1 Several sessions at this conference have been devoted not to fictional, but to factual discourse. They have dealt with such subjects as memoirs, testimonies, news reports, and medical interviews, asking about the function that narrative may play in this kind of text and interpersonal exchange. My purpose is to continue this conversation and treat a topic that has also been touched upon during the week: historiography. But I must start with a disclaimer: I’m not an historian. I have never done research in the archives, nor participated in archeological digs. My field is the poetics of factual discourse, by which I mean the study of the rules, codes, and conventions that shape that discourse, as they shape any discourse. I will deal today with historiography, that is, not with past itself (that’s the historian’s job), but with texts that deal with the past, claiming to make valid statements about it. Looking at the relations between historiography and narrative, I will first examine the debates which those relations have generated, both in English-speaking countries and in France. Then I will ask what contribution literary theory, and especially narrative theory, can make to these controversies. I will end with a few questions often posed about historiographic structures, specifically about their relations to the data on which they are based. As this outline makes clear, my “overview” will be incomplete. Largely limited to France, England, and the United States, it will ignore countries that have made significant contributions to the debates I am about to describe. It will also be restricted to the 20^th and 21^th centuries, ignoring the long history of the relations between narrative and texts that represent actual events and situations. My predictable excuse is that one cannot do everything, that “everything” anyway is a highly problematic concept, and that if I deal mostly with contemporary French- and English-speaking historians and theorists, it is because they are the ones with whom I am the most familiar.

2 From the 1930s to the 1960s, the problem of the relations between historiography and narrative had generally been formulated in normative fashion. The question was not, “do historians use narrative?” But taking for granted that they do use narrative, is narrative a tool that is suited for serious, scientific discourse? In English-speaking countries, discussions have opposed supporters and adversaries of narrative within the framework of analytical philosophy and philosophy of science. Assuming that historians necessarily rely on narrative, philosophers like Karl Hempel and Karl Popper have argued that history, measured by the standards of physics and chemistry, was an imperfect science. This position is exemplified in Hempel’s oft-quoted articles “The Function of General Laws in History” (1942) and “Explanations in Science and History” (1962). As their titles indicate, these articles focus on the issue of “explanation.” For Hempel, scientific knowledge is only valid if it is provided under the “nomological” model; that is, if it is provided by laws which “cover” the phenomena to be described, making it possible to predict how those phenomena will unfold in the future. If I drop my pen—you will excuse the crudeness of this example—it will fall because of the law of gravity. And I can safely predict that the same phenomenon will occur in Switzerland next week and in New Hampshire in a few months. History, for Hempel, does not enable its practitioners to explain through laws the phenomena that it describes; it can only provide “explanation sketches,” which must always be supplemented in order to account for the way those phenomena unfolded. History’s explanation sketches, moreover, do not allow predictions. Historians can explain retrospectively, using “retrodiction,” how certain events occurred; but they cannot predict whether similar events will occur again, nor when.

3 Starting in the 1960s, still within the analytical tradition, several philosophers attacked the thesis according to which there is only one model of scientific knowledge—the model of physics. More precisely, they attacked the idea that the only legitimate way of accounting for a phenomenon was to identify the law that “covers” it, in Hempel’s sense. Examining the epistemological status of history, William Dray, William Gallie, Morton White, Arthur Danto, Louis Mink, and others, have asserted what Mink (1966) calls the “autonomy of historical understanding” and in the same move, rehabilitated narrative. This position can be summarized in Danto’s (1985) statement: “To tell what happened and to explain why is to do one and the same thing” (p. 202). In other words, narrative can be regarded, in Mink’s formula, as a valid “cognitive instrument,” an instrument whose function is to place an action in a temporal continuum, relating it to previous actions that led up to it, as well as to possible future scenarios. Its function, therefore, is to allow for understanding how certain events occurred, when the “covering law” model would not. Philosophers in the analytical tradition hardly give examples when they deal with history, and I could return to the primitive situation I described earlier: if I drop my pen, it will fall because of the law of gravity. But if one asks the question, “Why did the lecturer drop his pen?” no law can provide an answer. Only a narrative will explain the lecturer’s gesture. William Dray (1964) would distinguish here between “explanation by causes” and “explanation by reasons,” “causes” referring to laws, and “reasons” to the grounds that an agent may have for undertaking a certain action at a certain time.

4 While analytical philosophers make a distinction between “law” and “narrative” as modes of explanation, they do not infer that history is devoid of regularities. Such regularities take the form of what some theorists call “law-like statements,” statements that resemble laws, but are not “covering laws,” because they do not always apply. In his article “Truisms as Grounds for Historical Explanations,” Michael Scriven (1959) has thus defended the thesis that perfectly valid historical explanations are based on commonplaces: statements that say nothing new but something true, like “Power corrupts,” “Proportional representation tends to give minorities excessive power,” and “Other things being equal, a greater number of troops is an advantage in a battle” (p. 465). These truisms, in historiography, are sometimes explicit, sometimes implicit. If most of us in this room understand the sentence “Louis XIV raised taxes and became unpopular,” it’s because we share with the historian of 17^th -century France the knowledge of the truism “people do not like tax increases,” and we assume that people in the 17^th century were not different from us. Scriven calls statements of this type “guarded generalizations,” and he notes that they often come with an adverb that modifies them, like “typically,” “usually,” “naturally,” or “probably” (p. 465).

5 Debates about the possible existence and the role of “laws” in historiography is no longer topical in English-speaking countries. Discussions about the nature and function that narrative may have in this discipline have of course continued, though within the framework of what has been called “narrativism”: the assumption that historians, when they organize their data, always give them a narrative structure. This assumption has been popularized by Hayden White (1973), who in Metahistory has argued that most historiographic texts fall under the four modes of “emplotment” described by Northrop Frye (1957) in Anatomy of Criticism: tragedy, comedy, romance, and satire. To be sure, many historians, philosophers, and literary theorists disagree with White’s pronouncement that there is no difference between history and fiction. Still, the view that historiography is basically a narrative genre is widely shared, even celebrated. At the January 2013 meeting of the American Historical Association, William Cronon titled his presidential address “Storytelling,” praising this activity as the most ancient and most essential of historical tasks. Similarly, the British theorist of historiography Alun Munslow (2007), a self-professed postmodern, has titled one of his recent studies Narrative and History, stating from page one in his introduction that his objective was to describe the “goals,” “procedures,” and “compositional techniques” that historians follow in order to turn “the past” into “that narrative about it” that we call “history.”

6 The historians and philosophers who still dispute the value of narrative as a mode of knowledge no longer pit that mode against the laws of physics; their objections bear on the coherence of historiographic narratives, a coherence that they hold to be repressive. This position is represented in the USA by Sande Cohen, in the UK by Keith Jenkins. In History out of Joint and other essays, Cohen (2006) has excoriated what he holds to be the artificial homogeneity that historiographic narratives confer upon their data. Such narratives, according to him, render “continuity out of discontinuity,” thus concealing the “cognitive dissonance” between the different moments of the past, as well as between the past and the present (pp. 246-47). Extending Cohen’s argument, Jenkins (2009) has added that the order that historians impose upon the past has ideological implication: it legitimizes “present interests,” obliterates injustice, and prevents any kind of social change (p. 283). At some point, Jenkins had advocated the development of a postmodern history that would be politically positioned on the left, and highly self-reflexive theoretically and methodologically. Because such a history never materialized, Jenkins now proclaims that he can live without history, “whether modern or postmodern.” That is, he can “wave history goodbye and look forward to a future unburdened by the historical past” (p. 17).

7 I must mention, to conclude this overview of the Anglo-American debate about the cognitive value of narrative in historiography, that some philosophers have questioned its very relevance. In Historical Knowing, for example, Leon Goldstein (1976) has explicitly attacked the “narrativist thesis” (title of his fifth chapter), insisting that the philosophers who take up the subject of history must focus on the discipline’s infrastructures, not on its superstructures. History, for him, is a way of knowing, a technical field with its own methods, not a mode of discourse. Avezier Tucker (2004) has argued along the same lines in Our Knowledgeof the Past. He explains in his introduction that he will pay “little attention to the superstructure of historiography,” and even “less attention to the debate [about] whether it has the structure of a narrative or not” (p. 7). Indeed, this debate is for him irrelevant: the real problem lies not in the “forms of historiographic explanations,” but in the “relations between historiography and evidence” (p. 8). Of course, historians interested in the epistemology of their discipline, such as Allan Megill (2007), had already contended that “simply telling a story” was not enough: historians also have to provide evidence, interpret their material, and explain why the story they are telling is better than othe r stories (pp. 96-98). Yet Tucker (2004) is more extreme; the way historians organize their data for him is peripheral, as philosophy of science has shifted its focus to “issues of validation: to asking “whether scientific theories are well founded and justified, and how they change” (p.8). I made room for Goldstein’s and Tucker’s theories because “we,” by which I mean “we at this conference who are interested in superstructures,” must be reminded that our interest is not universal: in some intellectual communities, it is thought to be misplaced, and even beside the point.

8 While controversies about the relations between narrative and historiography involved mostly philosophers in English-speaking countries, in France they first implicated trade historians. Starting in the 1930s, scholars who were to become members of the Annales School had attacked what they called with condescension histoire-récit (“narrative history)” and histoire événementielle (“event history”), studies published in the late 19^th and early 20^th centuries by historians whose focus was mainly the political, military, and diplomatic past of nation-states. Lucien Febvre had derided this type of research in his reviews, Fernand Braudel (1966) had dismissed it in the celebrated preface to his study of the Mediterranean, and François Furet (1982), in a programmatic article first published in 1975, had celebrated the shift from “narrative history” to what he called histoire-problème (“problem-history”), to a history “scientifically conducted,” whose purpose was to “pose problems” and “formulate hypotheses.” Furet did not specify which form this history was supposed to take, when its structure was no longer that of a narrative.

9 From the 1930s to the early 1980s, the anti-narrativist position of the Annales dominated the discussions that French historians had about the relations between their discipline and storytelling. Besides Febvre, Braudel, and Furet, renowned scholars such as Jacques Le Goff (1981) and Emmanuel Le Roy Ladurie (1975) also condemned narrative, both in theoretical essays and in the reviews that they devoted to the colleagues whose work did not conform to the Annales’ standards. Obviously, not all French historians belonged to the Annales School, nor to that School’s successor in the 1970s, la Nouvelle Histoire (the New History). But if they wrote studies that did not conform to the Annales’ way, they did not theorize their position. A notable exception, in 1971, was Paul Veyne’s Comment on écrit l’histoire: Essai d’épistémologie [Writing History: Essay in Epistemology]. Using provocative language, Veyne had the boldness to state that history, I quote, was “nothing but a true novel” (1971/1984, p. 10), “nothing but a truthful story” (p. 13). Relying on some of the British and American philosophers I discussed earlier, Veyne also stated that explanations, in history, were not provided by laws, but by the intrigues compréhensives (“inclusive plots”) through which historians textualized their data. Veyne’s intervention was so unexpected that the editors of the journal Annales had to farm out the review of the book to the philosopher Raymond Aron (1971), best known for his work on German epistemology of the social sciences. Aron, for that matter, was about to give at the Collège de France a course in which he introduced the theses of Anglo-American analytical philosophy of history, but neither this course, nor Veyne’s essay, had much influence at the time. Revealingly, the encyclopedia La Nouvelle Histoire, published in 1978, did not have an entry for récit (“narrative”), and entries like the one devoted to événement (“event”) only restated the Annales’ party line.

10 Things in France hardly changed before the early 1980s and the publication of Paul Ricœur’s (1983) Temps et récit [Time and Narrative]. While historians had not been impressed by Veyne’s Writing History and had largely ignored Aron’s work, they in contrast read Ricœur and were deeply influenced by his theses. Ricœur—his positions are well-known, so I summarize them very quickly—argues that some of the productions of the Annales School that are supposedly non-narrative fall in fact under storytelling. It’s the case, for example, with Braudel’s study of the Mediterranean, in which Ricœur identifies what he calls a “quasi-plot” (p. 298) or a “virtual plot” (p. 301): that of the Mediterranean’s decline, of the sea’s “withdrawal from major history” (p. 303). But all historiographic studies, according to Ricœur, eventually fall under the narrative genre: they necessarily involve a “plot,” by which Ricœur means a synthèse de l’hétérogène (“synthesis of heterogeneous elements”), that is, a synthesis that combines “goals, causes, and accidents” in the “temporal unity of a total and complete action” (p. 11). French historians who had not been convinced by Veyne’s view of history as a “truthful story” have, by contrast, adopted Ricœur’s thesis with surprising unanimity. Roger Chartier, for instance, long associated with the New History, has now become a strong narrativist. In the entry “Narrative and History” he wrote in 2006 for the Dictionary of the HumanSciences, he thus speaks of the “unanimous view that holds history as a narrative” (p. 969), of the “acknowledgment that history is narrative” (p. 970), and of the “membership, long ignored, of history in the category ‘narrative’” (p. 970). “Unanimous view,” “acknowledgment,” “membership”: these terms show that for Chartier, the problem is solved; history, whatever its practitioners might have stated at a certain point, inescapably belongs to storytelling. I don’t want to multiply examples, but several French historians interested in the theory of their discipline, for instance, François Hartog in his 1995 article “The Art of Historical Narrative,” have argued along the same lines: insisting that historians have indicted narrative for the wrong reasons. They now maintain that it is an integral part of the historiographic endeavor. Revealingly, récit now has an entry in encyclopedias of history, for instance in the 2010 Historiographies (Dosse, 2010b), where événement is also rehabilitated in a 13 page-long entry (Dosse, 2010a).

11 Can literary theory contribute to these debates? It seems to me that it can, provided that the conversation can be shifted from a prescriptive to a descriptive plane. While analytical philosophers and Annales historians raised the question, “Does narrative provide a legitimate kind of knowledge?” literary theorists will simply ask: “Do historians rely on narrative? And if they do, on what kind of narrative?” Answering these questions, of course, involves defining what is meant by “narrative,” something which philosophers and historians, who seem to take the term for granted, often fail to do. With Gerald Prince (2012), I will say that “An object is a narrative if it is taken to be the logically consistent representation of at least two asynchronous events, or of a state and an event, that do not presuppose or imply each other” (p. 25); and with James Phelan (2007), I will define narrative as “Somebody telling somebody else on some occasion and for some purpose(s) that something happened” (p. 203). Whether they treat narrative as an object or as a transaction, these definitions state basically the same thing: a text, in order to count as a narrative, must include at least two units located on a temporal axis, even if the first of these units may remain implicit. To take an example in this week’s commemorations: The mini-text “Franz Ferdinand was archduke of Austria-Este” is not a narrative, because it does not involve the telling of an event. But the mini-text “The archduke Franz Ferdinand was assassinated on June 28, 1914” is a narrative, because it represents a change with respect to a state, and could be parsed into “there was an archduke Franz Ferdinand” and “this archduke was assassinated on June 28, 1914.” If we use Prince’s and Phelan’s definitions to answer the question: “do contemporary French historians rely on narrative?” we can’t help but note that a large part of their production does not fall under storytelling. Some of the studies they have published develop a plot, but others do not, resulting in two main categories of textual organization.

12 The first one of these categories is the synchronic cross-section: the study that does not trace a change, but examines “what things were like” at a certain place, at a certain time. Synchronic cross-sections may take the form of the tableau, namely, of the comprehensive account of the political, social, and economic structures of a specific area during a specific period. The tableau was frequently used in the 1950s and 1960s, as it was a type of textual disposition that dissertation directors liked to impose upon their students. Several classics of French historiography belong to this subgenre, notably Pierre Goubert’s (1960) Beauvais andthe Beauvaisis from 1600 to 1730, Pierre Vilar’s (1962) Catalonia in Modern Spain, and Pierre Chaunu’s (1956-1960) Sevilla and the Atlantic between 1504 and 1560. Dissertations at the time were thèses d’état that came with unwritten length and completeness requirements, requirements that produced overdrawn, reader unfriendly studies. Chaunu’s work on Sevilla and the Atlantic, for instance, includes 8 volumes and 7343 pages, a record that is probably not about to be broken. In the 1970s, tableaux gave way to the more manageable format of the anthropological description. Whereas tableaux deal mostly with the political, social, and economic aspects of a community, anthropological descriptions are more concerned with cultural phenomena. They are also organized differently: their authors proceed from the outside to the inside, as if they were conducting a field study. The prototype of this subgenre is Le Roy Ladurie’s (1975) celebrated Montaillou, the study of a village in Southern France in the late 13^th -early 14^th centuries. As an anthropologist like Evans-Pritchard (1940) does, Le Roy Ladurie treats first what he calls the “ecology” of Montaillou: the physical environment, housing, and work; he then moves on to the village’s “archeology”: gestures, marriage, sexual life, as well as attitudes toward death, morality, and religion. Of course, both tableaux and anthropological descriptions include several short narratives. But those function mostly as examples; they are not episodes in a developing plot. Viewed in their overall organization, neither tableaux nor anthropological descriptions involve a plot, accounting for a transformation from point A to point B along a time sequence; thus, they cannot be regarded as narratives.

13 A second type of historiographic synchronic cross-section is the analysis. By “analysis,” I refer to the studies that focus on a theme or a problem, dealing not with the changes that theme or problem may have undergone, but with its various aspects at a specific time and place. Analyses can bear on many different domains. Military history, for example, is not limited to the report in narrative form of what happened on the battlefield. Annette Becker (1998), in The Forgotten of the Great War, does not only leave the fighting aside; proceeding analytically, she first identifies the group, “the forgotten,” then goes on to describe the communities that comprised it, the institutions that sought to help it, and the limits inherent in any humanitarian undertaking. In the totally different area of “connected history,” Romain Bertrand (2011) moves similarly in his recent History in Equal Shares. Focusing on the encounter between the Dutch and the Javanese in the 16^th and 17^th centuries, he asks, alternating standpoints, how the two cultures weighed merchandise, how they paid for it, how they measured distances, and what their ethics of commerce was like. Another field of analysis is metahistory: the histories which are about other histories: that is, which neither investigate new subjects, nor revisit already treated subjects on the basis of new evidence, but discuss prior studies. A recent example is Pierre Laborie’s The Sorrow and the Venom, published in 2011. As its title indicates, this book deals with The Sorrow and the Pity, Ophuls and Harris’s well-known documentary film about the period of the German occupation in France. Yet Laborie’s purpose is not to recount the reception of this film over the years. It is to discuss its main theses, as well as the theses of other historians of the Occupation, like Henry Rousso (1990), Philippe Burrin, and Robert Paxton. The structure of the book is thus not chronological, but analytical and rhetorical: Laborie surveys what he takes to be debatable versions of the Occupation, offering each time his own, supposedly better version. Like tableaux and anthropological descriptions, analyses may include brief stories. But those stories, again, do not add up to produce a well-formed narrative. They are used mostly to bolster a point, supplying examples of what is asserted in the analysis. The second large category of textual organization in current French historiography is the diachronic development. It can be divided into two subcategories.

14 The first one is the linear narrative: the narrative that proceeds from sequence to sequence on a temporal axis, the order of events in the discourse agreeing by and large with the order of events in the past, as documents have made it possible to reconstruct it. Condemned by the Annales, this type of textual organization had nevertheless survived and produced several important works from the 1940s to the 1980s: Pierre Renouvin’s multi-volume History of International Relations (1953-1972), Jean-Baptiste Duroselle’s studies of French politics, The Decadence (1979) and The Abyss (1983), as well as scholarly biographies like Jean-Paul Brunet’s Jacques Doriot: from Communism to Fascism (1986). Contrary to the Annales’ charges, these studies are not mere chronicles listing event after event; they are carefully emplotted. Titles like “Decadence” and subtitles like “From Communism to Fascism,” in this respect, are particularly revealing: they immediately point to the script, to the narrative scheme, which the text will then follow.

15 I did not mention as belonging to the category “linear narrative” the many studies that now focus on one specific event, because they generally do not fall under “narrative.” There is on this subject a misunderstanding that probably originates in the fact that members of the Annales School, in their critique of prior historiography, used the terms “narrative history” and “event history” as synonyms. Yet, events do not have to be represented in narrative form; they can also be described, or submitted to quantitative analyses. The French historians who have devoted studies to one event have thus generally steered away from storytelling. The prototype in this area, Georges Duby’s (1973) The Legend of Bouvines, only allocates 50 pages out of 300 to the battle itself; the rest of the book consists of a series of analyses, in which the historian accounts for entities like “peace,” “battle,” and “victory” as aspects of medieval culture. The same remark applies to other studies that focus on one event, like Olivier Chaline’s The Battle of the WhiteMountain (2000), Raphaëlle Branche’s The Ambush in Palestro (2010), and even The Assassination of Henri IV, by such a conservative historian as Roland Mousnier (1964). To be sure, these texts recount the battle, the ambush, and the assassination mentioned in their titles, but they do so briefly. Their authors are mostly interested in the issues, the conflicts, and the attitudes that are revealed by an event they have selected for its representativity rather than for its singularity.

16 The second subcategory of diachronically organized historiographic studies is stage narrative. Unlike linear narratives, stage narratives do not proceed from event to event, but from state to state, or from situation to situation. To return to terms I used earlier, one could say that they are made of a succession of tableaux, descriptions, or analyses, where the historian examines not a single moment but several consecutive moments in the evolution of a community, an institution, or a belief. Just like the other subgenres, stage narratives are found in most areas of historical research. It is one of the preferred modes of organization in cultural history, where it serves to show how practices or attitudes have evolved over time. The prototypes here are Philippe Ariès (1977) work on changing attitudes toward death, The Hour of OurDeath, as well as Georges Duby’s and Jacques Le Goff’s studies of the Middle Ages, such as The Age ofthe Cathedrals (Duby, 1976) and Birth of Purgatory (Le Goff, 1981). Stage narratives are also suited to more contemporary fields, like memory and gender. In The Vichy Syndrome, for instance, Henry Rousso (1990) distinguishes different phases in the memory of the Occupation. Similarly, Yvonne Knibielher and Catherine Fouquet (1982) show in A History ofMothers how attitudes toward motherhood have changed from the Middle Ages to the 20^th century. As for Christine Bard (2010), she traces, in her A Political History of Trousers, the steps of the progressive acceptance of this garment as part of women’s wardrobes.

17 While French historians resort at times to linear and stage narratives, the fact remains that a large segment of their production does not fall under narrative, at least not as this term is defined in literary theory. It is thus reasonable to ask why historians interested in the epistemology of their discipline, like Veyne, Chartier, and Hartog, should claim that historiography—to use Veyne’s phrase—is made of “nothing but truthful stories”: a claim that is all the more paradoxical since Veyne’s (1976) work on Rome, Chartier’s (1987) on reading, and Hartog’s (1991) on Greece belong to analysis, not to narrative. I see two possible explanations for this incongruity. The first one is that these scholars buy into Ricœur’s (1983) argument that histories, even when they do not explicitly tell stories, nevertheless are part of a virtual or underlying narrative. Goubert’s (1960) tableau of Beauvaisis between 1600 and 1730 could thus be viewed as a moment in the history of this province—a moment to which earlier and later tableaux of Beauvaisis during the Renaissance and in the 18^th century could potentially be added. Likewise, Bertrand’s (2011) analysis of the encounter between the Dutch and the Javanese in the 16^th and 17th centuries could be taken as an episode in a larger story, that of the relations between East and West during the period of colonial expansion. On this point, Veyne, Chartier, and Hartog would join with the British and American narrativists who hold that histories always come as stories: for instance with Hayden White (1973), for whom Jacob Burckardt’s The Civilization of the Renaissance in Italy, obviously a tableau, is in fact a narrative, though one which is “all middle” (p. 118).

18 Another way of accounting for the classification of all historiography under the category “narrative” is to see in this assignment a consequence of the “narrative turn.” In France, as in English-speaking countries, scholars in such areas as sociology, anthropology, law, and medicine have lent increasing attention to the fact that their inquiries often come in the mode of narratives. Yet, as Martin Kreiswirth (2005) has noticed in the entry he devotes to the “Narrative Turn in the Humanities” in the Routledge Encyclopedia of NarrativeTheory, these researchers “have come to the task with a set of instruments, texts, thinkers, presuppositions, and goals” entirely different from those of the narratologists. Kreiswirth gives as example the bibliography that follows the entry “Narrative” in Alun Munslow’s (2000) The Routledge Companion to Historical Studies: of the 43 authors listed there, only two (Chatman and Genette) are narratologists; the other scholars mentioned are philosophers and historians, and ones who are rarely included by literary theorists: Alex Callinicos, Frank Ankersmit, William Gallie, Robert Rosenstone, M.C. Lemon, Peter Munz, J.E. Toews, and Peter Zagorin, to name just a few. A consequence of what Kreiswirth calls narrative’s “migration” and resulting “ubiquity” has been a widening of the word’s definition from the definitions provided by narratologists like Prince and Phelan. As Prince (2012) himself has noticed, “narrative” may now be substituted for such terms as “explanation,” “argumentation,” “hypothesis,” “ideology,” “art,” and “message” (p. 23). Kreiswirth’s and Prince’s pieces bear on the narrative turn in English-speaking countries, but a similar phenomenon has taken place in France. Some of the contributions to a seminar held in the late 1990s at the Maison des sciences de l’homme, whose proceedings were published in 2001 as Models and Narratives (Grenier, Grignon, & Menger, 2001), show that the semantic extension of the term in France has been even more radical: their authors argue that “narrative,” in the human sciences, now refers to the texts, or parts of texts, which use natural language and natural logic. “Narrative” would thus contrast not with description and analysis, as I suggested earlier, but with “model”: with the texts, or parts of texts, which use mathematical language and formal logic. Although Chartier, Hartog, and the French historians who now endorse narrativism do not mention this seminar, they may implicitly agree with the all-embracing definition of narrative as “any text or part of text using natural language and natural logic” that was offered there. One may of course find that definition too capacious, or rather too powerful: it generates so many “narratives” that the term becomes “trivial.” I’m borrowing this latter adjective from Thomas Pavel (1986), who in Fictional Worlds (p. 5) had argued the same point, though in a different context, in his discussion of Greimas and the Paris School of Semiotics, who in the 1970s—that is, much before the narrative turn—had already offered the view that narrative is “the organizing principle in any discourse.”

19 To conclude, I would like to review quickly some of the issues raised by the “dispositions” I have described, so not by the facts that historians report, but by the ways they organize those facts in their studies. I will ask three questions, which are all related to the view of narrative as a “cognitive instrument,” as it was developed by Mink and theorists in his legacy.

20 The first one could be formulated as follows: are the dispositions that historians deploy found or constructed? In other words, do such dispositions originate in the data, or are they imposed upon the data by the historian? Philosophers of history have often posed this question, and their answers pitch “realist” against “constructivist” theses. The most eloquent representative of realism is David Carr. In his book Time, Narrative, and History (1986) as well as in several articles, Carr has defended the idea that since human actions unfold in time, they have a narrative structure that precedes the story the historian may tell and is independent from it: “Storytelling obeys rules that are imbedded in action itself, and narrative is at the root of human reality long before it gets explicitly told about” (2008, p. 29). For Carr, the historian’s ambition to “represent” is thus perfectly legitimate; far from differing by its form from the actual world, narrative can be regarded as homologous to the reality that it describes. The historian’s goal, according to Carr (2001), is eventually to “get the story straight”: to identify the story, and to tell it as it actually unfolded. One must point to the use of definite article “the”: it implies that there’s one story, one valid version of the past, which the historian has the task to identify on the basis of the available evidence.

21 The constructivist position is of course represented by Hayden White, but it had already been defended by Louis Mink in the articles I mentioned while discussing the issue of “laws” in history. According to Mink (1987), “stories are told, not lived” (p. 60). To put it otherwise, our experience of the world cannot be equated with a narrative: our lives do not take on a narrative form before we make them into the subject of a story. Similarly, the idea that the past is an “untold story,” a story that has not yet been recounted, is for Mink an “indemonstrable assumption” (p. 188). The job of the historian is to construct a story using the available data, not to uncover the story that lies hidden in those data. The Dutch philosopher Frank Ankersmit (1994) has argued along the same lines in his discussion of the concept of “time.” According to him, it is wrong to presume—as Carr does—that human actions unfold in a time that precedes the historian’s narrative and is independent of it. “Historical time,” for Ankersmit, “is a relatively recent and highly artificial invention of Western civilization. Thus, “temporal determinations” are not “expressed by statements”; they are expressed “in statements,” that is, in categories—the Renaissance, midnight, June 27^th —which we haven’t found, but have invented (p. 238). Although Richard Rorty, to my knowledge, has not contributed to this specific debate, I think that the constructivist position is best illustrated by his famous statement: “The world is out there, but descriptions of the world are not” (1989, p. 5). To put it otherwise, there’s something we call “time,” but that something is not out there; it is a description that we have devised, because it helps us organize the world and make sense of it.

22 The second question I’d like to ask proceeds from the first one: If historians construct their accounts of the past, are they free to go about constructing as they please? In other words, are they free to organize as they see fit the materials that they have gathered? Hayden White (1978), in his article “The Historical Text as Literary Artifact,” had stated somewhat carelessly that specific events do not call for specific modes of emplotment. “For example,” he writes,

Historians, according to White, are thus free to emplot their data as they wish, according to the meaning that they intend to lend to them: there is nothing in the data that constrains them to adopt one mode rather than the other.

23 White’s position has of course been challenged, notably at a colloquium held in 1990 at the University of California at Los Angeles, whose proceedings have been edited by Saul Friedlander (1992) as Probing the Limitsof Representation: Nazism and the “Final Solution.” One of the questions at this colloquium was to determine whether the Holocaust could be emplotted in any of the modes described by White, especially as a comedy, in the sense the term has in literary theory: that of a story with a happy ending. White was at the conference, and he sought to defend his thesis with a contribution titled “Historical Emplotment and the Problem of Truth” (1992/2001). Emplotting the Holocaust as a comedy, according to White, was indeed possible, and it is the mode of emplotment some German historians would probably have selected if the Nazis had won the war. As for representing the Holocaust as a comedy in today’s context, White maintains that this mode of emplotment would only be unacceptable if it was offered as “found,” not as “constructed,” as “inherent in the facts,” not “imposed upon them” (p. 377). White adds that no ethical or aesthetic requirement obliges one to represent the content “Holocaust” in noble forms, such as the tragedy or the epic. He raises the example of Art Spiegelman’s Maus: A Survivor’s Tale, a work that recounts some of the events of the Holocaust in the form of a graphic novel, Germans being represented as cats, Jews as mice, and Poles as pigs. Spiegelman, according to White, cannot be charged with trivializing his subject, because the form he has selected is highly stylized and allegorized. By representing momentous events in the form of a graphic novel, by associating a “low” genre with a “high” subject, Spiegelman for White manages to raise “all the crucial issues,” regarding not just the limits of representation of the Holocaust, but the limits of representation generally speaking (p. 378).

24 I’m now coming to my third and last question, which follows directly from the previous one: if historians are free to arrange their data as they see fit, are some types of organization inadmissible not from an ethical but from an epistemological standpoint? As White has shown, I think convincingly, the same events, or sets of events, can be represented in different ways. To take just one example—and return to commemorations: In the 10^th and last volume of the History of Contemporary France from the Revolution to the Peace of1919, published in 1922, World War I is emplotted as what White would call a comedy: France has won the war, Alsace and Lorraine have been reclaimed, and Ernest Lavisse (1922), the general editor of the series, can conclude optimistically with a comment titled: “Reasons to believe in the future.” But Marc Ferro (1969), who writes 40 years later, ends his own study of The Great War on a different note. In a chapter titled “The Illusions of Victory,” he explains that if France won the war on the battlefield, she lost it on the economic and demographic planes. Taking into account that Lavisse’s work was published in 1922 and Ferro’s in 1969, both plots seem to be equally acceptable from an epistemological standpoint. Indeed, Lavisse and his team did not know what Ferro knows, namely, that the physical devastations and the losses in human lives brought about by the war were to have dramatic consequences in the 1920s and 1930s: they caused France’s economy to stagnate and her population, in certain years, to have a negative growth rate.

25 That being said, and to return to my question, are there plots that can be viewed as wrong from an epistemological standpoint? The philosophers who have taken up this question, for example Mink, Tucker, and Ankersmit, propose to distinguish between the discrete statements that make up a study and that study regarded as a whole; or, as Donald Polkinghorne (1988) puts it in Narrative Knowing and the Human Sciences, between the “information contained in the sentences” and the “information generated by the specific type of coherence used to order the sentences into a discourse” (p. 61). According to these theorists, individual verifications can be made at the level of the statement or the sentence. For example, the statement, “The archduke Franz Ferdinand was assassinated on June 28^th 1914” can be verified in the archives. It can thus be regarded as “valid,” an adjective that several historians and theorists now regard as preferable to “true,” because it is less loaded with past philosophical arguments. But if the archduke was assassinated on June 28^th , does this event mark the beginning of World War I? Whether we answer “yes” or “no,” we are no longer in the area of individual verifications, but in that of emplotment. According to the entry “World War I” in the French-language version of Wikipedia, this assassination was the “spark that caused the war.” But according to the English- language version of that same site, the war began “on July 28^th ,” the day when Austria-Hungary declared war on Serbia and its soldiers fired the first shots. Is one version “better” than the other? And by what criteria should we decide, given the fact that Austria-Hungary’s declaration of war on Serbia on July 28^th was followed by several other declarations of war, respectively of Germany on Russia on August 1^st , of Germany on France on August 3^rd , and of England on Germany on August 4^th ? Literary theory is here of little help. At the most, it makes it possible to distinguish, in studies of the war, between the beginning of the discourse (the first lines of the printed text) and the beginning of the story (the moment when the events that the author is reporting are supposed to have started). But literary theory does not give us the means to determine when World War I “actually began.” To make a decision, we need both a definition of “beginning” and a theory of “war,” as we needed a definition of “victory” to decide between Lavisse’s and Ferro’s versions of the end of World War I. Discussing a modern war, we are thus back to the questions that Duby (1976) asked about conflicts in the Middle Ages: what does it mean, in a specific culture and during a specific period, to be “at peace,” to be “at war,” to “give battle”, and to “win a victory”?

26 Issues of this kind were discussed with particular vehemence in Germany during the 1980s, as part of the Historikerstreit: the debates that German historians and philosophers conducted in the press about the need to reassess the interpretation of the Holocaust and the legacy of the Third Reich. These debates bore, among other things, on cases of questionable emplotment, two of which I will briefly consider. The first one is part of the historian Ernst Nolte’s (1980/1993) article “Between Historical Legend and Revisionism.” In this article, Nolte argued that Auschwitz was “not primarily a result of traditional anti-semitism, and was not just one more case of genocide. It was the fear-borne reaction to the acts of annihilation that took place during the Russian Revolution.” In other words, the “so-called annihilation of the Jews by the Third Reich was a reaction, or a distorted copy, and not a first act or an original” (p. 14). Discussing Nolte’s plot in his article “The Impoverished Practice of Insinuation” (1986), another historian, Eberhard Jäckel (1986/1993), dismissed it on epistemological grounds. According to him, there was no evidence that “Hitler’s decision to kill the Jews was driven by fear,” specifically, that the Nazis “considered themselves to be potential victims of what Nolte calls an ‘Asiatic deed’” (p. 78). While the Nazis presented the attack against the USSR as a preventive war, they never, according to Jäckel, claimed that killing the Jews was a “preventive murder.”

27 Similar controversies surrounded Andreas Hillgruber’s (1986) book Twofold Fall: The Destruction of the GermanReich and the End of the European Jewry. As its title indicates, Hillgruber’s book is divided into two parts; the first one, “The Collapse in the East as Problem of German National History and European History,” treats the fighting on the Eastern Front in 1944-45, as the German army retreated from the territories it had occupied and sought to protect the homeland, helping populations in such areas as Prussia, Silesia, and Pomerania to flee before the arrival of the Russians. The second part, “The Historical Locus of the Destruction of the Jews” examines the genocide of the European Jews, in relation to other destructions. Commenting on the Historikerstreit in the article I mentioned earlier, “Historical Emplotment and the Problem of Truth,” Hayden White (1992/2001) explains that the problem in Twofold Fall does not reside in the events that Hillgruber reports, but in the way he emplots them, in the way he makes them into a tragedy, that is, into a genre in which “even villains are noble, or, rather, villainy can be shown to have its noble incarnation” (p. 379). Hillgruber’s decision, according to White, shows how the choice of a plot type can determine the kind of events and agents to be featured in the story, as well as, conversely, the kind of events and agents to be excluded from it. In this instance, the choice of the mode “tragedy” enables Hillgruber to include the deeds of the German army protecting civilian populations, while ignoring the atrocities committed by that same army, in the USSR and elsewhere. Similarly, White dismisses Hillgruber’s view of the Holocaust as an “incomprehensible enigma” (p. 378). The Holocaust, for White, can very well be represented, the difficulty being to determine which one of the available plot lines might best suit such a problematic subject.

28 One brief and I hope uplifting remark to conclude. I mentioned earlier that for philosophers like Tucker and Goldstein, the debates about the membership of historiography in narrative are largely irrelevant; history, according to them, is a way of knowing, not a mode of discourse, and the real issues concern the relations between historiography and evidence. What the controversies surrounding the Historikerstreit show is that questions of evidence, for historians and philosophers, are not the only ones worth considering; problems of textual organization also deserve to be scrutinized, as modes of discourse imply ways of knowing. To put it another way, the polemics between Nolte, Hillgruber, and their critics have legimitized the interest that we at this conference show for storytelling; in a word, they have made clear that “narrative matters.”