Volume 2, Issue 3: Naturalist Grist for the Philosophically Inclined Historians Mill

Naturalist Grist for the Philosophically Inclined Historians Mill: Paul Roth on Narrative Explanations and Their Mysterious Structure 

Paul Roth: The Philosophical Structure of Historical Explanation. Evanston, IL: Northwestern University Press, 2020. xvii + 192 pages. $99.95. Paperback ISBN 9780810140882. DOI: 10.2307/j.ctvt1sg91

Fashions come and go; intellectual trends are no exception; neither is the rational undertaking of so-called analytic philosophy. Although during its more than one hundred years of past, many different subjects and fields were discussed, if one were to concentrate on or name the most characteristic topics, then philosophy of language and philosophy of science were to be nominated with relative ease and security. But “philosophy of language” and “philosophy of science” are big and complex subjects, and especially the latter witnessed various trends and patterns after it was brought into the scholarly attention of many by the logical empiricists.

            In the field of “philosophy of science” what produced an enormous literature was the idea of explanation, especially with respect to history. After Carl Hempel raised the question of what form explanations and historical explanations shall take, one of the most productive and wide-ranging periods of analytic philosophy was on its way (at least this is the usual story of how explanation as a focus-topic emerged). Books, articles, conferences, workshops, and even a rare occasion to reach out to non-analytic philosophers and non-philosophers characterized the 1950s and 1960s; after a few decades, however, the subject became more technicalized on the field of metaphysics of explanation and became softened up among literary scholars, and slowly the whole field of “analytic philosophy of history” vanished from the foreground.

            This does not mean, however, that no one has worked on the subject of how philosophy of history could be envisioned with the tools of analytic philosophers; one of the leading philosophers to discuss the topic fruitfully and progressively was Paul Roth, who was even honored by a Festschrift before his present book on the subject was published.

            Roth’s new book, The Philosophical Structure of Historical Explanation, published by Northwestern University Press, appeared a few months ago just before the pandemic reached the shores of America and Europe. (Thus, if you got lucky before the quarantine, you had something really nice and engaging to read!) Roth’s non-hidden aim was to “revive analytical philosophy of history” from the ashes and by synthetizing the main results of other fields as well, to offer a new and synthetic account pointing ahead. It won’t be much of a surprise to tell already at the beginning that Roth got his Festschrift for very good reasons: his Structure offers fruitful insights, finds the middle-path between over-logicized norm-building epistemologies and indulgently subjective and often bias narrations.

            The book is easy to read generally; not just because of its language, style, and the beautiful typesetting that attracts the eye, but because it is structured quite transparently. Roth first sets the scene about the past and present of philosophy of history and notes those relevant points that deserve discussion. In fact, general philosophy of history is not on the table anymore; Roth knows well (especially as he was/is also part of the story) that even if the waves of fashion (or history?) wipe out certain issues from textbooks and classrooms, the significance of the given topic is not threatened. So what needs revival is not analytic philosophy of history, but a new perspective to show people that there is still something to be discussed explicitly.

            Philosophy of (historical explanation) in the twentieth century is often associated with C.G. Hempel (1942/1965). What Hempel has done in the 1940s and 1950s (thanks to Fons Dewulf (2018a, 2018b, 2020) we are now in the position to see a clearer picture of how Hempel himself has emerged from a very definite context), was to reconstruct the logical structure of explanations (applicable to historical issues as well). In his famous (or notorious?) hypothetical-deductive model of explanation, Hempel claimed that every acceptable explanation takes the form of a deductive inference (or as he later came to admit: generally to a logical form which might deductive or inductive as well); we simply subsume the explanandum under some general law, or law-like generalizations. To put it really bluntly, how do you explain that a certain event happened? Well, by calling attention to the fact that the given particular event is just one instance of a more general (law-like) issue. If this is how things always (or usually) happen, then it is no wonder, after all, that the given situation arose. If water boils at 100 Celsius, and your measurement shows that we are now dealing with 100 Celsius in the cooking pot, then do not drop your teeth! That water is boiling! This is a very oversimplified example, of course, but after all, Hempel’s idea comes down to the very simple idea: whenever we give explanations, we are falling back on generalities that either shed some light on the particular event or serves as, so to say, inference-tickets to structure our expectations. But I shall come back to this later.

            Hempel’s model (in fact, there were several, both deductive and inductive) came under fierce attacks right after its publication; both the philosophy of explanation and the philosophy of history became important and much-discussed topics of analytic philosophy for decades. But these are well-known facts; it is also known that after a while, Hempel’s logical model with regard history was, so to say, somewhat outdated by more informal accounts, emphasizing more and more the individual human factors instead of generalities and laws, and the strict logical structure was substituted also by narrative story-telling. Whether narratives could replace the logical model to explain historical events is another matter. As Roth notes quite soberly, “The use of narratives to explain is unquestioned; what is subject to philosophical dispute is whether this habit is to be tolerated or condemned” (p. 22). In that sense, Roth does not want to revise the customs of practicing historians: he tells a story about something “practitioners and readers already knew, that is, that narratives explain” (p. 139), what distinguished Roth’s book philosophically is the idea to dig deeper into how and why narratives explain and what their epistemic features are.

            Narratives are often conceived to be philosophically impotent as they do not have any stable and rigorousstructure and thus their general characteristics go against the basic assumptions of philosophers; narratives seemingly do not answer our basic and widely accepted cognitive norms. In fact, Roth’s main or general aim is to rehabilitate narratives as serious explanation-forms, to characterize their structure philosophically to show that we can rely on them when historical explanation is at issue. Moreover, Roth does not just takes the notion of ‘narrative’, but to legitimize it, he deals with all those fundamental epistemological and metaphysical assumptions that are usually used to exclude narratives.

            In Roth’s view, narrative explanations have three key features: (N1) the nondetachtability of the explanandum from the narrative; (N2) the nonstandardized character of events explained; (N3) the nonaggregativity of narrative histories. 

            (N1) means that in the case of a narrative explanation, the given narrative constitutes both the explanandum and its relation to the explanans. To put it otherwise: the way you forge your story, your narrative, influences the very event for which you sit down at first to provide an explanation. (There are no such things as revolutions; an event might be characterized as a revolution for one party and a counterrevolution by another, or just a simple aggressive fight on the street again by another; depending on how you describe your issue will change your explanandum and explanans as well.) (N2) means that elements of the explanation, e.g. the events as explanans, are not such standardized subjects as other well-defined, even axiomatized elements of the natural sciences. A revolution is not a sharp and rigid thing as one element of the periodic table. So to say, there is no such general “recipe” (p. 67) that tells you how an event shall look like in all its details and characteristics to count as a revolution, or a counterrevolution, or even a war. This connects to the third point (3), namely new elements in our knowledge and a new perspective might rewrite the nature and ‘essence’ of the event to be explained, then one might say that events in history cannot be aggregated to form one continuous, linear, and well-integrated story. To put it otherwise, while Hempelian explanation exemplifies a classic deductive structure, with all its truth-preserving monotonic characteristics, narrative explanation is non-monotonic; if you add another considered-to-be-true element to the structure, it might decompose the whole narrative. In a narrative explanation all points have the same epistemic status: all of them can be revised and all of them could be retained (note the quite Quinean point about reviseability). 

            All of these ideas point to the direction that narratives are bad candidates to form explanations because they are too soft, flexible, changeable, unable to figure as elements a well-defined deductive and monotonic structure. But Roth shows convincingly that since in history we are dealing with narrative sentences, we have to stick to narrative explanations as well. A narrative sentence is a sentence that expresses a truth about an earlier time that can be known only retrospectively, though it was true at the earlier time, that is, narrative sentences “will be sentences true at t but could not have been known at t” (p. 27). Such sentences include “The Thirty Years’ War started in 1618” and “The Holocaust started in 1933”, and so on. These are admitted now as true sentences, which seem to be true already in 1618 and in 1933, though they were not known in these forms in 1618 and 1933. No one could have known in 1618 that “The Thirty Years’ War” started that year and how it will unfold.

            A narrative explanation takes issue with such a narrative sentence to organize and sequence the available information in such a way as to be related to this narrative sentence. By knowing the closure of the event will we know also where to begin and how to structure exactly the temporal unwrapping of the concerned events that are related by their subjects and human interests. Historians start from morsels, pieces, almost from waste, something that does not have a traceable and detectable structure and meaning that points beyond itself. They put them into a coherent story where the end defines the beginning and where all the pieces fall into such a place that is determined from the end. 

            In fact, Roth seems to say often that in many narratives, we do not find new facts, but rearrange the available facts into a new structure, into a new narrative that paints in a novel way. But this seems to assume that there are facts to be arranged in a novel way in a new narrative. We do not get much direct information about how these facts came to be and what does “fact” mean here. But Roth’s relative irrealism points in these directions – as we will see facts are defined contextually, relying on historians’ actual practices and their choices.

            But given all of what has been said so far, what makes narratives explanatory? How to evaluate them? How to account for their use? That is, Roth has three major questions:

            (A) How does something become a narrative?

            (B) How does a narrative explain?

            (C) By what standards shall we evaluate a narrative explanation?

            To provide satisfactory, i.e. philosophically satisfactory answers, Roth has to dig deeper. The book is structured very cleverly: the first half is devoted to criticism, to philosophical rubble clearance (Chs. 2 and 3), while the second half (Chs. 4-7) gives Roth’s new program of naturalism that includes essentially narrative explanations.

            Chapter 2 and 3 takes the issue of all those implicit assumptions that Roth detects both in classical and in influential contemporary works. The most basic of these is the idea that there is a distinct, recognizable, stable, rigid, sharp, and discoverable past and the task of the historian is to discover the past as it is or as it was, to write such a story whose sentences are made true by past events; to use a contemporary metaphysical parlance (going back, of course, to Plato), the goal is to care nature (in this case: past) at its joints, the joints of the one true past. Historical realism is the view, which claims that historical knowledge is possible on the basic correspondence to definite past events.

            One might be surprised that Roth still has to fight against such a view, especially as after many turns (postmodern, linguistic, narrativist, historicist, micro-historist, and whatever not), it would be hard (though not impossible) to imagine such historians who still adhere to the seemingly positivist ‘scissors-and-paste’ method. (I say seemingly, as there are many studies to show that August Comte, the par excellence of the positivist method and view, in fact, the father of it, did not accept a purist and theory-independent view of facts.) This is an empirical question, of course, but Roth has a nice answer: even if many people claim to overcome the old positivist view, their actual work and framing still connotes the idea of realism. So, if made explicit, historians would deny their strict realism, but still, the metaphysical commitments is often lurking in the background, working as a presupposition.

            In fact, this assumption of existing a fixed past, goes directly against the idea of nonaggregativity (N3). For many historians, “histories, like maps, are guides over existing terrains” (p. 26), that is, the more historians working on the same subject will reveal the hidden facts and the unknown structure of events as they actually happened. Adding elements of knowledge to others, an “ever clearer picture of things as they actually were” emerges” (ibid.) and the realist smoke lifts, showing the landscape under its true lights. In a narrative form, this is simply impossible.

            But there is a problem to be addressed. Historical realism is more than intuitive and commonsensical. We just simply feel and know that there must be a definite past. We remember (in most cases) and know what happened an hour ago, or yesterday, or last week, or last year, or a decade ago. Even the last century might be a vivid memory for many. It is just foolish to say that the past does not have a concrete and recognizable shape! Of course, we might say that we do not know how the political structures and intentions shaped some of the events in the past (as we do not know how they shape our present), but that is just a contingent issue, even available to some chosen ones on the frontline. Perhaps the only way to weaken this more than a secured common sense view is to deny the distinct character both of present and any moment of personal experience. As we will see, Roth easily and proudly walks that line.

            An event, a historical unit, becomes an event only under a description, by subsuming it under a category. But as descriptions and categories are human creations and thus vary with time and place, events cannot be fixed once and for all. “Events simpliciter cannot be shown to exist; they are not known to be of nature’s making rather than of ours”, says Roth (p. 30). “Events exist only by proxy. [...] Knowledge of events is restricted to happenings isolated under descriptions provided by interested parties.”

            Instead of “the past”, Roth talks about “the pasts”; what to count as an event, what to count as a serious or significant event depends on the historian; and depending on the interests of the historians can we talk about various pasts; pasts worth to explain. But Roth is not an anti-realist (pace Dummett); he does not claim that there are events hidden from us, thus creating truth-value gaps. As realism is an extreme and indefensible metaphysical position, anti-realism provides similar commitments. Roth is an irrealist, which means two things: first, the whole realism and antirealism debate is meaningless in the sense of showing up a false dichotomy (p. 49); secondly, Roth argues that by denying realism and antirealism (or better, by overcoming them) one is not floating freely among relativistic and untouchable histories, but are rooted in a special socio-cultural and scientific community that determines her frameworks that determine thus her choices of historic events to be explained and to explain with. (At this point, regarding categorization and constitution of units, Roth relies on the writings of Hayden White, Louis Mink, but most importantly on Nelson Goodman’s and Ian Hacking’s work on categorization and natural/social kinds.) Thus while Roth breaks the chains of realism, he acknowledges historians’ social situation and the even more complex web of the scholars’ belief (just as we might expect from a naturalist!).

            Roth formulates a dilemma (or sort of a dilemma). Without our human and thus contingent categories, there are no events, and without events there are obviously no histories. (Let’s talk about histories, not about past, as past could be constituted by other entities, and anyways, if events turn out to be problematic conceptually, that really threatens our common-sensical picture of history, but does not necessarily break down the idea of past ontologically.) But if we assume that events are grasped via our categories, they cease to be ideal and rigidly real, objective, and independent from us, required by the realist. So we cannot have categorized and conceptualized events and realism at the same time; for Roth the choice is quite simple, as he painstakingly argues on long pages that categories are essential for our understanding of, well, everything. 

            From this perspective already, one might answer (A) and (B). Namely that in order to have a narrative, we need a narrative sentence, defined above, and we shall work with that narrative sentence. Whenever we have a narrative sentence, we can move on to the explanation-part; a “narrative explanation will be a presentation of a temporal series that answers why the explanandum [the narrative sentence] turns out to be as it is” (p. 68). A narrative explanation thus makes an order and provides sense to sporadic information, bringing coherence and meaning by subsuming units under a special description or category (“the past has not form until given one by categories imposed on what we take to be evidence”, p. 64). 

            But how to evaluate such narrative explanations? In the case of Hempelian-models we have many epistemic and logical issues to discuss: the formation of laws, the way one subsumes a particular event under the law, and draws the deductive inference (Hempel (1942/1965, 234) notes these as well, and talks about them as “objective checks”). The starting-point or problem with narratives was exactly that there are no such exact standards and norms to discuss them or to evaluate them in the light of objective standards. For Roth, that is not a big problem. Although he does not spend much time and pages on explicating the details of his strategy, he claims that we can compare and situate narrative explanations to each other only by a case-by-case method, weighing values, interests, and further human aspects of life and scholarship, that is, to see how they perform in contrast to each other given our backgrounds and aims. There is also a shorter allude to the “inference to the best explanations” strategy, but I will not discuss that here. (Just to note: at one point, Hempel (1942/1965, 239) notes that a certain prediction (similarly to explanations) is just “an ingenious guess” and works only by luck, but “cannot be arrived at by means of [the]theory alone”. One might wonder then, whether certain narratives are deemed better than others not by providing more compelling cognitive reasons, but only because of a match between the historian’s narrative and our wishful (or political) thinking. Though there is nothing eternal and external to play the role of the judge here, and we are always in the midst of our politically biased wishful thinking, this is an important point where further discuss shall reveal itself later.)

            Structure is way more detailed and rich in material than I noted it here. Perhaps that is one of the biggest weaknesses, or better, difficulties of the book; namely that Roth discusses many issues, even within a chapter, and sometimes the reader might be lost where are we staying at the moment. Of course, every chapter is closed with a nice and succinct summary of what was achieved and what lies beyond us, but still, it is often hard to keep track of all the achievements and details. Chapter 6 is, for example, a very interesting and novel account of Kuhn’s Structure of Scientific Revolutions, where Roth argues that Kuhn’s historic method points toward such a view on normal science that is highly relevant to other debates as well. His point becomes relevant with regard naturalism, the topic of Chapter 7.

            Roth argues that there is no such thing as “science” viewed from an Archimedean point; no conception of science could stand the waves of space and time, that is, science is a concept that becomes defined only from a certain point of view, after its norms, patterns, rules and conventions are settled after initial friction. That is why it becomes a narrative issue: normal science becomes normal science after a revolution but that requires a narrative to account for and to settle the events as we see them. As Roth notes:

[Structure of Scientific Revolutions] offers an essentially narrative explanation because that turns out to be the only form that can accommodate what needs to be explained – how what was not science came to be science. (p. 111, original emphases)

But if what counts as science is determined by a narrative, and thus narrative explanations are essentially involved in determining what is science and what are its legitimate and accepted methods, then narratives and narrative explanations become involved in naturalism as naturalism requires a definition and view on science as its subject. A naturalist does not accept any philosophy beyond science to provide foundations for science; science is founded (if at all) by science itself. Thus relying on what is science is essential for a naturalist, and defining science requires pieces and knowledge from history, thus narrative knowledge (and requires other human sciences). Nothing can be more Neurathian than such a non-reductive, dynamical (see below) naturalism that sees also what lies beyond evolutionary and cognitive psychology (this contrast between social scientifically affected Neurathian naturalism and cognitive scientifically determined Quinean naturalism was exploited before by Thomas Uebel – in fact, Uebel gets one note from Roth for calling attention at one point to Neurath’s achievement and roles in the story. Strangely, Roth has a very nice Neurath paper already from 1984 (thus from before the big English-speaking Neurath-boom, see Roth 1984) but for some reason Neurath plays no role in his book). 

            But one might raise the question: what science is and how it came to be are things that many working naturalists and scientists are just trained to know in practice, they are customized into science without making explicit any narratives about it. What is science is often provided for working scientists and naturalists either by their training, by textbooks or by their personal experiences; while these contain narratives as well, it is known or supposed to be known (mainly from Kuhn, but I would add here Philipp Frank too) to be faulty, oversimplified and misleading stories. This leads us back to the evaluation issue: how would be an alternative, more detailed narrative about science (along Kuhn’s lines) be better than a textbook narrative?

            Roth’s naturalism goes deep; he takes a broad step and embraces all the radical consequences. Historical knowledge is just empirical knowledge, “no more (or less) problematic than our account of any other” (p. 48), thus for Roth there is no distinguished difference between past and present (p. 56); his account “simply makes divides between theory and evidence [...] into a contingent fact and not a necessary or conceptual one” (p. 61, cf. p. 130); it washes away the normative/descriptive distinctions within the context of science and justification (p. 120); within his thoroughgoing naturalism the notion of science is constantly under revision (p. 121); even more radically, naturalism itself could be overcome by something else if that proves to be “more efficacious in furthering our goals” (p. 121), or as Roth says, “the preference for naturalism itself is result-driven” (ibid).

            We should admit and admire Roth’s radicalism at this point. For all the reasons people used to beat history and historical explanations with, apply as well to other forms of empirical knowledge and present times. And Roth explicates this so nicely, easily, and in a very understandable manner. I entirely agree with Roth’s radicalism about the epistemic status of the present and the past(s), about the washing away of the theory/evidence distinction, and with the idea that naturalism shall become a dynamic naturalism, meaning that it should be reflexive and sensitive to the differences and shape-giving character of science. But a few critical notices might in order here too.

            First of all, it is not seen entirely how something else can be “more efficacious in furthering our goals” than science. The idea is that our goals (whatever they are for Roth) are best served by science, and naturalism is the stance that puts science in the forefront. If science fails, naturalism fails with it. But how could science fail? According to the scientific world conception of the Vienna Circle, there are no depths, just surface, and its relations and structures, studied by science. Hence whatever escapes the arms of science will not be relevant in the fight against science. While science could be dropped temporarily in certain scenarios, naturalism still seems to be an attitude, less affected by empirical data. Roth claims that “naturalism is not a philosophical theory; it is empirical through and through” (p. 121, original emphasis). I agree with him that naturalism is not a philosophical theory (although note that Roth himself speaks of “philosophical naturalism” on page 113); not because it is not philosophical but because it is not a theory. It is an attitude towards certain matters of life and world. Or even if we would like to call it a theory (maybe we have here just a terminological bypass), it must be a second-order theory about how to handle first-order theories. In that sense, even if science seems to fail, a naturalist might change the rules and the field to bring science back into the game. If everything is ‘reviseable’, then everything is ‘keepable’ as well, and a naturalist can be a naturalist even if she extends the notion of science beyond rationality.

            There are two major and presumably interrelated morals or points of Roth’s book. He argues that narrative explanations are brought into the field to explain narrative sentences (remember, not available at t, but true at t). The feature of these narrative sentences is that they cannot be explained otherwise just with narrative explanations, hence these narratives explanations are essentially narrative explanations and history thus deals with essentially narrative explanations. “If a narrative explaining a narrative sentence has narrative form essentially, then there exists no nonnarrative way of explaining just that event” (p. 68); for Roth, a narrative explanation-form is not just one player in town but also the only one. As events get their status, role, meaning and significance by a narrative, they cannot be imagined outside of a narrative, they are essentially narrative. As such, they fall under the ideas of (N1), (N2), and (N3). This is indeed very important; not providing narratives thus is not just another option to grasp a dimension of history, but a mistake. But the question arises then: in order to provide explanations, we need to know what we are explaining; scientific entities have their own stories and histories. At point t, we have only a very “naturalistic” description of what we see on the monitor (point at x,y,z, or whatever), and that ‘thing’ becomes a thing at all only at a later point. Thus shouldn’t we say that physical explanations require elements from narrative explanations as well? Why did this thing boil at 100 Celsius? There is certainly a story to tell! Are all explanations narrative explanations after all?

            Or we can reverse this order! Let’s approach that from what Roth says. As events cannot have standardized forms, they cannot “utilize laws or law-like generalizations” (p. 68), no laws or any law-like connections link the events in a narrative through time (p. 146), only the categories and the story we fabricate from them. This seems to be another main moral of Roth, namely that as the events surfacing in narratives are not linked and joined together by laws/law-like generalizations, the typical Hempelian model of explanation cannot be utilized in the case of history – only essentially narrative explanations. We might foster some skeptic remarks that points to another direction.

            Roth seems to write about Hempel, or at least about his model that outlived Hempel in many senses, as by stating and requiring causal issues for the general laws, it states something about the objective and mind-independent world. By making that commitment, Hempel and his followers fall back on historical realism and the ontology of causes and objective events. In fact, at first glance, Roth might be right. We read the followings in Hempel’s classic “The Function of General Laws in History”:

An account of the development of an institution is obviously not simply a description of all the events which temporally preceded it; only those events are meant to be included which are “relevant” to the formation of that institution. And whether an event is relevant to that development is not a matter of evaluative opinion, but an objective question depending upon what is sometimes called a causal analysis of the rise of that institution. (Hempel 1942/1965, 241)

Given this terminology and way of speaking many people would reject or attack Hempel’s model and ideal. Roth’s own analysis and offered solution to the problem of historical explanation goes directly against the last sentence of Hempel. Though Roth would not say that what is relevant in the development of events is “a matter of evaluative opinion”, but what he says about politics, culture and human interests would fall into that category in Hempel’s eyes given his basic philosophical stance. But can’t we do otherwise?

            We have at least two options here. First of all, we change Hempel’s problematic terminology. Take the following version:

An account of the development of an institution is obviously not simply a description of all the events which temporally preceded it; only those events are meant to be included which are “relevant” to the formation of that institution. And whether an event is relevant to that development is not a matter of causal analysis but a question depending upon what is sometimes called a conceptual analysis of the rise of that institution.

Or if you like it, a “question depending on the chosen framework”, Roth’s own choice of words. But these are not just words. I would hesitate to claim that Hempel was committed to historical realism, or a general version of independent ontologies where “causality” carves at the joints in a necessary manner; but we are certainly entitled to reinterpret his words and ideas, and to claim that the laws that connects events in an explanation may reside (in a somewhat Kantian fashion) in us and not in the world. What matters for Hempel is empirical testability and tractability and not fundamental ontology. Thus all of Hempel’s law and law-like talk shall not indicate any such metaphysical conceptions that Roth rejects.

            But, and this might be the second option, we can retain the original formulation of Hempel (perhaps removing only the word “objective” because of its misleading connotations, though it is known now among logical empiricist scholars that “objective” meant for Carnap and perhaps to others as well only “intersubjective”), and claim that “causality” is not that dangerous as it might seem to be at first sight. After all, when we are moving within a “framework”, or a “worldview”, or if you like, a “paradigm”, then we have many accepted and internally well-established causal links. These could, of course, turn out to be false or oversimplified but in their everyday scientific practices scientists rely on such causal links and use the language of causality with relative easy, punctuality and prosperity. If we admit this, then as naturalist we might admit causality as well as did the scientist; even if we know that this is only a framework-relative notion of causality – a more contingent, flexible and human-interest relative issue. Roth did not say much about his own notion of “causality” in this context, but I am sure he would find such an account that could fit his conception and thus would strengthen his approach in general.

           Furthermore, narrative explanations are taken to be complex issues, especially in contrast to a purported instance of the hypothetic-deductive model. That complexity cannot be reduced to well-distinguished stand-alone units (see N1, N2 and N3) by their nature or essence; thus narratives cannot be evaluated by a simple, one-dimensional evaluative process pointing towards logical tractability supported by law-like generalizations. Roth is right here, of course; very complex narratives are indeed required because every historical event is a human event, or at least involves human interests and actions that are parts of nature. But even if you take out the human from the picture, we know from Neurath, even the simplest natural prediction requires a complex web of interdisciplinary beliefs. To note Neurath’s example; to predict whether a certain forest will born down, we have to know something about the weather (rain and wind), the biology of the trees, the chemistry of the water used to rake out the fire, the economy of using woods, the sociology of the closest fireman department, etc. etc. The same goes to explanation: why the forest did not burn down? In fact, Hempel noted the very same issue (the Hempel-Neurath relation is still to be taken down, though it is known that Neurath was outrageous because of the whole prediction, law, and explanation business of Hempel, but that is another issue). At the end of his article, Hempel notes that “the defeat of an army” requires many and various references. References to “lack of food, adverse weather conditions, disease, and the like” (1942/1965, p. 242); these are similar items of course, belonging to diverse scientific fields from Neurath’s perspective, but belonging to the auxiliary sciences of historical investigation from Hempel’s perspective. In a good narrative explanation, one does not have to cite such diverse reasons (a good narrative might be unfolded from one simple idea), but the most diverse issues can be covered the better. 

            For Hempel, these seemingly marginal issues became central in historical explanations, especially as they bring with themselves “usually tacit [...] assumptions of laws” (p. 243). He notes several times in his paper that one is not aware of laws in the case of historical explanations because they are implicitly assumed to be trivial. “The universal hypotheses in question frequently relate to individual or social psychology, which somehow is supposed to be familiar to everybody through his everyday experience; thus, they are tacitly taken for granted” (1942/1965, 236). In fact, whatever explanation involves social and psychologistic elements, it involves implicit assumptions about laws, law-like generalizations or simply general considerations that points towards Hempel, and not necessarily Roth. After all, why would we accept the narrative of Raul Hilberg (one of the basic examples of Roth) who claims that (using a narrative sentence) “the Holocaust begin in 1933” (see p. 72 in Structure). Hilberg explains the destruction of the Jews (as his book’s title suggest) by noting and emphasizing very detailed, painstaking and emotionlessly minute administrative and bureaucratic proceedings of the Nazis administrators. Why would this explain anything? Because we know generally, or at least assume, that at a certain point administrators got such an amount of power and influence that might have wide-ranging effects; or because many of us have the stereotype of German preciseness. After all, our explanation will explain something if we can subsume (or at least got the feeling of subsuming) the particular event under a more general consideration. It is of further question, what this general consideration might be. The late Hempel, for example denies that explanation amounts to “familiarizing” the given issue as most of our familiar issues are in need of explanation and the given physical explanations on the quantum and atomic level are everything but familiarizations. (Though it is a further question whether a nice equation of the physicists counts as an explanation for the laymen or not; one might debate the merits and misleading character of Sir James Jeans and Sir Arthur Eddington, but their popular books on dissemination scientific knowledge, or domesticating scientific knowledge, explained a lot to the days’ English readers.)

            This brings us back to the initial issue: whether every explanation amounts to general formulations or law-like generalizations. It seems to be quite dubious, as the old story goes, to find a suitable general law-like issue that explains (and predicts for later concerns) why a revolution has broken out. Usually either our law-like generalizations are too poor and thin or the particular event to be explained escapes the general issues. This is a problem, yes! Two things could be mad, however, in the case of these difficulties. Extend and thicken up the range of the law-like generalization so to include the particular event, or enrich the individual description in order to fit a law-like generalization. Both processes are familiar from the history of science (see Lakatos’ book on mathematical discoveries and explanations, Proofs and Refutations).

            I touched or scratched only the surface of this issue here and did not provide such arguments that Roth would deserve given his own detailed investigation. What matters for us now is that Structure is indeed such an illuminating and insightful book that invites the reader to presume even such radical ideas as Hempel’s original story. The idea of narrative explanations functioning as philosophically interesting and cognitively debunkeable items is attractive, amusing, but still controversial for many reasons. People might be afraid of slippery slopes; that happened already with constructivisms and other forms of antirealisms and irrealisms. As I said, Roth is radical, but he is right in his radicalism. The question is whether his radicalism could be extended even more to embrace Hempel’s radical idea as well. 

            One final, general note is in order. Roth’s book will be useful not only to those who are interested in the philosophy of history, but also to those who pursue the task of historiography regarding the history of analytic philosophy (one of Roth’s example is Michael Friedman, pioneer of the field). The history of analytic philosophy in the last two or three decades saw many different approaches towards its subject, good as well bad examples are flourishing as well. One of the most well-known example came from Scott Soames who has published now numerous volumes of his magnum opus. Soames was criticized by many philosophers and historians because of his quite ahistorical approach. For Soames, the whole history of analytic philosophy was just a long, muddled, bumpy road to the works and insights of Saul Kripke. In this approach to the history of analytic philosophy, all the achievements and failures of previous thinkers (like Russell, Moore, Carnap, Ayer, Quine, Davidson, and so many others) were interpreted in their relation to Kripke’s results. This approach was often described as “rational reconstruction”: subjects of historical investigation are put on the table with their arguments and concepts that are evaluated in our contemporary discussions. The problem is not ‘how they came to think what they did’, but ‘given what we know now, were they simply right or wrong’. This has led to many ambiguities, oversimplifications and simple failures. But the method of rational reconstruction could get perhaps another boost from Roth:

No one could say or predict at that moment that this was happening, or that a certain concept would come to have a settled use in a future scientific community. The point rather is to illustrate how historical events may be constituted and explained in terms of concepts in some sense present in but not known to those to whom they are attributed. (p. 61)

In a narrative explanation we utilize such concepts that were not present or were not known in that form by those who figure in the story. But, as shown by Roth, this process does not have to be purely anachronistic. In a narrative explanation we reinterpret some of the available information under new lights and provide new sense to old issues to see things differently. That does not mean, however, that this narrative explanation shall be the only player in town, or that this would represent things as they are. What might mislead many people in connection with Soames is that one cannot escape the feeling that Soames writes about history as something that inevitably leads to Kripke, and as something that represents some necessary underlining structures and developments, and not just an interesting story, fueled by a specific personal interest. Roth’s book thus might help us to understand how to write the history of analytic philosophy as well.

            To put it plainly, Paul Roth’s own Structure is an important book, filled with many insightful arguments, concepts and ideas. Roth knows how to equilibrate between spreading positions ingeniously and how to overcome fruitless distinctions and gauds. Nonetheless, Paul’s middle, synthetic way of philosophy and naturalism turns out to be more radical than the other options he resourcefully overcomes. Read the book, from time to time, as surely you will always find something new and fruitful, be it either the interpretation and commitments of naturalism, a novel interpretation of Hempel, Wright, Danto or even the new utilization of Quine. The Philosophical Structure of Historical Explanation invites to at some points to disagree, calls for further research and substantiation, but it is good in all that!

Adam Tamas Tuboly

Institute of Philosophy, Hungarian Academy of Sciences

Supported by the MTA Lendulet Morals and Science Research group and by the MTA Premium Postdoctoral Scholarship.

References

Dewulf, F. (2018a). A Genealogy of Scientific Explanation : the Emergence of the deductive-Nomological Model at the Intersection of German Historical and Scientific Philosophy. PhD Dissertation.

Dewulf, F. (2018b). “Revisiting Hempel’s 1942 Contribution to Philosophy of History.” Journal of the History of Ideas 79 (3): 385-406.

Dewulf, F. (2020). “The Place of Historiography in the Network of Logical Empiricism.” Intellectual History Review 30 (2): 321-345.

Hempel, C.G. (1942/1965). “The Function of General Laws in History.” In Aspects of Scientific Explanation and other Essays in the Philosophy of Science. New York: The Free Press, pp. 231-í243.

Roth, P.A. (1984). “On missing Neurath‘s boat: Some reflections on recent Quine literature.” Synthese 61 (2): 205-231.