Roland Barthes is a key figure in international intellectual life. He is one of the most important intellectual figures to have emerged in postwar France and his writings continue to have an influence on critical debates today.  When he died in 1981, he left a body of major work but, as many of his friends and his admirers claimed, with still more important work to come. I can't possible hope to do justice to the diversity of his various writings here - I can only point you in the direction of Culler (1983), Moriarty (1991) and Rylance (1994) where you will find good accounts of his career - so I will plunge straightaway into a discussion of Mythologies, which is one of his earliest and most widely-read works. Mythologies is one of Barthes's most popular works because in it we see the intellectual as humourist, satirist, master stylist and debunker of the myths that surround us all in our daily lives.

The formal description of sets of words beyond the level of the sentence (what we call for convenience discourse) is not a modern development: from Gorgias to the nineteenth century, it was the special concern of traditional rhetoric. Recent developments in the science of language have nonetheless endowed it with a new timeliness and new methods of analysis: a linguistic description of discourse can perhaps already be envisaged at this stage; because of its bearings on literary analysis (whose importance in education is well known) it is one of the first assignments for semiology to undertake.

This second level of linguistics, which must look for the universals of discourse (if they exist) under the form of units and general rules of combination, must at the same time obviously give an answer to the question whether structural analysis is justified in retaining the traditional typology of discourses; whether it is fully legitimate to make a constant opposition between the discourses of poetry and the novel, the fictional narrative and the historical narrative. It is the last point which gives rise to the reflections set down here. Does the narration of past events, which, in our culture from the time of the Greeks onwards, has generally been subject to the sanction of historical 'science', bound to the unbending standard of the 'real', and justified by the principles of 'rational' exposition - does this form of narration really differ, in some specific trait, in some indubitably distinctive feature, from imaginary narration, as we find it in the epic, the novel, and the drama? And if this trait or feature exists, then in what level of the historical statement must it be placed?(1) In order to suggest a reply to this question, we shall here be looking, in a free and far from exhaustive fashion, at the discourse of a number of great classic historians: Herodotus, Machiavelli, Bossuet and Michelet.

I. THE ACT OF UTTERING

First of all, we may ask under what conditions the classic historian is enabled -or authorized - himself to designate, in his discourse, the act by which he promulgates it. In other words, what, on the level of discourse - and not of language, are the shifters (in Jakobson's sense of the term)(2) which assure the transition from the utterance to the act of uttering (or vice versa) ?

It would appear that historical discourse involves two regular types of shifters. The first type comprises what we might call the shifters of listening. This category has been identified by Jakobson, on the level of language, with the term testimonial, according to the formula CeCa1/Caa2: in addition to the event reported (Ce), discourse mentions at the same time the act of the informer (Ca1), and the speech of the utterer which is related to it (Ca2). This form of shifter thus designates any reference to the historian's listening, collecting testimony from elsewhere and telling it in his own discourse. Listening made explicit represents a choice, for it is possible not to refer to it at all; it brings the historian closer to the anthropologist, in so far as he mentions the source of his information. Thus we find an abundant use of this shifter of listening among historian/anthropologists like Herodotus. The forms vary: they range from phrases of the type of as I have heard, or to my knowledge, to the historian's use of the present tense which testifies to the intervention of the utterer, and to any mention of the historian's personal experience. Such is the case with Michelet, who 'listens to' the History of France as a result of an overwhelming personal experience (of the Revolution of July 1830)and takes account of this in his discourse. The listening shifter is obviously not distinctive to historical discourse: it is found frequently in conversation, and in certain expository devices used in the novel (such as anecdotes which are taken from fictional sources of information mentioned in the text).

The second type of shifter comprises all the explicit signs whereby the utterer - in this case, the historian - organizes his own discourse, taking up the thread or modifying his approach in some way in the course of narration: that is to say, where he provides explicit points of reference in the text. This is an important type of shifter, and there can be many different ways of 'organizing' discourse accordingly; but these different instances can all be subsumed under the principle that each shifter indicates a movement of the discourse in relation to its matter, or more precisely a movement in relation to the sequence of its matter, rather like the operation of the temporal and locational deictics 'here is/there is'. Thus we can cite as cases where the shifter affects the flow of utterances: the effect of immobility (comme nous l'avons dit plus haut), that of returning to an earlier stage (altius repetere, replicare da piu alto luogo), that of coming back again (ma ritornando all'ordine nostro, dico come. . . ), that of stopping dead (sur lui, nous n'en dirons pas plus), and that of announcing (voici les autres actions dignes de memoire qu'il fit pendant son regne). The organizing shifter poses a problem which is worthy of attention, though it can only be lightly indicated here: this is the problem arising from the coexistence, or to be more exact the friction between two times - the time of uttering and the time of the matter of the utterance. This friction gives rise to a number of important factors in historical discourse, of which we shall mention three. The first relates to the many ways of producing the phenomenon of acceleration in a historical account: an equal number of pages (if such be the rough measure of the time of uttering) can cover very different lapses of time (the time of matter of the utterance). In Machiavelli's History of Florence the same measure (a chapter) covers in one instance a number of centuries, and in another no more than two decades. The nearer we are to the historian's own time, the more strongly the pressure of the uttering makes itself felt, and the slower the history becomes. There is no such thing as isochrony - and to say this, is to attack implicitly the linearity of the discourse and open it up to a possible 'paragrammatical' reading of the historical message.(3) The second point also reminds us, in its we, that this type of discourse - though linear in its material form - when it is face to face with historical time, undertakes (so it would appear the role of amplifying the depth of that time. We become aware of what we might call a zig-zag or saw-toothed history. A good example i Herodotus, who turns back to the ancestors of a newcomer, and the returns to his point of departure to proceed a little further -and the starts the whole process all over again with the next newcomer. Finally there is a third factor in historical discourse which is of the utmost importance, one which bears witness to the destructive effect organizing shifters as far as the chronological time of the history concerned. This is a question of the way historical discourse is inaugurated, of the place where we find in conjunction the beginning of the matter of the utterance and the exordium of the uttering.(4) Historical discourse is familiar with two general types of inauguration in the first place, there is what we might call the performative opening for the words really perform a solemn act of foundation; the model for this is poetic, the I sing of the poets. So Joinville begins his history with a religious invocation (Au nom de Dieu le tout-puissant, je, Jehan, sire, Joinville, fais ecrire la vie de nostre Saint roi Louis), and even the socialist Louis Blanc does not disdain the purificatory introit, (5) so evident is it that the beginnings of speech always carry with them a kind of difficulty, perhaps even a sacred character. Then there is a much more commonly found element, the Preface, which is an act of uttering characterized such, whether prospectively in so far as it announces the discourse come, or retrospectively in that it embodies a judgement on the discourse. (Such is the case with the Preface which Michelet wrote to crown his History of France, once it had been completely written and published.) Bearing in mind these different elements, we are likely to conclude that the entry of the act of uttering into the historical utterance, through these organizing shifters, is directed less towards offering the historian a chance of expressing his 'subjectivity', as is commonly held, than to 'complicating' the chronological time of history by bringing it up against another time, which is that of the discourse itself and could be termed for short the 'paper-time'. To sum up, the presence in historical narration of explicit signs of uttering would represent an attempt to 'dechronologize' the 'thread' of history and to restore, even though it may merely be a matter of reminiscence or nostalgia, a form of time that is complex, parametric and not in the least linear: a form of time whose spatial depths recall the mythic time of the ancient cosmogonies, which was also linked in its essence to the words of the poet and the soothsayer. Organizing shifters bear witness, in effect -- though they do so through indirect ploys which have the appearance of rationality - to the predictive function of the historian. It is to the extent that he knows what has not yet been told that the historian, like the actor of myth, needs to double up the chronological unwinding of events with references to the time of his own speech.

The signs (or shifters) which have just been mentioned bear solely on the very process of uttering. There are other signs which refer no longer to the act of uttering, but to what ]akobson calls its protagonists (Ta): the receiver and the sender. It is a fact worthy of note, and somewhat mysterious at the same time, that literary discourse very rarely carries within it the signs of the 'reader'. Indeed we can say that its distinctive trait is precisely that it is - or so it would appear - a discourse without the pronoun 'you', even though in reality the entire structure of such a discourse implies a reading 'subject'. In historical discourse, the signs of the receiver are usually absent: they can be found only in cases where History is offered as a lesson, as with Bossuet's Universal History, a discourse which is explicitly addressed by the tutor to his pupil, the prince. Yet in a certain sense, this schema is only possible to the extent that Bossuet's discourse can be held to reproduce by homology the discourse which God himself holds with men - precisely in the form of the History which he grants to them. It is because the History of men is the Writing of God that Bossuet, as the mediator of this writing, can establish a relationship of sender and receiver between himself and the young prince.

Signs of the utterer (or sender) are obviously much more frequent. Here we should class all the discursive elements through which the historian - as the empty subject of the uttering - replenishes himself little by little with a variety of predicates which are destined to constitute him as a person, endowed with a psychological plenitude, or again

(the word hasa precious figurative sense) to give him countenance.(6) We can mention at this point a particular form of this 'filling' process, which is moredirectly associated with literary criticism. This is the case where the utterer means to 'absent himself' from his discourse, and where there is in consequence a systematic deficiency of any form of sign referring to the sender of the historical message. The history seems to be telling itself all on its own. This feature has a career which is worthy of note, since it corresponds in effect to the type of historical discourse labelled as 'objective' (in which the historian never intervenes). Actually in this case, the utterer nullifies his emotional persona, but substitutes for it another persona, the 'objective' persona. The subject persists in its plenitude, but as an objective subject. This is what Fustel de Coulanges referred to significantly (and somewhat naively) as the 'chastity of History'. On the level of discourse, objectivity - or the deficiency of signs of the utterer - thus appears as a particular form of imaginary projection, the product of what might be called the referential illusion, since in this case the historian is claiming to allow the referent to speak all on its own. This type of illusion is not exclusive to historical discourse. It would be hard to count the novelists who imagined - in the epoch of Realism - that they were 'objective' because they suppressed the signs of the 'I' in their discourse! Today linguistics and psychoanalysis have made us much more lucid with regard to privative utterances: we know that absences of signs are also in themselves significant.

To bring this section which deals with the act of uttering to a close, we should mention the special case - foreseen by Jakobson and placed within his lattice of shifters, on the linguistic level - when the utterer of the discourse is also at the same time a participant in the process described in the utterance, when the protagonist of the utterance is the same as the protagonist of the act of uttering (Te/Ta): that is, when the historian, who is an actor with regard to the event, becomes its narrator, as with Xenophon, who takes part in the retreat of the Ten Thousand and subsequently becomes its historian. The most famous example of this conjunction of the I in the utterance and the I in the act of uttering is doubtless the he of Caesar's Gallic War. This celebrated he belongs to the utterance; when Caesar explicitly undertakes the act of uttering he passes to the use of we (ut supra demonstravimus). Caesar's he appears at first sight to be submerged amid the other participants in the process described, and on this count has been viewed as the supreme sign of objectivity. And yet it would appear that we can make a formal distinction which impugns this objectivity. How ? By making the observation that the predicates of Caesar's he are constantly pre-selected: this he can only tolerate a certain class of syntagmas, which we could call the syntagmas of command (giving orders, holding court, visiting, having things done, congratulating, explaining, thinking). The examples are, in effect, very close to certain cases of the performative, in which speech is inextricably associated with action. Other instances can be found for this he who is both a past actor and a present narrator (particularly in Clausewitz). They show that the choice of an apersonal pronoun is no more than a rhetorical alibi, and that the true situation of the utterer is clear from the choice of syntagmas with which he surrounds his past actions.

II. THE UTTERANCE

It should be possible to break down the historical utterance into units of content, which can then be classified. These units of content represent what is spoken of in the history; in so far as they are signifieds, they are neither the pure referent nor the discourse as a whole: their wholeness is constituted by the referent inasmuch as it has been broken down, named and rendered intelligible, but not yet made subject to a syntax. We shall not attempt to go deeply into the investigation of these classes of units in this article. Such an effort would be premature. We shall confine the discussion to a few preliminary remarks.

The historical utterance, just like the utterance in sentence form, involves both 'existents" end 'occurrents', that is beings or entities, and their predicates. Now an initial examination enables us to foresee that both of these categories, in their different ways, can form lists that are to a certain extent closed, and therefore accessible to comprehension: in a word, they can form collections, whose units end up by repeating themselves, in combinations that are obviously subject to variation. Thus, in Herodotus, the existents can be reduced to dynasties, princes, generals, soldiers, peoples, and places, and the occurrents to actions like laying waste, putting into slavery, making alliances, organizing expeditions, reigning, using stratagems, consulting oracles etc. These collections, in so far as they are (to a certain extent) closed, should observe certain rules of substitution and transformation and it ought to be possible to structure them - a task which is obviously more or less easy according to the historian. The units found in Herodotus, for example, depend largely on a single lexicon, which is that of war. It would be an interesting question to investigate whether, for more modern historians, we should expect to find more complex associations of different lexicons, and whether, even in this case, historical discourse would not turn out to be based, in the last resort, on strong collections (it is preferable to talk of collections, rather than of lexicons, since here we are discussing only the level of the content). Machiavelli seems to have had an intuitive understanding of this type of structure: at the beginning of the History of Florence, he presents his 'collection', that is to say the list of juridical, political and ethnic objects which will subsequently be mobilized and set in combination in his narrative.

In the case of less well defined collections (in historians who are less archaic than Herodotus), the units of content may nonetheless receive a strong structuring which derives not from the lexicon, but from the personal thematic of the author. These (recurrent) thematic objects are numerous in the case of a Romantic author like Michelet, but we can also find them without any difficulty in authors who are reputedly more intellectual. In Tacitus, fama is such a personal unit, and Machiavelli establishes his history on the thematic opposition between mantenere (a verb which refers to the basic energy of the statesman) and ruinare (which, by contrast, implies the logic of affairs in a state of decline).(7) It goes without saying that, by means of these thematic units, which are most often imprisoned within a single word, we can find units of the discourse

(and not of the content alone). So we come to the problem of the naming of historical objects. The word can convey with economy a situation or a sequence of actions; it aids structuring to the extent that, when it is projected on to the level of content, it forms in itself a small-scale structure So it is with Machiavelli's use of the conspiracy to save having to make fully explicit a complex datum, which designates the sole possibility of struggle remaining when a government has vanquished every form of opposition that can be displayed in the open. The very act of naming, which enables the discourse to be strongly articulated, is a reinforcement of its structure. Strongly structured histories are histories which give an important place to the substantive: Bossuet, for whom the history of men is structured by God, makes abundant use of substantives in sequence as a short-cut.(8)

These remarks are just as applicable to the occurrents as to the existents. The processes of history in themselves (however they happen to be developed through the use of terminology) pose an interesting question- among so many others, that of their status. The status of a process may be affirmative, negative or interrogative. But the status of historical discourse is uniformly assertive, affirmative. The historical fact is linguistically associated with a privileged ontological status: we recount what has been, not what has not been, or what has been uncertain. To sum up, historical discourse is not acquainted with negation (or only very rarely, in exceptional cases). Strangely enough, but significantly, this fact can be compared with the tendency which we find in a type of utterer who is very different from the historian: that is, the psychotic, who is incapable of submitting an utterance to a negative transformation.(9) We can conclude that, in a certain sense, 'objective' discourse (as in the case of positivist history) shares the situation of schizophrenic discourse. In both cases, there is a radical censorship of the act of uttering (which has to be experienced for a negative transformation to take place), a massive flowing back of discourse in the direction of the utterance and even (in the historian's case) in the direction of the referent: no one is there to take responsibility for the utterance.

To introduce another aspect, an essential aspect, of the historical utterance, we must turn to the classing of units of content, and the way in which they fall into succession. As far as a preliminary sample seems to indicate, classes of this kind are the very same as we have claimed to discover in the fictional narrative.(10) The first class comprises all the segments of the discourse which lead back to an implicit signified, through the process of metaphor; so we have Michelet describing the motley clothing, the garbling of coats of arms and the mixture of architectural styles, at the outset of the fifteenth century, as so many signifiers of a single signified, which is the disintegration of morality at the close of the Middle Ages. This particular class is therefore one of indices, or more exactly of signs (and it is a class very frequently found in the classic novel). The second class of units is formed by the fragments of discourse which are rational, or syllogistic by nature: it would perhaps be more accurate to call them enthymematic, since it is almost always a case of syllogisms which are approximate, or incomplete. (11) Enthymemes are not exclusive to historical discourse; they occur frequently in the novel, where bifurcations in the anecdote are generally justified in the eyes of the reader by pseudo-reasonings of a syllogistic type. The enthymeme confers upon historical discourse a non-symbolic intelligibility, and for this reason it deserves attention. Does it still exist in historical studies, where the discourse attempts to break with the class Aristotelian model? Lastly, there is a third class of units - which is no means the least important - comprising what we have tended to call, after Propp, the 'functions' of the narrative, or the cardinal points whence the anecdote may adopt a different course. These functions grouped together: they may be syntagmatically grouped in a closed succession, with a high degree of logical entailment or sequential order. Thus, in Herodotus, we can find on more than one occasion an Oracle sequence, composed of three terms, each of which presents an alternative (to consult or not, to answer or not, to follow or not); these may separated one from the other by other units which are foreign to sequence. The foreign units are either the terms of another sequence, in which case the schema is one of imbrication; or they are minor expansions (items of information, indices), in which case the schema as a catalyst which fills the interstices between the core elements.

To generalize - perhaps unwarrantably - from these few remark the structure of the utterance, we may offer the suggestion that historical discourse oscillates between two poles, according to whether it is indices or functions that predominate. When the indexical units predominate in a historian (testifying at every moment to an implicit signified), his is drawn towards a metaphorical form and borders upon the lyrical and symbolic. This is the case, for example, with Michelet. When, by contrast, it is the functional units which predominate, History takes on a metonymic form and becomes a close relation of the epic. An example of this tendency can be found in the narrative history of Augustin Thierry. There exists, it is true, yet another form of History: the History which tries to reproduce in the structure of the discourse the structure of the choices lived through by the protagonists of the process described. Here reasoning is dominant; the history is a reflexive one, which we might also call strategic history, and Machiavelli would be its best demonstration.

III. SIGNIFICATION

For History not to signify, discourse must be confined to a pure, unstructured series of notations. This is the case with chronologies and annals (in the pure sense of the term). In the fully formed (or, as we say, 'clothed') historical discourse, the facts related function inevitably either as indices, or as core elements whose very succession has in itself an indexical value. Even if the facts happen to be presented in an anarchic fashion, they still signify anarchy and to that extent conjure up a certain negative idea of human history.

The signifieds of historical discourse can occupy at least two different levels. First of all, there is the level which is inherent to the matter of the historical statement. Here we would cite all the meanings which the historian, of his own accord, gives to the facts which he relates (the motley costumes of the fifteenth century for Michelet, the importance of certain conflicts for Thucydides). Into this category also fall the moral or political 'lessons' which the narrator extracts from certain episodes (in Machiavelli, or Bossuet). If the lesson is being drawn all the time, then we reach a second level, which is that of the signified transcending the whole historical discourse, and transmitted through the thematic of the historian - which we can thus justifiably identify as the form of the signified. So we might say that the very imperfection of the narrative structure in Herodotus (the product of a number of series of facts without conclusion) refers in the last instance to a certain philosophy of history, which is the submission of the world of men to the workings of the divine law. In the same way in Michelet, we can find that particular signifieds have been structured very strongly, and articulated in the form of oppositions (antitheses on the level of the signifier), in order to establish the ultimate meaning of a Manichean philosophy of life and death. In the historical discourse of our civilization, the process of signification is always aimed at 'filling out' the meaning of History. The historian is not so much a collector of facts as a collector and relater of signifiers; that is to say, he organizes them with the purpose of establishing positive meaning and filling the vacuum of pure, meaningless series.

As we can see, simply from looking at its structure and without having to invoke the substance of its content, historical discourse is in its essence a form of ideological elaboration, or to put it more precisely, an imaginary elaboration, if we can take the imaginary to be the language through which the utterer of a discourse (a purely linguistic entity) 'fills out' the place of subject of the utterance (a psychological or ideological entity). We can appreciate as a result why it is that the notion of a historical ' fact' has often aroused a certain degree of suspicion in various quarters. Nietzsche said in his time: 'There are no facts in themselves. It is always necessary to begin by introducing a meaning in order that there can be a fact.' From the moment that language is involved (and when is it not involved?), the fact can only be defined in a tautological fashion: what is noted derives from the notable, but the notable is only - from Herodotus onwards, when the word lost its accepted mythic meaning what is worthy of recollection, that is to say, worthy of being noted. thus arrive at the paradox which governs the entire question of the distinctiveness of historical discourse (in relation to other types discourse). The fact can only have a linguistic existence, as a term in a discourse, and yet it is exactly as if this existence were merely the 'copy', purely and simply, of another existence situated in the extra structural domain of the 'real'. This type of discourse is doubtless the only type in which the referent is aimed for as something external the discourse, without it ever being possible to attain it outside the discourse. We should therefore ask ourselves in a more searching way what place the 'real' plays in the structure of the discourse.

Historical discourse takes for granted, so to speak, a double operation which is very crafty. At one point (this break-down is of course only metaphorical) the referent is detached from the discourse, becomes external to it, its founding and governing principle: this is the point of the res gestae, when the discourse offers itself quite simply as historia rerum gestarum. But at a second point, it is the signified itself which forced out and becomes confused with the referent; the referent enters into a direct relation with the signifier, and the discourse, solely charged with expressing the real, believes itself authorized to dispense with the fundamental term in imaginary structures, which is the signified. As with any discourse which lays claim to 'realism', historical discourse on admits to knowing a semantic schema with two terms, the referent and the signifier; the

(illusory) confusion of referent and signified is, as know, the hallmark of auto-referential discourses like the performative. We could say that historical discourse is a fudged up performative, which what appears as statement (and description) is in fact no more than the signifier of the speech act as an act of authority.(12)

In other words, in 'objective' history, the 'real' is never more than an unformulated signified, sheltering behind the apparently all-powerful referent. This situation characterizes what we might call the realistic effect. The signified is eliminated from the 'objective' discourse, and ostensibly allows the 'real' and its expression to come together, and this succeeds in establishing a new meaning, on the infallible principle already stated that any deficiency of elements in a system is in its' significant. This new meaning - which extends to the whole of historical discourse and is its ultimately distinctive property - is the real in itself surreptitiously transformed into a sheepish signified. Historical discourse does not follow the real, it can do no more than signify the real, constantly repeating that it happened, without this assertion amounting to anything but the signified 'other side' of the whole process of historical narration.

The prestige attached to it happened has important ramifications which are themselves worthy of historical investigation. Our civilization has a taste for the realistic effect, as can be seen in the development of specific genres like the realist novel, the private diary, documentary literature, news items, historical museums, exhibitions of old objects and especially in the massive development of photography, whose sole distinctive trait (by comparison with drawing) is precisely that it signifies that the event represented has really taken place.(13) When the relic is secularized, it loses its sacred character, all except for that very sacredness which is attached to the enigma of what has been, is no longer, and yet offers itself for reading as the present sign of a dead thing. By contrast, the profanation of relics is in fact a destruction of the real itself, which derives from the intuition that the real is never any more than a meaning, which can be revoked when history requires it and demands a thorough subversion of the very foundations of civilized society.(14)

History's refusal to assume the real as signified (or again, to detach the referent from its mere assertion) led it, as we understand, at the privileged point when it attempted to form itself into a genre in the nineteenth century, to see in the 'pure and simple' relation of the facts the best proof of those facts, and to institute narration as the privileged signifier of the real. Augustin Thierry became the theoretician of this narrative style of history, which draws its 'truth' from the careful attention to narration, the architecture of articulations and the abundance of expanded elements (known, in this case, as 'concrete details').(15) So the circle of paradox is complete. Narrative structure, which was originally developed within the cauldron of fiction (in myths and the first epics) becomes at once the sign and the proof of reality. In this connection, we can also understand how the relative lack of prominence (if not complete disappearance) of narration in the historical science of the present day, which seeks to talk of structures and not of chronologies, implies much more than a mere change in schools of thought. Historical narration is dying because the sign of History from now on is no longer the real, but the intelligible.

NOTES

'Le discours de l'histoire' was first published in Social Science Information (1967). See also the translation by Peter Wexler in Michael Lane, ed., Structuralism: A Reader(London, 1970), pp. 145-55.

1. Translator's note: Barthes makes frequent use in this essay of the linguistic terms: enonciation/enonce. While the latter denotes a statement or proposition, the former is used to designate the act of making the statement or proposition, in speech or writing. Since this distinction is central to Barthes' purpose, and cannot easily be conveyed in any other way, I have used the two terms: act of uttering/utterance.

2. R. Jakobson, Essais de linguistique generale (Paris, 1963), Ch. 9.

3. Following Julia Kristeva, we designate by the term 'paragrammatism' (which is derived from the anagrams of Saussure) forms of double writing, which involve a dialogue of the text with other texts, and call for a new logic (Julia Kristeva 'Bakhtine, le mot, le dialogue et le roman', Critique, 239 (April 1967), 438-65).

4. The exordium (in any form of discourse) poses one of the most interesting problem in rhetoric, to the extent that it is a codification of ways of breaking silence and combats aphasia.

5. 'Avant de prendre la plume, je me suis interroge severement, et comme je ne trouvais en moi ni affections interessees, ni haines implacables, j'ai pense que je pourrais juger les hommes et les choses sans manquer a la justice et sans trahir la verite' (L. Blanc, Histoire de dix ans (Paris, 1842)).

6 Translator's note: Barthes uses the term contenance, which combines the two sees' of 'content' and 'countenance'.

7 Cf. E. Raimondi, Opere di Niccolo Macchiavelli (Milan, 1966).

8 An example: 'On y voit avant toutes choses l'innocence et la sagesse du jeune Joseph; ses songes mysterieux; ses freres jaloux; la vente de ce grand homme; la fidelite qu garde a son maitre; sa chastete admirable; les persecutions qu'elle lui attire; sa prison et sa constance' (Bossuet, Discours sur l'histoire universelle, in Oeuvres

(Bibliotheque de la Pleiade) (Paris, 1961), p. 674).

9 L. Irigaray, 'Negation et transformation negative dans le langage des schizophrene', Langages, 5 (March 1967), 84-98.

10 Cf. 'Introduction to the Structural Analysis of Narratives', in Barthes, Image, Music, Text, translated by Stephen Heath (London, 1977), pp. 79-124.

11 Here is the syllogistic schema in a particular passage of Michelet (Histoire du moyen age, vol. iii, book vi, chapter I): (I) To distract the people from revolt, it is necessa to occupy them; (2) now, the best way to do that, is to throw them a man; (3) the princes chose old Aubriot, etc. (Translator's note: the term 'enthymema' 'enthymeme' has been used, from Aristotle onwards, to denote an argument based on merely probable grounds: i. e. a rhetorical as opposed to a demonstrative argument.)

12 Thiers expressed with great purity and naivety this referential illusion, or this confusion of referent and signified, thus fixing the ideal of the historian: 'Etre simplement vrai, etre ce que sont les choses elles-memes, n'etre rien de plus qu'elles, n'etre rien que par elles, comme elles, autant qu'elles' (quoted by C. Julian, Historiens francais du XIX siecle (Paris, n. d.), p. lxiii).