Narratives and Causal Explanation


In a previous post, I have discussed the connection between mechanisms and narratives. I tend to think that there are problems in linking historiographical explanation, narratives, and mechanisms.

However, even if the linking of narratives with mechanistic explanation has significant shortcomings, the idea that narratives – especially causal narratives – are central in historiography has been strongly supported. Before I delve into the discussion concerning causal narratives, a warning needs to be in place. The problem with the discussion about the nature and role of causal narratives in producing historiographical insights and understanding is that the notion of causality and explanation are usually left unanalyzed themselves. The suggestion does not appear to be that causal narratives are a unique class of causal explanations that are based on – or define themselves – some distinct notion of causality. The debate concerning the nature and role of causal narratives in historiography, therefore, may not provide insights on the conceptual core of the notion causal explanation in historiography, but needs to be discussed because of the insights it has produced on the issues intimately connected to causal explanations in historiography. There are important notions and phenomena connected to the notion causal explanation in historiography to which any account of causal explanation in historiography must be faithful. The discussion about causal narratives reveals exactly such issues and phenomena. This is remarkable since when it comes to the philosophical accounts of explanation, “history, and its explanatory narrative mode of argument, remains an outcast in the broader kinship of the sciences” (Morgan 2017, 86). An account of explanation is underdeveloped as long as it does not discuss the issues that causal narratives reveal.

However, it is one thing to leave some core notions unanalyzed and somewhat open, it is another thing to suggest that there is nothing to be analyzed. Philosophical work on causal explanation can be highly valuable even when it does not focus on the core notions but on conceptual and methodological phenomena surrounding these notions. However, it does not follow that defining the core notions is redundant or impossible. These two issues should be kept separate. For example, Roth asks “what makes narratives explanatory[?] [–] narratives typically seem to be descriptive” (2020, 70) but goes on to argue that the distinction between description and justification does not arise in the case of narratives. This seems to miss the point. Usually, a sequence of events is considered explanatory if it is causal. Roth is aware of this but argues that “The causal sequence, in turn, can consist only in this case of seeing facts as ordered and so related in a particular way” (2020, 73, emphasis original). Roth adds that “no functional distinction exists between describing that sequence and justifying causal links” (2020, 75). The problem is that if there is no stronger notion of causality in use than one that makes causality follow automatically from a description of a sequence of events, then causality does not add any explanatory import to the sequence. One cannot make causality carry the explanatory load without a notion of causality that distinguishes between causal and non-causal sequences.

The discussion above illustrates how we need to resist the temptation to eliminate analyses of causality for the reason that many of the goals, notions, and patterns of reasoning in historiography are confused, unclear, and ambiguous and, therefore, the notion of causality in historiography needs to be powerful enough to suggest how the confusions, unclarities, and ambiguities might be addressed. This does not mean that we should first provide a notion of causality and then build everything on that pregiven notion. On the contrary, the methodology must have two directions. On the one hand, we need to identify important issues and phenomena in causal explanation in historiography and see whether particular notion of causality and causal explanation can make sense of these issues and phenomena. On the other hand, we need to have a notion of causality and causal explanation to address confusions, unclarities, and ambiguities in the issues and phenomena in causal explanation in historiography. In what follows, I show how this two-directional methodology works with respect to narrative explanations, explanatory depth, and historical contingency.

Before that, I briefly sum up a narrative leading to Eddington’s observation of gravitational deflection in 1919, as this will serve as a source of examples in our discussion above.

The Case of Eddington

In May 1919, Arthur Eddington executed an expedition to observe how much light bends in Sun’s gravitational field (i.e., gravitational deflection). Later that year, Eddington announced that the observations confirmed Einstein’s prediction of the amount of deflection. Why did this historical event occur? Aspects of the narrative can be summarized as follows.

The Personal Background

Eddington was born in a Quaker family. However, he rejected the idea that Quakers are to keep distance to politics. Rather, the so-called Quaker Renaissance set a new set of ideas, among them the idea that pacifism meant that Quakers had an obligation to wage peace and not only to decline to fight. (Stanley 2019, 13-14.) Eddington’s Quakerism affected his will to unite the international scientific community during and after the Great War (Stanley 2003; 2019).

Eddington went on to study physics and mathematics. He started as the chief assistant to the Astronomer Royal at the Royal Greenwich Observatory in 1906. Eddington became the Plumian Professor of Astronomy and Experimental Philosophy in 1913. Along with the professorship. Eddington was named the director of the entire Cambridge Observatory. (Stanley 2019, 24 & 59-60.) Eddington had a unique position and set of skills that made the expedition possible.

The Theoretical Background

Einstein graduated in 1900 with a degree in physics and mathematics. The landscape of physics was dominated by Maxwell’s work in electromagnetism. Maxwell demonstrated that electric and magnetic fields travel through space as waves moving at the speed of light. There were also debates concerning luminiferous ether, a substance that was supposed to fill the whole space and serve as the carrier of light waves. The ether was supposed to be the absolute reference frame against which all motion could be measured. The ether theory had philosophical, conceptual, and experimental problems. In 1905, Einstein published his special theory of relativity which rejected the idea of ether and provided new conceptual grounds by unifying space and time. He then extended his work to gravity and started to develop the general theory of relativity. A consequence of his theoretical reasoning was that gravity should bend light. The development of the theory was a complex process into which we do not go in detail. However, it is worth mentioning that at certain point the theory included a mistake and predicted the amount of deflection wrong. The mistake was realized in 1915 and the theory was eventually corrected (Stanley 2019, 143-149).

It is important to notice that Einstein’s theory, once corrected after 1915, predicted a specific amount of deflection. Einstein’s theory predicted deflection of 1.74’’ while a “Newtonian” derivation predicted deflection of 0.87’’ (Earman and Glymour 1980). Eddington’s expedition and observations were aimed at deciding which of these predictions is true.

The War

In 1914, the Great War broke out. This had the immediate consequence that an expedition to Russia to measure the deflection in 1914 could not be carried out. As it turns out, this might have been an important turn of things, as the prediction Einstein made at the time was incorrect, as we saw. On a more general level, the war cut all connections between Allied and Central powers, including those of science. It even led to deep and long-lasting despise and suspicion between the scientists. In Britain, there were widespread plans to exclude Germany from science even after the war (Stanley 2019, 176-181). The war cut off both the concrete networks of communication and the cognitive kinship between scientists. Neither people nor information traveled through the borders.

Minor Occurrences

The war also had an effect on Eddington, in that he was not willing to go to the war due to his religious beliefs. Eddington was exempted due to the national importance of his work, but the exemption was far from certain (Stanley 2019, 232-233). Einstein, on the other hand, was an isolated pacifist in the scientific community of Germany who suffered stomach problems due to the quality of food in wartime Germany (Stanley 2019, 186.)

In 1915, Eddington wrote about gravitation and Einstein’s prediction but was not aware of Einstein’s work beyond 1911. In 1916, Eddington received a letter from Leiden, sent by Willem de Sitter, that summarized Einstein’s more recent theory and its implications. Eddington became interested in the theory but also in the fact that it was formulated by a pacifist in Germany. (Stanley 2019, 168-270. He wished to use the theory to bring British and German science closer again. (Stanley 2003; 2019.)

The Expedition

An opportunity to make Einstein’s theory famous was a solar eclipse in 1919 during which the deflection could be observed by taking photographs of stars around the sun. Eddington had to convince the British scientist about the importance of the theory, and, finally, was able to execute the expedition. (Stanley 2003 & 2019.)

The observation was attempted in two locations. Both had their problems, one with the technology and the other with cloudy weather (Stanley 2019, 272). Yet, they were able to get sets of photographs, and after analyzing the data from photographs, Eddington was convinced – and convinced others – that the observation had been made. (Stanley 2003; 2019.)

The Interpretation of the Photographic Plates

However, extracting data from photographs required judgments about the reliability of the observation procedures, and this selection has produced controversies in later philosophical and historical analysis (see Collins &Pinch 1993; Earman & Glymour 1980). At the heart of the controversies is the question of whether Eddington made biased choices when judging reliability issues. For example, it has been argued that the data could have been interpreted rather differently (Earman & Glymour 1980). A counterargument says that Eddington’s reasoning was sound, given his intimate knowledge concerning the instruments (Kennefick 2019).

To sum up, the process leading to the observation of gravitational deflection involved many types of factors and factors of different scales. In what follows, this book attempts to account for how such historiographical explanations are constituted and how they work. We will use elements from the Eddington case to illustrate causal explanation and related issues.

Explanatory Depth

As Currie (2014, 1165) states, “Narrative explanations account for particular events via causal sequences concluding with the explanandum.” Already, we do not have a definition of causality in the characterization of causal narrative. Currie seems to suggest that the difference-making notion is adequate (2014, 1179) but it is notable that difference-making is not unique to causal narratives and difference-making does not force the causal narratives account of explanation in historiography. One can build an account of explanation on difference-making without committing to the narrative form of explanations. Yet, with the notion of (causal) narrative explanation, Currie is able to recognize an interesting phenomenon, that explanatory depth is, at least sometimes, achieved by moving from simple narratives to disunified, complex narratives. Complex narratives involve more details as they involve many explanantia, more specific explanantia and those explanantia belong to different temporal and spatial scales. Complex narratives cannot be embedded, it is not an instance of a regularity and cannot be explained by a general model (Currie 2014, 1167-1169). For example, the Eddington narrative is a complex one. It involves many factors (theoretical background, the war, the personal stories, and so on); it includes specific details unique to the case, such as the influence of pacifism in a scientific development; and the factors are of different scales (the war lasted four years, the eclipse some minutes, and so on).

Notice that the D-N model is not able to capture Eddington’s case. What would a universal law that linked the theoretical background, individual motivations, and conditions of war to the observation of gravitational deflection look like? Currie provides the reason for the failure. The nature of complex narratives makes them such that they cannot be embedded, but complex narratives still provide deeper explanatory understanding in the cases in which they are used. An account of explanation therefore should be able to recognize the explanatory power of complex narratives. This is not a problem only for the D-N model.  As Currie points out, this phenomenon is something that the ephemeral mechanisms account seems unable to capture. “The explanans [–] are at many temporal and hierarchical grains, and it is not obvious whether such a disparate group is amenable to unified representation as a mechanism, no matter how contingent. Explanans are not ‘components’ but rather causal factors [–].” (Currie 2014, 1180; however, see Swaim’s 2019 defense against Currie’s conclusion.) Here we have a phenomenon, i.e., complex narratives provide deeper understanding in the cases in which they are used, that we should be able to make sense of within any acceptable account of historiographical explanation. The failure to make sense of the phenomenon speaks against certain accounts of explanation. This means that Currie’s analysis of narrative explanation reveals phenomena that are, prima facie, evidence against certain accounts of explanation. This fits the two-directional methodology mentioned earlier: our notion of causal explanation must be faithful to important phenomena in historiographical explanations, but it should also be able to shed light on the phenomena.

Ereshefsky and Turner (2020) have also argued that causal narratives provide conceptual tools to identify weaker and stronger historical explanations, i.e. explanatory depth. Following Gallie, they suggest that a historical explanation gets stronger when the narrative is thickened and tightened. “To thicken a historical narrative is to fill in gaps in that narrative. Tightening is different: it highlights the dependency of the outcome on earlier stages in the historical narrative [–]” (Ereshefsky and Turner 2020, 53). Again, to get the analysis off the ground, we need a notion of causal relevance: “the mere introduction of any factor does not make a historical narrative stronger. Only citing factors that are causally relevant to the outcome explained make a narrative stronger” (ibid.). While I completely agree, for reasons given elsewhere, that “historical narratives are stronger [–] when they articulate more factors the outcome depends on” (ibid.), it is not completely obvious that this means that stronger narratives “bring into sharper focus the path taken toward that outcome” or that explanatory depth should be analyzed in terms of narratives in the first place. Ereshefsky and Turner do not quite explicitly tell what counts as a narrative, but they seem to involve series of changes and events that are tied together. The problem is that many causally relevant factors in historiography are not naturally understood as events that belong to a series that constitute a narrative. The fact that the war cut off communications between Germany and Britain is not an event, it is more like a background or boundary condition that shaped the possible paths of events at any point of Eddington’s story between 1914 and 1918. In a similar manner, the theoretical background against which Einstein worked is not an event. Surely, the coming into existence of different pieces in that background are events that could be incorporated in a massive narrative about Eddington’s observation, but it is not merely the birth of the pieces that had a causal effect on the outcome but the fact that they constantly shaped scientific life. The war and the theoretical background both affected how things developed but they would also have an effect on whatever series of events would have been initiated. Instead of belonging to a series of events in the narrative, they might even make those events irrelevant details. For example, particular failures to communicate across the borders shaped the Eddington-narrative. However, even if different attempts to communicate would have been initiated, they would also have failed due to the background conditions in the war. The war is explanatory and does not belong to a narrative as a single event in any natural sense, and precisely because of this the war is explanatory at the cost of narrative details and “sharper focus on the path”.


The notion of historical path takes us to another central topic in the philosophy of history, that of historical contingency. Beatty has suggested that “narratives are especially good at representing contingency and accounting for contingent outcomes” (2016, 34). A narrative provides understanding about turning points in a sequence of events leading to the outcome (Gallie 1964; Beatty 2016). Beatty (2006), following Gould (1989), identifies two senses of contingency, contingency per se and contingency as causal dependence. An event is contingent per se if it is an unpredictable outcome that could have not occurred or could have occurred differently. Roughly, the fact that the weather improved during the eclipse is a contingency per se in Eddington’s case. An event is contingent in the causal dependence sense if its occurrence depended on the occurrence of an earlier event. Roughly, Eddington’s interest in the theory of general relativity was contingent upon de Sitter sending a letter.

There has been a lively discussion concerning the relationship between narratives and contingency and narrative possibility in general. While we cannot go to the details of this debate in the space of this post, it is worth mentioning that the discussion is informative about many issues related to explanation, especially about metaphysical and epistemic commitments that may or may not be associated with different accounts of historical narratives (see Swaim 2021 clarifying the issue). For example, some of the accounts understand possibilities in a strongly metaphysical sense while others define them in terms of our knowledge situation. These two strategies may have different applications. Moreover, not all metaphysical commitments are on an equal standing when it comes to their intimacy with the scientific outlook. (Swaim 2021.) However, in what follows, I will mainly focus on the questions about the relationship between contingency and explanation and how this relationship should be interpreted as constraining the accounts of explanation in historiography. These questions are the most relevant for our attempt to map the outline of the conceptual space that accounts of causal explanation in historiography must fill and clarify.

As the discourse on narrative possibility indicates, there seems to be a connection between contingency and explanation. However, this connection should not be overplayed. Contingency might not be in the core of historiographical explanation while still belonging to the conceptual space that an account of explanation needs to clarify. For example, it has been suggested that the connection between contingency and narratives is so tight that there would be no need for historical narratives in a deterministic world (see Ereshefsky and Turner 2020; Swaim 2021). This seems to be a mistake, as Ereshefsky and Turner (2020) argue. There is no need to tie together the analysis of historical narratives and controversial metaphysical thesis. Moreover, the link between narratives and contingency might be completely misleading. There must be something else in the narratives that provide explanatory understanding. Let me explain.

A narrative might provide novel understanding even if it did indicate the inevitability of the outcome – or for that very reason. We need to begin by noting that Beatty’s two notions of contingency both require that we are able to reason how things could have been. If it is the case that “[w]hat narratives are especially good for – what makes them worth telling, and renders them non-superfluous – are situations where history matters: where a particular past had to happen in order to realize a particular future” (Beatty and Carrera 2011, 491), the ability to provide narrative explanations guarantees that we are in the position to identify how things could have developed, had something been different. Once narratives demand this type of counterfactual reasoning, there is no guarantee that they will reveal to us that many different series of events would have led to a different outcome. It is equally possible that we are forced to conclude that different series of events would have led to the same outcome. The type of reasoning that makes narratives good at representing contingency can also reveal inevitability. For example, Sterelny has analyzed in detail robust process explanations. They “specify both the range of alternative initial conditions across which we would still see a qualitatively similar outcome and the limits on that range” (2016, 523). Robust process explanations can be found in human history. If it really was the case that a narrative is explanatory because it is contingency-indicating, then robust process explanations would not be explanatory. This judgement seems odd. If we have two narratives, both based on similar reasoning and both focusing on possible turning points, that differ only in the degree of contingency they associate with their respective outcomes, it would be completely arbitrary to say that only one of them really provides historical understanding. In both cases, we become to understand historical trajectories.

It seems that inevitability-indicating narratives provide similar understanding as contingency-indicating narratives. Or, to be more precise, the question of contingency is separate from the question of explanatoriness. It is one thing to provide an explanatory narrative, it is an another thing to evaluate how contingent the narrative is. There is a natural reason why contingency and explanatoriness have been often confused with each other. An explanation answers when an outcome would have been different. It always provides an alternative to the outcome. However, it does not follow that the outcome was contingent. Why am I not a famous footballer? Because my genes made me too slow. Had I had different genetic makeup, I would have been a famous footballer. However, it is not contingent that I am not a famous footballer because my having the right genes is far-fetched. In contrast, we can ask why I sit on a chair instead of a couch. Because I decided to sit on the chair. Had I decided differently, I would be on the couch. The outcome is contingent, as it could have easily been the case that I decided differently. Here we have two explanations that explain an outcome that would have been different from the actual world. Yet, only one of them is contingency-indicating.

Narratives, Epistemology, and Generalizations

The connection between narratives and contingency points towards a more general issue, that of epistemic demands in narrative explanation. For example, we have already seen that many scholars argue that narratives center around the issue of things that could have been and that this demands counterfactual (or, more generally put, modal) reasoning, and we also saw how Currie argued that complex narratives require that we sew together factors of different scales. This raises two questions that an acceptable account of explanation should be able to answer. First, how do these practices reflect the nature of historiographical explanation? In a sense, the scholars discussed in this post have revealed interesting phenomena about the construction of explanation. What we wish to achieve with an account of explanation are conceptual explanations for these phenomena. Secondly, how can we think clearly in the practices associated with historiographical explanations? Counterfactual reasoning is notoriously difficult, and not all counterfactual claims are explanatory relevant. Had a barometer reading not fallen, a storm would not have occurred; yet the reading is completely irrelevant for the sequence leading to the storm. Scales are also often difficult to handle. Had the Earth exploded, there would not have been an observation of the deflection. However, if we ask why Eddington observed deflection of 1.7 arc-seconds rather than 0.74, the non-explosion of the Earth is explanatorily completely irrelevant. An account of explanation should be able to make explicit how scales are tracked in explanatory reasoning.

However, it is not only that narratives require certain types of reasoning, but they can also support historiographical reasoning. Currie and Sterelny (2017) have argued that the construction of narrative explanations serves important epistemic purposes and that there is a partnership between idealized models and narratives. Narratives are not merely just-so stories. Rather, the “speculation” is an important step that guides the search for novel evidence. For example, if we do not understand why Eddington chose certain photographic plates as reliable data-source, we may provide two hypotheses. (A) Eddington was biased toward Einstein’s theory. (B) Eddington understood the instruments well. Given these hypotheses, we can search for evidence. For example, the fact that also other scientists publicly presented reasons why some of the data were to be disregarded, serves as evidence for (B). In this way, explanatory hypotheses guide the search and use of evidence. This illustrates a more general point. Explanations are rarely built on pre-given evidence. An acceptable account of explanation should be able to recognize how explanatory thinking can guide the search for evidence and how different explanatory claims suggest different evidential patterns.

At the intersection of epistemic demands and support is the question of explanatory generalizations in historiography. The topic has been sensitive in the philosophy of historiography. The failure of the D-N model made many scared to suggest that historiographical explanation involves generalizations in its core. The two main strategies are to deal with generalizations in history. The first one is to deny or remain agnostic about generalization in historiography. Roth has recently adapted this position: “narrative explanation will be a presentation of a temporal series that answers why the explanandum turns out to be as it is. For without reference to this retrospective stance, there would exist nothing to explain. And since what must be explained has no standardized format that explains it, a temporal sequence that cannot utilize laws or law-like generalizations will be required” (2020, 68). The second strategy is to suggest that generalizations are needed but these generalizations are not laws with a wide scope. Rather they are recurrent patterns of causal structure, with a much more limited scope of generalizability. The idea that limited generalizations provide explanatory understanding is widespread in the philosophy of science currently. In philosophy of historiography, Glennan’s definition of ephemeral mechanism is one way that such generalization can be used. The debate on causal narratives has been more meager with respect to the issue. However, it seems that Currie (2014; 2016) and Currie and Sterelny (2017) suggest that generalizations play an important role in historiographical explanations. Currie and Sterelny (2017, 5-6) argue that narratives are constrained by our knowledge of causal structures of the past and that modeling can support narratives (2017, 8). Moreover, Currie (2014, 1169) is “inclined to see [even complex] explanations as leaning on a patchwork of regularities” and Currie (2016, 939) suggests that “historical scientists do not rely on mere empirical correlations to bring out dependencies; rather, they posit causal models that connect past entities”. The debate concerning narratives and contingency also seems to suggest, at least implicitly, that explanatory understanding requires generalizations. Without generalizations, it seems difficult to get modal reasoning (what could and would have happened) off the ground.

It seems that narrative-building both requires and supports generalizations. For example, it is difficult to see how one could build a narrative connecting Eddington’s pacifism to his efforts without appealing to some sort of regular connection between worldviews and courses of action. It is equally clear that the Eddington-narrative suggests that the straightforward generalization “scientists test important theories” should be replaced with a more circumscribed generalization, which sees testing as a function of theories, values, and even politics. An acceptable account of explanation should be able to recognize the connections between explanations and generalizations and especially provide a characterization of what explanatory generalizations are like and how they differ from mere regularities. The bogeyman of historical laws has scared the philosophy of historiography since the D-N model was published, and we would finally wish to know what really is behind the mask.

To sum up this post, we can notice that the discourse on causal narratives has revealed interesting issues and phenomena in historical explanation. First, it indicates the need to clarify what is meant by causal or explanatory narrative. To understand how narratives, exactly, provide understanding, we need to clarify the core notions causality and explanation. Secondly, the discourse indicates that there is a connection between explanation and contingency/inevitability. Whatever explaining involves, it must have a modal dimension. Otherwise, the historicity and modalities would not be intimately linked. Thirdly, the discourse has shown that historical explanations can be stronger or weaker. Historical explanations have explanatory depth, and we should understand what features define the depth of an explanation. Finally, the discourse indicates that historical explanations are tightly intertwined with methodological issues, especially evidence and generalizations.


Beatty, John & Carrera, Isabel (2011). When What Had to Happen Was Not Bound to Happen: History, Chance, Narrative, Evolution. Journal of the Philosophy of History 5 (3):471-495.

Beatty, John (2006). Replaying Life’s Tape. Journal of Philosophy 103 (7):336-362.

Beatty, John (2016). What are narratives good for? Studies in History and Philosophy of Science Part C: Studies in History and Philosophy of Biological and Biomedical Sciences 58:33-40.

Currie, A. (2016). Hot-blooded Gluttons: Dependency, coherence, and method in the historical sciences. The British Journal for the Philosophy of Science.

Currie, A., & Turner, D. (2016). Introduction: Scientific knowledge of the deep past

Currie, Adrian & Sterelny, Kim (2017). In defence of story-telling. Studies in History and Philosophy of Science Part A 62:14-21.

Currie, Adrian Mitchell (2014). Narratives, mechanisms and progress in historical science. Synthese 191 (6):1-21.

Ereshefsky, Marc & Turner, Derek (2020). Historicity and explanation. Studies in History and Philosophy of Science Part A 80:47-55.

Gallie, W. B. (1964). Philosophy and the Historical Understanding. New York: Schocken Books.

Gallie, William (1955). Explanations in History and the Genetic Sciences. Mind, 64, 160-180.

Gould, S. J. (1989). Wonderful life: The burgess shale and the nature of history. New York: Norton.

Morgan, Mary S. (2017). Narrative ordering and explanation. Studies in History and Philosophy of Science Part A 62:86-97.

Roth, Paul (2020). The Philosophical Structure of Historical Explanation. Evanston: Northwestern University Press

Swaim, Daniel (2019). The Roles of Possibility and Mechanism in Narrative Explanation. Philosophy of Science 86

Swaim, Daniel (2021). What is narrative possibility? Studies in History and Philosophy of Science Part A 89 257-266

One thought on “Narratives and Causal Explanation

Leave a Reply

Your email address will not be published. Required fields are marked *