In Laws and Explanation in History, William Dray takes a journey through the issue of historiographical explanation. In this post, I will explicate the many different conceptions of historiographical explanation that Dray is able to find. While there are many important arguments in the book, the book is still centered around explicating the problems of the so-called covering law model of historiographical explanation (“the DN model”) to the point where many of the important arguments are lost. Excessive fear of some logical loophole from which some form of the covering law model could enter the historiographical explanations shadows and stagnates the development of the more fruitful ideas that Dray finds. Now, over 60 years later, these worries seem exaggerated, but we have to remember that the covering law model seems defeated only due to the efforts of its critics such as Dray. However, we can make better sense of Dray’s insights if we do not equate any possible use of laws (or generalizations or regularities) in historiographical explanation with the acceptance of the covering law model. In what follows, I will briefly introduce the covering law model (also known as the DN model [deductive-nomological]) and then discuss how Dray argues against it. I then proceed to the alternative theories of explanation that Dray explicates.
Dray’s Critique of the Covering Law Model
According to the covering law model (the DN model) (Hempel 1942), an explanation consists of
(1) a set of statements asserting the occurrence of certain events C1, . . . C, at certain times and places,
(2) a set of universal hypotheses, such that
(a) the statements of both groups are reasonably well confirmed by empirical evidence,
(b) from the two groups of statements the sentence asserting the occurrence of event E can be logically deduced.
The model is also called “deductive-nomological” (DN) because an explanation deduces the explanandum from initial conditions and nomological (law-like) statements.
According to DN model, the explanation of some occurrence of E (i) makes E expectable, and (ii) shows how E could have been predicted. There is symmetry between explanation and prediction, according to the model. It is important to notice that Hempel writes that “a universal hypothesis may be assumed to assert a regularity of the following type: In every case where an event of a specified kind C occurs at a certain place and time, an event of a specified kind E will occur at a place and time which is related in a specified manner to the place and time of the occurrence of the first event” (Hempel 1942, 35).
The DN model was initially supposed to be a general model in that it was intended to fit all fields of research from physics to historiography. The model is no longer accepted due to many problems – ranging from simple counterexamples to unclarities in the notion of lawfulness to the observation that explaining rarely involves deducting – it faces and because more powerful accounts of explanation have been developed. In this post, I set these later developments aside and focus on Dray’s arguments against the use of the DN model in historiography. We do not focus on the general problems of the DN model but on its difficulties to make sense of historiographical explanations. In this way, we can find from Dray issues that are central to historiographical explanations.
Dray argues that the DN model “is, in fact, so misleading that it ought to be abandoned as a basic account of what it is to give an explanation” (p. 18). Dray’s argumentative strategy appears to be two-fold. First, he criticizes the DN model by arguing that it does not provide necessary nor sufficient conditions for explanation in historiography. Secondly, he formulates alternative accounts of explanation that (supposedly) fit better historiographical practice. These two approaches are intertwined.
First, in order to show that laws are not necessary for explanation, Dray argues that explanation in terms of laws is, in most cases in historiography, either practically impossible or redundant and that there are alternative accounts of historiographical explanations that do not require general laws. This means that we cannot make sense of most historiographical explanations in terms of general laws but we can make sense of them in other terms. This issue is a rather complicated one because everything depends on what, exactly, is meant by the notion of law and how laws are supposed to be used in historiographical explanations, as we will see.
Secondly, in order to show that general laws are not sufficient for explanation, Dray argues that in most cases we require something else in historiographical explanation. He goes on to analyze what this “something else” is and sketches accounts of explanation in those terms. As a result of his discussion about the problems of the DN model, Dray provides a variety of suggestions concerning the nature of historiographical explanations where laws do not play the role that the covering law model supposes them to play.
So what are the arguments against the idea that historiographical explanation necessarily involves laws? We can begin from the obvious one: Historians rarely cite laws in their explanations. If laws really are necessary for explaining in historiography, there must be more than meets the eye in historiographical explanation; there must be some hidden structure in historiographical explanation. This puts the burden of proof to the supporters of the DN model. As Dray argues, the supporters must tell why such hidden must be postulated.
Dray notes that the supporters of the DN model often argue that while there rarely are explicit laws in historiographical explanation, those laws must be implicit in an explanation. For example, Hempel claims that actual explanations in historiography are only “explanation sketches” that “consists of a more or less vague indication of the laws and initial conditions considered as relevant”. How should we understand this idea of implicit laws? In some cases, according to the supporters of the DN model, the relevant laws are so trivial that they do not require mentioning, such as “[if] of two armies which are about equally well armed and led, one has a tremendous superiority in men, then the other never wins” (p. 24, citing Popper). If triviality was all there was to the laws in historiography, then we could understand the claim about the implicit use of laws: they are implicit merely for notational reasons. Every competent reader could easily explicate them if needed. However, such trivial laws hardly explain most historiographical explananda.
Dray discusses the following example: “Louis XIV died unpopular because he pursued policies detrimental to French national interests”. What does it mean that there is an implicit law in such explanations? Dray argues that there are two possible strategies for the supporters of the DN model to justify the postulation of implicit laws. The first one is based on logic of explanatory statements, the second one is based on methodological considerations.
First, one could argue that “The word ‘because’, and the many substitute expressions for it which are to be found in the historian’s explanations, [–] depend for their very meaning on some kind of related general statement” (p. 26). Dray points out that this begs the question. The debate concerning the DN model is about how to interpret the meaning of explanatory claims. One cannot argue that the DN model is correct because it captures the meaning of explanatory claims. The definition of “because” that the DN model provides does not exist, according to Dray, outside philosophical debates.
Alternatively, one could argue that explanatory statements commit historians to laws because it is unintelligible to assert an explanation and deny that there exists a law. However, historians seem to do exactly this: they explain and still deny laws. If a philosopher is to find a law that the historian cannot deny, the law must be extremely abstract. For example, a philosopher might not be able to convince historian that “farmers will always leave dry land when damper areas are accessible” is a law that a historian must accept if she is to explain a migration-pattern, but the philosopher might convince the historian that there is the law “people do what they think is best to them”. However, it seems that in this case the point of the DN model is lost. Dray argues that the extremely trivial abstract laws do not serve any methodological point. This is true and sufficient, in my opinion, to disregard the DN model. However, a supporter of the DN model might not be interested in methodology at all and while a philosophy of historiography that is completely cut off from actual historiographical practice surely is of little interest, the DN model could survive and have a Pyrrhonic victory if its only problem was the silence about methodological issues. The death blow would have been easy to deliver by Dray. For we can notice that the statement (S) “people do what they think is best to them” does not connect the dryness of land to migration even though the dryness is the explanatory factor that the explanation above clearly underlines. The abstract generalization (S) to which a historian might be committed is not the one that could make the explanatory work on the statement “farmers migrated from area A because A was dry”.
Dray also discusses the possibility of loosening the laws to the form “If C then usually E”. He asks: “Does the ‘law’, ‘Whenever C then usually E’, really explain the fact that in this case an E followed a C? Would not the same ‘law’ have ‘explained’, in the same sense, the non-occurrence of an E as well?”. Dray misses important issues here concerning probabilistic laws and explanation. Syphilis explains paresis even though not everyone with syphilis gets paresis. However, the theories that have serious power to account for probabilistic explanation are not in line with the DM model (see e.g. Salmon 1989). It would have been sufficient to underline the fact that if the DM model loosens the laws, it loses its original claim that explanations allow us to deduce and make thus the explanandum event one that could have been expected. Of course, there exists the so-called Inductive-statistical model (IS model) of explanation (Hempel 1965) where the probability of the occurrence of E can be inferred from initial conditions and statistical laws. However, we cannot require that Dray should have proved the model wrong in historiography in 1957. It should have been the task of the supporters of the IS model to show how the model can be naturally applied to historiography. My hunch is that the model would not survive the core of Dray’s criticism thus far that historians certainly do not appear to be committed to any sort of historical laws, whether universal or statistical in form, and those who demand laws in historiographical explanations have the burden of proof. Remember that we are not talking about generalizations that could have heuristic value in historiographical explanation. We are discussing models of explanation that make very specific claims about the role of laws in the structure of historiographical explanation.
Finally, Dray argues that the supporter could argue that once there is a nearly complete explanation that cites a set of factors F, then the historian is committed to the claim that “whenever F, then E”. We move from “Rulers who pursue policies detrimental to their subjects’ interests become unpopular” to “’Rulers who involve their countries in foreign wars, who persecute religious minorities, and who maintain parasitic courts, become unpopular” and so on. Dray writes that “[Despite unnaturalness] the eliciting of such a vacuous ‘law’ does show that the argument from meaning – the conviction that some sort of generality was logically involved in the original explanation – was not entirely an illusion”. I think this is a completely unnecessary concession to the DM model. In this case, the law has no other content than to claim that the set of factors F leads to E. The supposed general law does not perform any explanatory function. The case is similar to one where a physicist claims that he can explain the time it takes a rock to hit the ground when released from a tower by measuring the time and then specifying that “whenever the same rock is dropped from the same tower in same conditions, the fall takes the same time”. This is not a law, this is merely stating that initial conditions affect the fall in some way. In contrast, with Newton’s laws, we can calculate the duration of the fall by citing the important aspects of initial conditions and the laws independently of measuring how long the fall in fact takes. Newton’s laws tell us how the initial conditions affect the duration of the fall. In fact, this concession even suggests that Dray is a bit confused about the DN model. Surely, he is correct in arguing that there hardly is any absolute generality in historical explanations beyond the trivial “Whenever exactly the same causes occur, the same effect occurs” but this does not mean that there was some minimal truth in the DN model. A consistent supporter of the DN model would not think that this trivial generalization is a law in any relevant sense. If there is one correct insight in the DM model, it is the claim that if universal laws can be found, then we can explain because we know how the initial conditions shaped the effect. So either there is no concession to be made for the DN theory or Dray misrepresents the DN case. Later in the book Dray even writes:
“If covering law theorists were right in claiming that their model formulates a sufficient condition of explaining something, then reference to a covering law ought always to explain what falls under its apodosis clause. But it is surely not difficult to think of cases from everyday affairs which furnish evidence to the contrary. When puzzled by something, we do not ordinarily find it enlightening to be told: ‘That’s what always happens.’ Indeed, although such a remark appears to be just an idiomatic, incomplete way of subsuming what happened under a general law, we should often feel justified in protesting: ‘That’s no explanation at all.’” (61).
It should be obvious to anyone that the DM model does not suggest that the statement “that’s what always happens” is explanatory. What it claims is that we can explain an event when we can tell how the event follows from the initial conditions and laws. True, it also commits to the claim that “whenever the initial conditions are the same and given the laws, the same thing occurs”. However, it is not this latter claim that is explanatory. It is the initial conditions and the laws that are explanatory. If an event is explained by initial conditions and universal and deterministic laws, then it maybe follows that the same conditions always produce the same effect. However, the latter regularity is irrelevant for the initial explanation in itself. The only things that provide explanatory understanding are the laws and initial conditions. The latter regularity is also epistemically sterile. For example, we do not test Newton’s laws by applying them to similar conditions over and over again and checking whether the same effect occurs. Instead, we investigate different sets of conditions and check whether they work in accordance with Newton’s laws. That is what the universality and general scope of the laws are supposed to mean.
I wanted to point out this issue because, as we will see, for Dray (and some of his reviewers) the DN model appears to be a kind of bogeyman that might crawl from under the bed if even the slightest mention of laws and regularities can be found in historiography. It is important to notice, again, that the DN model is a specific suggestion of how laws are supposed to perform an explanatory function. A mere hint towards a regularity does not count as an explanation in the DN model. In order for the DN model to be successful and to show that historiographical explanations do not have any unique characters, the case needs to be made that historical events are explained by citing universal laws with a wide scope that tell how the effect followed from a set of initial conditions. Commitment to some regularities whatsoever at some stage of investigation does not bring the bogeyman alive.
Let’s move on and see how Dray deals with this issue. Dray argues that even if the concession is made, the supposed general law “whenever a set of factors F is present, E follows” does not, indeed, satisfy the requirements of the DN model. According to Dray, it is not an empirical law but “a principle of historian’s inference”.
“For if the logician’s statement ‘if p then q’, is to be understood in conjunction with the rubric, ‘we can infer that … ‘, rather than ‘we have found that … ‘, to say that the historian’s explanation commits him to the covering ‘law’ is merely to say that it commits him, in consistency, to reasoning in a similar way in any further cases which may turn up, since he claims universal validity for the corresponding argument, ‘p so q’.” (41)
The bogeyman is still there. As Dray notes, the statement “if F, then E” is without independent methodological use. “There is no point in saying that it is used, or functions, in the explanation; and there is no point in asserting it except to register one’s belief that the inference drawn was a reasonable one.” In other words, the historian must find out and justify that “E because F” independently of “if F then E”, and the latter becomes known and justified only once the “because” is justified. The statement “if F then E” depends on “E because F” and not the other way around like the DN model suggests. This means that when Dray says that “if F then E” is a principle of inference, it is not a methodological principle of inference. Rather it commits the historian, according to Dray, to the statement that “If there had been, or were to be, [F], then there would have been, or would be, [E]” (p. 41). Why does Dray bother to dig up this extremely weak commitment to a subjunctive general law? He says that it is this subjunctive law that might make the DN model persuasive. He feels that if the complete lack of methodological relevance of “if F then E” is not illustrated, the extremely weak but still possible commitment to a subjunctive law could lead us to accept the DN model. This procedure surely takes care of the bogeyman once and for all. The only problem is that if we think that any commitment to lawlike patterns leads to the DN model, we probably lose too many conceptual resources to make sense of historiographical explanation, as we will see below. We should not burn the bed to kill the bogeyman.
The problem becomes obvious when Dray wonders whether citing general laws is required to justify an explanation. He, correctly, argues that a law of the form “if the set of factors F occurs, E occurs” cannot provide any external justification for “E because F” since the former is dependent on the latter in its justification (see above). Dray says that is the historian’s judgement that allows to conclude that “E because F”. Fair enough. However, the problems begin when Dray notes that perhaps the supporters of the DN model would say that we need to cite a complex set of laws to justify the conclusion “E because F”. Dray’s arguments are not quite as convincing here as they have been before. He ends up writing:
“The covering law claim would now be, however, that a satisfactory explanation would have to specify conditions and laws such that from the conjunction of statements listing both conditions and laws the occurrence of what is explained could be deduced.
What does such an account leave out? The missing element is surely a ‘law’ or ‘rule’ which would inform the historian when such a group of ‘predisposing’ conditions becomes sufficient. Laws which allow him to regard each of a number of conditions as ‘favouring’ the occurrence of what is to be explained cannot simply be assumed to constitute a covering conjunction allowing the explicandum to be deduced from the explicans.” (p. 55).
This is problematic. First, Dray assumes a complex set of laws would tell how the individual initial conditions increase the probability of the explanandum-event. However, this surely is not a standard case of explanation. For example, building a huge cruiser ship increases the GDP. Without the building project, the GDP would have been smaller. The building project explains, although not alone, the GDP. However, it would be completely unnatural to suggest that the building of the cruiser only increased the probability of the actual GDP. Ceteris paribus, without the ship the GDP could not have been the same for sure.
Secondly, and more importantly, it certainly is misleading to require that we need meta-laws that tell that some laws determine a phenomenon. Once we explain the duration of the fall of a rock by using Newton’s laws, we have a real explanation. What would it even mean to suggest that we need some further law that shows that Newton’s laws made sufficient explanatory job?
Again, it seems that Dray loses his calm in the face of the bogeyman. It surely would have been enough to say that once there are very few know historical laws, using a complex set of general laws to justify an explanatory judgement cannot be necessary. Surely, we need all sorts of general knowledge in order to justify what would have happened without some cause, but this is different from what the DN model suggests. Again, we have to remember that the DN model is a rather specific suggestion about the role of general laws in bringing explanatory understanding. The whole motivation behind the DN model is its rigorousness; this rigorousness is the main thing that makes it so interesting analytic tool despite the fact that it does not fit the actual explanatory practices in most fields. The DN model is not equal to suggesting that we need general knowledge about the world when we go about our explanatory practices. The fear of the bogeyman makes Dray leave the nature of historian’s judgement opaque. I suspect that the judgements are based on historians’ understanding of how the world and historical societies generally work(ed). Dray does not take the risk of waking the bogeyman and this leaves his discussion about historian’s judgements unsatisfactory. As Strawson (1959) writes:
“But what makes such judgments acceptable? How is it that judgments coincide as much as they do? Surely something about general knowledge of human nature, about human experience of how things tend to go in the worlds of private and public affairs, might be mentioned here, not indeed by way of providing covering laws, but in the course of explaining what the exercise of judgment is, and how it can establish these satisfactory connections of which Mr. Dray’ speaks. As it is, ‘judgment’ appears as a slightly mysterious gift of nature to historians. I think there is a twofold explanation of Mr. Dray’s reticence on this point. Partly it is the result of that over-concern with his opponents’ folly, which will not allow him to grant them their grain of truth.”
We have now seen that, according to Dray, general laws are not necessary for historical explanation. But are general laws sufficient for explanation? In other words, even if we can explain without general laws, is it still the case that once we do have the relevant laws, we can automatically give a satisfactory explanation? Does the DN model guarantee that once we have an argument with the DN-form, we have a satisfactory explanation?
Dray begins by noting that there are, prima facie, clear counterexamples to the sufficiency: “What, for instance, is the explanatory force of the common-sense generalization, ‘Red sky in the morning is followed by rain’? Does the fact that the sky was red this morning explain the fact that rain fell before lunch? Surely not.” (p. 61.) In order to save the sufficiency of the DN model, it would be necessary to distinguish different kinds of laws. For example, we can distinguish between laws of theoretical sciences and mere inductive laws. However, this is not a sufficient defense of the DN model as both types of generalizations can be used in prediction, and prediction and explanation are symmetrical operations, according to the DN model. If both laws can predict, both can explain, given the symmetry. Moreover, there are not too many theoretically based laws in historiography and still historiography explains. Even if a distinction between theoretical and other generalizations could be established, the sufficiency of the DN model would not be established.
Next, we get to the most fruitful aspects of Dray’s book. He argues that the DN model is not sufficient because there are cases where a law can be provided but explanatory understanding is not achieved:
“The general law, ‘Whenever your oil leaks out your engine seizes up’, does not explain the fact that my engine seized up after my oil leaked out-in the context of puzzlement envisaged. But reference to a series of facts constituting the story of what happened between the leakage of the oil and the seizure of the engine does explain the seizure.” (p. 70.)
Now it is time to take a closer look at the alternatives that Dray proposes.
The Model of Continuous Series (MCS)
Dray suggests that we can, at least in many cases, explain some event E by tracing the course of events by which it came about. The idea is to reveal a mechanism. (p. 68.)
“[R]eference to a series of facts constituting the story of what happened between the leakage of the oil and the seizure of the engine does explain the seizure. Even if it were true that these smaller scale events were each covered by law in the sense that in every case I would be prepared to assent to a law corresponding to a sub-sequence, the laws involved would be, at most, part of the explanation of the gross event, not of the sub-event they cover; so that when they do function in an explanation they are not covering laws at all. Thus, although the engine seizure-the gross event-may be said to be explained by assuming many sub-laws like, ‘When the walls of a cylinder and piston are dry they heat and expand with motion’, the law mentioned would not in turn explain why the piston expands and heats up- if we were to go on to ask that question. The sub-law is part of the explanation of the gross event, although it does not cover it; the same law covers the sub-event, although it does not explain it.” (p. 70.)
In essence, Dray suggests a mechanistic account of explanation. While I am critical towards the general applicability of mechanistic accounts of explanation in historiography (see this post), the arguments of Dray are sophisticated. He considers two objections to the MCS.
First, it can be said that the series of events can be divided (almost) infinitely into subseries. How can we tell when we can stop the subdivision and find “real” understanding? Dray says that, in practice, an event E is explained by a series C1, C2, C3… E, once the constituents of the series do not “raise [–] further demand for explanation in [the given] context”. This is in line with the recent explosion in the literature concerning mechanistic explanation: “Nested hierarchical descriptions of mechanisms typically bottom out in lowest level mechanisms. These are the components that are accepted as relatively fundamental or taken to be unproblematic for the purposes of a given scientist, research group, or field” (MDC 2000, 5.1.) The potential divisibility of a series into subseries is not a series problem in the light of our current understanding or even in the light of Dray’s example.
Secondly, it can be said that we cannot get rid of covering laws because they describe how the events in the series are connected. We have already seen the answer to this worry in the quotation above: If we explain E by citing the series C1, C2…, then the law L that explains the connection e.g. between C1 and C2 would not cover the explanation of E; it would cover only the connection between C1 and C2. What provides explanatory understanding is not the covering laws involved themselves but the fact that we can represent a series of events leading to the explanandum. Moreover, even though we can cover individual links with laws, this does not amount to explanatory understanding about the link. If we wanted to understand that link, we should represent the series connecting e.g. C1 and C2.
I am not fully convinced of the details of the second argument. It does not seem to be true that we could never explain an event by using a covering law that connects it to a previous event, i.e. that we always need a connecting series. Be that as it may, the argument is convincing if understood as suggesting that, in many cases, there is much more to historical explanations than what the covering law model suggests. This establishes the insufficiency of the covering law model a general theory.
What is refreshing in the model of the continuous series is that the bogeyman seems to be defeated. Dray provides an eloquent discussion about different-level laws and their use in explanatory understanding and simply points out that covering laws might be found in historical explanations, but they do not function quite the way that the DN model suggests. We can have laws and generalizations without falling back to the DM model.
Finally, Dray connects the MCS with the idea that explanations have to make phenomena intelligible. We want a mechanistic story about the connection between C and E when the connection is unintelligible. (p. 73). Now, this is probably true, but there is the question of when something is intelligible. Dray makes a very unfortunate move concerning this issue. He argues that theories are not automatically explanatory because they might leave the phenomenon under investigation unintelligible (p. 79-80). More specifically, theories do not explain if they do not have an analogical model in well-known observable phenomena:
“My contention is, therefore, that in so far as the light ray theory explains shadow phenomena, it is because of its implicit reference to rays of light running tramlike along celestial rails from a certain source. Similarly, that the volume of a gas expands with increase of heat, is explained by the kinetic theory of gases, in that it allows us to think of gases as composed of little particles which increase the momentum with which they strike the sides of their container. Thus the role of theory in such explanations is really parasitic upon the fact that it suggests, with the aid of postulated, unobservable entities, a ‘hat-doffing’ series of happenings which we are licensed to fill in. The theory allows us to tell ‘a likely story’ behind the appearances. But if the travelling of observable entities along observable rails in a similar way would not explain a similar pattern of impact on encountering a wall, and if the jostling of a tightly packed crowd would not explain the straining and collapsing of the walls of a tent in which they were confined, then the corresponding scientific theories would not explain shadow lengths and the behaviour of gases.” (p. 80 [emphasis in original]).
My hunch is that this account of intelligibility cannot be defended. Analogies can be very misleading and even hinder understanding. It is difficult to find a concrete analogy for Newton’s conception of gravitation and Descartes’s vortexes hardly were more explanatory than Newton’s action-at-distance even though the former were “more intelligible”. Even though theory-driven explanation might not always be the golden standard, it is difficult to understand how they could be completely parasitic to common-sense mechanistic explanations. Dray seems to overplay his argument here and, again, the only reason seems to be the fear of the bogeyman.
The things get even more difficult for Dray when writes that
“The peculiarity of the historical case is that, normally, each event in the series will be established independently from evidence. There will be no general theory, even of the mechanical kind, to make detailed research into the actual course of events unnecessary. [–] For to explain with the aid of a theory is to do indirectly what the historian, perhaps painstakingly and piecemeal, does directly: reduce what is puzzling to what is not. (81).
This doctrine of “reducing to familiar” is debunked. There is nothing puzzling about falling bodies, but the theories of physics that have been used to explain the phenomena have gotten more and more puzzling and further away from common sense. Moreover, there is the general problem that Dray has not established sufficiently that historical explanations can be established from evidence alone. All the has established is that general laws do not function in all explanations (in the way the DN model suggests). The problem is similar to the one we faced with Dray’s notion of historian’s judgement. There might be some truth to what is written but the fear of general knowledge and generalizations as indications of the DM model makes Dray fall back to an obscure methodological position.
In chapter IV, Dray discusses causal explanations and the relationship between causality and explanation. His main argument is that “that causes seldom explain their effects by virtue of some implicit theory-indeed, that they need not explain their effects at all” (p. 86). However, he points out that it is a mistake to think that historians do not even attempt to provide causal explanations. “Now it is perfectly clear that, no matter what these theorists say, historians do commonly attempt to provide causal explanations of what they study. This is a fact which can be verified by the most cursory glance at one or two standard history textbooks.” (p. 90.)
He begins by discussing the suggestion that the DN model should be defined in terms of causal laws. The idea is that mere regularity between C and E is not explanatory. According to Dray, there needs to be some “ground” or “reason” why the cause is followed by the effect; the relationship between cause and effect has to “especially tight and intelligible” (p. 87). The suggestion is that the tight and intelligible connection is achieved through a theoretical grounding of causal laws.
Dray counters this suggestion by pointing out that causal explanations are often not achieved through the application of known theories. This is especially rare in historiography. He argues that causal relations are often known because we learn how to do something; causes are handles – like Collingwood suggests – that enable us to change the effect; they enable us to manipulate the effect. We can learn causal laws like “Dirt causes disease” by noting its practical utility. However, Dray argues that in such cases, we have not found an explanatory law because “no explanatory connexion between cause and effect is known. As Collingwood himself pointed out, the criterion of ‘the handle’ [–] is appropriate to the practical rather than the theoretical (i.e. explanatory) sciences.” (p. 97.)
It is difficult to understand, especially in the light of the current success of the interventionist theory of explanation (Woodward 2003), what is meant by “explanatory connection” if a relationship that allows manipulation is not explanatory. Prima facie, Dray seems to demand too much: We already distinguished causal generalization from other kinds of generalizations, the latter being presumably non-explanatory. To require further that we still need to tell some additional details about what makes the causal connection explanatory seems odd. As Dray said earlier with respect to MCS, explaining needs to stop somewhere. Why not on causal understanding? That seems kind of an obvious choice. It is even more odd that Dray seems to be implying that we need theoretical knowledge to achieve explanations. In the previous section, we saw that he argued that there are no general theories in historiography and that “[a] theory of the subject matter, as we have seen, may excuse an investigator from explaining a thing historically” (p. 84).
Thus far, Dray has not provided us with any insight on explanatory causal connections. He continues his inquiry into the issue of the selection of causal conditions. First, he discussed many practical dimensions that effect the choice of the explanatory condition. This is a good analysis for the most parts but does not answer the question ‘What counts as a cause?’. Only after the pragmatic discussion, he states the following:
“We must remember, as always, that [a historian] is talking about particular happenings in a quite definite historical situation. When he says that y would not have happened without x, he does not mean that only in situations where there is an x-type event can you expect a y-type. He means that in that particular
situation, if everything else remained the same, the y which in fact occurred would not have done so; or, at any rate, that it would have been different in important respects.” (p. 102).
Basically, we have a counterfactual theory of causal explanation in historiography here. X is the cause of Y if changing X would change Y when other things remain equal; had X been different, Y would have been different. Obviously, I think this the correct theory because I have defended it and developed it further (2019, 2020).
The only problem with Dray’s discussion is that he does not tell how we get to know that without x, y would not have happened. He points out that
“It is true that the historian must be certain that without x, y could not have happened, if he is to say without
qualification that x was the cause of y. But there is no need to assume that the only way he could arrive at such certainty is by knowing a law of the ‘only if’ form. As historical methodologists have often pointed out, what the historian has to do is ‘think away’ the suggested cause in order to judge what difference its non-occurrence would have made in the light of what else he knows about the situation studied.” (p. 104).
The choice is not between (non-modal) laws of the form “Y only if X” and some mysterious judgements. Even if the only-if laws are rare in historiography, we need some laws and generalizations to formulate the counterfactual scenario where the X did not happen in order to tell what would have happened without X.
Dray goes on to argue that “the relation of ‘support’ between laws and the particular connexions falling under them, is at times precisely the opposite of the one envisaged by covering law theory; for in many cases discovery of individual causal connexions precedes the formulation of causal laws, the laws-shocking though it may be to say if-requiring prior knowledge of the particular cases, rather than the cases requiring support of the covering causal law.” (106). Again, this is probably true but misses important elements.
First, the discovery of individual causal connections requires causal background knowledge. The question is not whether you apply general causal knowledge to individual cases but how you do that in order to get your individual causal judgements of the ground.
Secondly, causal reasoning is neither completely deductive nor completely inductive. The problem is not whether we should apply general knowledge to individual cases or whether we should build general knowledge from individual cases. The problem is how to combine these two approaches and balance the general causal knowledge and the study of the individual cases. General knowledge guides the case-study reasoning, but the case study reasoning can also shed new light on the accepted knowledge. Even if use a case C to justify a generalization G, we can still use G to explain C, unlike Dray claims (p. 109-110). What matters is that G has support from other cases than C.
Thirdly, generalizations are not mere summaries of what we know about individual cases. For example, generalizations may enable us to understand events in a new light. Take a toy-example: We wonder why British and French soldiers have not killed each other for quite some time. How come? Well, Britain and France are democracies. Moreover, soldiers kill each other mainly in a war. Why is this a relevant way of looking at things? Because we know that two democracies rarely go into a war with each other. In this way, generalizations suggest new perspectives on events and states of affairs. Moreover, two events may fall under different generalizations and the generalizations may differ in their explanatory depth (see Virmajoki 2019; 2020; 2021). For example, if one of the generalizations is more invariant with respect to changes in background conditions, it provides a deeper explanation. This also provides further support for the fact that a case C and justify generalization G even if G is arrived at by the study of C. If G is widely invariant, then it can be applied to many cases beyond C. The generalization does not merely repeat some features of C; it also tells us how those features would work in other contexts.
Again, Dray puts too much weight on historians’ judgements concerning causal explanation and too little weight on general knowledge behind the judgements. The fear of the bogeyman rises again. However, one does not need to kill all explanatory generalization in order to get rid of the bogeyman. It is enough to point out that the interplay between different types of generalizations and causal judgments goes far beyond what the DN model suggests. Yet, I am not suggesting that generalizations can make historical judgements redundant. Even when we have many generalizations, telling “what would have happened” is never a straightforward and completely objective task (see Virmajoki 2018).
Historically, it is interesting to notice that Dray does not contrast causal explanations in historiography with rational explanations. Let us look at the latter issue next.
Next, Dray direct attention “to a narrower range of cases: the kind of explanation historians generally give of the actions of those individuals who are important enough to be mentioned in the course of historical narrative. It will be my thesis [that] the explanation of individual human behaviour as it is usually given in history has features which make the covering law model peculiarly inept.” (p. 118).
Dray attempts to make sense of the doctrine that “is commonly expressed with the aid of a characteristic set of terms. To understand a human action, it will be said, its necessary for the inquirer somehow to discover its ‘thought-side; it is not sufficient merely to know the pattern of overt behaviour. The historian must penetrate behind appearances, achieve insight into the situation, identify himself sympathetically with the protagonist, project himself imaginatively into his situation. He must revive, re-enact, re-think, re-experience the hopes, fears, plans, desires, views, intentions, &c., of those he seeks to understand. To explain action in terms of covering law would be to achieve, at most, an external kind of understanding. The historian, by the very nature of his self-imposed task, seeks to do more than this.” (p. 119).
It should be remarked right way that, while this doctrine has intuitive appeal, it is quite obscure in comparison to the DN model, something that G. H. von Wright noted in the ‘60s. However, Dray has the advantage that he sees explanation of actions a “narrower range of cases” within historiography and does not define historiography to consist solely of such explanations. This does not cut off the relationship between historiography and other fields of inquiry and makes it possible to use a wider set of analytical tools to make historiographical explanation understandable.
Dray argues that “when we ask for the explanation of an action, what we very often want is a reconstruction of the agent’s calculation of means to be adopted toward his chosen end in t e light of the circumstances in which he found himself.” (122.) For example, to understand why Eddington executed an expedition to test Einstein’s theory, we need to understand how he (a) thought that the theory makes a plausible prediction that is possible (although) difficult to measure, (b) thought that the confirmation would unite international scientific community, (c) thought that using particular instruments in a particular spatio-temporal location could provide relevant data, (d) thought that it is his duty, as a Quaker, to unity international community. Eddington, in a sense, calculated that his scientific and ethical-political interests are best satisfied by executing the expedition.
Dray underlines the main idea: “The goal of such explanation is to show that what was done was the thing to have done for the reasons given, rather than merely the thing that is done on such occasions, perhaps in
accordance with certain laws (loose or otherwise)”. (p. 124, emphasis added). Explanation and justification of action are rather symmetrical. Given the agent’s circumstances, beliefs, and principles, his action is explained when it is shown to be the appropriate thing to do in the light of those circumstances, beliefs, and principles. However, this does not mean moral approval or anything like that. We can strongly disagree with the beliefs and principles.
Dray points out that rational explanations are not self-evident or easily achieved. Neither do they require or indicate any unique skill to penetrate through empirical evidence. A rational explanation is reconstructed by searching for equilibrium between the action and the principles and beliefs of the agent to the point that these match each other (p. 125). For example, I might play a game of chess and have the principle “do not sacrifice a knight for two pawns. However, in the game I do exactly that. My opponent plays pawns to h6 and g5; I capture g5 with the knight; he captures my knight with the h6 pawn; I capture his pawn with the bishop. To understand this, we might note that I needed a win, and a good way (although risky) to force a win is to expose the opponent’s king. A balance is found in my principles and the action of sacrificing the knight is explained.
Dray discusses a counterargument from the DN model that empathy is merely a methodological tool: it enables us to generate hypotheses about general laws concerning human actions (by generalizing from the knowledge of what I would have done) and to reach the beliefs and principles of the historical actors that can be then subsumed under general laws. Dray counters this argument by noting that “[it] is to obliterate a distinction between explanation types: a distinction between representing something as the
thing generally done, and representing it as the appropriate thing to have done” (p. 129).
This is somewhat puzzling. It seems that the distinction between “things that are generally done” and “things that are appropriate to be done” cannot divide actions into two groups, rational and general. If actions were thus divided, not all actions are rationally explained and the power of the account of rational explanation would be seriously limited. What is worse, the division would suppose that rational action is not something that is generally done; that it has some mystical element that enables the actor to overcome the limits of human nature that usually determine their actions. Even if we put the dubious metaphysics aside, it seems difficult to understand how we could be able to identify a rational action and confirm it by evidence if rational actions were a sui generis within actions. (See discussion below.)
Because the distinction (seemingly) cannot be about separating different types of actions, it has to be about different perspectives on all actions. Dray seems to claim that a rational explanation of action, which shows that the action was the thing to do, has special explanatory force. It is difficult to tell where the force comes from. Dray argues that there is a difference between a general empirical law and a principle of action. Even though reasons for acting have generality in the sense that “[i]f y is a good reason for A to do x, then y would be a good reason for anyone sufficiently like A to do x under sufficiently similar circumstances” (p. 132), this generality differs from the generality of empirical laws. An empirical law “if x then y” is falsified if x is not followed by y, but a principle of action “when situation c, y is the thing to be done” is not falsified if c is not followed by y. According to Dray, we can postulate that an agent A has the principle P even if it is not always the case that A acts in accordance with P; and we can say that P still explains A’s actions in the cases that are in accordance with P. (p. 132-133).
This is difficult to understand. If the distinction between rational and “general” explanation is merely in the perspective or “between the levels of language at which we talk about actions” (p. 136), how can it be that sometimes we lack a rational explanation because we are unable to subsume the case under a principle of action that is supposed to hold? Notice that we did not abandon and replace the “falsified” principle of action; we noted that it does not apply to a particular action. What, then, explains this action? How can its rationality be shown? If principles of action tell us what is rational and if an action does not fall under such principle, how can it be rational? If it still is necessarily rational, why is not the presumed principle of action falsified? If it is falsified, the principles of rational action need to be investigated empirically and cannot be achieved through any unique historiographical insight.
I have the feeling that the attempt to get rid of the bogeyman is obscuring the issue here. One reviewer seems to reveal what the nature of the bogeyman is:
“Are we not, having given our rational explanation, thrown back on the covering-law model, even if the only covering law we can cite is the portentious platitude that rational men who have worked out appropriate plans tend to act on them unless prevented?” (Nowell-Smith 1959).
In order to not fall back to a version of the DN model with the postulation of principles of action, Dray attempts to distinguish these principles from empirical laws. However, the way the distinction is made in terms of falsifiability leads to all sorts of difficulties, as we have seen.
Dray seems to notice the problem. “To say a priori that all actions must have a rationale, no matter how hard to discover, is just a dogma although we could make it analytically true by a suitable definition of ‘action’. In the ordinary course of affairs, rational and non-rational explanations of actions are alternatives -and alternatives sought in a certain order. We give reasons if we can, and turn to empirical laws if we must. Not only is this done in the ordinary course of affairs; it is done, too, in ordinary historical writing.” (p. 138.)
This means that we can explain actions even if we cannot reconstruct them as rational. It is not necessary to explain actions in rational terms. The only argument for rational explanation is that those explanations are especially intelligible. Notice that Dray explicitly says that the problem is not that in some cases we are not able to find a rationale but that there is none. This leads to difficulties.
Dray argues that, as we have seen, that we postulate beliefs and principles to a historical agent – even unconscious ones (p. 137) – in order to make it understandable. The principles and beliefs are chosen is a way that is consistent with the sources and that makes the principles and beliefs match the action. This heavy machinery becomes suspect if it really is the case that some actions do not have a rationale behind them. How can we justify such heavy machinery if it is not necessary for explaining actions? It seems that the machinery can go wrong in many ways. For example, how do we distinguish between rational action with unconscious motive and action without rationale? If it was the case that actions cannot be explained without finding a rationale, then we would naturally be ready to do whatever it takes to find some rationale because actions are too important to leave without explanation. However, now it appears that there is an alternative, the subsumption under empirical laws. Why take the risk of formulating a misleading rational explanation? Of course, we could say that if the rational explanation is correct, then we gain especially deep understanding. However, it seems too difficult to tell what indicates that a rational explanation is really correct, given the heavy machinery that is used to force them in place and given that there are no clear standards of testing the principles of action (see above).
Moreover, Dray argues that we can explain a “gross event” like colonialization by citing the reasons of individual people involved. This probably is not universally true. In many situations, different individual reasons would have led to the same outcome. The fact there is a certain number of students in medical school cannot be explained by the reasons of those students. Had they preferred something else, other people would have taken their place. Neither is the number explained by the reasons of those who made the decision concerning the number of the student. The decision is compatible with the possibility that no one applies. So there are complex dynamics that do not reduce naturally to individual reasons; to what someone thought was the thing to do. Reasons for individual actions can be irrelevant to the overall outcome.
All in all, it is difficult to give a verdict on Dray’s idea of rational explanation. It definitely has some intuitive appeal, but the machinery through which the explanations are produced is suspicious. The distinction between “what people tend to do” and “what is the thing to do in the particular situation” feels too strong. Probably Dray gives too limited an account of how a balanced view of human actions is achieved. When we explain why a new chess player plays e4; e5; Nf3; Bb5; Nc6; Bxc6, we might cite the fact that new players have the habit of capturing pieces just for no obvious reason; because they can; because its fun. When Magnus Carlsen plays Bxc6, we know there is an opening-preparation which tells that the move achieves this and that end. Same action, different perspectives, different explanations. Sometimes human nature explains more of our actions than in other cases.
Next, under the chapter concerning rational explanation, Dray considers the cases where human actions are given a dispositional explanation. “John hit you with a hammer because he is bad-tempered” is an example that Dray takes from the literature. He goes on to argue that these explanations do not fit the DN model. (Always: first kill the bogeyman…) He cites Gardiner (who cites Ryle) who argues that the DN model is not suitable because motives are not suitable antecedent events or processed to be connected by a law to the effect because they are not events or processes at all. This argument is not convincing all the way. The mass of the Sun explains the position of Venus at this moment. The mass, however, is not an event. If we say that the relevant antecedent condition that explains Venus’s position are events like “the Sun had such-and-such properties and position”, why could we not say that I got hit because there was Jones with such-and-such temper?
Dray admits that a disposition has some generality like laws because “it is at least partly hypothetical in what it implies; it can be satisfied bya wide range of behaviour” (p. 145). Dray notes that a supporter of the DN model could argue that in scientific cases, a dispositional property holds in virtue of physical laws and that, similarly, dispositional properties of human beings hold in virtue of laws (146). According to Dray, this analogy does not hold:
“For ‘ambition’ is not a general characteristic of men (or even, perhaps, of politicians) in the way ‘being brittle’ is of glass. To say ‘Disraeli attacked Peel because he was ambitious’ draws attention to the general pattern of action into which his particular action fits, but it implies nothing about the kind of men from whom this kind of action can be expected. It merely implies that action of this. general pattern can be expected from Disraeli; it subsumes his action under a regularity said to hold for a particular person, rather than a regularity said to hold for all persons of a certain type.” (p. 146.)
This argument is not convincing. We can classify things into “glass items” and find similarities in the dispositions of the items in the glass. In the same way, we can classify people into categories with respect to their levels of anxiety, intelligence, and so on. Surely, this requires an advanced conception of psychological testing, but the point remains: There are classes of people who share similar dispositions. Human dispositions are not unique to individuals.
Dray seems to understand this problem, as he continues as follows:
“Dispositional explanation thus falls short of law-covered explanation in its particularity [–]. It is accidental, not essential, to the explanation, that in the case of the glass we know that objects of this kind will have the dispositional property mentioned. The modification of the covering law theory represented by the recognition of dispositional explanation is therefore quite a major one.” (p. 146-147.)
This is not convincing. If a disposition of an entity was not shared with other entities of the same kinds, it hardly would be explanatory. If I say that my TV broke because Unto hit it and because it has the unique disposition “breaks-when-Unto-hits”, I would not have explained anything. I would merely have restated the fact that it broke because Unto hit it. This, of course, repeats the old virtus dormitive problem.
Be that as it may, dispositional explanations do not, according to Dray, account for all human actions. They cannot replace rational explanation:
“A pure dispositional explanation tells us that the person or thing under investigation tended to do things of (perhaps roughly) the sort done, under certain (unspecified) circumstances. It shows that what was done was the sort of thing we might have expected – it was the sort of thing that is done by this person or thing. But in most historical contexts, such an explanation would tell us scarcely anything we really wanted to know when we asked: ‘Why did he do it?’ For in giving the dispositional answer, the point of what was done tends to drop out of sight. To attempt to analyse explanations of the form, ‘A did x in order to achieve y’, as covertly dispositional simply ignores the question which we may reasonably assume the investigator to have had in mind when he represented this as an explanation.” (p. 148.)
There is a tension in the idea that some action can have both a weaker dispositional and a stronger rational explanation. How can it be that an action is at the same time part of a repeated pattern of a person’s life and still a deliberate action? Is the repeating pattern produced by repeated deliberations with the same effect? If I buy beer to my friends every time I finish a text about philosophy of history, is it natural to think that we can explain this both by (a) noting a disposition of becoming worry-free once finishing a text, and (b) some repeated deliberation (“I am happy; happiness is best when shared; my friends become happy if I buy them beer; therefore I buy them beer”). To me, it does not sound plausible that actions that follow repeated patterns are the result of case-by-case deliberation. When Dray writes that “There is a sense of ‘explain’ in which an action is only explained when it is seen in a context of rational deliberation; when it is seen from the point of view of an agent” (p. 150, emphasis added), I am willing to concede that if this is true, then many actions cannot be explained. My dispositions produce many actions that do not seem to have any “internal core” of strong deliberation. Reasons, habits, reflexes, dispositions form a complex interplay in human action. The category of rational action, as defined by Dray, is so limited that it cannot be in the center of historiography.
Causes and Reasons
What is the relationship between causes, dispositions and reasons? Dray writes:
“I do not think that the admission that ‘bad temper’ or ‘ambition’ or ‘ignorance’ can be a cause need give any comfort to those who [–] wish to reinstate the ghost in the machine. For there is no need to assume that because motives, intentions, habits, beliefs, and the rest can be causes, they are therefore to be regarded as mental events or processes after all. The error is to be located rather in thinking that only events or processes can be causes, whereas there would seem to be virtually no restriction whatever upon the type of thing that can qualify as a cause, provided it passes, in a particular context [the test that without the thing in question, the effect would not have occurred (see above)]. (p. 151.)
It seems that Dray does not think that reasons and dispositions cannot be causes. He seems to suggest that as long as something makes a difference to the outcome, it is a cause.
“The important point for our account of explanation in history is that the necessity of a causal connexion, when it is actions we are talking about, is very often rational necessity. In Chapter IV, in discussing the logic of ’cause’, I said that although there are various ways of arguing for a causal assertion, the cause had to be a necessary condition of its effect. But there is more than one kind of necessity; and in history the relevant kind will often be that found in action done for a good reason (from the agent’s point of view).” (p. 154.)
Without the good reasons, the action would not have been performed. Not all actions can be explained rationally, but when some action is thus explained, we can see how it depended on the reasons behind it.
There are only two worries with respect to this.
First, we have seen that an action can be explained either as “the thing generally done” or “rationally done”. Capturing a knight in a game of chess can fall into both categories, depending on who’s playing. In what sense do the actions depend on the reasons? The same action can be performed with or without deliberation. Notice that the argument cannot be that if there are good enough reasons, the action is necessarily performed. We saw above how Dray argued that the so-called principles of actions which “in C, y is the thing to be done” might hold even if C is not followed by y.
Secondly, we saw that the reasons behind an action are postulated in a way that makes the reasons to match the action. This means that the action guides to some extent what reasons can be justifiably attributed to the agent. One might wonder whether this makes the necessity of the connection between the reasons and the action something quite different from causal necessity. Is the necessity in the world like causal necessity or is it in our reasoning concerning what the reasons for a given action must be? There is a story be told here, the main character of which is the logical connection argument. However, that story must wait for another time.
Dray spends the final chapter of the book to discuss how-possibly explanations to give the DN model one more blow:
“I shall maintain, the historian need not show that what is to be explained happened necessarily in the light of the particular events and conditions mentioned in the explanation, and, a fortiori, need not show that it happened necessarily in the light of some covering law or laws. For the demand for explanation is, in some contexts, satisfactorily met if what happened is merely shown to have been possible; there is no need to go on to show that it was necessary as well.” (157.)
For example, we might wonder how it was possible that Anthony Joshua lost to Andy Ruiz. Once we are told that Joshua got overconfident and went for a quick knockout which backfired, we understand how the loss was possible. We do not deduce the loss from the overconfidence and decision to finish the fight quickly and we still have an explanation.
The modal framing is a bit unfortunate here. In the DN model, the only necessity is logical necessity in the deduction of the explanandum from laws and initial conditions. Still, Dray is correct. We do not need to give a complete DN explanation to resolve the puzzlement behind “how possible?” questions. However, this is only a corollary from the fact that an explanation cites what Dray calls “necessary conditions” without which the event would not have happened. When providing information about a condition that made some event possible, we are simply giving information about a condition without which the event would not have happened. In a sense, Dray’s account of causal explanation and his account of how-possible explanation are symmetrical. The only difference is perhaps that in how-possibly questions, one condition is more obviously cited as the explanatory condition. For example, Joshua would not have lost if Ruiz was slower, and Ruiz’s speed, therefore, made the loss possible. However, we already knew that Ruiz is fast. What we learned was that Joshua went for an early knockout, and this completes our picture of how the event came to be; it adds the final piece of information about a condition without which the loss would not have happened.
Because how-possibly explanations are symmetric to causal explanations and because the latter did not require DN-structure, how-possibly explanations do not have a DN-structure. However, this does not mean that laws and generalizations do not play any role in how-possibly explanations. We already saw in the section concerning causal explanation that judgements “had X not been the case, Y would not have been the case” require some sort of generalizations to get off the ground. The same is true of how-possibly explanations. For example, we have to understand that if a fighter attempts to finish the fight early, she exposes herself to the punches of the other fighter.
One does not have to kill the bogeyman. One needs to tame it.
We have seen that the DN model was, in the ‘50s, a bogeyman that stagnated fruitful discussion concerning generalizations and their use in historiography. We have seen that the fear was in most respects exaggerated. Covering law model was a specific suggestion about laws in historiographical explanation. It was, understandably, an unacceptable and even misleading model that needed to be get rid of. However, the fact that the DN model was unacceptable and misleading should not have meant that there is no place for laws and generalizations in historiography. By equating any use of laws and generalizations with the mistaken DN model, philosophers of history made a great analytical blunder.
On the positive side, Dray’s book is an excellent collection of insights on different ways in which historiography can be said to made understandable. However, the relationships between different accounts of explanation are not made clear enough in the book. There remain tensions between Dray’s suggestions, as we have seen. This might not be completely Dray’s fault because the different accounts may be objectively in tension: maybe there is no way to unify them. Yet, the book leaves open the questions concerning the possible explanatory unity or pluralism in historiography: (I) Do we need different approaches to explain different aspects of history (such as events and rational actions) or all history be explained in the same way (perhaps limiting the scope of historiography to e.g. rational actions)? (II) Can we sometimes explain the same historical explanandum in mutually incompatible ways; if so, what does this tell us about ontological and epistemic objectivity of historiography?
Dray, William H. (1957). Laws and Explanation in History. Greenwood Press.
Hempel, Carl G. (1942). “The Function of General Laws in History”. Journal of Philosophy 39 (2). 35-48.
MDC = Machamer, Peter ; Darden, Lindley & Craver, Carl F. (2000). ”Thinking about Mechanisms”. Philosophy of Science 67 (1):1-25.
Nowell-Smith, P. H. (1959). “Laws and Explanations in History. By W. H. Dray.” Philosophy 34 (129)
Salmon, Wesley C. (1989). “Four decades of scientific explanation”. Minnesota Studies in the Philosophy of Science 13:3-219.
Strawson, P. F. (1959). “Laws and Explanation in History”. Mind 68 (270):265-268.
Virmajoki, Veli (2018). “Could Science be Interestingly Different?” Journal of the Philosophy of History 12 (2). 303-324.
Virmajoki, Veli (2019) Cementing science. Understanding Science through Its Development.
Virmajoki Veli (2020). “What Should We Require from Accounts of Explanation in Historiography?”. Journal of the Philosophy of History.
von Wright, G. H. (1971). Explanation and Understanding.
Woodward, James (2003). Making Things Happen. A Theory of Causal Explanations. Oxford