Fragmentation is de rigueur¹. With the admirable modesty of perpectival displacement, Maurice Blanchot marginalizes his own discourse of «Fragmentary writing» — already subordinated to the Name of Nietzsche — by his reference to the group of contemporaries that includes Foucault, Deleuze, and Derrida². Citing the same authors and adding others, Martin Jay, concluding a recent work on totalization and Marxism, recognizes the post-structuralist challenge to western Marxism, whose struggles with the concept of «totality» are the very subject of his book³. But already within that tradition of western. Marxism, Adorno had inverted Hegel’s maxim «Das Wahre ist das Ganze», by standing it on its head, as it were, and affirming: «Das Ganze ist das Unwahre»⁴. The antinomy «fragment — vs. — totality» informs contemporary theoretical discourse in all its types: literary, textual, cultural, as well as historiographical and political. Its force cannot be recognized without reference… to its referent, however vague and disquieting it may be to the tradition of generic disciplinarities being transgressed. The referent is the existence of a dark and monstrous layer of human geology, a historical experience felt as present, always ready to be born again as the rough beast slouching towards our face in the mirror, which has transformed the image we can have of men and women. No cultural artefact transcends this moment of history, no chef d’œuvre of humanistically interpreted beauty over-rides the historical reference of theory operating in the shadow of totalitarianisms of the right and the left, the double shadows of Auschwitz and the Gulag. What work of color, tone, or language can claim to escape the annulment of the heroic occidental Subject, its values and epistemological aspirations, including that of a totalizing knowledge — brushing up against the felt but undefined margins of totalitarian knowledge? And should such a claim be made, who shall call it acceptable?

The methodological stakes in the binary opposition of my title comport political valences that are inescapable. However much one would wish to avoid them, one cannot do more than postpone their play in the discourse we hold. That postponement, however, is important. Nothing would be more totalizing, perhaps totalitarian, than an immediate reduction of the particulars of a given discipline to the broadest terms that over-ride all disciplines without the temporization necessary to account for the particularities of the field of knowledge concerned. These particulars must be attended, not as avoidance, but as an inclusion that avoids reduction. That inclusion is all the more valuable, since it focuses upon a moment that is, historically speaking, a Moment: the point of time in which textuality attends the borning of the State. Our trajectory will lead through the field of cultural and textual generality; abstract theoretical discourse; concrete text analysis; and an intertextual comparison between what the nineteenth and twentieth centuries have construed as «literature», and what we have differentiated as the «historical text» — distinctions of dubious pertinence to the medieval case. These moves, somewhat disquieting in an established disciplinary framework, follow the dialectic among theory, history, and the concrete textuality of the Middle Ages, along with its accretions of the scholarly tradition: the possibly unnerving shifts of focus are required by our present epistemological situation. But I leap ahead of myself: first, issues of medieval textual culture.

As usual in approaching a general problem in the discussion of medieval textuality, it is necessary to acknowledge Paul Zumthor’s precedence. In a brief article published a decade ago, Zumthor argues for «texte-fragment», or rather, for an «art of the fragment»⁵, recognizing that this fragment cannot help but call for a totality, located somewhere else, as its complement. Zumthor points to the great prose cycles of the XIIIth century, as indices of a prise de conscience extending into the XIVth century: «fragmentarity constitutes the nostalgic marker of a utopian totality» (p. 82). My argument will largely consist of searching out some implications of Zumthor’s perceptions. It will also expand the historical period affected back to the twelfth century, and perhaps quite early in the crucial turning point of European civilization. Above all, my argument will insist on the historical and political investments of the issue, and that from both perspectives: that of the period of the texts, that of our own period in the waning years of the twentieth century.

This historicity of our effort also defines its fundamental difference from Zumthor’s. That difference is dual. Its first step is the insistence that the «field» of textuality addressed has an independence, a self-identity, a cohesion, a density, that pre-exist and lie beyond any particular theoretical approach. Without for a moment denying the subjectivity of all cognitive endeavor, the presence of the epistemic Subject in the construal of the Object, it must be insisted that the object, before being constituted as an Object of knowledge, does exist quite independently of any knower. We may not «know» the thing per se, any knowledge of the thing is simultaneously reductive of the object in itself and constitutive of the object of knowledge: all knowledge necessarily comports perspectival effects of subjective predispositions. Nevertheless, the thing exists before the touch of our eyes, before the approaching flutter of our consciousness. As a result, the results of that encounter, of fluttering consciousness with field of knowledge, result in a dialectical modification of the initial subjectivity, of those predispositions we brought to the object in our initial approach and desire to constitute it as Object of knowledge. It is precisely the differentiae of medieval textuality that make of it a privileged interlocutor for contemporary theory, and that require it to play the role of an interlocutor with a sense of independent cohesion. It is only by its responses based on that sense of priority and anteriority that it can benefit contemporary theory as a real interlocutor rather than as parrot.

At the same time, these differentiae lead to a recognition of the identities that persist and traverse the distances of alterity. The theoretically informed construal of medieval textuality as an Object of knowledge constituted — in large part — by its differentiae from the values of the epistemic Subject, allows in turn — in dialectical turn — for the recuperation of those elements of the object, for those aspects of the field, in which the Subject re-cognizes those elements of the Object that are identical with its own subjectivity. The dialectical relationship between theory and objects produces both modifications of each and (re-) assertions of identity.

As a result of this dialectical interplay, theory today, even when it initially addresses what our cultural codes constitute as «literary» texts, cannot help but go beyond that epistemic, and hence historical, limitation. It is not only that theory recognizes its own historicity, its own ineluctable insertion within the linearities and frameworks of history: it is the fact that history is not somehow «outside» the supposedly «literary» text, but at its very heart. In dealing simultaneously with its own historicity, and in exploring that of the textual object, theory is political not as an epiphenomenal afterthought, but inherently. It is the nature of the historicity of theory at the end of the twentieth century to insist on its historicity, on its politicality, on the inherence of historical politics in any act of «reading» whatsoever, at the same time that this very historicity equally insists on textual integrity. It is that duality that constitutes our historicity, and the complex necessities that determine the bivalence (at least) of contemporary theoretical discourse. As a result, the study and discussion of medieval literature can no longer be a moralistic escapism from the conflictual ills of the present: as the present descends from and incorporates the Middle Ages in its sometimes horrifying presences, so it must recognize itself in the Middle Ages as Object of desire and study the sometimes repugnant traits that would make us seek out an avenue of escape.

II.

Medieval texts are often fragments. Sometimes, medieval fragments are «accidental» in their fragmentation: it is accidents of transmission that account for the fragmentary nature of texts such as the early epic, Gormont et Isembart, as well as both the Beroul and Thomas versions of the story of Tristan and Iseut. While such accidents of transmission tell us little about textual ontology, they are informative of the cultural attitudes of a civilization that readily fragmented the manuscripts containing its vernacular tradition: textual coherence per se was hardly an unquestioned value for the culture in question. This general cultural value (or its absence) works back to the domain of textuality, however, when we look at the production of a master of narrative such as Chrétien de Troyes. Two of his romances were left in a fragmentary state: the Lancelot had to be completed — apparently under Chrétien’s supervision — by a colleague otherwise unknown, named Godefroy de Lagny. Since the missing fragment was apparently a rather short bit of conclusion, this is hardly a major indicator. On the other hand, the incompleteness of his last work, the Perceval, impedes that interpretation of that work: there is no sense of just how much of a fragment we have, of just how much of a conclusion is missing. Nor is there any certainty that Chrétien himself would have known how much of a conclusion he might have given the text, had he completed it: we know so little about the process of composition, the mechanics of text production in the Middle Ages⁶!

A large category of texts, and major texts at that, enters into our next category. These are texts which are both fragmentary at one stage of their developpment, and totalized at another stage: at least, some effort was made to extend their syntagmatic existence into a textual structure responsive to a sense of narrative totalization. The Lancelot, which I have already mentioned, is a major example. Left unfinished by its author, it not only received a continuation that completes the narrative: that process of continuation is said to have been overseen by its first author, suggesting that some sense of appropriate continuation and completion was indeed at work. Chrétien’s Lancelot, of course, is from the second half of the twelfth century. Zumthor himself points to what was undoubtedly the most famous and the most sucessful job of text-continuation — both in terms of textual structure and popularity — in the Middle Ages: the Romance of the Rose. The delicate, sentimental, and perhaps somewhat cloying allegory of Guillaume de Lorris, dating from perhaps 1225-30, was continued two or three generations later by Jean de Meun, in a robust, intellectualist, and highly architectonic manner⁷. The coherence of the new text is obtained, neither as problematized narrative nor in the sentimental allegory of Guillaume de Lorris, but in ideational and rhetorical terms. What is unquestionable, it seems to me, is precisely that architectonic sense which allows for the integration of the earlier fragment into the new, larger, and more ambitious structure.

This last category is ambiguous. Its texts provide evidence of that fragmentary nature of a number of major texts; it also betokens, in their continuations, a sense at work as early as the twelfth century, of textual totalizations that answer to some felt cultural need. And it is this same need to which a large number of other narrative texts respond. With or without the Baligant episode, The Song of Roland is an intensely composed and structured text, no matter how difficult particular issues of detail may remain. And while other chansons de geste clearly reveal the unimportance of narrative closure, that is not the only kind of textual coherence available: it is late in the day to confuse form and structure, narrative completion and a structural principle of textual cohesion. Even at the level of narrative form, however, it is a fact that a large number of texts reveals a concern with the narrative formal completion that gives us — twentieth century readers — a sense of textual coherence. Early romances of antiquity like the Romans d’Eneas and de Thebes; Chrétien’s other romances, such as the Erec et Enide, which not only benefits from totalizing narrative structure, but is rhetorically «finished» to provide a proper heroic ending; the variation played on this kind of closure in the Cligès, indicating that heroic narrative was a culturally encoded norm; and the far greater complexity of the Yvain, which uses the same basic narrative pattern as the Erec while exploring a far more profound social problematic than the exaltation of royalty suggested by Donal Maddox in his excellent book on the Erec⁸. And one of the «traditions» (in the zumthorian sense) that Chrétien leaves to his successors is that of the «well-made romance», even if the concept of «well-made» is not that either or Aristotle or O’Henry: romances such as Le Bel Inconnu and the Partonopeu de Blois imitate the kind of narrative structure developed by Chrétien.

Early romances of antiquity such as the Eneas and the Thebes, exhibit a clear sense of formal narrative closure; and Chrétien’s first romance, the Erec et Enide, adds to the same self-evident narrative closure a verbal «rhetorical» finish which coordinates the narrative closure with the semantic closure of the text. But there is little room for argument on the point. The verbal conclusion of Chrétien’s Yvain overtly marks the syntagmatic ending and completion of his vernacular narrative, as follows:

Del Chevalier au lyeon fine

Crestiens son romans ensi;

n’onques plus onter n’en oi

ne ja plus n’en orroiz confer

s’an ne vialt mançonge ajoster⁹.

So does Chrétien conclude

the Knight of the Lion’s tale;

I never heard more told,

nor will you hear more told

except as added lies (my translation).

Any addition is fabrication in both senses: «making up» something that is stated to be a «lie», a falsehood additional to the narrative text. How further fiction added to fiction is a lying falsity can be problematic for contemporary theory, but the implication of such an added fragment of belonging to the preceding text signed by Chrétien would be a lie, so sayeth our auctor.

Indeed, it is perhaps not «straight» narrative closure that testifies most pungently to the encoding of narrative closure in twelfth century culture, but the possibility of playing with that cultural item in a number of texts whose syntagmatically final segments tease the reader. Chrétien’s Yvain, structurally more complex than the Erec which it parallels in narrative form, leads to and asserts a narrative closure while withholding the kind of rhetorical «finish» deployed in the Erec. Indeed, too great a reader investment in the notion of the «happy fictional ending» had already been undermined in the intervening text of the Cligès, whose concluding words reduce the fictional ending to a final pointe addressed as much, perhaps, to reader expectations as to its own «hero». But it is really in the work of a writer sometines cited as a member of a putative «school of Chrétien» that the solid implantation of a sememe of narrative closure in the culture of its time is most imperatively implied. Renaut de Beaujeu titillates his reader by textualizing three possible alternative endings to his own narrative¹⁰. Is more evidence required to make the point that the semantic element of syntagmatic narrative closure was an achieved, culturally encoded norm in the second half of the twelfth century?

As far as the totalizing structures of narrativity are concerned, it should be clear from these brief remarks that they abound in the twelfth century, without any claim being made thereby as to their dominance. Indeed, as far as this period of the Middle Ages is concerned, it is my impression that «dominance» itself is a rather doubtful methodological concept, in view of the remarkable dispersions of from and structure perceptible in the texts of the period. Suffice it to say that a substantial number of major texts reveal traces of an overall immanent textual structure that cannot be dismissed by an otherwise appropriate tactical stress — aimed against modern assumptions of the well-made plot — on the frequency, and indeed the normativity, of textual fragmentation throughout the Middle Ages.

The fragmental, however, is not just a matter of syntagmatic incompleteness, whether due to accident, ill health, death, or a rhetoricopragmatic strategic intention over-riding issues of narrative closure or textual coherence. On the contrary, the fragmentary inheres — as far as medieval literature is concerned — in the holistic. Even when the narrative is brought to an end, even when the text obtains its destined closure, the fragmentary remains at its textual heart as a condition and a means of wholeness. The medieval text is not only occasionally and more or less accidentally fragmentary, it is always and universally ontologically fractal, using fragmentary parts as elements of construction of a new entity¹¹.

A whole cannot be described except in terms of the organization of its parts. There is that, however, in the constitution of parts, which refuses mere subsumption by the whole, which does not allow for total recuperation of the part by its function within the integrative power of the text. Such integration can be achieved, and can even, by dint of careful attention, be adequately described by modern readers of medieval texts insistent on retaining awareness of the determined alterities in their relationship to those problematic texts. But the parts do not lose their dual identity as a result of their integration within a larger whole. The fact is that they retain the marks of their integration in two simultaneous wholes, even as they work their way into a particular narrative or other textual structure. Not only are they required to insert themselves into the system of a particular text and to become functional within that system, they are always already a part of another system, and remain a part of that other system, even as they enter the textual universe that functionalizes them syntagmatically. At the same time that a part enters its new textual world of the particular text, it remains what it was previously, and remains after its particular insertion, ready for further use, as one item in the paradigm of conventions out of which medieval texts are built. These may be considered in the context of a neo-Platonic philosophical tradition in which the concrete textualization represents the concretization of an Ideal Form, or merely as part of the paradigmatic dimension of textuality, very much on the model of the paradigmatic in Jakobson’s account of the process by which the syntagmatic selects and linearizes elements from the paradigmatic in actual language use¹².

In either case, the part is not defined simply by its insertion into the new whole. It remains a token of its type, a particular use and variation of the convention, for instance, of the description of personal beauty, of the town, or of hospitality in medieval romance. Such insertion is not only a functionalization within whatever textual structures now surround it, it is also an exemplification of the particular convention in question, an exemplification that is always inevitably a variation on the norm of that convention. Subject to all the techniques of either abbreviatio or amplificatio, modifiable by the resources of grammar, style, and rhetoric, the individual deployment of a convention simultaneously responds to criteria of integration within the single, concrete text, and the potentials for expansion, brevity, modalization, and aspectualization that are inherent within the particular convention in question. The criterion of littérarité applies, not only to texts as wholes, but to their constituent conventions as well. This dual integrativity may have some relation to what we experience as the markedly weaker power of architectonic subsumption in medieval textuality than in the modern: the ontology of the individual being already guaranteed by its appurtenance to the paradigmatic, its particular syntagmatic integration is less important. The «fit» of the individual component does not need to be as tight as in works produced under the modern principles of textuality, whose aesthetic of «originality» requires strong integration for ontological reasons. The weak integration of medieval textuality is made possible simply because of the convention’s always dual existence, as partial fragment of the particular text, where its integration in the overall structure makes of it a fractal, end as a token of the convention it also is.

It is its prior status as convention that makes the fragment available as a building bloc in the elaboration, affabulation, and constitution of a larger whole. It is its conventionalism that establishes it as a potential fractal. Its use as a fractal — written all over the face of all medieval texts — is what makes the fragmentary the necessary pre-condition of ontological existence for the medieval text. It is only in its use of the fragmentary that the medieval text attains holistic status, it is only in its reliance upon the priority of the fragmentary that the medieval text becomes whole. It is its fragmentary nature that leads, in at least a large and important number of texts, to the status of «totalizations».

That this paradoxical ambiguity in the relation between the fragment and the whole was sensed during the Middle Ages is suggested by a number of facts. The additive nature of an oral aesthetic obviously played into the multiplicity of relations possible between the part and the whole; so did the sense of the registre, explored more than once by Zumthor: its existence, and the individual themes developed under its aegis were warranted, as it were, by their belonging to the registre in the largely aleatory order of strophes characteristic of the grand chant courtois¹³. In romance, during the twelfth century, the equivalent dual play of part and whole inheres in the complex relations between individual episodes, their rhetorically unlimited number, and their organization, according to still unspecifiable schemes, under an overall narrative structure¹⁴. In the textual culture as a whole, however, it is undoubtedly the existence of what looks drolly like a sub-genre, that testifies both to the persistence of the acceptability of fragmentary texts, and the desire for completion that produces wholes. While it is certainly true that writers were remarkably ready to abandon texts in a state that strikes us as fragmentary, it is equally the fact that such fragments, perhaps enjoyed for their own sake, nonetheless gave rise to the sub-genre of the «continuation». This was the case, first of all, with Chrétien de Troyes’ last romance, the Perceval. By itself, this fragment called forth more than one example of the sub-genre known as «The Continuations of the Perceval», which is indeed the title under which they have been published: rarely studied, these texts and their affabulations should be a rich source or reflection on medieval narrativity¹⁵.

Such pairs of fragment + continuation suggest both the recognition of the fragmentary as a normal state of textual ontology, and some sense of dissatisfaction with the fragmentary, since it was felt to call for some sort of continuation and completion. For some period of time, including all of the twelfth and thirteenth centuries, «text» was an ambiguous notion, covering both the fragmentary and the whole, the anarchic and the coherent. It would appear that the twentieth century notion of «text» has to covers entities that are both fragmentary and yet frequent and normalized, as well as other entities that look strongly finished, syntagmatically completed, and that constitue coherent wholes of signification.

III.

Our discussion so far has been essentially historical: it is based on the observed characteristics of medieval textuality as it has come down to us. Even so, we have had to push against the limits of the common acceptions of the term «fragmentary», shifting over into a metaphoric use of a somewhat different form of the lexeme, one which has currency in the scientific discourse, of «fractals». It is time, however, to incorporate our preceding findings in a reflection on the fundamental terms of our discussion in a more directly theoretical manner, before returning to a concrete problem in the interpretation of medieval textuality.

The terms «fragment» and «totality» are dyssymmetrical. The concept of the /fragment/ is not the contrary of /totality/: rather, it is the latter’s negation. By definition and by etymology, the fragment is a broken piece, detached from a larger totality: it is the pottery shard, found by an archeologist at the site where, thousands of years earlier, a vase, a pot, a container of some kind or other was located and inserted into all sorts of social and economic activities. This fragmentary shard is usually broken off from the original totality in a manner that has nothing to do with the functions of the original parts of the object. The shard bears no trace of respect for function, nor for the shape of the constituent parts of the whole. On the contrary, the pottery shard’s outline traverses the lines and forms of the original object, whether of its aesthetic or functional rationality.

Rather than the contrary of /totality/, the /fragment/ is more likely the contrary of the /part/. The /part/ is a constituent element, subordinated to and presupposed by the /totality/. It is /totality/ and /part/ that are correlative and interdependent notions, caught in reciprocal networks of implications and presupposition. Totality cannot be imagined without recognizing its constituent parts, one does not understand how a part is part of a text as a whole without understanding its relations to the rest of the text. Indeed, understanding a text as text is precisely conceptualizing the relations implicit in the notion of the totality of the text: «understanding» is the competence to conceive the object as the integration of subordinated parts within a totality. To return to our vase or pot: to conceive of the bowl as the container for the storing of liquids, the beak as a means of delivering the stock of liquid to its actual use, and the handle as the form that allows the manipulation and transport of the container and the contained, that is what understanding the assemblage of functional forms that constitute the totality of the pot or the vase.

Whoever has studied a collection of archeological fragments knows perfectly well that the lines of breakage are entirely accidental in relation to the functional forms as well as to the aesthetic aspects of the object: these two rationalities are fractured by fragmentation. The fragment is the sign of the irrational, the irruption of the irrational into the world of design, hierarchies, and integrated units. Produced by violence, the fragment — literary or other — is inscribed in a lexical paradigm that includes items such as fragmentation bombs and grenades, as well as the fairly obscene term from the Vietnam war, «fragging». Fragmentation implies the denegation, not only of totality, but also of the principle that integrates the parts and the whole. That is why it is a mistake to call «fragments» the small forms used by Nietszche, for example, or Adorno in the Minima Moralia. The proverb, the apothegm, the aphorism, are by no means fragments. These very short forms of discourse refuse the development that can only take shape in the space of rationality, but this refusal does not represent what remains after the outbreak of violence unleashed upon a pre-existent totality. On the contrary, these forms are themselves aristocratic violences. They insist on their exemption from all need of justification by appealing to structures or codes that are larger and elsewhere — such as the domination of wit and wisdom over argument — and they refuse all appeal to external legitimation by claiming an inherent quality of artistocratic superiority. They are the aesthetic forms of an elite. It is hardly surprising that the first great artist of this form is the Duke of La Rochefoucauld in the century of aristocratic classicism, that it is a philosopher of the superman who employed the form in the nineteenth century, and that it is a brilliant representative of an elitist neo-Marxism who used it to assert the superiority of the monadic fragment over disintegrating European culture after the Second World War. In each of these cases, the utterance is presented with total autonomy, affirming that, whatever the subject broached, what is said in the particular utterance is all that is called for, all that is necessary, relevant, or adequate. Its content may be accepted or not, it does not offer a structure of rational argument, with its potentials of response and critique. Fragmentary discourse — to use Blanchot’s phrase — consists of something entirely different from fragments… or fractals, for that matter. The forms of discourse in question — the proverb, the apothegm, the aphorism — insist on their completeness, their self-sufficiency, their independence. They do so with pride and self-assertiveness that preclude the argument of a potential interlocutor. They are forms of totality that are no less authoritarian for being syntagmatically brief. Indeed, their brevity is the antithesis of fragmentation.

What is negated by the fragment is the very possibility of integration, of subordination, of subsumption, of Aufhebung. The fragment rejects both the notion of totality and the notion of the part, of integration, of coherence. Yet, a particular syntagm may both be subject to integration with the totalizing system of a text, and seem to protest against that very integration: some aspect of the syntagm, albeit integrated within the narrative structure of the text, may be refulgent with philosophical or political implications, particularly when revealed by a later reading, according to more fully developed codes of interpretation. A ready analogy in individual, personal terms occurs when a subject, integrated within a given social structure, is nevertheless alienated from that structure, consciously or not. However the splittedness of the Subject is specified, that position answers to the both/and position of the revised semiotic square expanded to cover six positions, as suggested by the logician Robert Blanche¹⁶. It is this position which answers best to Adorno’s monadic experience of rejecting the historical and cultural processes in which he found himself after the Second World War, told in the Minima Moralia.

The purely fragmentary, on the other hand, can only insist on its incompleteness and its constitutive irrationality. As fragment, the fragment always recalls its anterior status: refusing present integration, it does so in the name of an earlier totality, its originary status, more worthy of respect, more valorized than the present. The fragment — the word comes from the Latin frangere, to break — is always returning. It is the creature of the eternally desired return to its anteriority, in order to affirm its fragmentariness, its lost integrity. Negation of totality, the fragment is nonetheless a form of cathexis on totality, the totality of origins, of genealogical integration, a cathexis on totalities that have become impossible. Fragmentation, imposed by external violence, by historical cataclysm, can only recall an outmoded genealogy. The state of fragmentation may be a descriptive fact; as theoretical insistence, it is a form of sentimentaly, of nostalgia. The call for the state of fragmentation is an extension of kitsch: it is its theoretical form.

Its opposite, as is frequently the case with sentimentality, is violence. The insistence of totalization, particularly of new totalizations — such as are implicit in revolutions — as against the existence and will of the part called the individual entity, require abrogation of the latter’s ontological status. What is more complicated, is what happens when a new totalization is imposed upon the element or part that had earlier been integrated within an anterior totalization. How this turned out in concrete histories is aleatory, as the concrete details of history always are. But at the theoretical level, the rapid or sudden imposition of an anterior totalization onto a pre-existent totalization requires both fragmentation and violence. That is no less the case in the insistence on fragmentation. The demand that all texts (and other entities) be considered only fragmentarily is itself a call to violence against whatever bonds wend their way from the particular entities toward larger wholes. Furthermore, insistence upon the fragmentary cannot sustain itself vithout transformation into its opposite. To constitute only the fragmentary as acceptable object of knowledge is to totalize knowledge on the basis of fragments, it is to absolutize fragmentariness and hence to develop new totalities. Both exclusive insistence on totality and on fragmentariness require violences when narratively implemented: the relation between sentimentality and violence is not new.

If «fragment» is the contrary of the «part», if the «part» is implicit in «totality», a term is missing, one which would be analogous to «totality», in relation to the «fragment», one which would represent the true contrariety to «totality». This term, or rather, this semantic content, hovers behind and within several terms that are partly synonymous, partly overlapping: atom, unity, thing, entity, all belong to the paradigm in question. The concept must include the semes of /unicity/ and /particularity/. The two terms available in English that seem to me closest to this content are «entity» and «individual»: given the personal connotations attendant on the latter, I prefer «entity». If «totality» and «entity» are semiotic contraries, however, these must not be thought of as exclusive ontological categories. On the contrary, they are analytic categories, not «real» or «realistic» categories. The semantic axis by which they are related is that of /integration/: they are constituted by the relation of «integrator» and «integrated». «Totality» represents the power of integration that transforms the multiplex’s tendency to dispersal into a coherence which incorporates the multiplicity of individuals, or parts, of entities. If «totality» consists especially in the power of integration which provides the multiplex with coherence, then the «entity» constitutes the unit being integrated. The challenge before textual criticism is to carry out an injunction easy to formulate theoretically, but much more difficult to carry out in practice: to account for the integration of the part into the whole without thereby defiguring the particular shape of the part, and without disguising its simultaneous appurtenance to other integrative totalities. It should not be a surprise that the same problem occurs in the political dimension: how to account for integration of the individual person in larger, integrative social and political structures — such a the State, or class structures — without thereby erasing the person’s status as an individual, or that person’s simultaneous integration into a plurality of other structures and groups, such as ethnic, gender, or religious groups.

IV.

It is time to turn in the direction of empiricism, and the two concrete fragments that will be the final focus of these remarks. The first occurs at the end of the Chanson de Roland: Ganelon’s trial. In what sense is this to be considered a «fragment»? At one level, it testifies to the concern for narrative integration I have alluded to earlier. Not only does the traitor’s trial occur at the end of the epic song, it is the narrative result of earlier narrative developpments; indeed, its function is to resolve issues left open by the preceding narrative. At this level, it deals with the actor who can appear as the cause of the disaster of Roncevaux by putting him on trial, by judging him guilty, and by despatching him in a manner which, if somewhat cruel, certainly incorporates the anger and the pain of the survivors, including the primary victim’s uncle, Charlemagne. One reader in this grand tradition of interpretation may stress the effective presence of «the feudal Christian ethos» within a «feudal Christian setting»¹⁷; another may stress the subsumption of feudalism by the royal figure of Charles and the Christian monarchy he represents¹⁸; a third considers the Roland a song of propaganda for the Capetian kings¹⁹. All would concur, presumably, that the Chanson de Roland is a «historical metanarrative of Christian triumph and hegemony»²⁰, whose semiosis is monological in nature, however the balance of feudal, monarchical, and religious values is subtly and refinedly defined.

The assumption of such monologism is that a text (re) presents a coherent set of values coherently, a set of values that remains coherent from beginning to end, from one end of the text’s spectrum of value to the other. A desire to historicize the text therefore leads to associating the text with what is taken to be a particular position within the originating socio-political structure. It is this move which leads to such exclusive description of the text on one hand as a «masterpiece of feudal culture»²¹, and on the other as a piece of monarchical propaganda: contemporary historiography has taught us to recognize the concepts of «feudalism» and «monarchy» as antinomial, representing contradictory forms of social organization, and their fusion in a hypothetical feudal monarchy as a momentary syncretism at best²².

Rather than defining an exclusive categorical ontological opposition which determines the categorization of the text as one or the other, this opposition defines the central problematic of the text. The form it takes at the end of the narrative is juridical, it is the very substance of Ganelon’s trial, but that juridical substance is also political in nature. From our twentieth century perspective, it is possible to see that the issues are those which will occupy the French polity for a long time after the composition of the text. From that perspective, it is also possible to see Ganelon’s trial as a fragment in the Adorno sense: simultaneously integrated within a narrative structure and refusing that integration, the trial episode enters into an analytic series that far transcends its function within the particular text. This is not an aesthetic transcendence of historical givens: on the contrary, it is the transcendence of a material historiography over any attempted idealistic subsumption of politics.

Accused of treason, Ganelon argues that his accusers are correct on the facts, but wrong on their legal interpretations. His self-defense is based on the generally recognized right of the feudal noble to engage in private warfare, to assert one’s own value and interests without limitation, provided only that one’s antagonist is given fair warning: he is not to be attacked without prior notification²³. As the text demonstrates, Ganelon did give fair warning: he defied Roland quite openly and before witnesses. Not only were the assembled French barons present to hear him interrupt the condition of «friendly relations» with Roland; Ganelon manages the impressive rhetorical and juridical coup of invoking, as witness, the individual who is both his accuser and the ultimate executive of the judgement, Charlemagne. As Ganelon argues:

«Jo desfiai Rollant le poigneor

E Oliver e tuiz lur compaignun

Carles 1’oid e si nobilie baron.

Venget m’en sui, mais n’i ad traisun.»

«I challenged Roland the warrior

And Oliver and all their companions.

Charles heard it and his noble barons.

I took vengeance, but that’s no treason²⁴.»

The import of Ganelon’s self-defense must be stressed. Ganelon invokes feudal right and law in his favor: he was only doing what any noble had the right to do, and might well have done, in a similar situation. With this argument, Ganelon puts the feudal nobility on his side: small wonder that the jury of his peers decides not to condemn him, but to recommend to Charles the kind of compromise that, even with a fairly early dating of the text, was already becoming the norm²⁵. Ganelon’s behavior and Ganelon’s case are at one, in other words, with the system of values and the aristocratic practice that functions as the encoded norm during the longest part of the text: all of it, up to the trial scene. Ganelon’s position, in defending himself against the king, is that of feudalism; the opposition is that between the feudal nobility and the monarchical principle.

This is confirmed by the counter-argument put forward by Ganelon’s accuser, Thierry de Chartres. This counter-argument, which will carry the day, is a radically new theory of monarchical precedence. In fact, the king of France is one of the feudal nobility himself, far from the most powerful, and needs to scrabble for his own independence either as noble or as sovereign²⁶. To claim a practical, enactable superiority for kingship, as does Thierry, when he claims that being in the service of the king ought to have protected Roland, is to claim a right that does not exist, in the reality of the time as well as in the rest of the text²⁷. It is a transgression of the established order, cultural, political, and hence textual. In this transgression, it refuses recuperation by the rest of the Chanson de Roland, that is, by the system of values incorporated in the preceding narrative.

In juxtaposition, let us propose another text, another fragment. It tells the following story. A state of war exists between the king and a noble, a situation both frequent and normal under the conditions of the twelfth century. An attack is undertaken by a foreign power against the king concerned. The noble must therefore make a decision: continue fighting against the king, which would set the noble at least implicity in alliance with the foreign attacker; or interrupt the hostilities between himself and the king, in an alliance against the foreign attacker. In fact, it is the second alternative that is narrated by the text, with some amazement on the part of the enunciator. In spite of obvious differences in the narrative affabulation, the valorial identity of the two texts is remarkable: given the conflict between the feudal value and the value of royalty as a socio-culturally encoded norm, it is the latter that is chosen in both cases. The fact is that, given two fragmentary «takes» on the relation between two socio-political instances, the two texts not only repeat the structural opposition between the two instances, the two texts also make the same gesture of the subsumption of one instance by the other. The structure of political values in the two texts is identical. Both texts narrate the necessity of a political choice, a political choice which imposes a new totality on a prior set of social arrangements. The iterative character of the choice, however, itself establishes a new textual totality, that of texts narrating this choice. Each of the texts cited has been ripped from its original, textual narrative context, in disregard of its functional insertion within each of the originating texts. As a result, a new cognitive totalization has been achieved, one which subsumes the fragments in a category that might be labelled: «texts narrating the imposition of monarchical totalization on feudal fragmentation».

V.

A text is produced at a given historical moment, in specific insertions, social, political, textual, and aesthetic. The relations among these various instances are complex; while it may be possible to abstractly chart their relations for different types of social formations, the ways in which they play into concrete individuals — be they people or texts — are quite unpredictable. No theoretical structure can confidently predict the particular ways in which texts will incorporate, refuse, modify, or betray the semantic values of their historical insertions. This does not mean that the insertions and the values they imply are inexistent or irrecuperable. It only means that the relations among the epistemic instances deployed in reading — at this time — are not fixed, but subject to continual improvisation, an improvisation that may turn out to be a temporization dialectically induced: there is no reason to confidently predict the permanence of the present cognitive indeterminacy.

Two texts, taken out of their originating narrative contexts, have been juxtaposed and analyzed, to determine a fragmentary reading. A different analysis, more cognizant of narratological categories, would show that the two syntagms are not fragments but parts, by displaying the insertion of these partial syntagms within complex systems of textual coherence, both narratological and ideological. Such analysis, more sensitive to narrative structure semiotically perceived, should incorporate semantic values identified by the kind of fragmentary reading we have presented. The «double historicity of text and reader» has little meaning and less interest if that historicity is not taken to imply some isotopy, some dimension of meaning other than that which is denoted by terms such as «literary», «literariness», «self-referentiality», «fictional» or simply «textual». Granted that, in the concrete passages of real life, the two isotopies of /fictionality/ and /historicity/ cannot never absolutely be distinguished, that each is continually, permanently suffused with the other, they cannot simply be used indistinguishably: they «mean» differently, they refer to different elements of meaning, they produce different significations.

Recognizing the double historicity of text and reader, the chanson de geste gives itself as the narrative of res gestae, as a historical narrative, which we as twentieth century readers cannot help but take as fictional. Both the text and we are right: «to be right» has meaning only historically, since the criteria of knowledge vary from one epoch to another. In a broader sense (i.e., in the context of a larger totality), to be cognitively right as a «historical» act must include recognition of these differences. Thus, if the chansons de geste in general, and the Chanson de Roland in particular, give themselves as historical, the equivalent questioning must focus on the other text as well. It gives itself as historical (given all sorts of differences between the conception of historicity that separate the twentieth century from the twelfth), and it is precisely because it gives itself as a historical text that we twentieth century readers of Barthes, Derrida, Foucault, Louis Mink, and Hayden White, must take it as a fictional text. For it is a text with important components of hagiography, of adulation, and with strong political motivations. The narrative is the Life of Louis VIth, written by the Abbot suger of Saint Denis, the famous first theologian of the French monarchy, the first to verbalize in his texts the theme and theory of royal primacy within the feudal system²⁸. Presumably, he believed his biography to be of a different historical nature than the Chanson de Roland.

Whatever the Abbot’s opinion about the historicity of his text in comparison to that of the Roland, however, our perceptions of the two texts do make a difference between them. That The Life of Louis VI is «about» an individual living within Suger’s time-frame; that the text-enunciator knew his subject directly; that he advised the king repeatedly in multiple contexts; that in addition to wielding power as Abbot of Saint Denis, and being Louis VIth’s councilor, he also functioned as Regent during Louis VIIth’s absence from France during the Second Crusade; in other words, that Suger’s text not only reports on events but inscribes narrative dynamics with social and political effectivities means that his text necessarily questions us twentieth century readers in ways, to extents, that the Roland does not. That question is the question of the relationship between the textual and its «externalities»: that which it is «about», that to which it points, that to which it «makes reference». Alternatively, this may be phrased as the question of the text’s insertability: what are the various nontextual contexts which must figure as the text’s co-textuality? And this alternative phrasing of the question in turn becomes the question of the choices of insertion made by the Subject of phenomenological knowledge, choices with valences simultaneously methodological, epistemic, and political. These contemporary forms of the traditional question of referentiality are ineluctably posed by the Abbot Suger for a twentieth century reader in a way, with a force, which does not obtain in relation to his reading of the Roland. All texts are historical, but not all texts are histories.

Take Suger’s narrative of the invasion of France threatened by the Emperor Henry in 1124 near Metz. He recounts the agreeable surprise of Louis VIth who, having advised his vassals of the imminent danger, having perhaps recalled how few of them deigned to attend his royal court and that of his father Philip the First, nevertheless discovered an extraordinary number of those vassals assembled at the field of battle in response to his summons. Among them was the Count of Champagne Thibault IV, who put aside — for the occasion! — the war he had been waging, with the entire legitimacy of a great aristocrat, against his king. What surprises the narrator here is not at all the war between king and noble, which is stated matter-of-factly as a given of history, but the fact of the noble’s interruption of his legitimate war against his suzerain and sovereign in this moment of historical danger. Does this narrative, this story, claim only a textual consistency? Does its text not bear a particular weight, a specific density as well as a specific problematic, in the wake of its referential claims? Granted the likelihood of some hyperbolization, of some ideological shading of the reportage — of some degree of fictionalization, in other words — must one not nevertheless somehow «take seriously», in the common idiom, this image of men’s corporeal displacements, of bodies assembled on the field of battle near Metz, bodies of nobles, bodies of knights, in army corps, in the battlefield deployment of bone and flesh and the hazard of their survival, in the wager of a value other than that of the textual, even if, in lived reality as in textual theory, the two meanings of «(hi) story» are never entirely separable? At the moment when the corporeal isotopy interferes with the textual, isn’t there a necessity of giving a different kind of emphasis to the multiple insertions of these bodies, agglomerated in surprising number — says the text — that spring day of 1124?

After all, if Lyotard is right, and «speaking is fighting», so that «speech acts» and the linguistic games that are also «language moves» belonging to a «generalized agonistics», one is forced to ask the question of the respective identities of the combatants, the rules of their agonistic encounters, and the sakes at play in their games — one wants to ask that question²⁹. This passage, from Suger’s life of a sovereign he loved, is frequently cited by historians as the first historical evidence of a renascent feeling of French nationalism in the twelfth century. As textual evidence, it is perhaps no more demonstrative than the Song of Roland: does not the very long string of later texts incorporating, developping, praising and critiquing the forms and moments of nationalism in later (hi) story, with their sedimentation of human investment, make a claim to a kind of historicity that must be different in import from the claim we attach to the text labelled «epic» and hence «fictional»? And does not that claim imply (in the strong logical sense of that term) a recuperation of fragmentation in a totalizing epistemological category which pushes inleluctably against the fragmentation of knowledge?

It is at this point of the argument that the theoretical and methodological cautions developed by the conjunction of (post-) structuralisms and the Frankfurt School must themselves be understood as heuristic limits, quite possibly necessary at a given moment of history, but fated to be overtaken by the development of new disciplines, new forms of knowledge, which these cautions themselves help bring into particular existence within the general dynamic of the epistemological quest. As «post-structuralist», as «post-» or «neo-» Marxist, these schools abandon — with good reason — ancient theoretical positions with the unacceptable political effects they imply. Theory, in refusing theoretical positions because of their historical political effect, thereby re-instates the effectivity of the relations — however complex — between text and history. But the cognitive dynamics of comprehension — as the ambiguity of that word already suggests — are precisely to be even more comprehensive, to «comprehend» its object in new contextualizations that transform context into co-text, that establish new objects of knowledge through the development of new epistemic principles.

More concretely, in our time, the challenge is to «comprehend» the textual phenomenon, or rather, textual phenomena in their plurality and variabilities, as social activities, integrated within social structures, patterns, and practices, with valences that are always complex, often contradictory, simultaneously functional and protesting, sometimes even inventing new social and political modalities. In the particular case at hand, a crucial element of our cognitive historical position is indeterminate: we do not know the chronological relationship between the two texts that form the basis of our totality. The traditional dating, best represented by Joseph Bedier, places the Chanson de Roland at the turn of the century: 1100 or so, give or take twenty years or so in either direction. A more recent revisionist historiography tends to indentify the text more closely with its manuscript, and to date it in the second half of the twelfth century, say around 1150-1170³⁰. These two datings of the Roland frame both the event of 1124, and Suger’s redaction of Louis’ biography. As a result, it is uncertain whether temporal priority goes to the historical event or to the historically fictional text: whether ideological coding precedes political phenomenality, whether the symbolic precedes the material, or vice-versa, remains indeterminate within the information of our historical documentation. Historical chronology cannot help decide between the two directions of historical causation, from the mental to the material, or from the material to the mental. Temporal totalization is impossible in the sequencing of our two fragments.

What is certain is that the insertion of the text into a historical framework — since text and history are represented by different disciplines in the epistemic distribution of the academy — is a necessary step in the development of our comprehension of the text. One necessary, heuristic stage in that development is the fragmentation of the text, as its different moments are inserted into different historical Momente. The insertion of Ganelon’s trial into the same field of study as Suger’s narrative reveals (or confirms) its refusal of integration into the system of value that seemed to operate in the prior sequences of the narrative text. Its integration into the later textual series as the isotopy of /French nationalism/ — of which it is at least one of two initiating instances — defines the inadequacy of the preceding framework of /feudalism/ in accounting for this fragment. What seemed self-evidently a narrative sequence, hence a functional part, within the overall linear sequency of the totalized narrative text, is revealed to refuse that narrative integration, to insist upon its own fragmentariness in relation to that text. On the contrary, if integration is there, it is more probably an integration into future narratives yet undreamt of at the time of the Roland.

A tertium quid is perhaps possible between these two exclusive integrations of the individual concrete text into mutually exclusive semiotic systems of value, each of which produces an equally unsatisfying fragmentation. The assumption in that binary opposition is that the two systems of value are stable synchronic entities, and that the text must be located in the unifying force field of either one or the other of the two antagonistic value systems. An alternative theoretical view is that the text is constituted, not by aesthetic unity, but by the semiotic coherence of contrary value systems, simultaneous and conflicting with each other in the social body, and hence co-present in the text which we call a literary text but which is always already a social text. Ex hypothesis, the semiotic text is not constituted by a stable semantic coherence that is merely a intertextual transform of traditional aesthetic unity, but by the process of transformation which begins with one historical system of value and ends with another. According to this hypothesis, the text is not only the incidental incorporation of a particular socio-historical system of value: it may also be an actor in the historical field, incorporating certain historical givens, and determining certain social effects. A proper semiotic reading might then both account for the textual coherence of the text, and for its social and historical agency. These are not contradictory requirements, they are in fact the same requirement differentially represented by our epistemic distribution. This epistemic and methodological framework may help resolve the apparently contradictory effets de réel we produce through our words «fragment» and «totality».