2. An interpretive challenge for theories of explanation

In Aspects of Scientific Explanation (Reference Hempel1965), Carl Hempel used ideas from scientists about their own explanatory practice in two different ways. On the one hand, Hempel quoted scientists who apparently endorsed a covering-law model of explanation. For instance, in order to decide whether scientific explanations always reduce an unfamiliar phenomenon to a familiar one, Hempel quotes the sociologist George Homans, who claimed that to deduce low-level generalizations from a set of more general propositions is to explain them (Hempel Reference Hempel1965, 431–32). Hempel used Homans’s idea to support his own position that although the more general propositions are not necessarily more familiar, they can be explanatory. In this instance, the pronouncement of a scientist on his explanatory practice was used by Hempel as evidence that confirms the consequences of the covering-law model of scientific explanation.

On the other hand, Hempel also reinterprets the pronouncements of scientists so that they fit his model. Two pages after the use of Homans’s ideas, Hempel introduces a quote by Lord Kelvin: “as long as I cannot make a mechanical model all the way through I cannot understand” (Hempel Reference Hempel1965, 434). In this quote, there is no reference to a deduction from general laws. Yet, Hempel interprets the quote as if Kelvin demanded “not simply that an explanation somehow render a phenomenon plausible or familiar, but more specifically that it provides a model governed by the laws of mechanics” (Hempel Reference Hempel1965, 434). This leads to an interpretive question: Under what conditions can some pronouncement of a scientist be used as evidence confirming a philosophical theory of scientific explanation? Or, under what conditions should it be reinterpreted so that it conforms to the theory? Hempel never addressed this question.

This interpretive problem results from Hempel’s larger project of explicating scientific explanation. This project is shaped by two central assumptions, and these assumptions are taken over by most participants in later debates. One assumption deals with scientific knowledge and one with philosophical theory:

(E) is an assumption about the uniformity of the explanatory aim of science across time and across disciplines. It enables Hempel to use not only examples from the history of science but also more contemporary examples and examples from a varied set of scientific disciplines. If there were no such assumption, Hempel would have no reason to collect a varied set of examples as a set that expresses a single, uniform aim of scientific theories (i.e., explanation) that could be captured in a philosophical theory. (A) is an assumption about the task of the philosophy of science. Applied to explanation, (A) entails that the philosophical model of scientific explanation should be fateful to both actual scientific practice and the norms for explanation that are implicit in it. The philosophical model aims to clarify those implicit norms. Both assumptions are introduced in the opening paragraph of Hempel and Oppenheim’s (Reference Hempel and Oppenheim1948) article “Studies in the Logic of Explanation,” which is universally taken to be the foundation for all later debates on this topic:Footnote ¹

In 1948, the covering-law account still aimed at a definition of explanation in terms of necessary and sufficient conditions that could be expressed in first-order predicate logic and would fit all available examples (Hempel and Oppenheim Reference Hempel and Oppenheim1948, 157). In Aspects of Scientific Explanation, which was published seventeen years later, Hempel retracts this stringent goal and is satisfied with a model that explicates the relevant epistemic features of the explanatory aim of scientific knowledge (Hempel Reference Hempel1965, 412). Although both assumptions (E) and (A) are stated in the opening paragraph of the 1948 article, they are never made explicit as starting points of inquiry that require extra motivation, nor are they introduced as hypotheses that have to prove their fruitfulness. Hempel and Oppenheim take them for granted and initiate a research program starting from these assumptions. A theory of scientific explanation, given this program, will have two challenges:

1. 1. Selecting from scientific practice those examples (claims by scientists about explanation, purported examples of scientific explanations, disputes about explanation) that can be understood as expression of the explanatory aim of scientific knowledge

2. 2. Constructing a model of scientific explanation that is justified by the selected evidence

In the 1948 article, the first challenge proved difficult: because there were no well-established examples of scientific explanation within the contemporary philosophy of science, Hempel and Oppenheim had to introduce relevant examples themselves. They mentioned the explanations of the immersion of a mercury thermometer in hot water—the thermometer first drops and then swiftly rises—the appearance of a broken oar in water, and Galileo’s law for the free fall of bodies (Hempel and Oppenheim Reference Hempel and Oppenheim1948, 136). Unsurprisingly, in all three examples, a general law plays a crucial role in the explanation of the phenomenon. Later in the article, they also discuss “an explanatory argument from the field of linguistics,” an explanation of the evolution of the Latin word apis into the English word bee and the French word abeille. However, they decide to conceive this explanation as incomplete because it lacks clearly specified general laws concerning the phonetical evolution of words (Hempel and Oppenheim Reference Hempel and Oppenheim1948, 141). Deciding which examples are proper expressions of the explanatory aim of science, and especially how they are evidence for the offered philosophical model (incomplete attempts at explanation or outright counterexamples), is an interpretive problem that is similar to the use of scientists’ ideas about explanation, which I previously highlighted.

Another paradigm case in the explanation literature is the purported explanatory power of Newton’s law of universal gravitation: one can explain the position of Mars through information about the mass of the sun, the mass of Mars, Mars’s present position and velocity, Newton’s laws of motion, and Newton’s universal law of gravity. Hempel mentions this in 1948 (Hempel and Oppenheim Reference Hempel and Oppenheim1948, 135), and Norwood Hanson discusses it as the crucial evidence for Hempel’s model (Hanson Reference Hanson1959). Unfortunately, there is no more contested case about explanation in the history of science. Whether the universal law of gravitation explains at all was contested in Newton’s time, and it still is. Whether the examples that Hempel himself and later participants in the debate introduced are true expressions of the explanatory aim of science is a very difficult interpretive question to answer. Whatever the answers, it is crucial to see that the question only becomes possible because of a previous commitment. It is only because one assumes that explanation is an aim of science expressed in scientific practice that the choice of examples becomes a problem. Given (E), talk of explanation in the practice of science becomes a relevant and distinct aspect of the practice that one wants to understand on a philosophical level. But why assume (E)? Why do we think that scientific knowledge has an explanatory aim that is distinct from description? This assumption is part of the terms of the debate as they were set by Hempel.

Although most philosophers of science today believe that the covering-law model, at least in its Hempelian formulation, is mistaken, there is a widespread agreement that the model represented a major step in our understanding of science. It helped us approach questions concerning scientific explanation in a fruitful way. Wesley Salmon lauded the model as Carl Hempel’s most important philosophical achievement and as “one of the most significant pieces of philosophical progress in the twentieth century” (Salmon Reference Salmon and James2000, 316). In Salmon’s (Reference Salmon1989) Four Decades of Scientific Explanation, the 1948 article marks the distinction between the prehistory and the history of the modern discussion of scientific explanation. According to Salmon’s narrative, philosophers before the appearance of the DN model in Hempel and Oppenheim’s version “had no clear idea of what scientific explanation might be” (Salmon Reference Salmon1989, 10). Most commentators on the debate follow Salmon in opening their discussion of the topic with the 1948 article (McCain Reference McCain2016, 133; Woodward Reference Woodward and Edward2017; Skow Reference Skow, Humphreys, Chakravartty, Morrison and Woody2016). In their article, Hempel and Oppenheim show how assumptions (E) and (A) give rise to a fruitful avenue of research on scientific explanation. However, conceiving this article as the fountainhead from which the discussion on scientific explanation should start is not a neutral historical choice. Salmon’s decision to bracket all philosophical discussion on scientific explanation before Hempel as mere prehistory is problematic because it obscures a crucial question that Hempel’s work on scientific explanation never addressed, namely, how to decide whether scientific explanation is a distinct aim of scientific knowledge and, consequently, whether it is necessary to evaluate scientific theories on the basis of their explanatory power.

3. The prehistory unraveled

From a historical point of view, Hempel and Oppenheim’s article is a radical shift in logical empiricist philosophy. Whereas Reichenbach and Carnap, both important influences on Hempel’s work, are mostly silent about explanation as an aim of science, Hempel brings it to the center of attention in the philosophy of science when the field emerges as a professional subdiscipline in the 1950s and 1960s. What happened to the philosophical interpretation of scientific knowledge in the 1940s so that most American philosophers of science became convinced that explanation had to be accounted for as an aim of science? In answering that historical question and reframing the history of ideas on scientific explanation, there is, I believe, much to learn about the debate on scientific explanation in general.

In order to understand the historical significance of Hempel’s introduction of the DN model, one has to be familiar with the ideas about scientific explanation from the first half of the twentieth century. Contrary to Salmon’s historical judgment that philosophers before Hempel had little clue what scientific explanation could be, there actually was a thriving philosophical debate about scientific explanation, especially in the decades before Hempel’s article. Unlike the post-Hempel situation, this early-twentieth-century debate did not center on finding a comprehensive theory/model/analysis of scientific explanation that could fit all available examples and paradigm cases of scientific explanation. Instead, it questioned the validity of explanation as an aim of science and the role that explanation could play in understanding the nature of scientific knowledge.

In the logical empiricist circles around the younger Carl Hempel, most philosophers believed that explanation was not an acceptable concept in an epistemology of science. Philipp Frank identified three antiexplanatory philosophers as the intellectual predecessors of scientific philosophy: Mach, Poincaré, and Duhem. In Frank’s account, these philosophers opposed the idea that physical theories should be evaluated on their explanatory power over and above their descriptive adequacy. As a consequence, these philosophers attempted to account for supposed explanations in science as expressions of the descriptive aim of scientific theories. Frank quoted the following insight of Duhem as an important inspiration to the Vienna circle (Frank Reference Frank1949, 15):

In a similar vein, Ernst Mach conceived the aim of physical theories as the most economical and briefest description of natural events (Mach Reference Mach and McCormack1893/1974, 6). For Mach, the explanation that physical theories afford is merely a useful psychological effect of their descriptive power: the theories accustom our thoughts to a great variety of disparate facts (Mach Reference Mach and McCormack1893/1974, 7). According to this understanding of scientific knowledge, explanation is not taken as a goal of scientific inquiry that is distinct from description: whatever explanatory virtues are attributed to scientific theories, they can always be recast within descriptive terms. I will call this general position on the aim of science descriptivism. For Mach and Duhem in particular, descriptivism entailed that scientific theories are mere systematized descriptions of our experience of the world. If one believes that scientific theories can achieve insight beyond this systematization, then one has a wrong understanding of what a scientific theory aims at. Both Duhem and Mach took this to be a general lesson that they could draw from the history of science: they were capable of describing the historical progress in mechanics without any reference to the explanatory power of physical theories, even though at several points in time, past scientists had preferred several theories because of their purported explanatory value.

A descriptivist position has to be accompanied by an account of explanation-talk in the practice of science. Duhem is most explicit in this regard: that some “geniuses” of modern physics have built their theories in the hope of giving an explanation of natural phenomena is not a conclusive argument against a descriptivist position. As Duhem phrased it, “chimerical hopes may have incited admirable discoveries without these discoveries embodying the chimeras which gave birth to them” (Duhem Reference Duhem and Wiener1906/1991, 31). Although the explanatory aim and accompanying explanatory language of certain scientists is not a conclusive argument against descriptivism, it does present a challenge: to present a history of science without explanation as a driving force in its progress. Duhem’s historical analysis of physics is an attempt to meet this challenge. In The Aim and Structure of Physical Theory (Reference Duhem and Wiener1906/1991), he always distinguishes the representative (descriptive) part of any physical theory from its explanatory part, defending that a theory never owes its power or fertility to the latter. In the development of theories, Duhem claimed, the explanatory aspects of theories are always abandoned, whereas the representative parts of the theories are transferred (Duhem Reference Duhem and Wiener1906/1991, 32). Philipp Frank endorsed Duhem’s historical dialectic of scientific theories in the following way:

Even though scientists often believe that their theories “explain,” this aspect of a theory is taken by descriptivists to be of heuristic value only, as a driving force in the acquisition of new empirical phenomena. In Frank’s reading, “explanation” as a term can still be applied to scientific theories—but only to the extent that theories systematize experience. Explanation-talk by scientists should be understood within the descriptive aim of science. Philosophers in the logical empiricist movement broadly conceived also use the term “explanation” in that sense. Moritz Schlick maintained that “description by means of laws achieves all that can possibly be demanded of knowledge” (Schlick Reference Schlick and Wiener1953, 473). In his introductory work on positivism, Richard von Mises equally defended that “explanation is but a special form of description, namely, a description that is systematic, unified and, as far as possible complete” (von Mises Reference von Mises1968, 138). Science is only “a description of relations, of interconnections between phenomena” (von Mises Reference von Mises1968, 138). Rudolf Carnap also discusses the term “explanation” in this limited sense in 1931: it is the deduction of a singular scientific sentence from other singular scientific sentences with the help of a hypothetical general sentence. Thus, for Carnap, you can call connections between singular scientific statements through laws “an explanation” (Carnap Reference Carnap1931, 463). This does not imply that explanation is an aim of science that goes beyond its power to coherently describe the relations between empirical phenomena. In Hempel and Oppenheim’s early book The Type-Concept in Light of the New Logic (Reference Hempel and Oppenheim1936, 1), they maintain a descriptivist reinterpretation of explanations similar to Carnap’s: “Explanation is better called the deployment of laws that connect empirical data in a determined way. In the formulation of laws one uses concepts that describe the intertwined data.”

Among early logical empiricists, it was common to assume that explanation was not an aim of science distinct from description. In contrast to Mach and Duhem, logical empiricists did not write about explanation in a negative sense. It played little role in their analysis of science, that is, until Hempel and Oppenheim’s (Reference Hempel and Oppenheim1948) article, where they open with the statement that “scientific research in its various branches strives to go beyond a mere description of its subject matter by providing an explanation of the phenomena it investigates” (Hempel and Oppenheim Reference Hempel and Oppenheim1948, 135).

In the first half of the twentieth century, there were also prominent philosophers in both France and Germany who opposed the descriptivist understanding of scientific knowledge and believed that explanation was a preeminent aim of science, distinct from description. Émile Meyerson, an important figure in the French epistemological tradition, disagreed with Mach, Duhem, and the neo-Kantian Ernst Cassirer on the aim of scientific knowledge, claiming that description and explanation were both distinct, proper aims of scientific theories—henceforth, I will refer to this general position as explanationism.

Limiting the aim of scientific theories solely to the (systematized) description of phenomena is, according to Meyerson, in contradiction with a central habit of human thought, a certain instinct to look for the cause of things, to answer the question of why something occurred (Meyerson Reference Meyerson and Wiener1908/1953, 464). For Meyerson, the practice of science is a further development of this human mental instinct, the development of a pre-science already present in our minds—a desire for intelligibility (Meyerson Reference Meyerson and Wiener1908/1953, 468). Meyerson did not believe that an interpretation of science as the search for explanation was self-evident. It had to be grounded in a historical account showing how both the descriptive and explanatory aims of science have been applied simultaneously and continue to be so applied (Meyerson Reference Meyerson and Wiener1908/1953, 462). Meyerson’s aim was to produce a historical narrative of modern science as the expression of one and the same desire for understanding in his book Explanation in the Sciences. Meyerson’s narrative was to prove that if one has a descriptivist view of scientific knowledge, “the entire course of science, past and present, becomes an enigma, or rather a sort of gigantic and monstrous absurdity” (Meyerson Reference Meyerson and Loewenberg1921/1991, 34). Although Meyerson’s position diametrically opposed Duhem’s descriptivism, he acknowledged that references to the statements of scientists about the aims of their theories do not yield any decisive argument for his position. Scientists throughout the past have had both explanationist and descriptivist inclinations: scientists’ pronouncements on the principles that guide their thought should not be trusted (Meyerson Reference Meyerson and Loewenberg1921/1991, 465). In a critical review of Ernst Cassirer’s historical narrative of modern science in Das Erkenntnisproblem, Meyerson expressed this methodological problem for the philosopher of science in the following way:Footnote ²

Any argument for or against descriptivism or explanationism relies on an interpretation of scientific theories. No simple reading of scientific text could produce the meta-scientific concepts with which to evaluate scientific theories. For Meyerson, Duhem, Mach, and Cassirer, such interpretation had to be grounded in an account of the historical development of scientific knowledge. There were, however, also other possible routes.

In the German context, Heinrich Rickert, a notable neo-Kantian philosopher, defended an explanationist account of scientific knowledge. Rickert’s position did not rely on a historical account of scientific progress but was grounded in his logical understanding of the nature of scientific concepts in general. According to Rickert, concepts that merely systematized the available empirical data were not proper scientific concepts because they did not aim for universal validity, a validity that governed beyond the available empirical data. Rickert believed that there were two ways to think about scientific concepts. Concepts can make a phenomenon merely understandable. This implies that one uses classificatory concepts (Gattungsbegriff) to systematize objects or events of which one has experience (Rickert Reference Rickert1929, 102). In Rickert’s understanding, a descriptivist position, like Mach’s, would limit the aim of scientific concepts to this stage of concept formation. If this is the case, the distinction between explaining and describing collapses (Rickert Reference Rickert1929, 106). Rickert, however, distinguishes a classificatory use of concepts from a higher aim of concept formation, one that subsumes an event under a natural law. This subsumption shows that an event was absolutely necessary and could not have been different (Rickert Reference Rickert1929, 106). The search for natural laws was the highest possible aim of science. For Rickert, an explanation of an event entails knowing why an event had to occur and thus knowing how an event follows from a necessary principle. Explanation, in his account, follows from an understanding of the necessary coherence of events in nature, and in this sense, the natural sciences can go beyond the descriptive aim of scientific knowledge. Rickert’s refutation of descriptivism mainly relies on his understanding of concepts. The concepts produced by our understanding can strive for a universal validity over their objects that is not conditioned by the available empirical phenomena. Concepts that express such absolute, necessary relations, even though they may not yet have been reached by science in practice, are still possible superior objects of thought and thus a possible higher goal for scientific knowledge.Footnote ³

At stake in this European, pre-Hempelian debate is the following philosophical question: Is explanation an aim of science distinct from its descriptive aim, and what kinds of arguments can one develop in favor of one’s answer to the former question? Mach and Duhem claimed that there is an account of science possible without explanation as its aim, whereas Meyerson claimed that no such coherent account could be given. Crucial in this debate is that all philosophers recognized that a decision on the aim of science could not be motivated merely by referring to the “explanation-talk” of scientists. It is not because scientists have used the terminology of explanation to discuss their work that one has to implement scientific explanation as an essential concept in the epistemological interpretation of scientific theories. Because there are no given examples that are evidence for the explanatory aim of science without prior commitments concerning its aims, any motivation for such commitments should look beyond direct talk of explanation in science and should rest on other grounds.

4. Hempel’s assumption: Explanationism without motivation

Hempel and Oppenheim’s (Reference Hempel and Oppenheim1948) article shifted the balance in logical empiricist philosophy completely. In the opening paragraph, Hempel and Oppenheim introduce it as a given that explanation is an epistemic aim of science distinct from description, and they add that there is no reason to doubt this: “there is a general agreement” (Hempel and Oppenheim Reference Hempel and Oppenheim1948, 135). This statement is deceptive. Not only does it do no justice to the philosophical debates of the earlier decades in the twentieth century, but it is also in direct conflict with the previous position of most logical empiricist philosophers, who had consistently claimed that talk of scientific explanations could be interpreted within the descriptive aim of scientific knowledge. Hempel and Oppenheim abandoned this crucial aspect of the descriptivist position: it is the goal of their article to account for explanation as an aim of science in its own right. As a consequence, their work starts from the explanationist assumption (E). Given (E), talk of explanation in the practice of science becomes a relevant and distinct aspect of the practice that one wants to understand on a philosophical level. Just as the descriptivist has to interpret talk of explanation in science within descriptive terms, the explanationist has an equal challenge of interpretation, both in selecting which examples express the explanatory aim of science and in deciding how the offered model relates to these examples.

Based on the 1948 article, it is unclear how Hempel and Oppenheim came to have the commitment that scientific knowledge aims at explanation because they do not address this commitment at any length. In Aspects of Scientific Explanation, seventeen years later, Hempel only notes that human beings have a basic motive of “sheer intellectual curiosity,” “a deep and persistent desire to know and to understand themselves and their world” (Hempel Reference Hempel1965, 333). This seems to be some kind of anthropomorphic ideal for science: the human intellect has a common desire to understand the world, and science, as an expression of that desire, aims to explain the world. This type of anthropomorphic motivation is still prevalent in the literature. Michael Strevens, one of the most prominent contemporary explanationists, holds that “when science is pursued as an end rather than as a means, it is for the sake of understanding—the moment when a small, temporary being reaches out to touch the universe and makes contact” (Strevens Reference Strevens2008, 3). In fact, as is often noted, this motivation for (E) dates back to Aristotle’s opening of the Metaphysics: “All men desire to know by their nature,” which is specified later in Metaphysics Alpha as a desire to know not only what but also why things happen (Salmon Reference Salmon1989, 3). Meyerson started his history of science from the same anthropomorphic view: “science has been established with the tenacious hope that nature will manifest itself as intelligible” (Meyerson Reference Meyerson and Wiener1953, 462). However, this provides no real argument for (E). Even if human beings all have such a uniform desire to understand, this does not mean that scientific knowledge should be interpreted as the (successful) execution of that desire. To be responsibly committed to (E), more is required than a vague theory about the desires of the human intellect.

Because Hempel, throughout his career, never explicitly addressed what motivated him to take (E) for granted, it is an open question how to understand the philosophical rationale behind Hempel’s commitment to the idea that scientific knowledge aims at explanation. Given the lack of explicit argumentation, we can turn to the historical circumstances surrounding Hempel’s shift. This historical question finds no easy answer. Hempel’s diaries and letters in the 1940s contain remarkably little discussion of scientific explanation. There is no explicit eureka moment predating the shift of the 1948 article, nor is there evidence that Hempel suddenly found interest in the work of Meyerson or Rickert. Without addressing all relevant aspects of Hempel’s intellectual development here, there are several interactions in the 1940s that both shaped the path toward the 1948 article and yield an interesting perspective on its philosophical significance.Footnote ⁴

After Hempel’s doctoral defense at the University of Berlin in 1934, he had no academic position but was employed as the private research assistant of Paul Oppenheim in Brussels. In the academic year 1937–1938, Carnap hired Hempel as his research assistant at Chicago University but could not sustain his employment for more than a year. In the summer of 1938, Hempel returned to Belgium to work under Oppenheim. Yet, he was set to continue his career in the United States and managed to acquire a travel visa for the next year. When Hempel arrived in the States in February 1939 without any prospect of employment, he was greeted by Ernest Nagel at the docks of New York Harbor. Nagel immediately integrated him into the flourishing scene of New York philosophy. Searching for an academic position, Hempel was eager to become more familiar with the research of his American colleagues. Nagel welcomed Hempel to his Logic graduate course at Columbia University and introduced him to what was called the Nagel–Hook circle, a discussion group of New York–based philosophers who regularly came together to discuss their works in progress.Footnote ⁵

In his diary entry of February 14, 1939, Hempel noted that, to his surprise, Nagel’s coursework in logic had “many historical perspectives; critical discussion of Aristotelian and neo-Tomistic conception of science.”Footnote ⁶ Nagel’s course in logic comprised not only “a formal and highly technical study of principles of inference” but also “a study of the generic features of the methods used in acquiring knowledge, and an account of the nature of scientific systems of knowledge.”Footnote ⁷ For the discussion of the nature of science, Nagel did not start from logical empiricist ideas but discussed Aristotle’s account of scientific knowledge from the Posterior Analytics in great detail. For Nagel, the Posterior Analytics was not an obsolete relic from the history of ideas. It could still provide the starting point to discuss the very nature of science because the book is “based on a sharp contrast between scientific knowledge and common sense knowledge.”Footnote ⁸ In his course notes, Nagel noted that for Aristotle, scientific knowledge was “not simply knowledge of the fact, but of the reasoned fact, to know the reason why the fact occurs is the distinctive mark of science.”Footnote ⁹ Nagel continued to discuss Aristotle’s theory of the demonstrative syllogism in his notes as an attempt to capture the logical features of the distinctive mark of science, namely, to explain phenomena. Although Hempel only sat in on this course for the spring term of 1939, it could have been more influential on Hempel’s larger commitments than Hempel himself may have realized. Nagel introduced the logic of scientific explanation as a way to understand scientific knowledge in these classes not only to Hempel but also to a number of Columbia graduates, such as Morton White and John Hospers, who would take up this perspective on the sciences in their early writings. During the summer of 1939, Hempel got a job as a philosophy teacher at the City College of New York, and from 1940 onward, he was an associate professor at Queens College. Because of these appointments, Hempel’s intellectual environment became different from his previous logical empiricist surroundings in Berlin, Vienna, and Chicago.

In the Nagel–Hook circle, Hempel also gave a talk in 1941 on a topic that was characteristic of New York philosophy at that time: the validity of the distinction between historical and natural scientific knowledge. In this talk, Hempel defended the unity of science, arguing that the function of scientific concepts is the same in all scientific practices, including historiography.Footnote ¹⁰ Hempel does not yet mention that scientific concepts have an explanatory aim. Instead, they merely systematize the available phenomena through hypothetical generalizations (Hempel Reference Hempel1942, 36). This talk was published as “The Function of General Laws in History” in the Journal of Philosophy in 1942. It is often characterized as Hempel’s first article on explanation (Salmon Reference Salmon1989, 25), even though both (E) and (A) are absent from it. The reason why it is characterized as an article on explanation is that Hempel’s logical articulation of the function of laws resembles the logical sketch of an explanation from the 1948 article, and Hempel already refers to an instance of that sketch as something that might be called “an explanation.” Yet, the crucial assumption that explanation is an aim of scientific knowledge distinct from description is still absent.

Even if Hempel’s (Reference Hempel1942) article on history does not yet propose a theory of explanation, its prominent use of the term explanation is striking. Otto Neurath, who was a vehement proponent of Machian descriptivism, noticed this and launched a campaign from Oxford during the Second World War to halt the American influence on the younger philosophers within the Unity of Science movement. Because Neurath believed that all members of the movement were committed empiricists and descriptivists, he wrote multiple letters to Feigl, Morris, and Carnap warning about the danger of introducing explanation as a worthwhile concept in the logical interpretation of scientific knowledge. To Carnap, he wrote:

He was especially worried about the evolution of Hempel’s ideas and wrote to Carnap: “Hempel READING my papers LISTENING to my arguments did not REMEMBER afterwards these points, BECAUSE THEY ARE SO FOREIGN TO HIM.”Footnote ¹² Neurath also questioned Hempel’s overzealous attempts to analyze concepts through logic. “I like Hempel’s attitude very much, but I often suggested to him, he should deal a little more with scientific problems as such and then apply his analytical ability to them.”Footnote ¹³

Between 1934 and 1944, Neurath and Hempel had an extensive correspondence; a dominant theme of theirs was the use of formal logic to analyze science. Whereas Hempel thought it advantageous, Neurath considered it dangerous and warned Hempel:

Hempel generally did not engage with Neurath’s concerns, and in the 1940s, he paid little to no attention to Neurath’s empiricist warnings against supposedly metaphysically laden terminology, such as explanation. For Hempel, the ideas of his new American colleagues were more important than Neurath’s references to older European philosophers like Ernst Mach.

In 1943 Morton White, at the time a Columbia graduate student, continued to work on Hempel’s (Reference Hempel1942) article and the ideas on explanation introduced in Nagel’s graduate course. In the article “Historical Explanation,” White argued that explanations, which he understood as scientific answers to “why” questions, have a similar structure in both physics and historiography (White Reference White1943). John Hospers equally picked up on the topics familiar from Nagel’s classes. He worked on an account of explanation by analyzing its use in ordinary language. Hospers, a regular attendant of the Nagel–Hook circle, published this work in the Journal of Philosophy in 1946. Hospers’s article “On Explanation” is not an analysis of explanation as an aim of science per se. It is about “why” questions in general language use: “We are sometimes presented with a statement describing some observed fact, and when we ask ‘Why?’ we are presented with another statement which is said to constitute an ‘explanation’ of the first” (Hospers Reference Hospers1946, 337).

For Hospers, “why” questions and explanations are intimately linked. In the article, he used his linguistic intuitions to evaluate several possible positions on what constitutes a proper explanation statement, which he understood as an answer to a “why” question. He comes up with several candidates but reaches no definitive conclusion. Hospers’s article only attempts to form an ordinary-language model of answers to “why” questions. In Hempel and Oppenheim’s article, published two years later, “why” questions also play a central role: scientific explanations should have continuity with ordinary-language explanations. This continuity still forms a central aspect of the post-Hempelian debate on scientific explanation (Woodward Reference Woodward and Edward2017; Skow Reference Skow, Humphreys, Chakravartty, Morrison and Woody2016). It is very likely that the connection between “why” questions and scientific explanations originates from Hempel’s familiarity with Hospers’s work, which is also quoted as an inspiration in the 1948 article (Hempel and Oppenheim Reference Hempel and Oppenheim1948, 140).

Thus, the historical record strongly suggests that Hempel and Oppenheim’s introduction of the explanationist assumption (E) was not the result of an engagement with earlier debates. Most likely, for Hempel and Oppenheim themselves, the extent to which (E) departed from the logical empiricist tradition remained unnoticed. As I mentioned earlier, neither Hempel’s correspondence nor his diaries of the 1940s attest to a change of mind.Footnote ¹⁵ The 1948 article can be read as Hempel and Oppenheim’s way to extend the logical empiricist tradition with novel concerns from American colleagues. They present the DN model not only as an analysis of scientific explanation but also as a way to reconceive the concept of causality in science,Footnote ¹⁶ think about emergent properties in biology, and create a model for the systematic power of a theory. Although the DN model can be understood as a deflationary, empiricist view on these matters, Hempel and Oppenheim introduced and defended it specifically as an analysis of explanation, a distinct aim of science. By offering this analysis and introducing (E), they created the possibility for other philosophers to introduce nondeflationary models of explanations. Throughout this transition from deflationary to nondeflationary (often causal) accounts, (E) was generally accepted while glossing over important philosophical questions about (E) that had been central to the earlier debates. This raises important questions.

5. Reframing the debate: Descriptivism versus explanationism

How Hempel’s article on scientific explanation became so popular after its publication is a story in its own right. I cannot go into its details here—it is connected with the fact that Hempel’s university position changed after 1948, first to Yale and then to Princeton, and the fact that the 1948 article was reprinted in important volumes that would shape the budding field of the philosophy of science in the 1950s and 1960s. What is, however, crucial about the reception of the article is that none of its readers questioned (E) and (A). As Salmon noted, because of Hempel’s work, “the temptation to say that there is no such thing as scientific explanation seems to have vanished” (Salmon Reference Salmon and James2000, 315). Hempel’s assumptions became the bedrock on which a steady research program could be formed (Reck Reference Reck, Milkov and Peckhaus2013).

As the last eighty years of research have shown, both assumptions in tandem are fruitful. They give us many competing models of the explanatory aim of science, some models faring better than others. However, these models are all committed to the assumptions that made them possible, and those assumptions are not neutral. In Depth, Michael Strevens articulated the assumption (E) in the following way: “There is an underlying set of principles that has always determined what does and does not count as a good explanation for us. Any change in explanatory practice is due to some change in the parameters of those principles” (Strevens Reference Strevens2008, 37–38). This assumption is an explicitly historical one, that scientific practice has had a consistent explanatory aim as long as it existed. This is not a self-evident assumption; it requires a rich historical insight into scientific practice and a narrative of modern science that highlights a continuity that is not easily imposed on the record of intellectual history.

Strevens aims to take our explanatory practice as the most important evidence for a descriptive model of scientific explanation and defines our practice as the “sum total of explanations regarded as scientifically adequate in their day, together with an understanding of the background against which they seemed adequate” (Strevens Reference Strevens2008, 37). But as noted before, this only makes sense once one has already assumed that the search for explanation is a central, distinct aim of scientific practice both in time and across disciplines. Whether a reference to explanation in the practice of science, past or present, should be interpreted as an expression of the explanatory aim of science is itself something at stake, not something that can be taken for granted. Given (E), such references can be conceived as evidence for the philosophical model. Yet even then, any reference might also be interpreted as a mistaken expression of the explanatory aim of science. Even with (E) as the driving assumption, one still has to address the interpretive challenge of selecting which examples are expressions of the explanatory aim: there is no given, uncontested set of explanations regarded as scientifically adequate in their day. Any explanationist position shares this challenge.Footnote ¹⁷

In most discussions of scientific explanation, an explicit motivation for (E) remains absent. One recent author who addresses this issue is Kevin McCain. He points to important scientific institutions, such as the National Research Council, that advocate the position that “the goal of science is the construction of theories that can provide explanatory accounts of features of the world” (McCain Reference McCain2016, 134). According to McCain, the importance of explanation in scientific practice is related to science’s aim of increasing our understanding of the world. Quoting Erwin Schrödinger, McCain claims that the scientific worldview rests on the hypothesis that nature can be understood (McCain Reference McCain2016, 135). Yet, he never provides an argument that motivates such an interpretation of scientific theories and explains why other possible interpretations (e.g., descriptivism) should be excluded.

Currently, contemporary explanationists have no argument to support the assumption that enables their models of scientific explanation to become meaningful. In the search for such an argument, much might be gained: the presentation of explanation as a central aim of scientific theory and practice cannot be an epistemologically and metaphysically neutral image of science. Aristotle’s model of explanation in the Posterior Analytics explicitly relies on an essentialist account of natural events and its associated standards for the knowledge of nature. Similarly, the debates between Meyerson, Duhem, Cassirer, Mach, and Rickert are embedded within questions concerning the nature of scientific concepts, the standards of rationality, and the potential variability of those standards throughout the history of science. In the contemporary philosophy of science, Bas van Fraassen’s constructive empiricism develops a descriptivist position within such broader epistemological reflection.

According to van Fraassen, the acceptance of scientific theories involves only the belief that the theories are empirically adequate, that they yield correct descriptions of observable phenomena. Accordingly, explanation is not a distinct aim of science: van Fraassen rejects (E). Although van Fraassen’s ideas are often presented as a “pragmatic” theory of scientific explanation, this is a deceptive categorization. As a committed descriptivist, van Fraassen has to address explanation-talk by scientists. Just like Duhem, Mach, or von Mises, a contemporary descriptivist has to interpret scientists talking about the explanatory power of their theory within a descriptivist understanding of science. As a consequence, what van Fraassen offered was a pragmatic theory of answers to “why” questions, a philosophical account of a common linguistic practice, namely, that people ask “why” questions, including scientists. van Fraassen disambiguated two kinds of philosophical research programs that Hempel had tied together in 1948: the analysis of answers to “why” questions (part of ordinary-language philosophy) and the discussion about the aim(s) of science. In The Scientific Image, van Fraassen gives a pragmatic account of “why” questions, arguing that explanations are essentially relative to the context in which a “why” question was posed (van Fraassen Reference van Fraassen1980, 156). With such an account, van Fraassen can say that requests for explanation in science merely “satisfy certain of our desires; and these desires are quite specific in a specific context, but they are always desires for descriptive information” (van Fraassen Reference van Fraassen1980, 156). For van Fraassen, there is no single desire for understanding that is uniformly expressed within scientific practice. Because van Fraassen’s model of “why” questions does not aim to capture a uniform goal of science, it is deceptive to categorize it as a theory of scientific explanation, pragmatic or otherwise. It is best understood as a descriptivist position that uses a pragmatic model for answers to “why” questions to interpret explanation-talk by scientists.

Just like contemporary explanationist positions, van Fraassen’s descriptivism faces a challenge in interpreting the history of science. He must argue that episodes in the history of science that are regarded as scientific progress can be interpreted within descriptivist terms. van Fraassen has recognized this challenge and has given a descriptivist interpretation of Perrin’s experimental work, which is often taken as the paradigm case of a scientific episode where explanation plays a necessary role (van Fraassen Reference van Fraassen2009). At least since The Empirical Stance (Reference van Fraassen2002), van Fraassen also addressed the broader challenge of how to motivate a denial of (E) beyond the mere coherence of a descriptivist interpretation of scientific knowledge. His categorization of empiricism as a stance is an attempt to embed his descriptivism within a voluntarist epistemological project: the empiricist stance against the desire to achieve a uniform explanation of the world is motivated by a coherent set of empiricist values.