Past Facts and the Nature of History

We defend a realist account of history: past facts are discoveries not creations. We show how ‘moderate’ realists, who admit the critical role of perspective, while insisting on history’s metaphysical independence from historians, can accommodate Paul Roth’s arguments in favor of irrealism. Moreover, our position is consistent with a dynamic past: as history unfurls past events gain new properties. Realism is necessary, we argue, to capture substantive disputes within history. It also grounds history’s reflexivity: the point of the continual re-examination of history (and history’s history!) turns in part on there being mind-independent past facts to be had.

papers and a recent monograph.3 Roth draws on the insights of past philosophers, especially Danto and Mink, to argue that historical events metaphysically depend upon the actions of historians: A narrative traces a path of development, a path not defined or marked by any known laws or the like. The event emerges as an event only because our interests call it into being; events so constituted do not embody some natural kind.4 For the irrealist, past events depend metaphysically on historians, they 'call them into being' . Realists, by contrast, hold that narrative events are sometimes metaphysically independent of the actions of historians: past factseven dynamic past facts -and the relationships between past facts, can hold regardless of what historians have said about them. Although we defend realism, we share Roth's worries about naïve forms of it. Roth's exemplar realist is Mandelbaum.5 As Roth puts it: [for Mandelbaum] Historical pictures are successively filled in by collecting more evidence concerning the events of interest. The picture is always partial, but what history provides is an ever clearer picture of things as they actually were … The work of a historian, on Mandelbaum's conception, is more like that of a scribe than an author.6 This is in a relevant sense similar to the view developed by Hempel7 that casts historical narratives as "best explanatory sketches," wherein greater historical detail (and appeal to natural law) makes for better explanations. The job of the historian qua explainer is to continually "fill out" the sketch.
Following Roth, historians are not like scribes: history is too complex, too multi-faceted, and our interests too multitudinous to think of the historian as simply 'copying out' accounts from some pre-determined history. On our view, however, his is a point about the epistemic practices and difficulty of doing history, not the metaphysical status of past facts. Although we don't think he establishes his metaphysical thesis, aspects of Roth's views are friendly to our position. First, the past is dynamic: a class of past facts change and emerge as history moves forwards. Second, it is in capturing this dynamic process that narrative forms and structures become necessary for historical explanation. 8 We won't deny such claims here. 9 We'll begin (section 2) by arguing that the moderate realist need not worry about Roth's arguments, before turning to our positive project: reconciling moderate realism with dynamic facts. Beginning with an example from biology (section 3), we'll demonstrate how the past can be dynamic without mediation or intervention from historians. We'll then provide a positive argument in favor of forms of moderate realism (section 4). We'll claim that (1) historians have substantive disagreements which turn in part on past facts, and that irrealists cannot accommodate those disagreements; (2) in some circumstances, historians do influence temporally dynamic facts, but do not do so in a privileged or special way; it is simply that the practices of historians constitute one way in which historical events can take on new properties. Moreover, we'll argue, moderate realism can capture the motivations behind the reflexivity of historical practice.
A note on terminology. Throughout we'll refer to past 'facts' , 'properties' and 'events' . At base, we'll take 'events' as a catch-all term for entities, processes, and so forth. Events have properties; properties are what we might ascribe to them. We'll take a 'fact' to be a true description, sentence, or proposition about a property ascribed to an event. We don't think this way of speaking presupposes realism, anti-realism or irrealism. On Roth's view, there are facts about properties, but such facts hold only under historians' descriptions of events. That is, the historian categorizes a past event, and as such, creates new properties, and statements about them can only be said to be true or false relative to the categories the historian themself created. By contrast, on our view, the events and properties are (except under circumstances we'll clarify) independent of historians. Further, although (on some versions of minimal realism) 8 Roth, "Narrative explanations: The case of history"; P.A. Roth, "Essentially narrative explanations", Studies in History and Philosophy of Science Part A, 62 (2017), 42-50. 9 The reader should note that Roth's conception of narrative has it that historical narrative is both descriptive and explanatory; that is, description within a narrative and narrative explanation are essentially the same kind of activity. This idea is controversial, and has been criticized by F. Dewulf, "Paul A. Roth, The Philosophical Structure of Historical Explanation", OEconomia. History, Methodology, Philosophy, (10-2) (2020), 363-367. Our argument requires no stake in this.
facts might be relative to a description or sentence, their being facts turns critically on events and their properties. And on other versions, facts can be understood as propositions whose truth is entirely indifferent to token descriptions or sentences, uttered by historians or otherwise.

Irrealism
Roth's arguments begin with three points about narrative explanation, intended to underwrite his irrealism. Two, 'non-standardization' and 'nonaggregativity' are, we think, red herrings. That is, Roth's arguments do not distinguish irrealism from moderate realism (although they are telling against naïve realism). His third point concerning indetatchability, however, is critically important for understanding the dynamism of the past and the nature of narrative. But this still does not save irrealism from the realist challenges we develop in later sections.

2.1
Two Red Herrings Two of Roth's arguments for irrealism should not trouble moderate realists. Our aim here is not to argue in favor of realism (that is for later sections), but rather to demonstrate that many non-naïve minimal realist positions can accommodate his position (hence the arguments being 'red herrings').
Roth argues that narrative events are non-standardized. This is to say that they do not come 'typed' , treated as instances of regularities: they are not lawlike. Roth's arguments for the non-standardization of historical explananda do not threaten moderate realism. First, it is not obvious that all such explananda are non-standardized. Historians do craft at least some narratives around something like historical regularities.10 This is, for instance, the strategy of Peter Turchin11 in his work on the dynamics of empires. He points to regular causal processes arising from imperial frontiers as central to explaining these cyclical processes. This might fall short of "lawfulness," but it is nonetheless in some sense a "regularity." This should be familiar to philosophers of science. There is a long tradition of examining 'law-like regularities' , ceteris paribus laws and the like: a regularity need not be necessary for it to play legitimate roles 10 Sterelny, "Contingency and History"; A. Currie and K. Sterelny, "In defence of story-telling", Studies in History and Philosophy of Science Part A, 62 (2017), 14-21. 11 P. Turchin, "Population dynamics and internal warfare: a reconsideration", Social Evolution & History, 5(2) (2006). in science,12 and so the same goes for history.13 Second, even if historical events are non-standard, they're plausibly linked to processes that are, and this fact can help to underwrite the realist's explanatory aspirations. That some shard of ancient pottery is found in some location in a particular sedimentary formation might be non-standard, but archaeologists still make use of nearby regular, standardized causal processes to bolster their explanations.14 In short, moderate realists might agree that historical events are not always 'typed' , but this doesn't require conceding that those events are metaphysically dependent on historians.
Onto the second red herring. Drawing on Mink,15 Roth emphasizes the non-aggregativity of historical narratives. Different narratives often characterize events in non-consistent ways. In virtue of this, their accounts cannot be woven into a unified narrative. You can't just capture the past in terms of a set of narratives joined via conjunction without inconsistencies. This is because past events are sensitive to description; how the events are characterized (what they 'are' if you want) are partly determined by our explanatory interests in them.16 The 'same' event might, for one historian, count as the beginning of a new movement; for another it could be the middle of an unfolding process. The argument here seems to be that realism requires that events be strung together in a maximally coherent way: that there is one 'grand narrative' to be had. But realism need not assume any such thing.
Philosophers of science have long considered the critical role of distortion and omission -abstraction and idealization -in scientific work, particularly models.17 Idealization is often presented as a challenge for scientific realism: if science aims at truth, why is it that so many scientific representations 12 S. Schiffer, "Ceteris paribus laws", Mind, 100 (1)  Mink, "Narrative form as a cognitive instrument." 16 Roth might deny that 'sensitivity to description' is the right reading here, after all, he thinks that the events are metaphysically dependent on description! Our reply is simply that sensitivity to description captures just as well the phenomenon he is interested in, and moreover doesn't beg the question against realism or irrealism, as the latter reading appears to. 17 M. Weisberg, "Three kinds of idealization", The Journal of Philosophy, 104 (12)  contain untruths? Answers are varied and typically pluralist. Some idealizations function as something like approximations: they are true enough for our purposes. Others mitigate a lack of understanding or a lack of computational power. Others isolate causal dynamics thought to have a privileged explanatory role and aim for unification across systems with those dynamics. In light of this, some philosophers adopt what might appear to be anti-realist positions: the view that models are 'convenient fictions'18 or that they primarily aim for understanding rather than truth.19 But crucially, such views are not anti-realist in the relevant way. All claim an important relationship between the worldly systems the models aim to represent and the model's success. Faced with scientific idealizations, philosophers of science do not then claim that facts metaphysically depend upon scientific representations. If Roth is right about the consequences of non-aggregativity, this is surprising because differently idealized models seem of apiece with historical narratives: their targets are sensitive to description, they capture the events in strikingly different, non-consistent ways. So idealized models, like narratives, do not aggregate into a unified system. Philosophers of science are not irrealists about scientific facts in the face of non-aggregativity because it simply doesn't follow from our making idealizations -even ineliminable idealizations -that the facts are not 'out there' to be had. There is an important sense in which non-aggregativity does not actually lead to inconsistency. Non-aggregativity is due to events being sensitive to description, that is, the events are not treated as 'bare' , but as events-qua-some perspective. It is perfectly consistent to capture some event qua perspective a, and the 'same' event qua perspective b, even if event-qua-a and event-qua-b would be inconsistent if treated as either perspectiveless or from the same perspective. It does follow that there is no consistent, single, non-perspectival 'god's-eye-view' to be had.20 The full picture will be irredeemably pluralistic. But pluralism is not in conflict with realism insofar as it amounts to the denial of those events only existing because of scientists taking those perspectives.21 Why will become clearer below, when we consider substantive, cross-perspective To put things in another way, realists need not (and should not) be committed to the possibility of a kind of 'ideal chronicle' which in a unified, non-perspectival way, captures all the past facts.
debates between historians. The inference from an event being sensitive to description to that event being brought into being by that description is invalid. Relatedly, Roth argues that events themselves do not come pre-carved into categories, thus it is our categorization practices which bring those categoriesand thus the events themselves -into existence. "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."22 This is a non-sequitur: it doesn't follow from our not being able to show that events exist simpliciter to them only existing via proxy, that is, their only existing because historians have done their categorization work. At most, it speaks in favor of agnostism about 'events simpliciter' . However, as we've already seen, realists need not be committed to the existence of events simpliciter, nor of non-perspectival, privileged 'god's-eyeview' events. One way, then, to gloss a realist response to Roth's argument for non-aggregativity is to say that nature (and a fortiori history) is structured or patterned. A particular phenomenological pattern corresponds, roughly, to one potential carving of the world's natural (or political, economic, etc.) history. Some of the ways that historians carve the world are more similar to true patterns of events than others, and this fact is what underwrites the substantiveness of historical debate (as we will argue below). The moderate realist, then, can deny that these potential carvings are the inventions of historians, they are instead the events, processes, and patterns that historians seek to find and understand.

2.2
Indetachability So, neither narrative events failing to aggregate nor their being nonstandardized provide routes to irrealism. At best, these arguments hold against the kind of realist who thinks there is a single privileged description of the past. But moderate realists commit to no such thing. Roth's stronger argument comes from what he calls non-detachability. This is expressed both as a feature of narratives and the nature of past facts. Let's begin with Danto's notion of a narrative sentence.23 A narrative sentence defines some past fact in terms of some later fact. For instance: The 40th President of the United States hosted General Electric Theatre in the 1960's.
The structure of a narrative sentence is indetachable in the following sense. The earlier fact (the subject hosted General Electric Theatre) is defined in terms of later facts (the subject's being elected president in 1980). As such, one cannot make sense of the earlier event understood in those terms without 22 Roth, The Philosophical Structure of Historical Explanation, 30. 23 A.C. Danto, "Narrative sentences", History and Theory, 2(2) (1962), 146-79.
its being related to that later fact. This matters for Roth (and Danto) in part because it tells us something about the structure of narrative explanations. Where for some explanations the explanans and explananda may be logically decoupled, thus allowing for the explanation to be represented in terms of a deductive argument, this is impossible for a narrative argument as the content of the explanans and the content of the explananda overlap. If represented as an argument, narratives would be circular. This, for Roth, marks a syntactic difference between 'scientific' explanations and historical narratives (though it is worth noting that explanatory relations, as understood in much contemporary work on scientific explanation, need not be in any sense deductive).24 We're happy to go along with these points about narratives for the purposes of this paper. However, there are also metaphysical conclusions drawn. The indetachability of narrative sentences derives from the connection between the content of a past event and a later event. The fact that a T.V. host became president is only true in light of the electoral process that unfolded at some later time (namely, 1980). This is not a mere feature of linguistic practice: in a very real sense it wasn't true that the 40th President hosted General Electric Theatre until he won the later election. That is, the truth of the sentence, which refers to the 1960s, depended on later events occurring as they did. Roth captures this in terms of historical practice: it is in virtue of historian's categorization practices that the events are the events they were. As he puts it: [narrative sentences make] vivid and logically explicit why retrospective characterizations of the past add truths to past times not knowable at those times.25 Note that Roth infers from the claim that some past truths depend on later occurrences (what we'll soon call 'temporal dynamism') to the claim that it is in characterizing past truths that the truths are 'added' . That is, the past's dynamism depends upon the historian's characterizations; they act as a kind of truth (or event)-making mediator. But as we'll see there are bountiful cases of temporal dynamism where no mediation is required (indeed, Reagan's case is one! This point has been raised elsewhere. L. Tsilipakos, "Descriptive Accuracy in History: The Case of Narrative Explanation", Philosophy of Social Science, 50(4) (2020), 283-312. Argues, along a similar vein, that the temporal dynamics of past facts is not really captured by our adding truths to the past. Rather, we enhance our epistemic position with respect to past Roth's central point regarding the dynamism of the past concerns the formation of new concepts which couldn't have been known/applied by past actors. We might want to say that, for instance, "ritual sacrifice" is an example of a "social control strategy" even though that concept was not available to any of the historical actors under the event description as given by historians. As such, for Roth, the new concept's generation entails the generation of the past fact. The reasoning here is fairly common in the philosophy of social science, particularly that drawing on Hacking's work.27 In short, what options are open to us as agents are in part constrained and enabled by what conceptions we have of what we might be or do. There is, then, a feedback between the conceptual environment and social patterns and ways of being. This is how some categorizations can be 'self-fulfilling prophecies': the very act of categorizing changes the behavior of those in the society. However, it doesn't follow from our legitimately categorizing using a non-actor's category that historians create the events that are so-legitimately categorized. Or at least the realist need not acquiesce to the inference.

Temporal Dynamism
As we've seen, Roth argues from non-standardization, non-aggregativity and indetachability to irrealism. However, for moderate realists his arguments involve a series of non-sequiturs. The inference from an event's sensitivity to description to its being created by that description doesn't follow; inferring from the past being dynamic to those new facts being created via the mediation of historians doesn't follow. The same can be said, mutatis mutandis, for any argument that takes event description to be ineluctably linked to the something like a "framework" as might be the case with certain kinds of positivism, especially Carnap.28 We see no substantive difference between the notions of "sensitivity to description" and "framework dependence" (or similar), given facts when we come across new, relevant findings. There is, then, a historical dynamism to historical facts, but a relatively metaphysically modest one. that each points essentially to a kind of linguistic determination of the status of events qua metaphysical entities, which we deny. At best, these positivistic arguments speak in favor of agnosticism: we simply haven't enough evidence to pick between a past with historian-created-events and one with independent events. In this section, we'll turn to a biological example to demonstrate how temporal dynamism occurs without historians' inventiveness. We'll then analyze the case in terms of a distinction between temporally dynamic and static facts. Looking forward, this distinction will underwrite our argument in favor of moderate realism.

A Turn to Biology
Marc Ereshefksy has argued that biological species are 'path-dependent entities' .29 That is, what makes a species the species it is turns not on its intrinsic properties or details of its origin, but on its unique trajectory. Importantly for our purposes, he argues that for a speciation event to be the event it is depends upon downstream occurrences.
… prominent theories of speciation imply that speciation is a pathdependent process. They imply that whether a branch (on the Tree of Life) is a species is determined by events in the path of that branch, not merely at its initial branching event.30 On allopatric models of speciation, speciation occurs when a population is divided, blocking interbreeding and allowing variation between the subpopulations to accumulate. Classic examples are geographical: some segment of the population makes it to an island, or some natural event (the rising of a new mountain range, say) cuts off subpopulations. This spatial division removes homogenizing processes like interbreeding. Over time, the two subpopulations accumulate differing traits as differing mutations arise and have differing successes depending on the environments they find themselves in. Island gigantism or dwarfism are classic examples. A subpopulation arrives on an island, finding itself in an environment profoundly different from the mainland: potentially with different flora and fauna, different niches, and different resource availability. Over time, the subpopulation adapts, individuals becoming smaller in response to resource availability, or larger to occupy new niches. 29 M. Ereshefsky, "Species, historicity, and path dependency", Philosophy of Science, 81(5) (2014), 714-726. 30 Ibid., 717.
Downstream, due to reproductive isolation and accumulating mutations, the subpopulation becomes increasingly divergent from its mainland cousins. Looking back, we can identify the populations splitting -the subpopulation arriving on the island -as the beginning of a speciation event. Critically however, that the branching event is a speciation event is not knowable until the two populations have diverged.
… we see that a branching event, a unique origin, does not make for a new species. Whether there is a new species at that branching event depends on what happens later. It depends on the historical path of that branch.31 To see this, consider circumstances where a subpopulation arrives on an island but fails to thrive: perhaps the right mutations do not arise, or they fail to compete in the new environment. More dramatically, say the island falls prey to a natural disaster: a volcano erupts, the island sinks, etc. Here, despite the branching event, we have no speciation event as the subpopulation did not evolve into a new species.32 The upshot is that the branching event -the population's splitting -at t 1 is not, from the perspective of t 1 alone, a speciation event. It is only from the perspective of a later time, t 2 , where we have two different species, that t 1 counts as speciation. Speciation has the same structure as narrative sentences. As with historical narratives, in speciation we posit that the event of the population's division gains a new property once the speciation has occurred. The populations' initial isolation being a speciation event is indetachable from their diverging genetically and phenotypically. But from where does the new property (and fact) emerge? On Roth's model, we would say that it is via the intersession of biologists. When a biologist points at a population splitting and names it a speciation event, a new fact comes into existence. But this is implausible: the population division is a speciation event because it led to two new species. The species do not care whether biologists notice them. 31 Ereshefsky, "Species, historicity, and path dependency", 720. 32 We may note here, however, that as regards the issues of indetatchability and nonstandardization, it is plausible to think that a great deal of explanatory information, in this evolutionary case, is contained at a high level of generality. Though a particular speciation may be path dependent, the processes that generate species, and so explain the event, will cover a range of type-level mechanisms.
Consider Roth's discussion of the notion of a 'career' , such that of the American President Roosevelt: Roosevelt's career does not exist until constituted by a historian. The grouping represents an artifact, a colligation by historians studying a particular person or period.33 Consider an analogous statement: the speciation event did not exist until constituted by a biologist. The event being a speciation is an artifact, a colligation by biologists studying that lineage at that time. We might agree that the event is an artifact insofar as biologists represent it in various ways, and that it has a life through biological practice.34 But recognizing that representations are artifacts doesn't require claiming that the existence of the event or property -that the speciation is a speciation -requires the mediation of biologists. In fact, it is the representational activities of biologists that allows us to uncover past facts of interest, such as speciation events. Similarly, we'll argue, neither does Roosevelt's career require intercession by historians to exist, and neither does it take our construction of a narrative sentence for it to be a fact that the 40th President of the United States hosted a television show sponsored by General Electric.

Dynamic and Static Facts
We've seen that both in human and natural history earlier events gain new properties as later events occur. Two populations being isolated becomes a speciation event once they diverge into differing species. But this doesn't seem true of all past facts: that the populations became isolated, or the population's phenotypic and genotypic makeup at some time-slice, do not seem to have this dynamic property. Let's capture this difference conceptually: some facts (or properties) are dynamic while others are static.
Static facts can be understood as facts which are not sensitive to future events; while dynamic facts are those which are sensitive to future events. In virtue 33 Roth, The Philosophical Structure of Historical Explanation, 52. 34 It is even possible to agree that this is, in fact, a kind of "colligation" on the part of biologists (or historians) without making the further claim that this threatens realism. of what might a fact or event be dynamic or not? This, we think, depends on whether the facts turn on temporally spread processes. A process is 'temporally spread' when it takes time to occur, and thus has space for defeaters. What could turn out to be a speciation event might not turn out to be such if, say, the island in question sinks, the right mutation doesn't arise, or so forth. And indeed: Reagan could have lost the 1980 election. All, or almost all, events require temporally spread processes to occur. Because things take time to happen, whether or not a fact is dynamic is indexed to a temporal scale. More carefully: Some property p of an event e is dynamic at time t 1 just in case e's having p at t 1 depends upon the occurrence of another event e* at a later time, t 2 .
Once e* occurs at t 2 , e having p at t 1 becomes static.
Note our account refers to properties, not facts, but as we take facts to refer to properties, this is an innocuous difference.35 Compare the populations being isolated at t 1 and the isolation being a speciation event at the same time. If we take our temporal index to be t 1 , then the population's isolation is a static property, as the temporal processes of isolation have occurred. Once we take the temporal index to include t 2 , once the speciation event has occurred, then the speciation becomes static. But this doesn't mean that the speciation at t 1 is detachable from the two species' existence at t 2 . The defining feature is that the event was not the event that it was (that is, possessing the properties it does) except in light of the later event. Dynamic properties can be expressed using narrative sentences, but there is no challenge for realism here. We can, then, identify two kinds of static properties. The first concern properties detectable at a time-slice. These might include, say, the genotypes and phenotypes present in the populations at some time; or the number of individuals in a population; their locations and so on. The second concerns properties which were dynamic at some earlier time, but have since become fixed as the temporally-spread processes have completed. The former we might call timeslice properties, the latter completed-process properties. Whether a property is of the time-slice or completed-process variety may turn on certain sensitivities 35 Further, we make no particular commitments concerning any particular ontology to underwrite our characterization of dynamic facts. That past facts are in some sense dynamic we take to be a broadly empirical claim; one need only examine some case studies in order to see that this is the case. This can be made compatible with an ontology that takes processes as fundamental, entities as fundamental, events as fundamental, etc. This is an interesting question in its own right (one worthy of further exploration), but we do not attempt to answer it here.
to description (e.g. explanatory interests). But, according to moderate realism, they are no less properties of the world for that fact. "Population number" may be a time-slice property relative to one description and a completed-process property relative to another, but even so propositions about these properties are truth-apt and mind-independent. Concerning dynamic events, note that there are many -potentially infinite -new properties that may arise as time passes. This is because events accumulate downstream causal influences.36 Say that our population-splitting event is the seed for a great radiation: the two new species themselves speciate, spreading into many new forms and niches. By a later time, t 3 , the progeny of the original population now dominate a variety of ecosystems. Now we might say that the event at t 1 was both a speciation event, and the beginning of a macroevolutionary radiation. At t 2 the event at t 1 gains the property of being a speciation event. At t 3 the event gains the property of being the beginning of a radiation. Further, we might pick out different aspects of the event. Say that the success of the radiation is due to some novel trait. Now, more and more completed-process properties are added to the event at t 1 .
So, we can understand the speciation case in terms of temporally dynamic properties becoming stable. At t 1 , the population splitting event is not a speciation (although it could turn out to be such). But by t 2 , when the speciation has occurred, the event at t 1 becomes a speciation event (that is, the event takes on the property of being a speciation). Assuming the process is complete, at this point the speciation becomes a completed-process event. So, as time goes by, new facts are added to past events as they gain new properties.
A crucial aspect of temporal dynamism is the openness of past events to acquiring new properties. Although many processes are complete, it doesn't follow from this that the event itself has therefore ceased, or will therefore cease, to acquire new properties as new events occur. We've already seen this in our biological example. Although (say) at t 2 the population isolation at t 1 has gained the property of being a speciation event, it is not therefore closed: as at t 3 it gains yet another property; being the basis of a macroevolutionary radiation. So, on our account the past isn't simply dynamic, it is in principle open-ended. As we'll see, this open-endedness comes to the fore especially in light of the reflexivity of human investigation of the past. 36 C.E. Cleland, "Methodological and epistemic differences between historical science and experimental science", Philosophy of Science, 69 (3) (2002), 447-451.

Dynamic Historical Realism
We've thus far seen that Roth's arguments do not conflict with moderate historical realism, one which happily makes perspectival or ecumenicist claims about history -there is no one best, maximally unified, account of historybut nonetheless insists that historical events are discovered, historians do not bring them into being. What we've not seen yet is a positive argument for taking on such a view. Our aim in this section is to do exactly this. To begin, let's reconsider Roth's argument. Roth argues that historical events are the events they are in virtue of historians categorizing them into narrative explanations. On this view, past events gain new properties (or facts) or, more strongly, new events come into existence, via the historians' actions. As such, historical facts (events, etc.) are metaphysically dependent on historians. But in the last section, we saw biological examples wherein past events gained new properties without the biologist as an intermediary. Events like speciation are temporally dynamic: the event being the event it is depends upon later events. Such events, then, transition from dynamic to static as the temporally extended processes they depend upon run to completion. This at the very least suggests that appeal to the powers of historians can only partially explain the temporal dynamism of historical facts.
On our picture of dynamic properties, there is a dependency relation between the dynamic property and the outcome of a temporally extended process. Until the process has completed, whether the dynamic process holds is undecided. How might Roth's historian intermediaries fit here? Roth must either replace a dependency between dynamic property and temporally extended process with one between the dynamic property and the historian, or make the dynamic property rely on two things: the process and the historian. The first cannot make sense of historical practice, the second adds a conceptually unnecessary extra ingredient.
We'll first make explicit how our account of dynamic past properties underwrites realism, second turn to a historical case study to defend realism.

4.1
How to Be a Realist What is required to be a realist about historical events? If our contrast is Roth's irrealism, then the realist must claim that history's events exist independently of historians. If you want, we might say that history comes 'pre-carved' . But this is a highly misleading metaphor, as the realist need not say that there is a single, unified, privileged carving. Instead, the 'carving' will be multi-faceted, sensitive to description and potentially open-ended. There's no need, for the moderate realist, to appeal to any special sense of 'carving' . Historical processes lead to a patterned history, some highly contingent, some more robust, and these patterns and patchiness are the targets of historical discovery.
A crucial aspect, and a point of agreement between ourselves and Roth, is how past events afford, or are amenable to, a multitude of characterizations and complex interrelations with other facts.37 Let's go back to our toy biological example. We discussed a population becoming isolated at t 1 , becoming a speciation at t 2 , and a macroevolutionary radiation at t 3 , and how the temporal relations holding between them determine what kinds of events they are. Now let's imagine that this radiation was shortly followed by an externally-caused mass-extinction event at t 4 (an asteroid impact, say). From one perspectivethat of the mass extinction -the events from t 1 to t 3 may be insignificant: they made no difference (or at least very little) to how that mass-extinction played out. But from another perspective (that of the radiation, say) the events from t 1 and t 2 are critically significant. There is an apparent contradiction here: t 1 is both significant and not significant. But as we've seen this is innocuous: significance is relative to description, indexed to a perspective. Whether it is true that t 1 mattered from the index of the radiation turns on what actually happened: whether in fact that seed of the radiation turned on the population's being isolated. And whether it is true that t 1 didn't matter from the perspective of the mass extinction turns on what actually happened. On the realist take, then, sensitivity to description just marks off the way different kinds of facts are connected to each other relative to particular questions and the like. The connections (or carvings) already exist, but different ones will be selected given a different set of questions and explanatory concerns.
There is, then, a set of realist positions fully compatible with a dynamic past. All the realist needs to concede is that some events being the events they are depend upon later events, and the in-principle open-endedness of past events. The historian's narrative, then, typically aims to pick out a static processcompleted event: that, say, Reagan in fact did win the 1980 US election. In this sense, speaking of narratives as 'describing' past properties and events rather than creating them is happily consistent with those events being dynamic. 37 For discussion of the relationship between complexity and pluralism in historical explanation see: Førland, "The Ideal explanatory text in history: A plea for ecumenism"; Currie, "Narratives, mechanisms and progress in historical science"; A. Currie, "Simplicity, oneshot hypotheses and paleobiological explanation", History and philosophy of the life sciences, 41(1) (2019), 10; Currie and Walsh, "Frameworks for historians and philosophers"; K. Sterelny, "Explanatory pluralism in evolutionary biology", Biology and Philosophy, 11(2) (1996), 193-214; T.A. Grantham, "Explanatory pluralism in paleobiology", Philosophy of Science, 66 (1999), S223-S236.
Further, when the historian generates a new narrative, or new categorization, they do not thereby 'create' a new event or fact. They rather describe (or at least attempt to describe) a static process-completed event. The truth-conditions of the historian's analysis are rooted in what actually occurred in the past. Let's situate this view. We can pick out flavors of historical realism and anti-realism by considering what factors constrain historical narratives. Naïve realists -monists -hold that there are a set of static facts and of these there is a privileged set more or less irrelevant of explanatory interests. A more minimal kind of realism (we think this a plausible reading of Danto) claims there are a set of past facts which constrain narratives (a chronology) but in constructing narratives, historians make significance attributions to various events in that chronology.38 On this view, the events picked out must have occurred, but there are no empirical constraints on significance-attributions. Roth's antirealism goes further, denying that there is a chronology to be had in the first place.
On our account, chronologies are multi-faceted and perspectival, but the admissibility of an historical narrative is partially constrained by those chronologies (by what actually happened). Thus, it is a form of realism. However, it is stronger than Danto's because we also think there are constraints on significance attributions. That is, there is often a fact of the matter to be had about whether some past event is significant, given some explanatory interest. To see the importance of these constraints on significance properties, we'll turn to an example of a substantive historical dispute: barter economies.

Substantive Historical Disputes & Barter Economies
One of realism's critical advantages is that it can make good sense of the claim that historians have substantive disputes about what occurred in the past. Irrealism, however, cannot.
Irrealism situates historical truth in the models and frameworks historians employ. The inference, which we've claimed is a non-sequitur, is to then argue that the events and truths metaphysically depend -are 'created by' -those models and frameworks. But then historical disputes appear to involve warring created facts. Historical disputes, as Roth describes them, turn on non-empirical 38 See Gallie's discussion of narrative "turning points" for some helpful background. Turning points within a narrative are, for him, something like the "crucial moments" that give a developmental narrative its distinctive shape. Gallie  factors concerning frameworks, their fruitfulness, say.39 However, there are many historical disputes that are substantive, and explicitly involve denying that some 'created facts' are facts at all. Some of these are mundane: disagreeing about temporally static properties such as the date someone died, the actual population at a time, and so forth. But others concern what were once temporally dynamic properties. Here, historians argue about whether some categorization of an event gains purchase: how the other historical narrative characterizes the event turns out to be false or inapt. This sense of false characterization is a deeply empirical activity and, we think, requires a realist treatment.
In substantive historical disputes the rubber hits the road: the aptness of competing narratives turns on what actually happened in the past. This, we argue, necessitates realism because the independence of the past to historians is required for the rubber to hit the road. To see this, we'll examine a case study in which conceptual and empirical aspects are related in deep and interesting ways. As we'll see, the debate is interwoven, which is to say empirical, conceptual and interpretive issues are bought into iterative contact. But it is also substantive, which is to say the debate critically turns on new empirical and material discoveries about the actual past. This latter feature underwrites an argument for realism.
In a barter economy I exchange something I have and do not need or necessarily want -but you do -for something you have and do not necessarily want -but I do. Let's say Angela has finally finished working her way through her copy of Hilberg's The Destruction of the European Jews, and is interested in reading some relevant philosophy of history. Jang-Mi has just completed Roth's The Philosophical Structure of Historical Explanation (less daunting length-wise than Hilberg!) and Roth's discussion has piqued her curiosity about Hilberg's book. Happily, Angela has something that Jang-Mi wants, and Jang-Mi has something that Angela wants. The two can now negotiate and agree on an exchange.
The concept of a barter economy is popular largely to Adam Smith, who used it in The Wealth of Nations similarly to how Hobbes used his 'state of nature' (although the concept's history is long, looming large in discussions of political justice in Plato and Aristotle). For Hobbes, the pre-state nature of 39 Or, if there are empirical disputes for Roth, they are about empirical facts that are entirely "theory laden" (see Roth, The Philosophical Structure of Historical Explanation, chapter 3). But this doesn't help Roth, since disputes over empirical facts will have to involve appeals to the theories (or models) which constitute the empirical facts. Thus it would appear such disputes still lack the kind of substantiveness we claim to be present in such debates. the human world is marked (derogatorily) by general anarchy; a struggle of all against all. Contractual obligations -and eventually the state -emerge necessarily as a response to this chaos, which we are rationally obligated to avoid.40 Similarly, for Smith in some prior state humans exchanged via barter and nearinevitably developed money.41 The crux of the problem with barter is that it requires what economists call a double-coincidence of wants. In order for the barter to be successful, I must want what you have, and you must want what I have. Jang-Mi might reflect on the almost 1400 pages of Hilberg's book and decide she doesn't want it after all; without the coincidence of wants, the barter collapses and Angela (assuming she's nothing else Jang-Mi wants) will have to look elsewhere for Roth's tome. That is, unless she offers to pay Jang-Mi. As Smith puts it: But when barter ceases, and money has become the common instrument of commerce, every particular commodity is more frequently exchanged for money than for any other commodity. The butcher seldom carries his beef or his mutton to the baker or the brewer, in order to exchange them for bread or for beer; but he carries them to the market, where he exchanges them for money, and afterwards exchanges that money for bread and for beer.42 Smith's model is neat: because exchange in a barter economy is fundamentally limited by people wanting each other's stuff, the invention of money opens up economies by endowing a set of tokens with general exchange-value. So long as someone wants something, they'll likely be willing to sell their surplus goods for cash. Later economists, most prominently Robert W. Clower,43 employ transaction costs to explain the emergence of money from barter economies. Finding folks who have what you want and who want what you have can be tricky, so people will begin to gather together for such purposes, and whatever the most common item exchanged is will, he argues, inevitably become monetized. We can, then, identify two ways of understanding barter. First, barter as a model incorporating transaction costs and the double-coincidence of wants. Second, barter as historical claim that economies develop from barter to monetized systems. Barter as a model of exchange dynamics still looms large in economics, and is apparently taken seriously as a real historical claim in popularizing and pedagogical contexts (economics text-books for instance).44 It also still turns up in serious economic work: Modern monetary theory shows that monetary exchange takes place if individuals are sufficiently specialized in consumption and production in the sense that they frequently end up in situations where the double coincidence of wants does not hold, and the good serving as money does not lose its value too quickly over time. or bargaining in exchange. This latter notion opens the door to a 'barter' system, whereby even cash can be on object of barter. "By definition, barter is a complementary exchange in which each participant bargains until he or she is satisfied."54 These claims about the nature of barter economies were popularized in David Grueber's History of Debt (chapter 2). There, Grueber draws on the kinds of historical and anthropological evidence we've discussed in an attempt to overturn the Smith-based model.
The economist Jualo Huato's response to Grueber is illuminating: … if it is plausible to argue that barter imposed large opportunity costs on transacting parties (by requiring from them an improbable "double coincidence of wants"), then barter cannot be expected to have existed as a regular, stable, or dominant social practice in any well-defined historical period -and, therefore, to be readily observable in the historical record.55 Huato is pointing out that according to Smith-like models of barter economies, we should expect them to be fleeting -and thus invisible in the historical record. It might be tempting to read this as an attempt to accommodate the anthropological and historical data into the pre-existing account but this is too quick. Rather, the argument demonstrates that Smith-like models are still helpful in discussing the far-more complex reality of market dynamics than the barter-then-money picture has it. In short, folks like Huato retain the barter model (perhaps heuristically), while denying the veridicality of Smith's historical claim. Let's highlight some important features of the debate, before demonstrating how Roth's account cannot accommodate it and moderate realists can. First, the debates are interwoven. That is, historical inferences, conceptual machinery and explanatory models are heavily interlinked. The argument that barter economies are not an early stage of the development of economies relied upon (1) empirical observations such as the lack of barter in ethnographic studies, (2) conceptual developments, such as Humphrey's negotiation-based conception of barter, and (3) explanatory models, such as Huato's suggestion that Smith-models can be put to work in explaining the transient nature of barter economies. Roth's account is well-placed to accommodate these interwoven aspects. However, the debates are also substantive, which is to say they turn crucially on new empirical discoveries -on confronting the conceptual and theoretical with the historical and ethnographic records. Although the conceptual and the empirical are interwoven, the rubber hits the road with the empirical record. Responses like Huato's do not involve sticking to theoretical guns, but rather showing how that conceptual machinery can nonetheless retain utility given the new empirical information. Humphrey's conceptual innovations are due to her interacting carefully with ethnographic information. It is this substantiveness that irrealism cannot accommodate. Roth argues that debates between narratives are primarily theory-driven, not evidence-driven: [testing hypotheses] will primarily be a function of assessing competing explanations, and so draw on evaluative criteria more akin to theory appraisal than to hypothesis confirmation.56 Or, more strongly: The significance of 'the empirical' disappears on the assumption that theories either determine what counts as experience or explain away any apparently discordant evidence. What comes to be termed 'empirical' can readily become instead an artifact of theorizing. The empirical so understood then ceases to have a determinate function in the assessment of theories under consideration.57 It is true that the empirical and theoretical interweave, but it does not follow from this that the empirical disappears. But this isn't to commit to a naive empiricism, either. A steady diet of Kuhn over the last several decades has disabused all parties of the notion that theoretical and empirical entities can be plausibly treated as independent of one another. But the examples above, we argue, show that the relation of determination can't flow strictly from the theoretical to the empirical; the recalcitrance of the empirical data is what invites so much reworking of the same problems, so the empirical facts seem to be in the driver's seat, even if there is partial determination flowing in both directions at once. The empirical data appears to be the stronger force vector, in some cases, at least. On Roth's account, why or how this should be so is unclear: if historians are merely comparing different theoretical frameworks in terms of "how they focus and shape subsequent inquiry and debate" (81), then the role of empirical data is left mysterious. If irrealism is true, then we're 56 Roth, The Philosophical Structure of Historical Explanation, 66. 57 Ibid., 127.
not sure why historians should hunt down original texts, attend to the material and ethnographic record, and generally have concern for the veridicality of their claims, after all, their debates are not substantive, merely turning on theoretical virtues. Moderate realism makes sense of substantive historical disputes. They do not merely turn on theoretical preferences, but our evidence for what occurred in the past. There is a fact of the matter about whether past economies were originally barter economies, and whether such economies formed a basis for monetary economies. No doubt determining this is tricky and involves conceptual innovation, but this in itself is no reason to deny the existence of those past facts prior to historian's discussing them. And no doubt the past is complex: it may be that the barter economy model applies better in some instances than others, but determining this in itself partly turns on the empirical facts historical work uncovers. Empirical data turns out to do what nothing seems to be able to do on Roth's account of narrative evaluation: dislodge some historical narratives in favor of others. It is only from some form of a minimally realist perspective that this is sensible. Realism provides the necessary anchor for explaining the substantiveness of historical debate.

4.3
Historian-Created-Facts An advantage of Roth's approach is its emphasis on the reflexivity of historians' practices: history is often not built directly from interaction with inferred past facts or primary texts, but from interaction with the work of other historians. If past facts are constituted by the actions of historians, this is unsurprising. However, realists can also make sense of this reflexivity, in fact, we'll argue, they do so better than irrealists.
In our example involving barter economies, it seems relatively clear that the earlier categorization of certain kinds of economic activities as "bartering" (as in the case of Smith) opened up a set of interesting conceptions for future historians and social scientists. The conceptual refinements that served to clarify our picture of early economic activity was, in some sense, made possible by the extended dialectical process in which historians are engaged. Humphrey could only turn the classical model of "barter first, then currency" on its head, because the introduction of the initial model itself opens up a dialectical space where particular questions can be asked, concepts refined, and, consequently, new facts introduced (or created).
How might the iterativity and reflexivity of historical practice threaten realism? Roth (2020, chapter 3) argues (roughly following Hacking58) that there is no stable way to characterize human actors into natural kinds (especially regarding things like social behaviors and categories). The reason for this instability of kinds, from a synchronic perspective, is relatively clear. As we've already said, the behavior of human agents can turn out to be sensitive to the way that such behavior is categorized, and this can be a source of kind instability.59 It's far less clear how this kind instability is epistemically significant from a historical (or diachronic) perspective. While past human actors can't turn out to be sensitive to description in anything like Hacking's terms (economic actors in the depths of human history can't be sensitive to being described as "barterers," mainly because they're dead), Roth does think there is a kind of instability of kinds at play in the characterization of past actors in the sense that descriptions we now give can't be true of past actors prior to our descriptions, since they could not have conceived of themselves as we now categorize them. Roth's position, then, involves a kind of Kuhnian Incommensurability of Historical Kinds. Roth understands this as a major threat to realism: because past actors would not understand themselves in terms of our categories, it follows that statements about past actors are not true or false, but only true or false relative to a historico-conceptual model, which we are not entitled to think of as even an approximately true description of the past.
The first thing to say in response is that there are many plausible counterexamples to the thesis of historical incommensurability. Consider a rather armchair counterexample: Let's say we were, through some technological marvel, able to resurrect some early human, and find some means of communicating information to them concerning the work of historians and archaeologists on primitive economic behavior. While we would certainly be missing quite a bit of information on cultural context and economic milieu, it seems entirely plausible that such a person would be able to discriminate between explanatory models that possess more descriptive adequacy from those having considerably less. This wouldn't involve anything much different than, say, unpacking the concepts philosophers use when talking about theories of knowledge with first year undergraduates. At first the concepts seem foreign, as if in another language. But, through careful dissection, a clearer picture emerges over time, and we can then position ourselves to have substantive discussion and disagreement over which ones have epistemic purchase. We see no reason why this should not also be true in hypothetical cases involving past actors and imagined conversations with them. 59 Although for an interesting development of Hacking see: J. Laimann, "Capricious kinds", The British Journal for the Philosophy of Science, 71(3) (2020), 1043-1068.
More pressing, though, are cases where historians do seem to create facts. Historians' traditions are often built upon falsehoods. Is it not true that the concept of the barter economy is crucial for understanding the history of economics (and how modern economies have been shaped) even though it was based on a falsehood? The realist responds: yes, it is. But the relevant historical event here is not the existence or otherwise of barter economies, but the previous claims and interpretations of folks like Smith. Similarly, but more subtly, historical disagreement often turns on the inaptness of previous models: although they get many of the facts right, they problematically over-emphasize some factors over others, that is, they get the significance wrong. To see this play out in realist terms, let's dip our toes into a final case, taken from the history of philosophy.
The Rationalist-Empiricist Distinction, although still extremely common as a framing device in pedagogical contexts,60 is increasingly either abandoned or significantly complexified by historians of philosophy.61 In a simplistic form, the distinction frames the early modern period as characterized by a canon of works and figures: Hume, Locke and Berkeley on the Empiricist side, and Descartes, Spinoza and Leibnitz on the Rationalist side. Further, disputes are understood as centered on the foundation of knowledge: whether it depends on experience or not. Historians of philosophy have pointed out that the actors at the time would not have recognized the dispute along those lines, in fact the narrative and the canon was constructed by Kant's students in order to emphasize Kant's importance as resolving the dispute, a narrative that was only really codified and accepted in the late 19th Century. 62 How might a realist characterize this dispute?
The realist can happily say that, for instance, Descartes' Meditations is significant because it forms part of the Rationalist canon. This is a narrative sentence: later philosophers' construction of the canon and its becoming solidified through the 20th Century completed a dynamic process, adding a new property to that work. Kant's followers, indeed, created a new fact. However, there is nothing metaphysically mysterious about this. Just as the speciation event added a new fact, so did the Kantian interpretation of the Early Modern Period: historians, after all, are part of unfolding causal history just as much as species are. But now historians challenge this orthodoxy (as Humphrey and others did for barter). Aspects of their arguments are non-empirical: the canon reinforces particular conceptions of philosophy that ought to be challenged (that, for instance, epistemology should start with questions of knowledge's foundations). But much of the arguments turn on interpretations of the original texts and their historical context: they turn on facts of the matter about Early Modern debates. That there are such facts to be had -independently of historical practice -makes sense of this reflexivity on the part of historians.
We can say that understanding 20th Century philosophy requires understanding the significance of, say, Meditations. But it might be that 21st Century philosophy takes a different turn, emphasizing different aspects of philosophy's history as the canon fragments, expands and is perhaps abandoned. And those events might well take on new properties -become significant in new ways -as past works and figures take on new significance: history's dynamism is in principle open-ended. But understanding these processes requires seeing that past facts are not only multitudinous, complex and sensitive to description, but also independent of us in a critical sense. The fact that Meditations is significant for 20th Century philosophy is indifferent to what 21st century philosophers think; that a speciation has occurred does not depend on biologists knowing about it; that barter economies did not predate monetary economies is independent of economic historians; the fact that the 40th President of the United States hosted television shows cares not a whit about how future historians characterize the event.
Only a minimally realist take on history, then, can accommodate the substantive and reflexive nature of historical practice.