In this post, I will discuss E. H. Carr’s views on historical causality. Although some of his views are infamous, I attempt to defend and make a good case for some of the controversial views. I argue that his discussion about causal patterns, accidental causes, causal hierarchies, values behind causal selection, and future-oriented causal selection deserve more credit than they have received. I argue that the notions discussed by Carr are so central to the theory of history that they cannot be ignored simply on the basis of excessive historiographical cautiousness towards big pictures and historical coherence.
The basic assumption
In his book What Is History?, Carr argues that “[the] study of history is a study of causes. The historian [–] continuously asks the question ‘Why?’” (p. 87). This may seem like an empirical claim about the nature of historiography, and one might wonder whether the truth of the claim is only accidental: If historians simply refuse to discuss causality, then the claim no longer holds. So why introduce such claim? The answer is obvious: The claim is a normative one. Carr argues that insightful and relevant historiography is a study of causes.
Here comes our problem. As we will see, Carr makes seemingly controversial claims concerning the historiographical search for causes. For example, he argues that relevant causes are the ones that fit into a “rational explanation of history” and Oakeshott argues that Carr is therefore committed to a priori ordering of history. According to Oakeshott,
“Mr Carr’s view is that historical enquiry must be governed by what he calls a ‘philosophy of history’. And [–] the expression ‘philosophy of history’ stands for some broad understanding of the ‘meaning’ of the course of events postulated in advance of historical enquiry itself.” (Oakeshott 2004.)
Moreover, Carr has been mocked about his claim that historiography requires an idea of historical progress and future course of events. “He goes on to assert that, as he understands it, historical enquiry is impossible without a belief in a ‘progressive’ direction in the course of events (that is, a belief that we are moving from worse to better) and without some vision of what the future holds in store for us” (Oakeshott 2004).
Given that the search for causes presumably leads to controversial outcomes, it seems that Carr’s claim that historiography is a study of causes must be abandoned. That historiography is a study of causes cannot justify the controversial claims if one is skeptical towards the claim that historiography is a study of causes. The problem is: Why should we accept Carr’s normative claim about historiography as a study of causes if it leads to controversial outcomes?
There are only two ways to avoid the problem. Either one should provide some independent reasons to accept the normative claim or provide reasons to accept the controversial claims. While I am sympathetic towards the first option (I would argue that causal relationships not only provide genuine understanding but also must be assumed in historiography to make sense of the inferences in the field), in what follows I will follow the second option and attempt to argue that the controversial claims stand their ground, at least when modified a bit. I will argue that Carr’s position is a rather sound one and that the ideas mocked by Oakeshott are actually defendable, if understood in depth (unlike Oakeshott does).
Laws, Generalizations and Causality
Carr notes that “we no longer speak of historical ‘laws’; and even the word ‘cause’ has gone out of fashion, partly owing to certain philosophical ambiguities into which I need not enter, and partly owing to its supposed association with determinism” (p. 88). Given the intimate connection between laws and causality and given that historical laws – universal generalizations – are difficult to find (to say the least), the notion of historical causality becomes unclear and obscure. How can we know historical causal links if we do not have access to laws that govern those links?
Carr points out that the model where scientists discover laws by “induction” and then straightforwardly use them in explanation and prediction is old-fashioned. The picture is more complex and involves feedback loops. “It is recognized that scientists make discoveries and acquire fresh knowledge, not by establishing precise and comprehensive laws, but by enunciating hypotheses which open the way to fresh inquiry” (p. 59). All enquiry, according to Carr, is theory-laden: “All thinking requires acceptance of certain presuppositions based on observation, which make scientific thinking possible but are subject to revision in the light of that thinking.” (Ibid.) In the same way, historians can formulate hypotheses concerning historical causality and proceed in their inquiry in a theory-driven and fallibilistic way.
There is, however, a possible problem. Even if we agree that historiography provides insights in a theory-driven way, we could still wonder whether the hypotheses can be about laws or generalizations and therefore relevant to our causal understanding. If there are reasons to question the existence of laws in historiography, then we have, ipso facto, reasons to question causal relationships in historiography as long as laws underlie causal relations.
Carr argues that historiography requires generalizations in many ways. First, historiography is committed to generalizations by “the very use of language” (p. 63). Secondly, historiographical trends and research traditions are based on shared understanding concerning the relevant generalizations:
“In the 1920s discussions by historians of the causes of the war of 1914 usually proceeded on the assumption that it was due either to the mismanagement of diplomats, working in secret and uncontrolled by public opinion, or to the unfortunate division of the world into territorial sovereign states. In the 1930s discussions proceeded on the assumption that it was due to rivalries between imperialist powers driven by the stresses of capitalism in decline to partition the world between them. These discussions all involved generalization about the causes of war, or at any rate of war in twentieth-century conditions.” (p. 63.)
Thirdly, historians fill gaps in evidence by relying on generalizations. “If the evidence is not clear whether Richard murdered the princes in the Tower, the historian will ask himself – perhaps unconsciously rather than consciously – whether it was a habit of rulers of the period to liquidate potential rivals to their throne; and his judgement will, quite rightly, be influenced by this generalization.” (Ibid.)
However, we should not think that “generalization permits us to construct some vast scheme of history into which specific events must be fitted”. Rather, “[historiography] is concerned with the relation between the unique and the general”. (64-65.)
It seems that the claim about the relationship between the unique and the general has at least two readings in Carr’s philosophy. The first reading is an ontological one. He quotes Marx: “Events strikingly similar, but occurring in a different historical milieu, lead to completely dissimilar results. By studying each of these evolutions separately and then comparing them, it is easy to find the key to the understanding of this phenomenon” (p. 65). I take this to mean roughly that patterns of causal dependencies between two factors can be tracked down only by taking into account the effect of other factors in any given system or situation and by comparing many historical situations with each other. I come back to this below.
The second reading is an epistemic one: We have to balance our trust in generalizations that are implicit in language, research traditions, and in the filling of the gaps in evidence with what we learn about the objects which the generalizations are applied to. For example, we might learn that two conflicts, both called “wars”, were very different and the term “war” should not be applied to one of them. Moreover, the fact that it is unclear whether “Richard murdered the princes in the Tower” is true should (to some extent) increase our suspicion towards the generalization “it was a habit of rulers of the period to liquidate potential rivals to their throne”. The epistemic reading can be understood as a reformulation of Carr’s analysis of the role of hypotheses in historiography that we discussed above.
Both the ontological and epistemic reading imply that the generalizations that historians use are restricted, open ended and imprecise rather than universal and well-defined. Given the epistemic reading, the scope and adequacy of the generalizations need to be assessed case-by-case. Given the ontological reading, the existence of causal relation between some variables X and Y does not mean that we can determine the value of Y if we know the value of X; we have to take into account other factors in the situation.
The idea that causal explanation in many special sciences is based on restricted and local generalizations rather than universal laws is strongly supported these days. We have seen (in an earlier post) how Stuart Glennan makes a strong case for a mechanical analysis of causal explanation in historiography. In the mechanistic account, we explain a historical event or phenomena by describing a mechanism that produces the event or phenomena. Such ephemeral mechanism (as Glennan calls them) is a collection of interacting parts where “the interactions between parts can be characterized by direct, invariant, change-relating generalizations” (2010, 60) without a requirement for universality. Moreover, the widely influential interventionist account of causal explanation (Woodward 2003) analyzes the notion of explanation supporting generalizations in terms of test-invariance and is explicitly skeptical towards the requirement that explanations require laws (ibid. Ch. 4). To put it simply, we are able to achieve explanatory understanding even in the absence of knowledge concerning laws, so knowledge of laws cannot contribute to the explanatory import of every explanation.
The fact that we can achieve causal explanations even in the absence of universal laws means that Carr’s claim about the study of history as a study of causes does not suffer from the fact that there hardly are any historical laws. We can connect historical factors by using generalizations that “may hold only in limited domains, have a rather open-ended set of exceptions that defies any simple characterization, and may be imprecise and vague” (Woodward 2003, 180).
However, there is a long way, according to Carr at least, from raw and unprocessed causal knowledge to insightful historiographical accounts. Let’s take a look on how Carr understands the construction of insightful causal narratives in historiography.
Hierarchies of causes
History has been a complex process. This complicates the construction of causal narratives:
“The first characteristic of the historian’s approach to the problem of cause is that he will commonly assign several causes to the same event.
[The] second characteristic of the historian’s approach [is that the] true historian, confronted with this list of causes of his own compiling, would feel a professional compulsion to reduce it to order, to establish some hierarchy of causes which would fix their relation to one another, perhaps to decide which cause, or which category of causes, should be regarded ‘in the last resort’ or ‘in the final analysis’ [–] as the ultimate cause, the cause of all causes.” (89-90).
Carr’s notion of ultimate (or true or relevant) cause is quite ambiguous. It can refer to
(a) A causal factor that underlies other relevant factors (“the cause of all causes”). In his text, Carr uses as an illustration a story where Robinson crosses a road to get cigarettes and is killed by a drunk driver (104-105). Carr argues that Robinson’s desire for cigarettes is not a relevant cause of the death while the drunk driver is a relevant cause. One way to understand is that without the drunk driver, the desire for cigarettes would not have led to a death. In contrast, even in the absence of the desire, the drunk driver could have killed someone else. Of course, the difference between these cases is a matter of degree: Robinson could have hit a broken electric wire (for example) and died even in the absence of the drunk driver; and the drunk driver could have made his journey without killing anyone. However, a scenario where Robinson would have died in the absence of the drunk driver is much less plausible than the one where the drunk driver would have killed someone even in the absence of Robinson. The drunk driver is the “true cause” because his drunk driving made the causal structure of the city such that many possible reasons and actions (such as the desire to get cigarettes) could lead to a death.[1] This brings us to the next suggestion:
(b) A causal factor that was not accidental. Carr writes: ”in so far as [some causes are] accidental, they do not enter into any rational interpretation of history, or into the historian’s hierarchy of significant causes” (p. 103). The question, then, is how to define “accidental cause”. A clue is provided by the notion of rational interpretation of history. Below, we will see that true causes are the ones that belong to rational interpretations. However, the notion of non-accidental cannot, in itself, define true causes because the notion of accidental cause refers to the residue of a real explanation and is therefore logically dependent on our ability to identify true causes: “[F]rom the multiplicity of sequences of cause and effect [a historian] extracts those, and only those, which are historically significant; and the standard of historical significance is his ability to fit them into his pattern of rational explanation and interpretation. Other sequences of cause and effect have to be rejected as accidental, not because the relation between cause and effect is different, but because the sequence itself is irrelevant.” (p. 105.)
(c) A factor that is generalizable. “There, too, we distinguish between rational and accidental causes. The former, since they are potentially applicable to other countries, other periods, and other conditions, lead to fruitful generalizations, and lessons can be learned from them; they serve the end of broadening and deepening our understanding.” (p. 107). We should be careful here. This claim should not be interpreted as saying that “rational causes” are the ones that produce the same phenomena in different contexts. If this was the case, accidental causes would not affect the outcomes of historical processes which they clearly do, according to Carr: “The shape of Cleopatra’s nose, Bajazet’s attack of gout, the monkey-bite that killed King Alexander, the death of Lenin – these were accidents which modified the course of history” (p. 103). Moreover, we saw above his quotation from Marx: “Events strikingly similar, but occurring in a different historical milieu, lead to completely dissimilar results.” Rational causes, then, work in many situations but do not produce the outcomes alone.
(d) A factor that is relevant with respect to some goal that historiographical investigation has. Carr writes:
“When we recognized certain explanations as rational, and other explanations as not rational, we were, I suggest, distinguishing between explanations which served some end and explanations which did not” (p. 106).
“Accidental causes cannot be generalized; and, since they are in the fullest sense of the word unique, they teach no lessons and lead to no conclusions. But here I must make another point. It is precisely this notion of an end in view which provides the key to our treatment of causation in history; and this necessarily involves value judgements” (p. 107).
In other words, Carr suggests that the ultimate (or true or rational) causes must be chosen on the basis of values. He also suggests that the choice is intimately connected to kinds of generalizable patterns we wish to learn.
(e) Some combination of (a)-(d). This seems to be the correct reading, given the interdependencies between (a)-(d).
Here is how I understand Carr: His view is that a historiographical explanation is a theory-driven construction that attempts to capture the most invariant features of the phenomena under investigation.
Invariance
Notice first the connection between (a) and (c). If, on the criterion provided by (a), X is a relevant cause of Y and Z is not a relevant cause of Y, then there are more scenarios where X leads to Y in the absence of Z than there are scenarios where Z leads to Y in the absence of X. (There are more scenarios where a drunk driver kills a person even if that person did not have a desire for cigarettes than there are scenarios where the desire of cigarettes lead to the death of the person in the absence of drunk drivers). Given this and given that X occurs repeatedly in actual history, many of the scenarios where X is present in the absence of Z have a counterpart in actual history, and some of them lead to Y. (The scenarios where the intoxicated person is driving in the absence of Robinson’s desire for cigarettes has many counterparts in actual history, and some of them lead to a death of a person). If we want to effectively prevent Y (or make it happen), we need to focus on the causal relation between X and Y, not Z and Y, because X is more invariantly connected to Y (i.e. there are more (actual and possible) situations where drunk driving (X) leads to death (Y) than there are (actual and possible) situations where a desire for cigarettes (Z) leads to death). Given the invariance, the relationship between X and Y is more generalizable than the relationship between Z and Y.
Ends
In order to get the detailed reasoning of the previous paragraph off the ground, a historian has to decide which phenomena are such that they require attention. Carr suggests that the relevant considerations are pragmatic and future-oriented – a topic to which I return below. While such pragmatic reading is probably the most natural one, it is not the only one. It is conceivable that a discouraged historian who no longer beliefs in social change still studies the causes of inequalities in societies. The historian might very well be interested in invariant relationships between some socio-economic conditions and inequality in order to understand the phenomena deeply. It might often be the case that pragmatic considerations drive the construction of historiographical explanations. However, we do not have to commit to the pragmatic value of explanation in order to formulate “hierarchies of causes”. It is enough if we commit to claim that the choice of explanandum in historiography is driven by present-day interests, whether they are pragmatic, intellectual or whatever. The notion of invariance discussed above is independent of pragmatic considerations in the sense that the relative invariance of a causal relationship is independent of our pragmatic purposes: some causes produce phenomenon P in wider variety of (actual and possible) situations and are therefore more invariantly connected to P than other causes. The usefulness of knowledge of some causal relationships is a derivative from invariance, not the other way around. Of course, we cannot explain all of the history and therefore our interests drive what explanation-seeking questions we ask. However, once we have asked an explanation-seeking question, the goodness or depth of our explanatory answers depend only on the relative invariance of the answers.
Theory-drivenness
However, historians never begin their investigation by looking at some historical case without any presuppositions. When they investigate the causal structure of a historical process, they suppose that certain generalizations may hold with respect to the case at hand. In other words, the historians suppose that certain types of causes are more invariantly connected to the phenomenon under investigation than other and attempt to apply these causes to the phenomenon. For example, when we find a body that has been hit by the car, one of the first hypotheses is that the driver might have been drunk. We saw above how Carr thinks that historiographical explanations are theory-driven. In the light of the consideration presented here, this means that historians attempt to apply the causal generalizations that, thus far, appear most invariant to the new cases they face. This is a perfectly natural way of approaching and does not imply any suspicious a priori -driven historiography that Oakeshott suggests (2004). We have also seen that, according to Carr, it necessary to balance “the general and the unique”. Invariance-centered considerations do exactly that: invariance of a causal relationship can be assessed only by investigating many cases where the causal relationship might exist.
Accidents and patterns
There is still a great puzzle to be solved. As we have seen, accidents do affect the course of history, according to Carr. Given this, how can they be excluded from explanations?
First, all causally relevant factors cannot be realistically incorporated into an explanans. There are just too many factors that affect historical event. There is even an easy formula to produce causal factors that could be mentioned in some complete (or rather: ideal) explanation: Take some cause (like drunk driving) of the explanandum-event (death). Then imagine a factor that would have cancelled the cause (friend stopping the driver from reaching the car). The absence of this imagined event friend stopping the driver is then causally relevant: had a friend stopped the driver, there would not have been a death.[2] There is no mystery per se in excluding factors from an explanans.
Second, accidents make sense only against the causal background that is provided by the more invariant causes. Robinson’s desire for cigarettes was deadly only because there was a drunk driver in the city. Non-accidental causes that are more invariant and belong to historiographical patterns are logically prior to accidental causes. One cannot identify an accidental cause and its effect without prior knowledge about non-accidental causes. If we find Robinson dead,[3] we could think he had a heart attack or some other medical condition. If we do not know that there is a drunk driver or some other danger outside Robinson’s home, we might easily infer that he would have died even if he did not go to get cigarettes. However, once we know the non-accidental causal structure of the city, including the drunk driver, then we can infer that his desire for cigarettes was a cause of his death. This means that understanding history and the effect of accidental causes depend on prior knowledge of invariant patterns. In this way, it becomes possible “to establish some hierarchy of causes which would fix their relation to one another, perhaps to decide which cause, or which category of causes, should be regarded ‘in the last resort’ or ‘in the final analysis’ [–] as the ultimate cause, the cause of all causes” (p. 89-90). Carr might exaggerate when he excludes accidental causes from explanations altogether but his commitment to the centrality of non-accidental causes is extremely important and can be strongly supported.
Thirdly, there is an interesting analogy between the so called best-system-account (BSA) of laws of nature and Carr’s philosophy. Imagine that we know everything about the universe. Then imagine that we attempt to build a deductive system that enables us to derive from a set of axioms true statements about the universe. We could build a system where all true statements are included as axioms. This would be a strong system because it allows us to deduce every true statement. However, the system would be hopelessly complex. On the other hand, we could have a system with only one axiom “everything happens because god wants so”. This would be a very simple but hopelessly weak system because it would not allow us to infer anything that happens. So the best system would be one where the simplicity and strength are optimally balanced. Then, according to BSA, a generalization is a law of nature if it is an axiom or theorem in the optimal deductive system. Given that laws are axioms and theorems in the best system, it becomes easy to understand why they are explanatorily powerful in contrast to more accidental generalizations: They are applicable in many cases (strength) and relatively easy to operate with and coordinate with each other (simplicity). In analogous way, Carr’s approach attempts to balance simplicity and strength: He wants historiography to find invariant generalizations that can capture many historical causal chains. The demand to balance general and unique in Carr’s philosophy is perfectly sound against this background: There can hardly be anything more rational than to cope with events and phenomena in as simple and powerful way as possible. “Rational interpretations” that Carr stresses are quite literally rational when they are based on a system of generalizations that is simple and strong.
Future and history
Finally, we need to recognize the essential role that future-oriented considerations play in Carr’s philosophy. He writes that
“It is at once the justification and the explanation of history that the past throws light on the future and the future throws light on the past.
What, then, do we mean when we praise a historian for being objective, or say that one historian is more objective than another? Not, it is clear, simply that he gets his facts right, but rather that he chooses the right facts, or, in other words, that he applies the right standard of significance. When we call a historian objective, we mean I think two things. First of all, we mean that he has a capacity to rise above the limited vision of his own situation in society and in history – a capacity which, as I suggested in an earlier lecture, is partly dependent on his capacity to recognize the extent of his involvement in that situation, to recognize, that is to say, the impossibility of total objectivity. Secondly, we mean that he has the capacity to project his vision into the future in such a way as to give him a more profound and more lasting insight into the past than can be attained by those historians whose outlook is entirely bounded by their own immediate situation. [–]
Historiography is a progressive science, in the sense that it seeks to provide constantly expanding and deepening insights into a course of events which is itself progressive.” (p. 123-124.)
In order to make sense of this view, we have to remember that Carr thinks that a cause is significant if knowledge about that cause makes it possible to achieve some end. Given some end E, historians attempt, according to Carr, find generalizable causal relations that make it possible to achieve E. Now, a historian can be objective in two ways:
(a) The historian is able to recognize some end E that could be desirable even though achieving E is not an immediate interest of the current society.
(b) The goal E that the historian recognizes is such that it will be desirable in the future.
When a historiographical explanation tracks down causal patterns P that are relevant to the achievement of an end that belongs to a desirable future, the explanation in terms of P is more lasting than explanations that serve some transient present end.
The problem is that there is no guarantee that there is any progress in history, at least if we do not trivialize the notion of progress by equating it with whatever the the future holds. This means that even if some end E is recognized as desirable in the future, we still have to ask whether the willingness to satisfy E is ethically sound. Willingness to achieve ethically unsound E does not count progress even if E is an end that the future generations in fact are willing to achieve. This, in turn, implies that some end E that a historian recognized can be desirable and progressive even if the future generations do not, in fact, adopt E.
A related problem is that we all, including historians, have an impact on the future. Progress cannot be achieved if we do not drive towards better futures. If a goal E is such that a future is desirable only if it adopts E, then we can achieve the desirable future only if we recognize E. This means that we have to be able to judge the ethical soundness of the goal E independently of whether the future generations adopt E or not. If a future is desirable only if E is adopted in that future, we have to work towards the adoption of E now. However, our desire to achieve a future where E is adopted is rational only if we are able to recognize the desirability of E now.
The considerations above mean that the actual course of events in the future cannot be the ultimate standard that tells us which historical explanations identify significant causes. If progress is not guaranteed and if progress depends on our current activities and ability to find desirable goals (as it clearly does), then the significance of a set of causes that an historian identifies must be judged independently of whether the end (against which the causes are judged as significant) is adopted in the future or not.
In order to avoid a suspicious commitment to progress in history and in order to understand our role in creating desirable futures, we need to modify Carr’s philosophy in order to make sense of objective judgements of significance. We have to say that a set of causes C is significant if knowledge about C enables us to achieve some desirable end. However, our conceptions concerning the desirability of certain ends are fallible. Therefore, we have to say that C can be judged as significant if a good case can be made about the desirability of an E that knowledge about C enables to achieve.
Given that (a) historiography is a study of causes, (b) a study of causes requires the identification of significant causes, and (c) the significance of causes depends on the desirability of the end they enable us to achieve, the italicized sentence implies that historiography requires that historians reflect on the reasons why they judge some end desirable. Given that these reasons are based on values, the construction of a historiographical explanation requires that we reflect on our values. Therefore, the construction of historiographical explanations is a value-laden practice. It is also a value-laden future-oriented practice, as the values and goals are intertwined. There can be no historiographical explanations without visions of the future. Sound historiography is essentially future-committed practice.
Conclusion
If the study of history is a study of causes, we have to have tools to organize causal knowledge. This organizing is possible only if we commit ourselves to historiographical generalizations, invariance, theory-drivenness, and historical significance. While these commitments appear unnatural when rejection of big-pictures and historical coherence revolve around sowing destruction, we have seen that the commitments are defendable and make sense. We could even go further and say that those commitments provide a rather balanced and powerful picture concerning historiographical explanation. Therefore, the study of history is a study of causes.
Yet, there is one final point to be made. Carr thinks that historical counterfactuals do not belong to serious historiography. According to Carr, “the ‘might-have-been’ school of thought – or rather of emotion” (p. 96) is not serious historiography. “[O]ne can always play a parlour game with the might-have-beens of history. But they have nothing [–] to do with history” (p. 97).
However, our analysis of Carr’s philosophy used an essentially counterfactual notion of invariance and Carr also seems to be committed to the counterfactual difference-making conception of causality. For example, he writes “[T]wo distinguished gentlemen [–] burst into the room and begin to tell us, with great fluency and cogency, that, if Robinson had not happened to run out of cigarettes that evening, he would not have been crossing the road and would not have been killed; that Robinson’s desire for cigarettes was therefore the cause of his death; and that any inquiry which neglects this cause will be waste of time, and any conclusions drawn from it meaningless and futile. Well, what do we do? As soon as we can break into the flow of eloquence, we edge our two visitors gently but firmly towards the door, we instruct the janitor on no account to admit them again, and we get on with our inquiry. But what answer have we to the interrupters? Of course, Robinson was killed because he was a cigarette-smoker. Everything that the devotees of chance and contingency in history say is perfectly true and perfectly logical.” (p. 104-105 [emphasis added]). The problem is not that “Robinson had not happened to run out of cigarettes that evening, he would not have been crossing the road and would not have been killed” cannot be known or asserted. Rather, the problem is that the statement is does not identify a non-accidental cause but an accidental one. And if the assertion captures a causal relation (even if only an accidental one), then there is some intimate relationship between counterfactual might-have-beens and historical causality.
It seems that Carr’s skepticism towards historical counterfactuals must be abandoned in order to keep his ideas concerning historical causality alive. To me, that does not seem like a great price to be paid as the motivation behind Carr’s skepticism seems to be that counterfactuals imply that the course of history might have easily been different or could be reversed (p. 97-98). Of course, counterfactualists do not have to accept these claims (Virmajoki 2018; 2020). Given this, we can conclude that Carr philosophy of historical causality makes good sense, if only we inject some counterfactuals into it.
References
Carr, E. H. (1961 [1990]). What Is History?. Penguin Books.
Glennan, Stuart (2010). “Ephemeral Mechanisms and Historical Explanation”. Erkenntnis 72 (2):251-266.
Oakeshott, Michael (2004). What Is History and other essays.
Virmajoki, Veli (2020). ”What Should We Require from an Account of Explanation in Historiography”. Journal of the Philosophy of History.
Virmajoki, Veli (2018). ”Could Science Be Interestingly Different?”. Journal of the Philosophy of History 12 (2).
Woodward, James (2003). Making Things Happen.
[1] If this sounds unclear, consider the setting of traps. The whole practice of setting of traps is directed towards changing the “causal environment” in the way that makes more possible actions deadly in the environment.
[2] Notice that there are many cases where absences explain without any tricks. That the aircraft carriers were not at Pearl Harbor explains why the attack was not decisive.
[3] Assume that Robinson does not have any obvious signs of injury because if he does, then it becomes possible to make an inference from these signs to the non-accidental causes and therefore identify the non-accidental cause against the background of non-accidental causes.