier.

When examining Barthes' example reading in S/Z on Sarrasine, one can see that, besides choosing good descriptive terms, Barthes felt constrained to often give more detailed information in his Notes. This should be taken as a sign that perfection in this area of precision is not going to come easily, if at all. Arriving somewhere near the target is probably the best that can be done. The instability of meaning is the source of this troublesome fact, while ironically enough it is the very foundation for the method that it ultimately blocks from conquering.

In summary, determining precise descriptive terms relative to the signifiers that they are meant to describe is, like completeness, not perfectly possible, at least not ultimately. However, with certain aids, such as thesauri, dictionaries, and encyclopedia, particularly in computer-based versions, one can at least reasonably aim at being as precise as is technically possible.

Barthesian System Practice Problematic: On Consistency

Consistency of application of the method is a problem at all levels of the process. However, the primary concern is that the same signifiers are recognized as such and are tagged the same no matter where they are in the text. A computer-based tool can be of help through automating methods, that, while capable of being manually executed, are more easily executed using a computer.

In a manual-based Barthesian reading, the reader could keep a list of all the descriptive terms used and all the lexias with which they were used. When determining applicable descriptive terms for the signifiers in a lexia, the reader would then peruse the list of terms, compare the terms to the signifiers they had used to label in previous lexias, and determine if any were applicable to the current signifier in the current lexia. This could be very tedious work, and prone to error due to that very tediousness.

A computer-based Barthesian reading tool could automate the searching of the descriptive term list and the comparison to the signifiers in the lexia. Both the signifiers and their associated descriptive terms could be kept in such a way as to make the search and compare procedure simple, straight forward, and reliably accurate. This would relieve the tediousness to a large degree, thus lowering the risk of inconsistency through errors introduced through the level of effort required in the manual system.

In summary, consistency of application of the system is largely a matter of paying attention to details. With a computer to help keep track of those details, and automating the process of searching through the text and the related descriptive terms for the embedded signifiers, one has a better opportunity to successfully keep one's attention focused where it needs to be. Ultimately, a computer cannot do the work; but it can make the work less difficult. The next section will discuss why this is true for the Barthesian system as a whole.

The Barthesian System: Addressing Its Problems

A tool designed to support a Barthesian reading should either eliminate or greatly alleviate the empirical-based problems of the Barthesian system that were discussed in the previous section. This section will focus on the functions and qualities that such a tool should have, particularly, if it is a computer-based tool. Theory-based problems will be addressed first, followed by those of practice. However, these two kinds of problems are interdependent, and where necessary, their interrelationships will be foregrounded.

Addressing Barthesian System Problematics: Its Theory

If extensions or refinements to the Barthesian system are necessary, or just desired, by a reader using a Barthesian reading tool, the tool should be sufficiently flexible to accommodate such changes. The problem for the tool designer is in anticipating the types of extensions and refinements that are likely. Clearly, the four areas that have been identified above as problematic, from a theory-based perspective, have a high probability for receiving such attention. But what form such changes might take is a serious question for the tool designer.

Extensions and refinements that address the precision, consistency, and completeness of the system will, with high probability, center on adding to, subtracting from, and modifying the vocabulary and the grammar of the system, as Barthes originated them. A tool that only allowed the reader to work within the confines of the original grammar and vocabulary is clearly insufficient. A tool that allowed certain kinds of changes, such as adding to, subtracting from, and modifying either one or the other would be more useful, but, unfortunately, it would still be insufficient.

It is possible for a computer-based tool to support vocabulary and grammar extensions, but the designer must build the flexibility into the tool from the beginning. Computer programs can be, and often are, written very specifically to a set of detailed requirements. For attributes like the grammar and vocabulary of the Barthesian system, this means that they would, quite likely, be fixed in the program as they existed at the time that the computer code was written, or, as computer programmers say, these elements would be hardcoded into the program. This is because it is simplest to do so.

Programmers taking this route are the philosophical equivalent of those philologists of Barthes' who would declare all texts to be univocal.¹⁰⁰ The tool designer, however, must take the more Barthesian approach to the writing of texts and allow for a plurality of meanings, of operations of the code in order to allow the vocabulary and grammar that the tool uses to undergo additions, deletions, and modifications.

¹⁰⁰Ibid., 7.

The other area of theoretical concern, Barthes' view of the relationship of wholes to parts, is not something that a tool needs to address. If a reader pursues a Barthesian reading, whether using a tool or performing manually in the Roland Barthes' tradition, by definition, such a reader has accepted, at least tentatively, the Barthesian philosophy on this matter: readerly texts have wholes and parts, even if the limit-text, the writerly, does not. Such a reader is being very empirical; if the system works in a specific instance, then that is what matters. Whether it works for all texts across all times, or whether there are fully plural texts which have no totality, no sum of parts that equal a whole, does not matter to such a reader and the tool need not take the problem into account.

Addressing Barthesian System Problematics: Its Practice

Using a computer-based tool for supporting Barthesian readings is, currently, the optimal approach to performing such readings. As discussed above, there is a great deal of effort, as well as certain kinds of difficulties, involved in applying the Barthesian system manually, without automated support. There are a number of functions a computer-based tool can provide that would make the reader's performance a simpler activity than would, otherwise, be the case using manual means.

Practice Problematics: Level of Effort

In terms of level of effort support, a computer-based tool can easily reduce the problem of handling large amounts of text repetitively, as well as aiding in a productive tracing of the complex relationships of a text's signifiers. The size and complexity of a given text cannot be arbitrarily or systematically reduced without semantic loss. Thus, any computer-based tool must avoid the reduction of textual size and complexity. What such a tool can do is provide functionality to make the actual handling of the text's size and complexity reasonable.

This kind of support could come in several ways in terms of the Barthesian system. One, the text can be restructured automatically, that is, the lexias can be produced by the tool. Two, the restructured text and the accompanying Barthesian labels can be stored in multiple machine-readable formats for later use. Three, the lexias and labels can be automatically ordered as the reader requires. Four, the lexias and labels can be retrieved for analysis according to the reader's specification. And five, the statistical and graphic pattern-based relationships among the lexias and labels can be analyzed using sophisticated analysis tools that are very easy to apply in a computer environment.

While the Barthesian definition of the attributes of a lexia is not directly computable, since there is no computable definition of a signifier that would allow a computer program to recognize one, there are possible paradigmatic analogues for signifiers. For example, signifiers generally are composed of noun phrases, so recognition and counting of noun phrases could substitute for actual signifier identification for lexia generating. Alternatively, since signifiers commonly occupy a space defined by major punctuation marks, a lexia generator could count such marks and produce a lexia each time a particular number, defined by the tool user, have been counted. In any case, automatic generation of lexias should relieve a Barthesian reader of a great deal of pre-signifier labeling effort, thus making it a worthwhile function.

The ease with which large amounts of textual data can be converted and stored in useful forms in a computer environment is very important, especially when considering the large number of lexias and associated labels a Barthesian reading can potentially produce. Keeping these in printed form in a readily accessible format could be very difficult, requiring multiple printed indexes of the lexia labels as access points. Machine-readable text is readily manipulated by currently available computer programs and methods, with little or no effort on the part of the computer user. Thus, a computer-based Barthesian tool can potentially save a reader a great amount of manual effort just in terms of text storage and manipulation.

Directly related to this is the ease with which machine-readable text can be automatically ordered or arranged, as specified. Thus, the lexias and their associated labels can be easily placed in any order necessary for analysis. For example, if the reader, for analysis purposes, wished to order the lexias by number or type of signifiers, this could be quickly done, even if there were several thousand lexias and associated signifiers. Manually, this could be a very time consuming operation.

Taking this idea of looking at the lexias in a particular way a step further, the ability of a computer to retrieve text based on a user's specification of relationships is very important to any Barthesian tool. For instance, if a reader wished to view only those lexias that had either a SEM label of "Wealth" or a REF label of "Banking Practices" associated with them, machine-readable text can be easily ordered, indexed, searched, and manipulated to support this need. One reason why this is important, is that this ability would directly support a Barthesian reader's need to control precision, consistency, and completeness of the application of the five codes to the text. In the example given, the reader might be checking to see if the SEM had been used consistently across all lexias that dealt with banks. To do the same operation manually, especially on a text divided in a great number of lexias and having an equally large number of associated labels, such a performance could actually be quite problematic as to its success, and, most certainly, would be extremely tedious.

Finally, computer-based data analysis tools can be excellent labor saving aids. By automatically generating a variety of statistics and pattern-based relationships, and displaying these in convenient and productive forms, such as tables and graphs, the manual effort required for similar operations is largely negated. These operations, while not necessarily part of a Barthesian reading per se, are indisputably useful to any later critics, availing themselves of the Barthesian reader's labor, looking for patterns of signifier usage. Certainly, the reader could generate basic statistics and pattern graphs as part of the original reading, if the reader so desired. To do this manually would be a prohibitively expensive effort.

It is not clear that these are operations that Barthes considered doing. However, he did do at least one such operation in preparing his reading of Sarrasine for S/Z. In an appendix, he displays all of the ACT labels in order: first, by act sequence type ("To Enter", "To Decide," etc.); and second, by order of appearance in the text of a given label in each sequence. When Barthes decided to retrieve, order, and display the ACT label sequences as an appendix to S/Z, while excluding any of the other label types, part of his reasoning was probably predicated on perceived effort required versus perceived value gained. His own explanation was simply that "since actions (or proairetisms) form the main armature of the readerly text,"¹⁰¹ he would place them in a readily accessible form.

A reasonable conclusion would be that he saw the need for later critics to peruse the lists for clues to the text's plot or action structure. He refused "attempting to structure them further"¹⁰² than he had already done by associating an ACT label with each lexia that required such. Certainly, if such a retrieving and ordering process was useful on ACT labels, it could also be useful on the others.

¹⁰¹Ibid., 255.

¹⁰²Ibid., 255.

The discussion above has focused on how a computer-based Barthesian reading tool can ease the level of effort required to handle the problems related to the large amounts of data associated with a Barthesian reading. However, those same five attributes, as analyzed above regarding level of effort, can also be specified as aids in the easing of the problem of handling the complexity of the data.

Automatic generation of lexias, storing data in machine-readable format, and automated ordering, retrieval, and analysis, all aid the reader in overcoming the complexity of the relationships that exist among the lexias and their signifiers. By allowing the reader to easily order the lexias and associated signifiers, the computer-based tool is significantly reducing the chance of overlooking, or otherwise missing, relationships across lexias. Further, the ability to retrieve only those lexias that match a particular relationship definition reduces the possibility of lost information even more. Finally, the automated generation of statistics and pattern-relationships in the form of tables and graphs using data analysis tools reduces the complexity as far as current technology can manage.

Practice Problematics: Difficulty of Effort

The discussion above focused on how a Barthesian reading tool might ease the level of effort required to perform such a reading. At this point, a refocusing of the lens brings to light the related, but different problem of level of difficulty. The level of effort metric, as defined here, largely affects whether or not someone is even willing to undertake a Barthesian reading. The level of difficulty metric is correlated with the ability of the reader to obtain productive results. Simply put, the former is related to the amount of sheer labor involved; the latter with the ease of doing the labor well.

These metrics are related in that the complexity of text site relationships, as represented by signifiers, underlies both. They are different in that the sheer amount of text only necessarily effects level of effort, not difficulty. That is, a large text possessing only a few unique, simple, signifiers would be easy to achieve productive results with, while still necessitating a large expenditure of effort. On the other hand, a small text, with many complexily related signifiers, could be very difficult to read usefully, and also require a great amount of actual labor.

A computer-based tool that eases the burden of effort should also lower the level of difficulty, as defined here. This is true because, the tool's functions which are necessary to making it simple to manipulate the text and associated meta-data as represented by the Barthesian labels also make it simpler to achieve precision, consistency, and completeness. The reasons for this are two-fold.

The first is straight forward and highly empirical. Having eased the manual labor involved in the purely tedious, repetitious operations of storing, ordering, relating, and analyzing, the mind is free to pay attention more closely to the actual effects of the operations, especially, the last three. Once one pays more attention, it becomes easier to make certain that operations are done precisely, consistently, and completely.

The second reason is that computer-based operations, by definition, are generally given to precision, consistency, and completeness. A well-designed, computer-based Barthesian tool, while performing its functions precisely, consistently, and completely, will encourage the reader to do the same in terms of the intellectual effort. In fact, a well-designed tool will make it hard not to be precise, consistent, and complete in applying the system.

Addressing Barthesian System Problematics: In Sum

In the end, a computer-based Barthesian tool can only do so much of the labor. It can do the simple, but tedious and repetitious, and therefore, distracting labor. It can do these precisely, consistently, and completely, without error, by omission or commission. However, the reader remains responsible for the intellectual effort needed.

For instance, decisions will need to be made in order to identify and label the signifiers. Evaluations will need to performed and conclusions will need to be drawn regarding the results of any statistical or graphical analysis operations performed by using the Barthesian tool on the text and its associated labels.

Computers and programs do not replace, but only augment, the capability of human intellectual effort. A computer-based Barthesian reading tool cannot be an exception, including, Xebra, the tool developed for this dissertation.