Now we arrive at the third tenet of narrativism: holism. The natural starting point here is the narrativist demand to treat and study works of history as wholes, that is, not to cherry-pick historical claims from a historiographical text, but to try to understand what kind of story, message or thesis a work of history as a whole amounts to. White emphasized that historians produce unified texts, and Metahistory is a case in point. In it, many classics of Western historiography and philosophy of history receive a comprehensive reading in an attempt to uncover what tropes and story modes govern them as whole texts. White's talk of the structures of texts and the relationships between story elements is also symptomatic. Cutting stories or narratives into smaller pieces, except as an analytical exercise, means that they lose their primary identity and communicative function.
The same idea of historical texts and theses being distinct and qualitatively dissimilar from the elements out of which they are composed characterizes Ankersmit's thinking. 'History and historical debate is holistic in that the universally shared assumption in historical writing is that only the whole of the text conveys the historian's cognitivist message, and to which the parts only contribute', writes Ankersmit (2008b, 92; similarly 2012, 159). But while White is content with a structural examination of texts, Ankersmit analyzes their philosophical status at length. In Narrative Logic, Ankersmit makes an important link between three closely related theoretical notions: images/pictures (of the past), colliga- tory concepts, and narrative substances. Narratios or narratives thus amount to images or pictures of the past that a historical text articulates (conceived of as proposals about how to view the past and not as being copies of the actual past). Ankersmit writes that the term 'colligatory concept' is well suited to replace the terms 'image' or 'picture' of the past, but that it in turn is subsequently replaced by an even more fitting term: 'narrative substance' (1983, 93). Although the concept narrative substance came to be replaced by other notions in Ankersmit's later writings, the idea of the holistic nature of historiography remains the same. In what follows, I characterize narrative substances in more detail in order to understand the narrativist view of the holistic nature of historiography.
The first important observation is that there are two kinds of narrative entities that need to be kept separate: narrative subjects and narrative substances. The previous are an integral part of the narrative realist's world-view, and also accepted by the narrative idealist, while the latter belong only to the narrative idealist's universe. Suppose we have a biography of Napoleon, containing a large number of statements concerning him and his life. The narrative subject in this case is the historical Napoleon, the person who lived between 1769 and 1821 and became the emperor of France. Ankersmit's narrative idealist sees that, in addition to the narrative subject, there is also a narrative substance, which is the 'image' or 'picture' that the author presents of the life of the historical Napoleon, and which is an outcome of all individual statements of the narratio combined (Ankersmit 1983, 90; cf. 2001, 61).
The difference between narrative subjects and narrative substances is like flipping the perspective on individual statements of a narratio. If the narratio is seen merely as an enumeration of singular statements, then we are talking about narrative subjects and derivatively of 'real' entities in the past to which these statements refer. But if we view the narratio not just a list of these statements but as the totality that these statements altogether amount to, and importantly, as a specifically ordered set, then we are talking about narrative substances.
If the claim is that 'Napoleon conquered Russia', the primary function in historiography is to make a particular statement about the past, that is, that Napoleon conquered Russia. In principle, it seems that we can decide whether this is true or false. Its truth requires that there was a person called 'Napoleon' and that this Napoleon actually conquered Russia. To the best of my knowledge, the French Emperor Napoleon did not succeed in his attempt to conquer Russia, and therefore, the statement is false. In its second function, this very same statement contributes to the image of Napoleon that is detailed in a work of history. The statement is part of the whole picture provided (and a 'property' of a respective narrative substance). Ankersmit remarks that it would be easier to differentiate which entity is in question if we were to reserve the name 'Napoleon' for the historic Napoleon and distinguish it from images offered by historians by using such locutions as, 'my Napoleon' or 'the Napoleon of the historian H' (1983, 95). And the following would be even better expression: 'Louis XIVh1, narr', 'Renaissance h2', 'the emergence of a new social elite in the 19th century h3', and so on, where 'h1', 'h2', and 'h3' denote different historians' interpretations or different works by those historians. There is also an addition 'narr' in the case of Louis XIV in order not to confuse the narrative substance with a narrative subject Louis XIV, as otherwise the danger is that we multiply the number of historical Louises (1983, 124-125).
Further, narrative language is self-referential. The name the 'Cold War', for example, refers to the the Cold WarNs,15 which expresses the narrative meaning of the whole but is composed of a set of statements. If the narrative substance itself is seen as a name of something else, it refers to that set of statements it is composed of, that is, to itself. Further, it is necessary to express the narrative meaning in the form of 'N1 is p ... N1 is pn', and not just as a sequence of individual statements 'p1 ... pn', because otherwise one misses the most important feature of historical texts, which is that they bring forward some general picture or thesis of the past, identified as narrative substance and here expressed as 'N1'. We thus construct this new holistic entity, narrative substance, in order to see and recognize this aspect of historiography (see Ankersmit 1990, 278-281).
Ankersmit writes: 'The thesis that all statements expressing the properties of Nss are analytical is, perhaps, the most fundamental theorem in narrative logic' (1983, 127; similarly 1988, 220; also 1995b, 225-226).16 In light of what has been said of narrative substances as holistic entities, this thesis is now entirely understandable. All the parts of a whole belong to it inherently; we might say that they are necessarily parts of it, and are true of the entity due to this membership alone. It would be possible to express the analyticity of narrative substances as follows. If a specific narrative substance N1 is composed of the set of statements, s1, s2, and s3, then to say, for example, that 'N1 is s1' does not bring any new information, but is necessarily true due to the definition of this narrative substance. The requirement to make a complete enumeration of all properties of Ns as the only way to individuate (e.g., 1983, 110) can also be explained by the idea that narrative substance is a whole composed of all its parts. And as a further consequence, the identity of an analytically defined or true entity requires that all parts remain unchanged: 'Whichever Ns we may choose, none could ever be different from what it is, without ceasing to be the Ns it is' (1983, 213). If there is any change in its constituents, it will be a different entity: 'as soon as one statement is omitted or added we have to do with a different Ns' (1983, 213).
A consequence of this position is that a narrative cannot misdescribe its object, such as 'Renaissance', because it creates it: RenaissanceNs 'is nothing more and nothing less than what individual historians tell us that it is' (1983, 201). And there is, for example, no such thing as the 'Fall of the Roman Empire', strictly speaking. Narrative substances are not 'shorthands' that enable us to speak about things in historical reality. As a consequence, narrative historical knowledge is not knowledge proper, but 'an arrangement of knowledge' (1983, 227). There is no fact of the matter regarding what the 'Fall of the Roman Empire' is, but its identity is totally up to the historian's stipulation.
Although there is a gradual terminological change from 'narrative substances' to 'representations', all the elements of holism are still an integral part of the later Ankersmit's philosophy. In his latest writing, holism is applied to 'representation' in the three-place representation schema of representation, presented and represented reality. In general, one may say that three central features characterize holism in the narrativist philosophy of historiography: undecomposability, analyticity and unfalsifiability.
Undecomposability means that an entity cannot be decomposed into its constitutent parts without losing its identity. Undecomposability can be derived from the meaning of 'whole' or 'holism' itself, but is seen also in claims like 'none of the statements which constitute the text is ... irrelevant to the text's presentation of the past' (1995b, 225; my emphasis) or earlier, in the one pointing out how no statement can be omitted. It amounts in the later writings to the idea that one cannot make a distinction between the attribution of predicates and reference, an example of which was the image of a black cat above, where 'cat' and 'is black' constitute one unified representation. Or compare the following view of portraits: 'We do not experience it as a composite of bits of information about hair, nose, color, form of the eyes, etc. (all of them corresponding to statements about the sitter's hair, nose, and so on) but rather as a representational whole' (2012, 98).
Analyticity is a very consequential feature of historical representations: all that can be said of representations derive from their definitions. If claims are true, they are necessarily so, because they constitute the very meaning of those representations. For example, think about historians writing about 'Renaissance'. They all come up with a different definition of it, and as a consequence, all that they say about their respective 'Renaissances' is analytically true of them and can be deduced from their meanings: 'Each historical account of the Renaissance is true, since it can be derived logically from how the historian in question proposes to define the Renaissance . .. what is then said about fourteenth- and fifteenth-century Italian civilization is, admittedly true by definition - but true it is' (Ankersmit 2001, 38).
We notice then that each representation is strictly unique, like an individual person. In actuality, representations are even more unique than human beings. An individual could lose a hand or a leg and still be the same person. By contrast, if any part of a holistic entity is changed or removed, it is necessary to talk about a different object altogether, about a different representation.
The third characterizing feature of holism is unfalsifiability. Given that representations are wholes, one cannot try to falsify or corroborate any part of the narrative but instead needs to focus on entire narratives. But narratives are definitionally and analytically true, which means that they are immune to any empirical challenges. The classic example of an analytically true sentence in philosophy is 'A bachelor is an unmarried man'. One cannot possibly falsify this linguistic stipulation by empirical information on the numbers of married and unmarried men.17 It is naturally possible to come up with an alternative definition, but this has nothing to do with empirical standing. Whenever linguistic definitions, constituent parts of representations, are changed for whatever reason, a new entity is defined and created, but the old one is not falsified. Alternatively, its epistemic status remains unchanged. Compare White's words on this: 'history progresses by the production of classics, the nature of which is such that they cannot be disconfirmed or negated. ... It is their nondisconfirmability that testifies to the essentially literary nature of historical classics' (1978, 89). This has to be so, because the classics create their own 'worlds' and their own analytically true linguistic creations. How could anyone hope to corroborate or falsify novels? From a logical point of view the message is that historical 'facts' do not yield precedence to any specific narrative or to any narrative form. They are all equally justified and immune to falsification. Further, representations cannot be compared to each other in terms of their empirical adequacy.18 That is, since they are autonomous and 'analytic entities', they cannot be ranked in terms of their fit with empirical evidence. There may be of course other criteria with which they could be ordered, such as their originality or aesthetic/literary appeals, but empirical adequacy is not one of them.
My discussion on holism later in this book (Chapter 5) focuses on the question: how feasible is the assumption that historical theses cannot be undecomposed? That is, I will examine some such theses and their contexts of origin and see whether they could be seen as composable. In this chapter, I hope to have shown that three concepts: represen- tationalism, constructivism, and holism characterize and define the philosophy of narrativism. In the following chapter, I concentrate on evaluating the tenability of one of them: representationalism.