Wesleyan University

On the Nature and Role of Narrative in Historiography Author(s): W. H. Dray Reviewed work(s): Source: History and Theory, Vol. 10, No. 2 (1971), pp. 153-171 Published by: Blackwell Publishing for Wesleyan University Stable URL: [Link] . Accessed: 21/02/2012 13:19
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . [Link] JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@[Link].

Blackwell Publishing and Wesleyan University are collaborating with JSTOR to digitize, preserve and extend access to History and Theory.

[Link]

ON THE NATURE AND ROLE OF NARRATIVE IN HISTORIOGRAPHY

W. H. DRAY

The question of the extent to which an examination of narration is central to the analysis of history as a form of inquiry or a type of knowledge has received a good deal of attention from philosophers of late. It has been held that history simply is narrative; or that it is essentially narrative; or that a history must contain some narrative elements; or that one form of history, at any rate, and perhaps the most important one, narrates. It has been held, too, that it is through narration that historians achieve whatever is specifically historical about historical understanding; or that historical explanations get their distinctive structure by reason of their occurring in the course of historical narratives. It has even been held that narratives can themselves be explanatory in a special way; or that narrative is per se a form of explanation, if not indeed self-explanatory. Claims such as these have added interest to recent attempts to clarify the "logic" of historical narrative - although, as Mink has remarked, there has been a certain amount of skepticism, too, as to whether narrative is the sort of thing which could very appropriately be said to have a "logic" of its own.' Whether they talked the language of "logic" and "structure" or not, those philosophers who have emphasized the centrality of narration in history have not failed to generate opposition to their claims. Thus, it has been argued against them that narratives, by their very nature, preclude satisfactory explanation of the events which occur in them, and that they necessarily oversimplify the past in characteristic ways.2 It has also been held that attention to the narrative aspect of historical works diverts the philosopher's attention from those features of history which make it a form of inquiry or knowledge, rather than a form of art, or even of entertainment. The narrative, it has
A version of this essay was presented at the Philosophy of History Symposium held at the University of Kentucky on April 10 and 11, 1970. 1. "History and Fiction as Modes of Comprehension,"New Literary History I (196970), 541-558. 2. See for example M. Mandelbaum, "A Note on History as Narrative," History and Theory 6 (1967), 416-417. 3. Mandelbaum, ibid., 414.

154

W.

H. DRAY

been claimed, is just a way in which historiansoften "writeup" what they have discoveredin the course of historicalinquiryproper. In this articleI shall try to thread my way throughsome of the chief issues, arguments, and positionscenteringupon the questionof the relationof historiography [Link] own general positionwill be that the narrato tivists- amongwhom Gallie, White, and Danto have been especiallypromto inent- have made fresh and illuminating contributions the contemporary My quarrelswith them will be philosophicaldiscussion of historiography. chiefly on points of detail. I regret that in trying to make this clear, I shall sometimesseem to be going over groundwhich Mink, especially,has already traversed.I regretit the more because he has done it with an elan which I cannothope to match,and with a thesis I have no greatinclinationto attack.

Let me look first,althoughnot at greatlength,at the questionof the necessary connection,if any, between the very ideas of history and of [Link], at times, such a necessaryconnectiondoes seem to be assertedby some [Link] Gallie, for example, historyis "a species of the genus is Story";and becauseof this, the idea of historicalnarrative "logicallyprior" to almost all other questionsof critical philosophyof history.4Accordingto narrative,the story form providingthe hisDanto, all history "presupposes" scheme"just as theory provides one for the scitorian with an "organizing entist.5Morton White is more cautious, saying only that narrativeis "the typical form of discourseemployed by the historian." But Glenn Morrow, commentingfavorably on White's views, declares roundly that "history is and narration";7 A. R. Louch has placedhimselfsimilarlyamongthe extreme with the claim that narrative narrativists techniquesare "essentialto the business of historical explanation,"not just "an incidental, stylistic feature."8 Even Maurice Mandelbaum,who has recently expressed some of the most serious misgivingsabout philosophicalaccounts of history as narrative,has been shown by an opponentto have earlierimplicitlyequated historicaldescriptionwith historicalnarration.9 as If we interpretthe claims of the extreme narrativists implyingthat no
4. Philosophy and the Historical Understanding(New York, 1964), 66. 5. Analytical Philosophy of History (Cambridge, 1965), 142, 137. 6. Foundations of Historical Knowledge (New York, 1965), 4. 7. "Commentson White's 'Logic of Historical Narration'" in Philosophy and History, ed. Sidney Hook (New York, 1963), 286. 8. "History as Narrative,"History and Theory 8 (1969), 54. 9. R. Gruner, "Mandelbaumon Historical Narrative," History and Theory 8 (1969), 287 n9.

NARRATIVE IN HISTOROGRAPHY

155

work which fails to have the overall organization a narrativeis a work of of history,then their positionis [Link] Rolf Grunerhas reminded readersof History and Theory, the "cross-section" "portraitof an age" or varietyof historiography a perfectlyrespectableone.10To use his own exis ample,it would be strangeto have to say that Huizinga'sThe Waningof the Middle Ages is not a work of history because it lacks an overall narrative form of organization. For it is clearly not a work of social science: it is not intendedto exemplifyor test any [Link], course, often of advancea somewhatmore limited and seeminglymore plausiblethesis. They insist only that narrativehistories are the principal,essential, or more importantof historicalworks. Thus Gallie sometimessays only that the "key members"of that "familyof researches" called history are narratives."But on what groundscould it be said that the non-narrative works are merely "ancillary"? Gallie himselfoffersnone. I can think of only two likely reasons for assertingGallie's conclusion, neither of them [Link] first is the argumentthat non-narrative histories "presuppose" narrativeones. It might be claimed, for example, that the very notion of the Middle Ages is meaninglesswithoutsome background conceptionof a transitionfrom the ancientto the modernworld- the kind of thing it would requirea narrativeto [Link] as a "still"might be said to have no meaningas a still withoutreferenceto a movingpicture,so a cross-sectional slice of historicaldevelopmentmight be said to requirereference to the development from which it was abstracted. a similarargument But could surelybe elaboratedfor the converseconclusion:that narrativehistory the sort. As Mandelbaum pointed out in this has presupposes cross-sectional connection,any attemptto stick to pure narrativehistoryruns into methodologicalproblems.12A narrative historythat never pausesfor a cross-sectional harder to follow; and the need to preface becomes progressively "breather" an historicalnarrative a sketch of the context in which the action develops by I is is also commonlyacknowledged. think that Mandelbaum correctin representingthe needs of continuingnarrativeand of adequatedescriptionand explanationas to some extent pulling in opposite ways in historicalreconstruction.I doubt, therefore,that we shall find good reason in this direction for regardingnarrativehistories as more fundamentalthan cross-sectional ones. conclusionmight be sought in some for A second argument the narrativist familiar claim that all historicalworks must ultimatelybe variationon the And history-in-the-large, to or regardedas contributing history-in-the-large.
10. Ibid., 284. 11. Gallie, 71. 12. Mandelbaum, 417.

156

W. H. DRAY

universal history, it might be argued, necessarily extends through time, so that a narrative treatment of it might be thought to be required. Now I would agree that, in some important sense, every history is offered as a putative contribution to universal history, and comes under judgment as such. As Collingwood put it, the unity of the historical world, the existence of "one history," is a presupposition of historical inquiry13- something which might perhaps be compared to the presupposition of physical scientists that the laws of nature form a single system. But the sense in which this is true, I submit, is not the one required by the narrativist argument. The sense in which all history might constitute one universal history is not the same as the sense in which it might yield one super-story, whether of the heady sort constructed by speculative philosophers like Hegel, or the more mundane but still linear kind traced by world historians like H. G. Wells. It is not a presupposition of every historical work that there exists a final narrative account of historyin-the-large to be discovered. What is required is only that every account, narrative or otherwise, be consistent with all the rest. There are nevertheless two points which might be noted on behalf of the narrativists before leaving the untenable claim that all history, or even every important work of history, narrates. The first is that, although there may be works of history which lack the narrative form, we do not usually call these "histories." And when one looks at the examples typically cited by narrativists, it seems clear enough that it is histories they were primarily interested in. Morton White is quite explicit about this: what he has to say about the "logic of narration," he tells us, is to be taken as applying only to those works which offer the history of some continuing central subject, such as a nation.14 The second and related point is that, even if it is both possible and desirable to write non-narrative works of history, we have to recognize that both narrative histories and historical works which at least contain narrative elements do exist. They are facts of historiography requiring philosophical explication. If the question is seriously raised whether it is at all important for historians to write histories, a narrativist could perhaps point out, for a start, that people commonly ask questions of historians which can only be answered naturally by a narrative - questions like "How did that come about?" or "What consequences followed from that?" Or he could involve himself - as Gallie almost does at some points - in such deeper philosophical issues as the importance for a human sense of identity of a consciousness of the past under narrative forms - especially, although not exclusively, the past which one regards as peculiarly one's own. But these are not the questions I propose to pursue further in this article. I shall proceed on the assumption that the
13. The Idea of History (Oxford, 1946), 246. But see also Gallie, 56-59. 14. White, 221.

NARRATIVE IN HISTORIOGRAPHY

157

construction of narratives is an admissible and prominent, although not universal, aspect of historiography. And I shall ask what historians can hope to achieve through it.

II

One recurring answer to the question what historical narration can achieve is "explanation." And this often seems to mean more than just that historical narrative may contain explanations. Narrative itself, according to Louch, affords a "distinct kind of explanation."''5 When ideally formulated, Gallie declares, a narrative is self-explanatory.1 Danto, too, claims that historical ;17 narrative is itself "a form of explanation" and he apparently intends this in a strong sense which would imply that "non-explanatory narrative" is somehow an incoherent or contradictory notion. This becomes clear in the way he treats a well-known distinction, earlier drawn by W. H. Walsh (following Croce) between "plain" narrative or chronicle, which would tell us no more than what exactly happened in the past, and a "significant" narrative, which would also explain to us why it did.18 Danto attacks this distinction with great ferocity, as expressing a complete misunderstanding of the nature of narrative in history. It may be helpful for our examination of narrativism as a whole if I try to show why I find this position of Danto's unconvincing. Danto's central point is that by a narrative we mean much more than simply a list of disconnected facts set forth in chronological order. In some sense, those facts have to constitute a "story." And by implication, at least, Danto seems to agree with Mandelbaum, White, and others that this requires at least their being about something; the story must have what White called a "central subject."19 The kinds of central subjects which characteristically interest historians are often suggested by the titles of their books: they are such things as the developments and declines of social movements and institutions, reigns, wars, revolutions, climates of opinion, and the like. In some cases, the notion of a central subject may be as loose as "what happened to the English in the nineteenth century." But even this would impose a degree of unity, a kind of structure, upon what would otherwise be just a miscellaneous collection of facts, and it would yield a principle of relevance for their selection and rejection.
15. Louch, 58. 16. Gallie, 108. 17. Danto, 141, 251. 18. An Introduction to Philosophy of History (London, 1958), 18, 32-34, 61. See also his "'Plain' and 'Significant'Narrative in History," Journal of Philosophy 55 (1958), 479-484. Danto discusses both statements, 116-142. 19. White, 235-236.

158

W. H. DRAY

Now Danto may be right that, at any rate in the indicatedand quite minimal sense, a narrative would have to have a degree of connectednessto be a narrativeat all. I am not quite sure myself that it would alwayshave to have this precise kind: it may well be that other sorts of connectednesswould sometimesbe sufficientto sustain a narrativeeven where nothing that could plausiblybe called a continuoussubject could be [Link] let us suppose that a central subject would invariablybe needed. It seems clear that in this would still not ensure that all narrativeswere "significant" the sense which interestsWalsh;it would still not requiretheir showingwhy what hapof pened in their [Link] requirement a centralsubject betweennarrativesand mere chronomight providea basis for distinguishing logical reportsof just anythingat all. But it would not exclude the possibility between explanatoryand non-explanatory narratives. of distinguishing A quite differentargumentfor refusingto recognizea distinctionbetween plain and significantnarrative,and one which does concern itself with significancein the explanatorysense, has been advancedat least tentativelyby Danto, who points out that historianshave to establishor argue for the facts they narrate.20 This, if nothing else, Danto maintains, will guaranteethat what they eventuallyreport about a central subjectwill be more than a disconnected series of facts about it. The argumenthere is reminiscentof a by famousparadox,crypticallypropounded Collingwoodin The Idea of History: the claimthat when the historiantrulyknows what happenedhe already of necessarilyknows why it happened.2'On one quite naturalinterpretation Collingwood's paradox,what he is tellingus is that the very processof reasoning throughwhich the historianmust go in order to establishhistoricalfacts why they occurred;and although requireshim at the same time to understand his conceptionof historicalreasoningis differentfrom Collingwood's since it is basically Hempelian, or "scientistic" I think Danto can be read as makingan analogousclaim. And I think both of them are clearlywrong. In Danto's case, the claim seems to rest at least partly on the role in historical thinkingwhich he assigns to what he calls "conceptualevidence."22 Historiansmay use their knowledgeof human affairsto argue,not just from evidencewhich receivesno mention as such in their final narrativeaccounts, itself. One of the ways of certifying but also from otherdetailsof the narrative as the facts of a narrative facts, in otherwords,is to representthem as having been necessitatedby other constituentfacts of the story being told. And this would representthem as explicableby referenceto those other facts. Insofar as Danto's point is simplyone about a type of verificationopen to
20. Danto, 140-141. 21. Collingwood, 214. 22. Danto, 122, 125.

NARRATIVE IN HISTORIOGRAPHY

159

historians,I have no objectionto offer. But I fail to see the relevanceof this point for the questionwhethernarrativeis, as such, a form of explanation. The possibilityof a degree of internal verificationin a narrativedoes not show that a narrativecould not be entirelynon-explanatory, while still meeting the requirements being a narrativeat all. There seems to be nothing of incoherentor logicallyimpossibleabout the notion of settingforth a series of facts about a centralsubject,none of which explains any of the others- a notion which other narrativists, like White, have used to clarify the sense in which they think actual historicalnarratives,by contrastwith their so-called "entailedchronicles," may sometimesbe [Link] may be right to maintainthat we can never establishthe occurrenceof a given event without establishing"connections between this event and other events in the past."23
The point is that such other events need not be in the narrative.

A third argumentofferedby Danto is open to a combinationof the sorts of objectionwhich have been urged againstthe first two; althoughthis time it is indeed a logical or structural featureof narrative,ratherthan a methodological point about the verification its constituentfacts, which is called to of our [Link] and indeed all narration,Danto points out - characteristically makes use of a distinctivetype of descriptivestateAn ment.24 example, for an historianwritingabout the events leading up to the outbreakof war in 1914, might be: "And so, at Sarajevo,the first shot of the First WorldWar was fired."Such a statement,althoughabout an event at one time, refersforward,not just to somethingenvisagedby the historical agentsconcerned,but to the actualsubsequentcourse of events. It is a judgment which could be regardedas typically historical,since it introducesa kind of retrospectiveintelligibility into the account of what [Link] it clearlydoes this by connecting facts or events at differenttimes. Danto's discussion of the role and implicationsof "narrativesentences" as he calls them, is intrinsically interestingand important perhapsone of the most fruitfuland originaltheoreticalideas in recent philosophizing about [Link] is hard to see, however,that it providesus, any more than do the with a rationalefor sayingthat narrativeis itself a form precedingarguments, of explanation. Firstly,the use of such sentencesis not necessaryfor the construction of [Link] a narrative will be more intelligible or significantif it employs such devices; but a narrativewhich abjures them, possibly attemptingto representthe past entirelyfrom the standpointof the is participants, so far from being inconceivablethat it in fact representsthe which regardit as a re-enactideal of all those conceptionsof historiography ment of past [Link],the connectionsassertedby narrativesen23. Ibid., 140. 24. Ibid., ch. VIII.

160

W. H. DRAY

tences where they are employedneed not be explanatoryones; they at any rate need not be explanatory the sense of showingwhy the directlyreferin enced (i.e., earlier) event [Link] point of a narrativesentenceis not to show why somethingcame about, but to show what its significancewas. And thirdly,althoughthe significance an event with respect to a later one of may be locatedin the fact that the earlierexplainsthe later, this is surelynot necessarily case. The shootingat Sarajevocan be judgedthe beginningof the the First WorldWar without also being judgedits cause. Danto's "narrative sentences"do show how assertionsof connectednessmay enter quite surreptitiouslyinto [Link] they hardlyrequireus to concludethat narrative itself a form of explanation. is Danto sometimes argues, quite independentlyof any role he assigns to narrativesentences, that there simply could not be a narrativewhich was "plain"in the sense of failing to indicatecauses and [Link] he agreesthat there is a differencebetween saying that Napoleon lost at Waterloo and saying why he did, the point does not seem to be the utterly one that we couldn'tnarratewithoutmentioningcauses and conimplausible [Link] seems to be ratherthat takingall the causal statementsout of a narrative would not necessarilytake all the impliedcausal significanceout of Danto maintains,is "a way of organizingthings";25 there and it. A narrative, is more to the narrativeway of organizingthings than the referenceof every detail to a centralsubject.A narrativeof such a subject,it seems, claims implicitly to select what is importantabout it, and to order what it selects in such a way that the importance,if not stated, is at least shown. An account in details. loses story statusif it gets submerged inconsequential I thinkDanto is on sound groundin insistingthat the notion of recounting important facts, from the standpointof some interestor other, is part of what is involved in the idea of [Link] I cannot follow him in regarding consequential importance as [Link] puts his position very he strongly."Everynarrative," declares,"must spell out some consequences of some events."26 And again: "If an earlier event is not significant[in this sense] with regard to a later event in a story, it does not belong in that These claims appearto me to be [Link] are at any rate not story."27 entailedby the generaladmissionthat historicalnarrativesmust select what is [Link] as Danto himself conceded, there are many kinds of importance at work in historicalthinking;indeed, Morton White has recently found no difficultyin distinguishingseven such kinds, all different from Danto'snotion of what is "consequential."28To take a single example: events
25. 26. 27. 28. Ibid., Ibid., Ibid., Ibid., 140. 138. 134. 238 ff.

NARRATIVE IN HISTORIOGRAPHY

161

may sometimes deserve a place in an historical narrative simply because of their intrinsic human interest. This is not to suggest that any actual historical narrative will get very far with this conception of importance alone; but then, actual historical narratives are, as Donagan has put it, "characteristically" explanatory.29 Our question, however, is whether the very concept of narration or story telling requires the kind of causal inter-connection which would allow Danto to deny the intelligibility of Walsh's distinction. Once again, I fail to see that he has shown this. A scattering of further arguments can be found in Danto's writing, but none of them calls for any lengthy consideration. We are reminded, for example, that some descriptions are "very nearly" explanations.30 That is true, but not to the point. We are asked to note that historians would not normally establish a fact without "presupposing" a narrative, and that facts which are established will probably eventually enter into some narrative.8 This may also be true, but it is similarly irrelevant. We are told that a narrative that failed to explain would "very likely" turn out to be little more than a mere list, and hence no narrative at all.32 But this ignores the possibility of other modes of organization than the explanatory one. Danto even argues, against Walsh, that since an explanatory narrative must say what exactly happened, and since the complete description of what exactly happened will include the explanatory connections between facts - these also being facts - the distinction between plain and significant narrative must disappear.33But Walsh did not mean by "plain" narrative just any account that told what exactly happened; he meant what did only this. And a plain narrative would surely not have to give a "complete description" in Danto's special sense of telling everything that was true about its subject, including, perhaps, all those connectional truths that it would require narrative sentences to express.

III For the reasons indicated, I think that Walsh is right to distinguish between plain and significant narrative. I think that the distinction between history and chronicle remains similarly viable: a useful device for the philosophical analysis of historiography. But whether these distinctions are recognized or not, it is usually agreed that historical narratives can be, and (if there is any choice) ought to be, explanatory. It is often thought, too, that narratives con29. 30. 31. 32. 33. Review of Analytical Philosophy of History, History and Theory 6 (1967), 432. Danto, 130. Danto, 140, 142. Danto, 130. Danto, 140-141.

162

W. H. DRAY

stitute the specificallyhistoricalway of [Link] an hisoffer torian for an explanation,Danto observes, and he will "spontaneously writes Mink, is communiWhat the historianunderstands, us a narrative."34 style of one-thing-after-another."35Whether narrative catedin "thenecessarily distinctiveabout such explanation there is anythinglogically or structurally is a matterabout which there is less [Link] Robert Stover, peculiarto the scheme of intelligibility for example,there is no "fundamental often speak as if they think there is. Yet narrativists narrativeform itself."36 The questionsI now want to raise are thereforewhat sorts of narrativesare explanatory,and by virtue of what structureor logical form. This may give us a basis also for decidingwhetherthere is anythingdistinctiveabout a narrative explanation. Narrativistliteratureyields a numberof accounts of what, in its skeletal narrativein historywould be like. I should structureat least, an explanatory on several of these accounts,beginningwith one like to note and comment between As proposedby MortonWhite.37 indicatedearlier,Whitedistinguishes an explanatorynarrativeand a mere chronicle on the principlethat in the latter no causal assertionswould occur. He thereforefinds the explanatory efficacy of a narrativeentirely in the causal connections it traces. A very narrative,accordingto White,would go somewhat simplecase of explanatory as follows: The King of Englanddied, which led the Queen to grieve, which led the Princess to worry, and so on. A correspondingchronicle, or plain would go: The King of Englanddied, and then the Queen grieved, narrative, after which the Princess began to worry, and so on. More formally, White says, what makes an historicalaccountexplanatoryis its reiteratedclaim that what was true of its central subject at one time was the cause of what was true of it at a latertime. Perhapsthe sense in which White'sexampleconcerns states of a centralsubjectmay be thoughta trifle thin. But as an account of the family problems of the royal house of England, it might be allowed barelyto fall underthis notion. White'saccountof explanatorynarrativesets up as the ideal for historical the understanding model of a causal chain. And it may well be felt that this Whiteis, of course, awareof this; he conrequirement. is a prettyunrealistic cedes that actualhistorieswill seldom achieve a high degree of what he calls The they achieve can never"causalintegrability."38 degree of understanding
34. Danto, 201. See also 11. 35. "The Autonomy of Historical Understanding"in Philosophical Analysis and History, ed. W. H. Dray (New York, 1966), 188. 36. The Nature of Historical Thinking (Chapel Hill, 1967), 70. 37. White, 221 ff. For a clear contrast between White's and Danto's models, and some further comments, see R. G. Ely, "Mandelbaumon Historical Narrative," History and Theory 8 (1969), 276-279. 38. White, 239, 224.

NARRATIVE IN HISTORIOGRAPHY

163

thelessbe measuredonly againstthis ideal; Whitecan offerus no other. What is especiallystrange about this is the fact that White's own full account of what he called "the logic of historicalnarration"contains elements which virtuallyguaranteethe non-achievement this very ideal.39What I have in of mind particularlyis his avowed "pluralism" about the principleson which facts may justifiablyfind a place in an historical narrative the various concepts of importancealreadyreferredto. What makes a narrativea history of its subject at all, White says, is its claim to be composed,in the first instance, of "basic"facts about that subject, which have been selected on appropriate principlesof interest and importance.40 basic facts of a historyof the United States,he suggests(and The his intentionis purelyillustrative),might consist of the Revolution,the Civil War, the First WorldWar, the Depression,and the Second WorldWar. It is around such salient facts as these that an eventual narrativehistory of the Americannationwill be [Link] separatefrom this historyact constituting of selection,however,is what mightbe called the explanationinjectingfeature of narrativeconstruction:the attempt to account causally for, and indicatethe consequential significanceof, all those facts which have status as basic ones. This explanatorytask involves the selection of still furtherfacts, called by White "derivative" ones, on the groundthat they are causes and consequencesof those chosen as basic.41 But why should we expect the productof two operationssuch as these to which is causallyintegrated i.e., one in which, as in White's be a narrative lugubrious historyof the Englishroyal house, every event mentionedis both consequenceof the precedingone and cause of the next? There is no reason at all to expect this, if the basic facts of a historyare selectable,as for White they quite clearlyare, on principleswhich are logicallyunrelatedto the overall purpose of constructinga causal chain. The demands of causal integrability, and of what White (in a Herodoteanmoment) sums up as "memoracould coincide only by some kind of miracle. And this is not just bility,"42 because, as White himself laments, "interesting facts, like interestingpeople, Doubtless there are may have boring antecedentsand boring offspring."43 causal chains to be discovered,in history, and the analysisof their structure is a legitimate and somewhat neglected philosophicalexercise. There may even be historieswhich take causal chains as their central [Link] it is hardto see, if we can find any plausiblealternative, why we should represent
- 39. White, 219 n.1. For his pluralism see 257 ff. 40. White, 231. 41. White, 232-237. 42. White, 266, 256. I owe the point to D. R. Newman's "Significancein History," a thesis submitted for the M.A. degree at the University of Toronto in 1968. 43. White, 239.

164

W.

H.

DRAY

such chains as formulatinga logical paradigm for historical explanation. Certainlythere will be few explanatory narrativesif we do. An alternativeaccountof what makes an historicalnarrativeexplanatory, althoughstill basically a causal one, is in fact offered by Arthur Danto.44 Danto'sconceptionof the causal patternitself is less restrictivethan White's. The way he thinksof an explanatory sequencedevelopingcould be illustrated somethinglike this. Becausehe noticedthat the enemy line was wavering,the generalorderedthe cavalryto advance;becausethe cavalryhad to traversean open space rakedby cross-fire, was unableto penetratethe enemyposition; it because enemy reinforcements arrivedbefore it was possible to regroup,the chance of a quick decision was lost. Here again we have somethingwhich could be called a causal series; but it is not a causal chain in the sense explicatedby White. For causes now enter the series, as it were, from the outside. Less figuratively,what is called a cause at each stage is itself left unexplained,although it explains the relevant change in the state of the centralsubjectat that [Link] be said to offer a causalchain be model of explanatory Danto's mightmore appropriately referred narrative,
to as a model of causal input.

One interestingfeature of Danto's analysis is his insistence that, strictly speaking,it is not a final event or state which the historianexplains,but the process of changewhich it [Link], at the second stage of the series just noted, what is explainedis the transmutation the cavalryfrom a state of of splendidpotency to one of [Link] explains that [Link] the basis of this change- the cause of it - is the intervening sort of analysis,Danto representsan historicalcause as invariablya "middle" in a beginning-middle-end-structurea narrative,in other words.45 And he thereforeargues, in the end, not only that all historicalnarrativesare explanatory,but also that all causalexplanations historyare narrational. in The causal questionitself is seen as narrative-generating. If I have difficulties with Danto's accountof the structure a causalinput of minor points. The first, which has been narrative,they are on comparatively noted by MichaelScriven,is the apparentdependenceof this accounton the that historians,in seekingcausalexplanations,are invariablyconassumption cerned with accountingfor changes.46 No doubt they generallyare; but they sometimesask also why thingsremainedthe same; and it is not entirelyclear how Danto'singeniousargument connectingcauses with story-middles would lies in what seems to me a very misfare in such cases. Another difficulty leading way in which Danto repeatedlyputs his doctrine in contrastingit
44. Danto, ch. XI. 45. Danto, 235-237. 46. Review of Analytical Philosophy of History, The Journal of Philosophy 63 (1966), 504.

NARRATIVE IN HISTORIOGRAPHY

165

with the standardHempeliantheoryof explanation. The Hempelians,he says, in their anxietyto cast all explanationinto the form of a deductiveargument, incorrectlyassignedthe initial state of explained changes to the historian's
explanans. But "the earlier event," Danto objects, "is part of what has to be

accountedfor";"boththe beginningand end are part of the explanandum."47 This could wronglygive the impressionthat Danto thought a causal middle explainedthe existenceof the initial state itself, which, of course, it doesn't. But the problemmay be no more than one of unhappyformulation. The real difficulty with Danto'saccount,from the standpointof our present concerns (and one which it shares with White's), is that no sufficientargument is in fact offeredfor concludingthat an historicalnarrativewill be explanatoryonly if it offerscausal [Link] in effect definesexplanaso tory narrative that its explanatoriness [Link] does not; is but his chief efforts are directed toward showing that causal explanations generatenarrativeratherthan the converse. Although more accommodating than White's,Danto's model must still appearhighly prescriptive when taken to the details of historicalwork- Danto himself had to concede that actual includeda considerable amountof what he called "narratively narratives inert We must,therefore,surelystill ask whetherhistoricalnarrative information."48 in cannot also be explanatory other ways.
IV

The third accountI want to look at is that of W. B. Gallie, who claims unequivocally that further alternativesare possible.49Galle's theory of ex even with planatorynarrative, however,is much more difficultto summarize, respect to its centralidea, than the models of causal chain and causal input. For one thing, it bringstogethera numberof ideas, not all of which seem to be equallyrelevantfor a contrastwith the causal models. Gallie also makes or contrastdifficultby refusingto speak directlyof "narrative explanations," even of "explanatorynarratives."What we properly call explanationsin historicalwork,he says, are normallygiven at those pointswhere an historical narrativebecomes difficultto follow.50Explanations are intrusive, correcwouldhave no need for them. However, tional;an ideallyintelligiblenarrative form is the vehicle of what he prefers since Galliedoes hold that the narrative and since the accounts he gives of to call the "historicalunderstanding," as explanationsrepresent them, characteristically, simply re-establishinga
47. 48. 49. 50. Danto, Danto, Gallie, Gallie, 235, 234. 250-251. chs. 2-5. 22, 89, 105, 110.

166

W. H. DRAY

type of intelligibilitywhich good historical narrativeshave as a matter of course, it does not seem to me that we would misrepresent him seriouslyif we said that what he offersis an accountof what makes a narrativeexplanatory. Indeed, as alreadynoted, he sometimeshimself slips into speakingof narratives self-explanatory; he talksof the need for explicitexplanation as and from time to time, as if the rest of a narrative could be regardedas implicitly
explanatory.5'

If Gallie is askedfor a generalaccountof what a self-explanatory narrative is like, his replygoes somethinglike this: it consistsof a sequenceof incidents, actions, states of affairs,and the like, which catch our interestsufficientlyto make us want to follow them to some vaguely indicated but unpredictable to conclusion,and whose relationship each other is such that we can accept them in succession,however surprisingand unprecedented, plausible and as relevant developmentsof the theme or subject matter under consideration. Just as Danto mightexpect an occasionallink or two in an explanatory series than that of causal input- perhaps the to manifest a tighter relationship causal chain relationship so Gallie would not rule out the possibilityof occasional connections in a narrativebeing of a tighter sort than those suggestedby the generaldescription just given. But his claim is that the links of an explanatorynarrativedo not need to be any tighter than this to be An as intelligible. explanatory narrative, he sometimesputs it, can incorporate His account might thus be referred to as the followable "contingencies." contingencymodel.52 Mink, especially, will probably have qualms about my use of the word "model"here. And his own suspicion that, in the approachof an extreme like Gallie, we shall find nothing strictly analogousto the logical narrativist schemasof a White or a Danto, may well appearto be vindicatedby the very summarywhich I have just [Link] calls attentionto the degree to criteriaof historicalnarration whichGallie concernshimselfwith extra-logical and of story-telling generally;and rightlyso. Nevertheless,I think that something like a competingtheory of the so-called "logic of narration"can usefully be extractedfrom what Gallie has to say. And as at least a preliminary move towardextractingit, I shouldlike to make a few remarksabout several ideas which I believe to be central to Gallie'sview of explanatory structural [Link] ideas all seem to me, in various ways, to bring him into logical conflictwith the claims of the causal models. The first is the idea of contingencyitself. Mink has said that Gallie is about this, attendingmore to the nature of the "stoutlyphenomenological" experienceof being confrontedwith a surprisingturn of affairs- and the
51. Gallie, 23, 22. 52. Gallie, 29-30, 32, 96 ff.

NARRATIVE IN HISTORIOGRAPHY

167

experience, often enough, of a naive or ignorant reader at that - than with anything like an objective relationship.53 And Gallie certainly encourages such interpretations, both in the detailed account he gives of the requirements for "followable narrative," and in such general complaints as that his predecessors have made of critical philosophy of history just "so many exercises in applied logic."54 But at several crucial points he makes it clear that, in his view, an intelligible historical event or sequence can be contingent in the strong sense of being unpredictable in principle, unpredictable even "after the fact," because not deducible from antecedents and laws.55 Unlike science, Gallie maintains history lacks the "fundamental aim" of removing all the contingency from its subject matter. This could, of course, be interpreted as meaning that historians are generally satisfied with half-doing the explanatory job. I take it rather to mean that, although in the ordinary case the question simply won't arise, historical understanding is such that a concomitant assertion of real contingency would not necessarily undermine it. And this, I think, is right. Gallie nevertheless concedes that the contingent and unpredictable is unintelligible per se.56Thus although contingencies are "intellectually acceptable" as items in an historical narrative, they contribute nothing to its intelligibility unless further criteria are satisfied.57 One such criterion that bulks large in Gallie's account is being a necessary condition of some other items of the history - and this is the second structural idea I want to note. According to Gallie, the necessary condition relationship provides "the main bond of logical continuity in any story"; it gives it its "logical texture" or "ground of intelligibility."58 This doctrine is reminiscent of Gallie's earlier work on the alleged similarities between explanations in history and the genetic sciences.59 And now, as then, it is important to notice that there are in fact two different ways in which necessary condition relationships are seen as introducing intelligibility into a narrative. There is first the way a story gains in intelligibility when we see it as having the kind of continuity - a third structural idea - which is fully discernible only when we consider earlier episodes from the standpoint of later ones.60 A contingency which is unintelligible per se, Gallie says, can become intelli53. "Philosophical Analysis and Historical Understanding," Review of Metaphysics 21 (1968), 684. 54. Gallie, 19. 55. Gallie, 91-92, 88. 56. Gallie, 41. 57. Gallie, 31. 58. Gallie, 26-27. 59. "Explanationsin History and the Genetic Sciences" in Theories of History, ed. P. Gardiner (Glencoe, Ill., 1959), esp. 387-388. 60. Gallie, Philosophy and the Historical Understanding, 106, 40-41, 43, 33.

168

W. H. DRAY

gible through its having made a contributionto the developmentwhich a historyis concernedto trace. To achievethis status,he maintains,it need not have been a cause; it need only have been a condition but for which the would not have [Link] envisageshere, it should development and be noted, is the necessaryconditionitself achievingintelligibility, thereby he is talking as of to contributing the intelligibility the narrative a whole. What The chief but about, in other words, is not "explanatoriness" "significance." gap in Gallie'saccountat this point, as in his earlierwork, is his not saying seen to have been a clearlywhetherevery conditionwhich is retrospectively necessaryone therebyachieveshistoricalsignificance,or whether only some do. If only some (which is the more likely), then we need to know how the significantones are discriminated. The other way in which the necessary condition relationship may be is, thought to underwritehistorical understanding of course, the necessary condition'sbeing taken as explainingwhat it was necessaryfor. We may well ask, again, whethereverythingthat a conditionwas necessaryfor would be explainedby it. But this time Gallie offers us at least the beginningsof an [Link] sorts of non-causalexplanatoryconnectionswhich he generally mentions- and these constitutethe fourth and the fifth of Gallie'sstructural ideas- are between what people in fact do and conditionsprovidingthem for with reasons and opportunities acting in those ways, the historianpresumably applying whatever limiting criteria are generally accepted in the languageof practicefor calling antecedentnecessaryconditions"reasons"or Even when we would claim no more aboutcertainantecedent ''opportunities." conditionsthan that they "evoked,occasioned, or made possible"what was done, Gallie maintains,they may neverthelessenable us to understandits being done.61He goes on to commend various non-Hempeliansfor having formulatedviews of explanationin terms of agents'reasons or "the logic of the situation,"which, he says, highlightconcepts which are ideally suited to into an explanatorynarrative. incorporation Does Gallie recognizestill a sixth structuralidea - some notion of purposive order?This may perhapsseem to be suggestedby such dicta as that or is historicalunderstanding a "teleologicallyguided form of attention,"62 that a story's conclusion, although usually unpredictable,nevertheless not only "guidesour interest almost from the start,"but even pulls us forward One mightobject that the actual conclusioncould "almostagainstour will."63 are readingan historicalaccount, we don't know hardlydo this if, while we what it is; and Gallie (probablywrongly) seems to think we usually don't.
61. Ibid., 22, 26. 62. Ibid., 38, 64. 63. Ibid., 28, 22.

NARRATIVE IN HISTORIOGRAPHY

169

But he often, in any case, appears to be telling us that a "presentiment" of a conclusion, or a sense of "possible" conclusions, would do just as well.64 I am myself inclined to write off most of this - along with much else Gallie has to say about the role of interest, directed feelings, and emotional involvement as belonging to what Mink calls the "rhetoric" of history.65 If anything structural is being hinted at by such language, the five ideas already elicited, especially the idea of retrospective continuity, seem already to have made Gallie's legitimate point. Few stories which historians would bother to tell would fail to emphasize the relevance of earlier conditions which were necessary for what eventuated. In that sense, but surely not in any stronger one, a "teleological order" could generally be expected. It might be added that Gallie is occasionally tempted to develop the notion of explanation by necessary condition (of whatever further-specifiable kind) into chain-like analogues of the causal models. Thus he tells us at one point that a series in which each incident was seen to be a necessary condition of the one following would clearly be a self-explanatory one.66 If such a series were indeed his ideal of explanatory narrative, then although its individual sequences would be logically different from those of the causal chain, the whole construction might well encounter the same difficulties about integrability that White's model did. The causal input model would provide a more likely analogue for the ideal he might be expected to recognize: indeed, for all Danto knows, any number of his inputs may be contingent. But even this, as a general requirement, would hardly square with the flexibility Gallie often displays about what can be incorporated into an intelligible narrative. His minimal demand seems to accept the possibility, at least periodically, of sheer contingencies. We can follow a narrative through and across contingencies, he insists, provided they do not preclude further relevant developments.7 The element of contingency in history, he adds, can be expected to match that experienced in life.

V There is one more structural idea which I think is at least embryonic in Gallie's account of historical understanding. This is the idea of a narrative's intelligibility arising out of its delineation of a whole of nonsimultaneous parts. Let me call this the model of synthetic unity - with apologies to Mink. It is tempting to say that such an idea introduces nothing new. For what makes
64. 65. 66. 67. Ibid., 25, 42, 67. Mink, "Philosophical Analysis and Historical Understanding,"686. Gallie, Philosophy and the Historical Understanding, 109. Ibid., 29, 67.

170

W. H. DRAY

the narrated whole a whole must surely be the interrelationship of its parts. And what could these relationships be if not the kind already discussed by White, Danto, and Gallie, or some combination of them? But to say this would, I think, be a mistake. As Mink has argued, the ultimate goal of much historical inquiry may be to delineate a complex of relations; and it may take the historian's whole work to say what that complex is - to state his "undetachable" conclusions about it.68 There are anticipations of this sort of claim, perhaps, in Gallie's notion of story-incorporation as a criterion of significance - his notion that whether or not a contingency has introduced "radical and unacceptable" discontinuity is retrospectively decidable in terms of a criterion of relevance which develops along with the narrative itself. There may also be anticipations of it in Danto, who, apropos of his account of causes as story-middles, said that the historian's task, from one point of view, was to constitute "temporal wholes."69 But it is Mink who has said it best. The claim that narratives may achieve the intelligibility of synthetic unity deserves more attention than it has yet received from philosophers - and more, too, than I can give it in the final paragraphs of this essay. I shall offer just three concluding remarks. I would emphasize, first, the non-equivalence of the present claim with the one noted earlier that a narrative must have a "continuing central subject." For the model of synthetic unity envisages a kind of intelligibility which a mere chronological account, even of what happened in a central subject, may lack; and the unity at issue would, in any case, be the conclusion of a narrative inquiry, not its precondition. (The difference is the sort which might lead an Hegelian to talk about concrete rather than abstract wholes.) I would emphasize, too, the need to distinguish between the sort of "continuity" envisaged by this model of understanding and an apparently quite different sort which crops up from time to time in philosophical writing about history: a sort articulated mysteriously, but not untypically, by Michael Oakeshott when he says that the only explanation necessary or possible in history is a full description - an account of change in which, at the ideal limit at least, every "hiatus" is overcome.70 (Oakeshott would consider even a distinction between cause and effect as opening a "hiatus.") A similar sort of structureless holism may be suggested by Louch's talk of "making continuity visible,"7' and even by Mink's remarking, in an earlier essay, that if we are to grasp the nature of historical understanding, we may need to take "process categories" seriously.72
68. 69. 70. 71. 72. Mink, "The Autonomy of Historical Understanding," 180-181. Danto, 248. Experience and Its Modes (Cambridge, 1933), 128-130. Louch, 56. Mink, "PhilosophicalAnalysis and Historical Understanding,"681-682.

NARRATIVE IN HISTORIOGRAPHY

171

Finally, I think it is worth pointing out explicitly that the model of synthetic unity, as it has been presented, is, in an important sense, not in competition with the other three models - at any rate, not in competition with them in the sense in which they are in competition with each other. For if the unity of a story is grasped through the perception of all its constituents' interrelationships, synthetic understanding might be said to presuppose the other kinds. One consequence would be that a philosopher who conceded the indispensability of such an ultimate, synthetic phase of historical understanding would still have to decide what counts as piecemeal intelligibility at the lower level: he would still have to make up his mind, for example, whether detailed explanatory connections would all have to be Hempelian. Gallie seems to me to be telling us that we could expect a number of quite different sorts of connections to be story-constituting -perhaps many more than have been in this essay. And on this point he seems to me to be entirely considered correct. Trent University