Is Historical Knowledge Philosophically Interesting?


Review of The Philosophical Structure of Historical Explanationby Paul Roth

Evanston, IL: Northwestern University Press, 2020


For those of us invested in defending the epistemic value of history, especially when facing arguments made by skeptical analytical philosophers, Paul Roth’s work will already be well-known. So far, it has arrived in gleaming, landmark papers—like “The Pasts” (2012)—but now, finally, as a result of the publication of The Philosophical Structure of Historical Explanation (2020), it can be approached as a whole. And approaching it as a whole is revealing: both about the positive picture Roth paints of the importance of narrative to historical knowledge (and, relatedly, irrealism about the past), and about his critique of analytical skepticism as to the capacity of narrative to explain and thus to be subject to rational evaluation (and thereby, on some views, to be philosophically interesting).  

In this review, I focus on the first part of Roth’s book—namely, the positive account, which Roth develops in chapters 1 to 4, of historical knowledge, namely as narrative explanations of narrative sentences. It is an important part of Roth’s book to turn then (as he does in chapters 5 to 7) to what version of naturalism is compatible with this account of historical knowledge, which also involves arguing (in large part by reference to Thomas Kuhn) that what counts for science, at any particular time, is itself only recountable by history. Arguably, however, the core argument is in the first part: namely, that historical knowledge consists in giving narrative explanations of narrative sentences. I therefore focus on that. In conclusion, I offer some thoughts as to how this argument may be significant for legal theory.   

Why might a philosopher think historical knowledge has no real explanatory value, is not subject to rational evaluation, and therefore is philosophically uninteresting? As Roth recounts it, the story begins with Carl Gustav Hempel’s influential paper, “The Function of General Laws in History” (1942). According to Hempel, historians offer, at best, “explanation sketches.” What is missing from history are (as found in scientific explanations) covering laws or law-like explanations. In the absence of such laws, there is no way to evaluate the inferences that historians make. As a result, history cannot be rationally evaluated, does not belong to the legitimate sciences, and is philosophically uninteresting (on a certain understanding, of course, of what philosophy is).  

One of the virtues of Roth’s book is that it clearly articulates the presuppositions that such an attack on history evinces—these being presuppositions that are often not made explicit in these attacks. One key presupposition is metaphysical: the commitment to a past that is simply there, waiting for a historian, or many historians, to describe it “as it actually was.” The accumulation of historical knowledge, on this account, is both possible and desirable: the more historians describe what actually happened, the more we know about the past. It is like accumulating maps of the same, existing terrain: the maps may differ in “scale, scope and detail” (26), and the more of such maps we have, the more we know of the terrain.  

Roth has a simple but powerful riposte to such views, and it builds, as he acknowledges, on the work of Arthur Danto and Louis Mink. The key is to see that there cannot be any sense in which the past is waiting for us to describe it because what we know about what occurred in the past changes over time: some of what we can describe, and find interesting and relevant to describe, about the past only emerges later. For example, it is only once, in 1981, Ronald Reagan becomes President of the United States, that we can say that a President starred in the pre-war film, Love is in the Air (1937). In fact, it is only after the Second World War had begun that we could describe that film as a “pre-war” film. Both of these “narrative sentences,” as Danto called them, might be said to be important and relevant aspects of our contemporary historical knowledge. But they were not known, or knowable, when they occurred. Therefore, there is no sense in which there is a past—the past—waiting for historians to describe it. Rather, historical knowledge moves with the times: the writing of history is always dynamic, and more like a relation with many pasts. It is thereby also an aesthetic activity—e.g., of framing and selecting what is important and relevant—and therefore also a deeply ethical and political act.  

The past, then, is not fixed. Indeed, there is no such thing as “the past.” Another way to see to see this is to observe, as Roth does, that there are no “ideal events,” which the “Ideal Chronicler” (another Danto invention) would necessarily describe. There are no “basic units” of the past. “Events,” as Roth says, “may be sliced thick or thin; a glance may be identified as an isolated event or as an instance in an event. What the unit-event is depends on the telling of it” (29). We make pasts by telling and re-telling them: if you like, there are as many pasts as there are ways of making them, which themselves continually move, as waves do. The past, if one insists on the singular, is always necessarily incomplete. As Roth puts it:  

Events simpliciter cannot be shown to exist; they are not known to be of nature’s making rather than of ours. Events exist only by proxy. This is why one cannot presume that there are any ideal events for our erstwhile chronicler to chronicle; knowledge of events is restricted to happenings isolated under descriptions provided by interested parties. (30) 

Notice how, on this view of history, one thing that emerges as both significant and fascinating about historical practice are the disagreements between historians as to what count as events, as well as events worth describing. Indeed, on this view, there are many different kinds or degrees of historical disagreement: e.g., first-level disagreements as to what to treat as a unit of description (e.g., whether to talk at all of “events” or “actions”); second-level disagreements as to whether to describe something as an event (or action); and third-level disagreements, i.e., on just how to characterize and describe the events (or actions) in question, even when there is second-level agreement on what the relevant events (or actions) are. The history of history, on this view, becomes a fascinating aesthetic, ethical, and political battle: the work of historians now appears as a series of emotionally charged disputes about what is relevant and important to describe, and how.

To read history, then, becomes an adventure in reading different kinds of disagreements over many—indeed, infinitely many—pasts. Where those disagreements are third-level ones, as above, reading history is also reading about what was agreed to (e.g., that it was thought important to describe “events” and “actions,” and what particular events were described as events, and what particular actions were picked out as actions), and thus also considering the reasons for those agreements at any particular time and place (ethical, political, aesthetic, affective, material). This may seem very obvious to historians, but, I wager, it is not the way that many philosophers approach the history of history.  

Of course, all this may be grist for the philosopher’s mill in another sense. Does this account of history, philosophers might ask, not sweep under the carpet any claims to historical knowledge? Does not history, on this view, just become entirely subjective and relativist? Has not the defender of historical knowledge here thrown out the baby with the bathwater—defeating, perhaps, the great enemy of Universal History, and the fixity of the past, but only at the cost of descending into an endless array of monadic, entirely contingent, snapshots of a particular, non-repeatable pasts? 

In fact, the reference above to contingent and conventional first- and second-level agreements about what counts as events and/or actions already shows us that we have at least one kind of answer to that question: there are, within certain communities at certain times and places, conventions as to what are the units of historical knowledge, although these conventions are contingent and also subject to contestation and change. This already shows that historical knowledge is unlikely to be—or simply cannot be—entirely monadic, subjectivist, and relativist. As Ludwig Wittgenstein taught us, none of us, historians included, can speak an entirely private language. None of us is ever entirely in control of what we can see and tell about the past: we are social creatures through and through, and our modes of seeing and telling are shot through with more convention (and interested convention, with much at stake, including economically) than we can ever be reflexive about. But that is only one possible, and incomplete, answer to the above questions.  

Part of what is significant, and delightful, about Roth’s book is that it is not defensive about historical knowledge: rather, it is upbeat and positive. There are no apologies for history here. Rather, there is a clear statement as to why history is valuable epistemically, and thus also philosophically interesting (though in ways that may also—or, more strongly, should also—transform certain views about what philosophy can find interesting). The key concept for Roth is narrative, and in particular, how narratives explain.  

Recall that “narrative sentences” were descriptions that were only retrospectively knowable, e.g., that a President of the United States starred in the pre-war film, Love is in the Air. Narrative explanations, Roth claims, are uniquely suited to explaining these narrative sentences. A narrative explanation “will be a presentation of a temporal series that answers why” the narrative sentence “turns out to be as it is” (68). There is no way of providing a temporal series in non-narrative form, e.g., by covering laws. After all, there was no way of predicting that the person who starred in Love is in the Air would go on to become President of the United States. The narrative sentence—that a President starred in a pre-war film—only becomes possible, and knowable, retrospectively. Narrative explanations allow us to explain those narrative sentences: taking the end (that Reagan is President) they connect that end back to what it constitutes as a beginning of that end (that the actor who starred in Love is in the Air is a President). Narrative explanations, if you like, are frames without which there would be no narrative sentences, or if they were there, they would be floating aimlessly in some unrecognizable ether. Narrative explanations, which of course also depend on conventions about what counts as telling a story—conventions theorized by many, perhaps most emblematically, within historiography, by Hayden White—are what make history possible, and what constitute historical knowledge.   

Roth turns to a number of different historical works in fleshing this out: they include Fernand Braudel’s The Mediterranean and the Mediterranean World in the Age of Philip II (1976); Michael Friedman’s A Parting of the Ways; and Kuhn’s The Structure of Scientific Revolution (1985). But the example he turns to most is Raul Hilberg’s The Destruction of the European Jews (1961). Roth does so in part because of the “paradigmatic status of Hilberg’s history,” but also because “of the role that debates about the Holocaust have played in discussions of constructivist positions in historiography” (140).  

Roth finds the following narrative sentence in Hilberg’s book: “The destruction of the Jews was thus no accident. When in the early days of 1933 the first [German] civil servant wrote the first definition of ‘non-Aryan’ into a civil service ordinance, the fate of European Jewry was sealed.” In Roth’s summary, this narrative sentence becomes: “the Holocaust began in 1933” (72). On Roth’s view of historical knowledge, Hilberg’s book, in turn, can be conceived as a narrative explanation, or a collection of narrative explanations, that would explain that narrative sentence. For example, Hilberg’s narrative explanation places an emphasis on the administrative and bureaucratic processes of the Nazis. It does that rather than, say, presenting the events and actions of this period as one great pogrom, or a descent into barbarism (which would constitute alternative narrative explanations, or perhaps simply be less convincing narrative explanations of that narrative sentence). Hilberg, for instance, does not focus on the victims and their experiences, but on the Nazi machinery (that language is, in itself, significant) of destruction, and how it came to be mobilized. His narrative explanation is also dependent on certain conventions, e.g., he employs statistical tables, and writes in an austere style.

The entire package—the book that is The Destruction of the European Jews—thus offers a narrative explanation for the above narrative sentence, namely that the Holocaust began in 1933. For Roth, crucially, this explanation cannot be provided in any other form: it both constitutes and explains the history that Hilberg offers, i.e., that only-retrospectively-knowable narrative sentence mentioned above. As Roth puts it: 

The Destruction of the European Jews fits the mold of a narrative explanation…Its conclusion can be stated as a narrative sentence, one that the text explains by providing a beginning-middle-end structure that presents a story line detailing the causes of that event, but where “causes” can be identified only by offering specific steps in any extended developmental sequence. No laws underwrite this sequencing…[N]o functional distinction exists between describing that sequence and justifying causal links. The event explained…moreover cannot be detached from the narrative that presents it. (74-75) 

Narrative explanations explain narrative sentences. These narrative sentences cannot be explained in any other way. Therefore, history operates with essentially narrative explanations.  

Very well, you might say, but this still leaves some questions open. I mentioned earlier that there were different kinds of objections to historical knowledge made by skeptical analytical philosophers. One of those was metaphysical—and that objection is answered by Roth’s showing that its presuppositions are untenable, as well as by his positive account of how narratives explain. There is, however, also a methodological objection. Thus, the skeptic might still ask: how can narrative explanations be evaluated?

The answer, it seems, hinges on the relationship between narrative sentences and narrative explanations. Recall, again, that narrative explanations are explanations of narrative sentences. Thus, it would appear, in order to evaluate narrative explanations—and, similarly, in order to compare different possible narrative explanations—we must first have agreement as to the narrative sentence that is being sought to be explained. Once we have that, we can evaluate whether the narrative explanation offered is a good explanation, or which of the alternative narrative explanations is the better one. One consequence of setting evaluation up this way is that it may be that there is no general “test” as to what constitutes a good or better narrative explanation: it will depend on what the narrative sentence is, which might then offer up different kinds of relations between it and possible, plausible, and desirable narrative explanations. Further, these relations—or any such particular “test”—will not be evaluable on a single, simple basis. Instead, any such evaluation is likely to involve complex aesthetic, ethical, and political considerations. For instance, Roth recalls, and endorses, Hayden White’s insistent reminder that “any decision on how to write history will represent a moral choice on the part of the historian” (142). Roth discusses these difficulties and challenges in other essays (e.g., see his “Hearts of Darkness”).  

It is, it must be said, somewhat of a pity that a more extensive discussion of these issues (as to evaluation and its complexities) does not feature in the book, for it does seem interesting to ask: what makes a narrative explanation an alternative explanation of the same narrative sentence? It seems that there is a tension between two poles here. On the one hand, there is, on this view, a tight (at times, Roth appears to suggest a constitutive) relation between a narrative explanation and a narrative sentence (one cannot “detach,” as Roth puts it, the narrative explanation from the narrative sentence); and, on the other hand, there is the possibility of evaluating narrative explanations, which—at least in my reading—implies some level of choice between different narrative explanations of the same narrative sentence. Roth needs that relation between a narrative explanation and a narrative sentence to be tight—otherwise, room is possibly made for arguing that narrative explanations could detach from their object of explanation (namely, narrative sentences), and that perhaps other kinds of explanations might be offered of the objects of historical knowledge.

However, if evaluation is to be possible, that relation cannot be so tight as to leave no room for alternative narrative explanations. If anyone can walk that tightrope, Roth can. Still, it would have helped to include some more discussion of how that might be done. It is, once again, understandable, especially given Roth’s version of naturalism, that he withholds from giving any general evaluative criteria—for Roth, there can be no a priori way of deciding what those criteria (of rationality, say) ought to be. Perhaps the best we can do, he says in a few brief pages at the end, is appeal to the explanatory strategy of “inference to the best explanation” (see 144-48). In general, with evaluative matters, we must proceed case-by-case. The reader will not find an examination of such a case in the book—for that, one needs (happily) to read more of Roth’s essays.  

I mentioned earlier that Roth’s is a positive, upbeat account of historical knowledge, rather than a mere critique of analytical philosophical skepticism about history, or an apologia for historical knowledge. A large part of that is down to what Roth calls his “historical irrealism.” This emerges clearly in his chapter 3, which is a revised version of the above-mentioned paper, “The Pasts.” Here, Roth first articulates the significance of Ian Hacking’s view of “styles of reasoning” and related kinds, and then expands it (in part by reference to Nelson Goodman as well as Leon Goldstein) considerably, generalizing it to offer precisely a thesis about the irrealism of the past. This irrealism is a key reason why history is so philosophically interesting. At stake, as Roth himself puts it, is how we think and constitute “ourselves and others, as agents and as beings with a past” (37).   

Part of Hacking’s contribution, as Roth describes it, is his category of “interactive kinds”: these are kinds that are responded to by the things to which these classifications are applied. The clearest example is that of people: people respond to how they are classified, sometimes very strongly (e.g., think of the response of persons to being classified as either male or female, or either as heterosexual or homosexual). The same point can be given temporal orientations: when a new kind (or classification) is offered to persons, some may well respond by adopting that kind, and coming to conform to it (e.g., classified as having multiple personality syndrome, some persons identify with it and come to live it), it thereby becoming a sort of self-fulfilling prophecy; similarly, when a new kind is created, one can look back upon the actions of persons in the past, and come to characterize those actions in entirely new ways.  

It is this past-oriented capacity of new kinds that animates Roth. If we take actions to be so only under certain descriptions (as famously argued for by G.E.M. Anscombe), then the provision of new descriptions (in the form of new kinds, such as multiple personality disorder, or being transsexual), then new actions in the past are also created. The past, in that sense, is literally made with descriptions (or kinds), or remade with the help of new descriptions (or kinds). For Roth, the conclusion to be drawn is that the past is “indeterminate,” because “changing classifications…literally alter[] the actions ascribed and so reconfigure[] accounts of causes and consequences of such actions” (46). There is an ethical and political imperative here too: we must, says Roth, drawing on not only Hacking but also Michel Foucault, “exercise constant vigilance over invitations about how to think about ourselves” (46), and not only because such invitations may become self-fulfilling prophecies, but also because they re-make our past, with all the power and consequence that that entails.  

It is at this point that Roth generalizes Hacking to offer his “irrealism” about the past. This irrealism “eschews the metaphysics of both realism and anti-realism” (49). Each of these has its own problems: realism requires the positing of a fixed past, waiting to be found (rather than made) by a historian (i.e., the myth of Universal History, which Roth discredits, as above), and anti-realism makes historical knowledge too easy (or empty), because “nothing on this account appears to remain by which to drive a wedge between representations of the past and its putative object—‘the Past’” (54). The trick is to maintain a wedge, but to characterize it as contingent and not necessary or conceptual. This is what irrealism provides, yielding as it does what Roth refers to as “a multiplicity of pasts”: 

Irrealism denies an assumption that both realism and antirealism turn out to share: an imagined view from nowhere…Given alternative modes for structuring what happens, changes in descriptions can alter relations among events imputed to a past, and so how a past thus structured impacts what becomes possible going forward. A plurality of pasts results because constituting a past depends to some degree on socially mediated negotiations of a fit between descriptions and experience. Even what we take to distinguish what can change and what cannot itself depends on the descriptions deployed. Unless for reasons now unknown there ceases to be a possibility of descriptive change or reclassification, human histories will continue to reveal a multiplicity of pasts. (64) 

For Roth, it is conventionality all the way down (indeed, in all directions), and only irrealism manages to hold on to this. One consequence is that there is no meta-discourse – e.g., a philosophical one – that has a monopoly on creating the descriptions (or kinds, classifications) that any other discourse or discipline can deploy. Put another way, the creation of descriptions, used to narrativize the past, is just as philosophically interesting as philosophy itself, and it has just as much capacity to change philosophy, as innovations in philosophical language have the capacity to change historical practice. With this also comes responsibility: historians have as much responsibility for the language they use to make and re-make our past, and equally our present and future, as philosophers do. This responsibility is not just, or even mainly, fancy metaphysics: how we constitute ourselves and our pasts and our futures is the very substance of ethics and politics. That is both the gift and the burden of recognizing the epistemic value of historical knowledge.  

In concluding, let me briefly observe how Roth’s account might be important for legal theory. For a start, surely it suggests that legal theorists should read more legal (and also not necessarily just legal) history. After all, as above, historical knowledge is just as generative of invitations to think about ourselves as theoretical knowledge is. It is, one might say, a particular mode of theorizing: one that deserves attention, but also calls for critical scrutiny. The candidates here are countless: the most recent work of history I have been reading is Renisa Mawani’s Across Oceans of Law (2018). In it, Mawani critiques the effects and the stranglehold—analytical, ethical, and political alike—that a distinction between the free, lawless sea and lawful, possessable land has had. Mawani shows, through a narrative of the voyages taken by the Komagata Maru and its charterer, Gurdit Singh, how such a distinction is unhelpful, and does not allow us to notice many law-filled encounters on the seas, as well as many ways in which events and actions on sea and land relate to and affect each other (e.g., the ideological uses made, both by colonialists and anti-colonialists, of appeals to Indigenous peoples, with resulting, and devastating, effects on both Indians and Indigenous peoples).

Mawani creates new ways of thinking about ourselves, and draws on them to re-make the past, now gleaming anew: oceans, now seen to be connected by currents, are spaces of fierce political contestation, creating new legal regimes (on land as much as on the sea), and mirrors of a kind to stories of citizenship, race, and empire never before told. Further, the oceans, so conceived, challenge habits of conceiving law as a unity – a habit itself possibly generated by the above distinction between land and sea – and replace it, or at least sideline such neat (territorial) unities, in favor of the ever-changing pluralities and contests of jurisdiction (arguably more visible on the seas). It is as if Mawani asks: how would legal theory look like if written from the ocean rather than from land, or from a vista that does not accept a rigid distinction between ocean and land? Legal theorists ignore such historical work at their peril.  

But Roth’s philosophy of history is possibly even more significant than that. For instance: how might Roth’s irrealism, and thus his critique of the myth of a fixed past, speak to debates over legal and constitutional reasoning? How does, say, originalism measure up against this philosophy of history? What import might Roth’s account of making and re-making the past have for how legal theorists understand precedential reasoning? These are questions beyond the scope of this essay, but they raise the prospect of a revival of historical jurisprudence, which is informed by, and contributes to, contemporary historiography and debates in the philosophy of history. After all, legal practices—such as precedential reasoning precisely—themselves might offer their own kind of historiography, if not philosophy of history, that might inform present debates. For instance, we might ask: does not common law reasoning put into practice the kind of historical irrealism that Roth discovers philosophically?

A final possible link with contemporary debates would lie at the intersection of law and literary theory. Given his emphasis on narrative as the touchstone of historical knowledge, it is a little surprising not to see more appeal to narrative theory in Roth’s book. (Admittedly, though, Roth has a good excuse: as he says, his book does not aim too “contribute to any theory of narrative as a literary form. Rather, the argument has been to establish that narrative has a place as well within epistemology and the theory of explanation” [140].) Narrativizing has, of course, its own complex histories, to which, once more, legal practices have a lot to offer: after all, one can see law as a more or less connected series of spectacles of narrative jousting. Law, like literature, has experimented with different modes of narrative, and thereby also made and re-made the past—or many pasts. These innovations are shot through with pragmatism, nowhere more evidently than in the law, though, as James Phelan has argued with his rhetorical approach to narrative, this may be the best approach to narrative in general.  

All this, and more, remains to be unearthed but has also been opened up by Roth’s great new book. One thing I think we can say having read Roth: there is hardly a doubt that history is philosophically interesting, and if it is thought not, so much the worse for philosophy.  


Posted on 2 April 2020

MAKSYMILIAN DEL MAR is Professor of Legal Theory at Queen Mary University of London.