History and Theory, Vol. 28, No. 1 (Feb., 1989), 43-58.
THE PARADOX OF HISTORICAL CONSTRUCTIONISMMICHAEL E. HOBART
I. CONSTRUCTIONISM: KNOWLEDGE BY CONVENTION
We owe the term "constructionism" as a philosophical doctrine to Jack W. Meiland, whose Scepticism and Historical Knowledge explicated and defended history as, roughly, "a fable agreed upon." Meiland asserted that historians "must be regarded as constructing or creating the past rather than as reporting the past" in order to develop a critique of idealist theories of history, such as those of Oakeshott, Croce, and Collingwood, and to present the skeptic's case for doubting any putative objectivity in historical knowledge. Although he employed constructionism to ground a skeptical perspective, Meiland denied the doctrine's necessary ties to any particular philosophical persuasion (for example, idealism or realism) or to any specific methodological program of historians. Rather, he sought to show that because objective knowledge of the past is impossible, historians should be considered as constructing their accounts for "other purposes" than discovery of the past per se.' In short, history should be understood as a product of the perspective-laden conventions of historians.2
Since the appearance of Meiland's book in 1965, two families of conventions, rhetorical and conceptual, have steadily gained the attention of philosophers and historians alike, and have served as foci for discussions about constructionism.3 Rhetorical constructionism stresses the figurative or rhetorical use of language as the key to understanding historical accounts and is most frequently associated with the tropistic analysis of Hayden White.4 For White, histories are narratives
that "emplot" events from within the dominance of one or another mode of emplotmentromance, tragedy, comedy, or satire. To grasp the meaning of an emplotted history one must, accordingly, understand the historian's underlying rhetorical structure. In turn, this requires the fourfold analysis of figurative language as metaphor, metonymy, synecdoche, and irony. Any truth of narrative is to be conveyed in "figurative terms,"5 and any discussion of "real" events, historical causes, explanations, or referents is reduced as a consequence to the trope which the historian has selected. On this argument there can be no extra-tropistic, cognitive "touchstone" to which one might appeal for resolving disagreements among historians. No "one conception of historical knowledge is more 'scientific' than another," and to claim so begs the question, "prejudicing the problem of what a specifically historical or social science ought to be."6 Consequently, histories from different tropes are epistemically incommensurable. One accepts an account for aesthetic, moral, or political reasons, but not because, ultimately, it is true, or any truer than its rivals.
A second family of constructionist conventions attends less to figurative language than to the conceptual framework through which the historian approaches his subject. This "framework of knowing," in the phrase of Leon J. Goldstein, closely resembles Thomas Kuhn's notion of a paradigm, understood here as a "disciplinary matrix" within which a community of scholars or researchers formulates queries, identifies evidence, and presents findings, either in a narrative or some other form of expression.7 By framework of knowing, then, Goldstein means the historian's "infrastructure," which describes the methods that "define the discipline of history." These methods cannot be understood apart from the actual, specific instances in which historians practice their craft. There is "no way to reach outside such frameworks," nor any need to do so. Disagreements
between historians on substantive points over what the past was like are resolved piecemeal as the "community of scholars" comes to accept some "historical constructions" as "better than others," the latter being simply "dropped from consideration."8 Deciding which is "better" can only be done on a case-by-case basis, as historians generate evidence for their accounts from available documentary (and other) sources; no distinction between fact and assertion can serve as a control of external reference. There can be no appeal to structures or relationships that are, in Maurice Mandelbaum's words, "inherent in the evidence."9 There is no touchstone, no "real" past to which the scholarly community has access, only the constituted or constructed past. Recently Goldstein has even gone so far as to argue that little, if anything, remains for epistemologists to write about with regard to historical knowledge. The "practice" of the discipline develops its "own criteria of truth and objectivityas well as of factuality and reference."10 Only the scholarly community can determine collectively which historical accounts are cognitively commensurable.
Whether with tropes or frameworks both families ofconstructionists have added their voices to a chorus singing the death of "epistemological foundationalism." Both deny the central assumptions of what R R. Ankersmit terms the "epistemological philosophy of history": that historical accounts can "mirror" the past;
that one can distinguish between a constructed and a real past; that one can promulgate independent "criteria for the truth and validity of historical descriptions and explanations"; or even that one can legitimately distinguish "levels" of analysis between the past itself, allegations about the past, and philosophical critiques of the allegations. Rather, historical accounts and their epistemological justifications are collapsed into the perspective-bound practices of inquiry as history provides a "belvedere" from whose vantage point one metaphorically envisions, and hence shapes the past. Within and through its constructions the perspective of historians permeates and governs their enterprise as they formulate and organize the objects of historical inquiry, generate accounts about historical objects, and proffer evidence in support of accounts." Historical knowledge thus justifies itself as its practitioners engage the conventions of their craft; if one does it well, White suggests, works of history will be "reborn into art."12
For some working historians this sounds good. It comforts to read that "each historical interpretation is essentially the proposal of a criterion for how we are to set about if we want to understand a part of the past.'"3 Others though will find it difficult to shake off two persistent doubts. The first of these derives from the fact of historical disagreement. Major bodies of historiography, such as those
pertaining to the English Civil War or to the French Revolution, continue to be rent by extensive, seemingly irresoluble debates over interpretations. For many historians the commitment to a criterial court of appeals to help resolve, or at least clarify, these debates dies hard, and the tendency to invoke epistemic criteria implicitly or explicitly lingers. It seems specious to contend that because interpretations and criteria are yoked in historical accounts they neither can nor ought to be distinguished for the sake of sorting out arguments. It is as though, having acknowledged their linguistic and institutional yoking, we were to claim that wives and husbands were identical, and that we could learn all about the one by studying the other. This doubt grows in the face of an assertion, such as made by White, that history provides "another kind of knowledge" than does science.14 That other kind of knowledge depends largely on the point of view adopted by the historian, and arguments on its behalf seem in the end to take an apologetic form: "Try it, you'll see." But while apologetics work to defend religion and pesto sauce, they avail little in historiographical debate. Thus the first, twofold suspicion: either the tendency to distinguish between an interpretation and epistemic criteria encourages a return to the arena of epistemological philosophy of history, from which constructionists counsel flight, or resolving historiographical debates becomes largely a matter of taste, and de gustibus non disputandum.
Constructionism cannot be so easily dismissed as the above polemic indicates. But the impulse to appeal to epistemic criteria for resolving debates, to slide from constructionism back into epistemology, suggests a second and even more profound suspicion about the coherence of constructionist claims, namely that predicating historical knowledge on an analysis of rhetorical or conceptual conventions looms as problematical as articulating its epistemological foundations. For it would appear that any attempt to resolve a dispute in historical interpretation within a convention of self-contained criteria of confirmation by appealing to justificatory criteria outside the conventionto wit, the theory of constructionismis self-defeating.
In the following remarks I wish to explore this second, deeper suspicion by showing that constructionism harbors at its core just such a self-defeating supposition, one whose truth implies its own falsity. To develop this argument I shall address some problems pertaining to the question of historical reference and the related theme of historical objectivity. I think it possible to show constructionism self-defeating by examining other topics, such as its ontological assumptions or its various accounts of fact, explanation, causation, and the like. However, the question of reference bears most directly on historians' concerns about the point of view one adopts and particularly on the issue of whether there can exist any reasons, beyond agreed-upon conventions, for adopting one point of view over another as a more persuasive historical account. Thus by exposing the paradox of constructionism through an examination of reference, I shall also be claiming that we cannot coherently abandon historical objectivity, understood most generally as ways of resolving disputes in interpretation.
II. HISTORIANS IN A VAT: THE CONSTRUCTIONIST DILEMMA
The central premise of constructionism is, to repeat, that the process and product of historical inquiry are both functions of the historian's perspective-laden conventions. To examine this premise it is useful to introduce Hilary Putnam's striking metaphor of brains in a vat, which he exploits to argue that "mental state (in either the individualistic or the collective sense) does not fix reference."15 If we extend this "science-fiction possibility" and imagine historians in a vat, we capture the premise. Forsaking any escape from a "point of view," constructionists maintain that the historian's perspective, his mental state as it were, fixes the terms of historical reference, that about which history purports to provide knowledge. Although historical reference differs in significant ways from Putnam's examples, as we shall see, examining this extended metaphor will enable us to show why the constructionist premise is self-defeating, and thus to disclose what historians presuppose when taking refuge in a point of view. But first, the fiction.
The image of the "mad scientist," Descartes's evil genius in contemporary guise, allows us to concoct the story. Our mad scientist has been able to remove an historian's brain from his body and place it in a vat of nutrients to keep it alive. He has connected the nerve endings of this brain to a "super-scientific computer," which enables the historian to have the illusion of normalcy. Moreover, our scientist has read his Arthur Danto, and knows that not only must he program the historian's illusions to include the normal objects of vision and other sensations, but also to include the illusion of durationthis to contend with Danto's critique of the normal skeptic's reflection that history itself is illusory. Because we shall have later occasion to refer to Danto's discussion of duration a brief look at this critique is warranted here.
The normal, non-vat skeptic holds, as Bertrand Russell once wrote, the "logically tenable, but uninteresting" hypothesis that the "world sprang into being five minutes ago, exactly as it then was, with a population that 'remembered' a wholly unreal past." Danto's challenge to this skeptical hypothesis, which he sees as untenable but interesting, hinges on exposing its "sheer arbitrariness." Assume the world sprang into being five minutes ago; then only some statements about the past are true or false, namely those about what has transpired within the past five minutes; the others lack referenda and are "either false, or their truth or falsity cannot arise, or else questions of truth or falsity are irrelevant since these statements have the status of theoretical sentences." But nothing in the skeptic's hypothesis precludes our assuming the world sprang into existence three minutes ago, rather than the five arbitrarily posited. In this case there would be fewer genuinely past-referring predicates (for example three-minute, but not four-minute eggs). As we move the moment of creation closer and closer to the present the set of such predicates becomes increasingly smaller, with nothing to prevent the skeptic from sliding into "instantaneous skepticism." But to speak of instants is to speak of points in time, not about time or parts of time itself. In-
slants presuppose duration, just as geometric points presuppose length. It follows that instantaneous skepticism cannot ground doubts about duration, and hence must be abandoned, and with it the skeptic's hypothesis of a world that started five, or any arbitrary number of minutes ago.16
Our scientist incorporates Danto's criticism by programming the illusion of successive instants, allowing the historian in a vat to understand duration and to formulate, at least apparently, a whole range of past-referring predicates. (This diners from Putnam's evil scientist, who had obliterated memory, and is quite imaginable if we picture something like separate, successive photographic frames stimulating the afferents connected to the brain. The brain retains these "mathematically dense" frames and identifies as duration the intervals between any two or more of them.)17 It would seem possible then, within the vat, to accept Danto's refutation of normal skepticism and to pursue what appears to be normal historical inquiry, at least in "vat language." Finally, to complete the fiction, assume that all historians, as all human beings, are merely brains in a vat, that the universe consists of automatic machinery tending the brains and their nervous systems, and that, in Putnam's words, "this automatic machinery is programmed to give us all a collective hallucination, rather than a number of separate unrelated hallucinations."18 Collectively these vat historians talk with one another, criticize one another's vat work, evaluate the vat documents before them, and arrive at their conclusions about the vat past on the basis of their best vat conventions. Working within the disciplinary matrix provided by the vat, some vat historians are even artists.
Setting up the fiction in this manner immediately poses well-known puzzles about the extension of some of the terms employed. For instance, in what sense can we talk about collective hallucination as an empirical generalization? If we all hallucinate together then we have no means of establishing this fact because we have no way of stepping outside our hallucination to identify it as hallucinatory rather than non-hallucinatory. Without this possibility the very term collective hallucination falls quickly into nonsense. We recognize here the standard form of argument used against extreme skepticism. The evil genius deceives everyone by making everything illusory, including the ability to detect illusion. But then, failing this ability, how can we make sense of illusion itself?" The supposition that everything is illusory (or hallucinatory, relative, and so on) becomes either self-refuting or leads to what Mandelbaum has termed the "self-excepting
fallacy" in which such general claims are true of everything except, arbitrarily, themselves.20
We would scarcely need an elaborate science fiction merely to note that as an empirical generalization about historical inquiry constructionism either is self-refuting in the above-mentioned sense or commits the fallacy of self-exception (that is, "everything rests on a point of view, except, of course, the claim that everything rests on a point of view"). Were this the only objective, we would simply infer that in positing such a generalization we have mis-spoken ourselves and used key words in an inappropriately generalizing way.21 Something more profound lurks in the constructionist premise, a genuine antinomy of the sort Quine discusses, which goes to the heart of how we think about history.22 The self-refuting supposition at the core of constructionism is a logical one, whose own truth implies its own falsity. It takes the following form (the example is Putnam's). Consider the statement "all general statements are false"; this is a general statement, so either it is false or, if it is true, then it must be false; therefore it must be false. To assert the statement is to deny it. Such is the sort of incoherency constructionism harbors. To see this, let us look more closely at what happens when we ask whether our fiction might be true. If we were historians in a vat, could we say that we were historians in a vat?
The answer is no, and its defense will be presented in three stages. First, although too technical for extended commentary, we shall need to consider briefly Putnam's general claim that brains in a vat cannot refer to objects in the same way non-vatters can. This will yield both the form of the argument and observations about reference as an activity. Second, we shall note how reference to historical events diners markedly from reference to objects. Third, from these two points, we shall then be able to discern how Putnam's general argument can be brought to bear on historians in a vat, the constructionist premise. In so doing we need to keep in mind that by stipulation vat historians can do anything non-vat historians can. They have the same experiences, think the same thoughts, speak the same way, analyze documents, write histories, and so forth as their counterparts outside the vat. It will be shown, however, that the historians in this possible world cannot refer to past events any more than to present objects. Not being able to refer, they cannot refer to themselves as historians in a vat, thus refuting their contention.23
Putnam's general claim turns on his criticism of three characteristics generally associated with referring, each of which has served as the basis for an explanation of reference, and each of which he finds inadequate. These three characteristics are: qualitative similarity; conceptual representation; intentionality. Qualitative similarity between thoughts and extra-mental objects can be shown inadequate to explain reference, Putnam proposes, with a counterexample. Picture two computers "talking" to one another. Together they have passed the "Turing" test for consciousness and, by extension, reference; they give appropriate responses to the statements each offers. They "discuss" trees, apples, orchards, fields, and the like. Properly programmed, the two computers may continue indefinitely in this conversation without ever actually referring to any real trees, apples, orchards, fields, even though there is a similarity between the term "tree" of computer a and that of computer b (and even, one can add, a real tree). They could go on fooling one another, even were the world to disappear. Their discussion certainly resembles intelligent discourse about the objects mentioned, but such resemblance does not constitute referring to the said objects, any more than an ant's chance tracing of a caricature of Winston Churchill in the sand actually refers to Churchill. Something is missing, for our normal talk of apples and fields is "intimately connected with our non-verbal transactions with apples and fields."24 The syntactic play between the two computers lacks these nonverbal transactions; consequently, qualitative similarity cannot suffice to account for reference.
A second possibility Putnam considers is whether concepts are words (or images, sensations, or, in general, mental phenomena) that "intrinsically" represent something extra-mental. His answer, in brief, is that they cannot be so considered. In an argument derived from Wittgenstein he attacks the phenomenological notion that, when one has described the inner expression of thought, particularly in its referring capacity, one has understood it. Understanding, he counters, is an "ability." We demonstrate that ability when we use concepts, and indeed all expressions of thought, as signs in order to "produce the right phenomena in the right circumstances." This ability is social, not individual, to be sure, and can be generally stated as the capacity to act in a certain way, based on the knowledge one has. Thus merely describing the intrinsic referents of concepts cannot be adequate to account for reference.
Finally, Putnam takes up the question of whether the thinker's "intention" can be considered as the determining factor to explain reference. Here again this proves inadequate. For instance, believing there is water on the table presupposes that the word "water" actually refers to water (or to something). Were such reference not assumed the belief would be impossible. "Believing presupposes the ability to refer. And in exactly the same way, intending presupposes the ability to refer!
rial is provided by lan Hacking in Why Does Language Matter to Philosophy^. (Cambridge, Eng., 1975), and Representing and Intervening: Introductory Topics in the Philosophy of Science (Cambridge, Eng., 1983). Richard Rorty's Philosophy and the Mirror of Nature (Princeton, 1979) has emerged as a book one can argue for or against, but not without. 24. Putnam, Reason, 8-12.
Intentions are not mental events that cause words to refer: intentions (in the ordinary 'impure' sense) have reference as an integral component. To explain reference in terms of... intention would be circular."25 Therefore, he concludes, even though brains in the vat exhibit the features of qualitative similarity, conceptual representation, and intentionality in their statements, they cannot be thought of as referring in the same way people outside the vat do. Because they cannot refer, they cannot refer to themselves as brains in a vat "(even by thinking 'we are brains in a vat')."26
Historical referents, in or out of the vat, have a significantly different property from the objects involved in the cases Putnam analyzes. To begin with, historians have no possiblity of any direct acquaintance with the object of their inquiry. This is so not merely because those objects are past per se, but more importantly because they are durational. Whatever else events possess, they must possess or presuppose duration. (Of course historians examine states of affairs, institutions, and so on, in frozen-frame analyses, but such inquiry is a function of accounting for what happens over a passage of time.) An historian cannot directly observe, say, a revolution, even when he lives through it. A direct seizure of power, for instance, might not last, in which case the event would not be a revolution, but something else, a failed revolution or rebellion. Events, to generalize, are always constructed objects of inquiry. The problems of determining what counts as an event are legion, facing those inside as well as outside the vat. Moreover, as Danto has shown, there is an important, intrinsic connection between the passage of time attributed to the event and the passage of time pertaining to the historian's consideration of the event. The historian's perspective itself changes with the passage of more time, and what were previously understood as events of one sort (or even events at all) are often subsequently grasped in a different manner, as events of a different sort.27
Constructionists have often derived their claims from the observation that all events are "constructed" by the historian. But the derivation does not necessarily follow. Even a realist like Murray Murphey can contend that Napoleon is an historical "construct" in the same sort of way that theoretical entities are constructs for physicists.28 The issue is not whether constructed events ground construc-
tionism, but whether, given their perspectivist conventions, constructionists can refer to events at all. Can our historians in a vat refer to events, understood as constructed objects of inquiry, without falling into the same sort of self-refutation that faces brains in a vat as they "observe directly" the objects inhabiting their world? There are two possible lines of argument for showing that they cannot.
The first of these is to try to reduce claims about events to statements about direct observations, which are then taken to be the events' constitutive components. This move draws on the type of distinction Russell made between knowledge by acquaintance and knowledge by description. The former he characterized as a "direct cognitive relation" between a mind and an object present to it (such as sense-data, some universals, perhaps ourselves). The latter, descriptive knowledge, he argued, is composed "wholly of constituents with which we are acquainted."29 In the present context descriptive knowledge is what we can construct with the logical and linguistic tools available to us, and the claim is that our constructed events could be in effect analyzed into direct observations about instanteous components, call them event-slices.30 If correct, this strategy reduces the historian in a vat (referring to events) to a brain in a vat (referring to present objects), and Putnam's demonstration of the incoherence of "brains in a vat" prevails.
Except this strategy will not work. In order to analyze events into their composite instants, we would have to assume it sensical to talk both about a non-durational event-slice and a non-durational relation between the event-slice and its observer. But recall Danto's refutation of the normal skeptic. The refutation rests on the supposition that one cannot use instantaneous claims to ground doubts about duration because one cannot speak of instants without presupposing duration. Not only can we not base our doubts about duration on instants, but by the same token we cannot base our assertions about events on them either. If events were composed of event-slices, then we would face the same, essentially arbitrary composition in moving from the slice to the event that the skeptic encounters with the proposition that the world sprang into existence an arbitrary number of moments ago. This is merely another way of stating that while tense-lessness can be a proper characteristic of the truth-conditions of some propositions, it cannot be so of referring. Referring, rather, falls into the class of what Danto calls "project verbs," those talking about activities that lead to results.31 (Even were referring to be considered an "achievement" verb, following Ryle's division between achievements and other activities, it should be understood that achievements, although they happen at an instantfor example "winning" as opposed to "running" a race assume the activities that led to the results.)32 Project
verbs thus entail temporal structures. Accordingly, statements about reference between propositions (or their components) and events must be assumed as tensed.
The more general epistemological context for this conclusion stems from twentieth-century developments in science, best expressed by Whitehead's dictum, "there is no nature at an instant." Behind this claim lie the notions that any talk of natural objects must resolve substance into function and that function presupposes time-lapse.33 In other words, objects themselves can only be understood as events. Similarly, event-slices cannot be understood as instants of historical occurrence, but as durations of varying quantities. To say that events are composed of event-slices is thus to say only that "smaller" events make up "larger" ones. Danto summarizes these thoughts in their application to history as he concludes his long discussion of future and past contingencies with the remark that "we cannot give timeless equivalents for sentences which include temporal circumstances amongst their truth-conditions."34 Because referring assumes a temporal structure, descriptive statements about reference must include temporal circumstances among their truth-conditions. It follows at least for historians that the set of claims about direct observations between the historian as viewer and the object as event-slice turns out to be empty.
But if the strategy of reducing events to event-slices cannot work, we still need to examine the question of whether Putnam's description of referring can apply directly to events themselves. For if his arguments hold as well in the case of events, the general claim that we are all historians in a vat remains logically self-refuting. Although he eschews talk of duration in characterizing reference (his brain in a vat has no memory), Putnam does contend nonetheless that referring must be understood as an activity, one "intimately connected" with "non-verbal transactions" (between mind-dependent and mind-independent objects) and relying upon an ability of the thinker, among other things, to "produce the right phenomena in the right circumstances."35
Are historians in a vat capable of such activity? Recall, they certainly have the appearance of this activity. They are programmed with a mathematically dense
range of photographic frames (event-slices), within which they "understand" duration as intervals between any two such frames. They argue to the best case from their documents and construct an account, which then "refers" to the "event(s)" in question. Yet this "reference" does not and cannot have the aforementioned features of referring activity. For when historians in a vat "refer" to an "event" (say a revolution) they are not referring to anything external at all with which they can possess intimate connection through nonverbal transactions, with which they have the capacity to interact. They use the word "revolution," just as those outside the vat might, to designate a constructed series of occurrences, but a "revolution" in vat language could never be "about" something in the actual world, even though discourse about a "revolution" might be identical to discourse about a "real" revolution.
If this seems puzzling, imagine that the historian in the vat does have all sorts of connections between words, and between words and nonverbal inputs and outputs. These latter are provided through the efferent and afferent nerve endings that connect the brain with the computer program, which gives the historian his basic system of data, even considered as durational. But the data produced by the automatic machinery have no connection to anything outside their own program. From the fact that vat language is connected to the program through sensory inputs and outputs which do not referintrinsically or extriniscallyto anything external, one cannot infer that such language of the vat historians has the possibility of referring. Neither qualitative identity, nor conceptual representation, nor intentionality can serve to connect the program to anything external. Even the addition of "language entry rules" and "exit rules" will not help, for these rules mediate the system of sense-data and verbal (or conceptual) thought of the vat dwellers, not the basic program and external events. The rules tell these inhabitants when they are uttering a vat word correctly, when it "fits" with the programmed sensa of the vat itself. Within the vat referring activity is carried out between the program and the historian's brain.
The vat historian can thus say "a revolution occurred in France, beginning in 1789," and justify this construction by saying, further, "this statement is warranted because of the written (and other) documents before me" (each of which conveys a "smaller" event). But in justifying his construct by appeal to the "documents before me" the vat historian refers to events or documents "in the image," or to the electronic impulses that have caused such images, as Putnam suggests, not to actual events or documents. That is, there exists a close, causal connection between the use of the word "document" or "event" and the image of a document or presence of electronic impulses in the machine's program which produced the image. Reference within vat language is thus correlation of word and image, and the truth conditions for the phrase "documents before me" can be understood, correctly, as the (tensed) presence of the image or the electronic impulses when the words are uttered. But by extending the same argument the terms "vat" and "historian" refer in vat language to vats or historians in the image or perhaps the "program features" that produce the terms, not to real vats or to real historians.
The same is true with every other utterance by the historian in a vat. From which it follows that if we are historians in a vat, what we mean when we say "we are historians in a vat" is "we are historians in the vat in the image" or something comparable. But the hypothesis under question here is not that we are historians in a vat in the image, but actually historians in a vat. So if we are historians in a vat, then the sentence "we are historians in a vat" says something false (if anything at all). To sum up: if it is true that we are historians in a vat, then "we are historians in a vat" is false, so it must be false that we are historians in a vat.36
III. THE DILEMMA'S SIGNIFICANCE
The foregoing analysis has invoked a possible-world argument for the sake of exposing part of a predicament of historical knowledge, namely that the more historians insist on the primacy of their point of view as governing the product and process of historical inquiry, the closer they are led to espouse a self-contradictory supposition. At the most abstract level, the dilemma informing the above science-fiction involves an extension of Russell's paradox concerning classes or sets. The general form of the paradox is well-known. Some sets are members of themselves, for example, the set of all sets having more than five members itself has more than five members, and hence has itself as a member. Other sets do not, such as the set of revolutions, since the set is not a revolution, and thus not a member of itself. The paradoxical set is the set of all sets that are not members of themselves. Because its members are the non-self-membered sets, it qualifies as a member of itself if and only if it is not a member of itself. It both is and is not such a set.
Apply this to historical constructionism. The set of all perspective-laden conventions (history in a vat) contains itself as a member (assuming that our way of thinking according to sets is itself an historical convention, something the constructionist must maintain). This set entails as its logical contradictory the possibility of empty sets, ones containing no historical conventions. The paradox
36. This formulation paraphrases Putnam. Constructionists are among those Putnam refers to as "internalist philosophers." They deny the dilemma by arguing that, as articulated, the hypothesis of historians or brains in a vat is itself only a story, a linguistic construction, and not a possible world at all, for it presupposes an independent observer, a God's Eye point of view (or, better, a "No Eye view of truth"), one outside all frameworks, which is impossible. By contrast, externalists claim the hypothesis cannot be so easily dismissed, that reference and truth must be considered in some sense as corresponding to the "world as it is in itself," independent of the observer. This assumption makes the hypothesis possible. (Ankersmit, "Dilemma," 19-27; Putnam, Reason, 50-51). Since Kant the meaning of "world as it is in itself has been the nub of the issue, though Kant himself was not always clear on how to address it. Here I am following P. F. Strawson's reading of the Kantian "principle of significance," that there can be no legitimate or meaningful employment of concepts or ideas that does not "relate them to empirical or experiential conditions of their application." Without this "externalist" principle, the constructionist paradox reveals that in relying on constructionism, to adapt Strawson's salutary remark, "we shall not merely be saying what we do not know; we shall not really know what we are saying." P. F. Strawson, The Bounds of Sense: An Essay on Kant's Critique of Pure Reason (London, 1966), 16, and generally 15-44.
arises when we ask whether the set of these empty sets with no conventions includes itself as a member. As with the general form, the members of this set are the non-self-membered sets (the empty sets with no conventions); therefore it qualifies as a member of itself if and only if it is not, the "familiar face" of antinomy.37 The antinomy in the contradictory extends to the affirmative, and with it falls the claim that we are all historians in a vat.
The relevance of such an abstract, possible-world argument for an understanding of historical inquiry, which is conducted in the looseness of ordinary language, may be illustrated through three implications. First, we need to recognize and insist upon diffferent levels of abstraction both in historical writing and in the justification of its claims. Constructionists deny these levels, claiming as Goldstein and Ankersmit do that historical inquiry is basically self-justifying, that it can make no appeals outside its own framework of conventions for help in resolving interpretative disputes within it. But once the levels are collapsed, the set paradox demonstrates that Goldstein's above-cited charge that one cannot "reach outside" a "framework of knowing" is itself incoherent. For, just as vat historians cannot say we are all historians in a vat without refuting themselves, so neither can Goldstein claim we are all inside a framework of knowing. Failing this possibility, the charge that we cannot "reach outside" a framework becomes meaningless. A similar incoherency pervades White's tropistic account of the historian's framework.
Goldstein and other constructionists respond, in part, by contending that there are different frameworks of knowing; for White there are different tropes. In their view, to make a "meta" argument, showing a framework of frameworks as incoherent, confuses the point that all historians have some perspective-bound frame of reference with the supposition that the frame of reference must be the same for all. From the former one cannot infer the latter. Nor would anyone who examines the practice of history want to claim that all historians are bounded by a single frame of reference. But on this view if everyone has some frame of reference, each of which is self-justifying, then there are no possibilities of resolving disputes in historical inquiry when accounts from different frameworks conflict with one another. The result is incommensurability of perspectives.
It will not do to say that the scholarly community itself decides which interpretations are to be "dropped from consideration" without begging the question of why this is so.38 On what grounds does this occur (for example, change of interest, a more encompassing explanation, development of new theory, discovery of new documents)? Whenever constructionists maintain that some interpretations are "better" than others (that they yield a better account than their rivals of what the evidence obliges us to believe), then there must be some manner for arguing why this is so. Invariably this argument will be forced to address the question of the strengths and weaknesses of competing frameworks. To do this involves a criterial appeal, either to a "higher-level" framework that itself can be
defended as more persuasive, or to some "cognitive" criteria traversing frameworks, by virtue of which knowledge claims within each of them can be distinguished, some assertions being deemed more warranted than others. In either case the constructionist is forced to step outside his own framework of knowing to an appellate level of resolution and clarification. To hold that one cannot do this is not only incoherent in itself, it overlooks the practical ways in which historians in fact do try to defend their claims at different levels of abstract argument. Thus the practical dilemma the constructionist cannot escape is the following: either all frameworks are of equal epistemic value in governing historical inquiry, in which case historical knowledge is significantly reduced to irresoluble opinion and preference, or some criteria may be evoked (either as a higher level framework of knowing or as cognitive conditions which must belong to all frameworks), in which case their claim of the self-justifiability of a framework of knowing falls.39
Second, applied to history the set paradox reveals not only the need to differentiate levels of abstraction when considering historical inquiry, but also the need to recognize that at the most abstract level of consideration we encounter inescapable incoherence. In this regard constructionists are quite correct, ironically, to argue that one cannot produce one set of epistemological "foundations" that will serve to legitimize historical claims. In other words, the notion of an ideal category that can be adduced to ground historical knowledge is not only empty of content; it is formally self-contradictory. This point, I would suggest in passing, applies as well to W. H. Walsh's effort to lend constructionism, in particular Gold-stein's version, the helping hand of Kant. Building on Kant, Walsh distinguishes between the constituents of particular historical judgments made by historians (individually or collectively) and "judgment proper," the subject of which is the impersonal "Kantian unity of apperception." Walsh introduces judgment proper to account for how errors in particular judgments may be detected and corrected as historians move toward ideal objectivity, understood as "not what is acknowledged or concluded to by any actual thinkers, but what would be acknowledged or concluded to by thought in its ideal form."40 He thus answers the question of how constructionists might resolve disputes with the idealist appeal to a category of universal objectivity. But if the foregoing analysis has been correct, not only would this universal category be empty, in the sense that we could never have adequate grounds to claim whether our particular judgements accorded with it or not, it would be self-refuting as well. An ideal framework of judgment proper is a set of all sets with no members.
Third, we need to recognize that in some important sense reference must be completed in the world, regardless of the problems in characterizing this sense philosophically. As Putnam's paraphrase of Mill suggests, the "substance itself
in this case our actual durational world of the past in its continuity with the present and future"completes the job of fixing" reference.41 Furthermore, consideration of this requirement bears directly on the manner in which we can resolve disputes in interpretation, or demonstrate their irresolubility, what we mean by historical objectivity. This is no easy task. In recent years, distinguishing science from pseudo-science has proven far more difficult than once imagined, and the line between history and pseudo-history remains murkier still. Yet from the premise, or fact if one will, of a blurred division between categories of knowledge, we lapse willy-nilly into incoherence when we infer either that there is no such division or that we ought not seek its articulation. In sum, the paradox of historical constructionism reveals not only the negative constraint against the self-defeating abandonment of objectivity, but also the positive challenge of fixing objective historical reference in the past.