Loading field map
Mathematical explanation is the branch of philosophy of mathematics concerned with a deceptively simple question: what does it mean for a mathematical fact to explain something? The question has two distinct faces, and the subfield is defined by the tension between them. On one face, mathematicians themselves speak of explanations all the time—a proof that "explains" why a theorem holds, a concept that makes a result "clear," a rearrangement of definitions that reveals "the real reason" something is true. On the other face, philosophers ask whether mathematics can explain facts about the physical world—whether, for instance, the fact that you cannot divide a cake into three equal parts with straight cuts is explained by a theorem of geometry, or whether the mathematical fact is merely a description of a physical constraint. The subfield of mathematical explanation is the sustained attempt to make sense of both uses of the word, to determine whether they are the same phenomenon, and to specify the conditions under which a mathematical fact genuinely explains rather than merely accompanies or describes.
The subfield is organized around two questions that are often run together but are logically distinct. The first is the intra-mathematical question: within mathematics itself, what distinguishes an explanatory proof from a non-explanatory one? Mathematicians routinely contrast a proof that merely establishes a result with one that shows why the result holds. The classic example is the difference between a brute-force verification and a conceptual argument. The theorem that the sum of the first \(n\) odd numbers is \(n^2\) can be proved by induction, which checks the formula step by step; it can also be proved by arranging dots in a square, which makes the equality visually and conceptually immediate. Most mathematicians would call the second proof explanatory and the first merely verifying, but articulating what the second has that the first lacks is a genuine philosophical problem.
The second question is the extra-mathematical or applications question: can mathematics explain facts about the non-mathematical world? When a physicist explains the period of a pendulum using a differential equation, is the mathematics doing explanatory work, or is it just a convenient language for describing physical causes? When a biologist explains the hexagonal shape of honeycomb cells by appeal to the isoperimetric theorem—that a hexagon is the most efficient way to partition a plane into equal areas—is the mathematical theorem itself an explanation of a biological fact? Some philosophers have argued that mathematics is always merely a descriptive framework, and that genuine explanations must cite physical causes. Others have argued that some mathematical facts are genuinely explanatory of physical phenomena, and that a complete account of scientific explanation must accommodate them.
These two questions are connected by a deeper issue: what kind of thing an explanation is. If explanations are arguments, then a mathematical proof is a natural candidate for an explanation, and the intra-mathematical question becomes central. If explanations are answers to why-questions, then both mathematical and physical explanations might be answers of different kinds. If explanations are causal, then mathematics—which seems to describe necessary, timeless relations rather than contingent causal processes—faces a special problem. The subfield is thus not merely a taxonomy of examples but a sustained engagement with the nature of explanation itself, using mathematics as a test case that is both unusually clear and unusually difficult.
The modern subfield has its roots in two older discussions. The first is the ancient and medieval tradition of analysis and synthesis, in which a proof that works backward from a desired conclusion to its grounds was distinguished from a proof that works forward from assumptions. This distinction was not exactly the modern explanatory/non-explanatory distinction, but it carried an intuition that some proofs reveal structure while others merely demonstrate truth. The second root is the nineteenth-century debate over the nature of proof, particularly in geometry and analysis, where mathematicians began to distinguish proofs that rely on a particular representation (say, a diagram or an algebraic formalism) from proofs that capture the essential structure of the subject. The German mathematician Hermann Grassmann, for instance, explicitly distinguished a proof that shows a proposition to be true from one that shows why it is true, and he argued that the latter requires the right conceptual framework.
The immediate precursor to the modern subfield is the work of the philosopher and logician William Dray and, more importantly, the philosopher of science Carl Hempel. Hempel's deductive-nomological model of explanation—according to which an explanation is an argument whose conclusion is the fact to be explained and whose premises include a law of nature—was developed for science, but it raised an obvious question for mathematics. If a mathematical proof is a deductive argument, is it an explanation? Hempel himself was cautious, suggesting that mathematical proofs explain in a derivative sense but that the paradigm of explanation is causal and empirical. This left the intra-mathematical question largely unaddressed and the extra-mathematical question unresolved.
The subfield as a self-conscious area of research emerged in the late twentieth century, driven by two developments. The first was the rise of philosophy of mathematical practice, a movement that turned away from the foundationalist questions that had dominated the field (What are numbers? Are mathematical truths necessary?) and toward the actual activities of mathematicians: proving, defining, visualizing, generalizing. Within this movement, the question of what mathematicians mean by "explanation" became a natural topic. The second development was the growth of scientific explanation as a major area in philosophy of science. As philosophers of science refined their accounts of explanation—moving beyond Hempel's model to causal, mechanistic, and pragmatic accounts—they began to ask whether mathematics could fit any of these models, and whether the failure to fit was a problem for mathematics or for the models.
Within the intra-mathematical question, several distinct approaches have emerged, each addressing a different aspect of the problem.
The most straightforward approach holds that explanatory power is a property of proofs. A proof is explanatory if it reveals the "reason" for the theorem, where "reason" is understood in terms of the structure of the proof itself. The most influential version of this view is due to Mark Steiner, who proposed in the 1970s that an explanatory proof is one that generalizes: it shows that the theorem holds not because of accidental features of the particular objects involved but because of a general pattern. Steiner's criterion was that an explanatory proof is one that would still work if the objects in the theorem were replaced by other objects of the same kind. A proof that the sum of the first \(n\) odd numbers is \(n^2\) by arranging dots in a square is explanatory because it shows the result depends on the structure of odd numbers and squares, not on any special property of the particular numbers involved. A proof by induction, by contrast, works for any sequence of numbers that satisfies the recurrence, so it does not isolate the specific structural reason.
Steiner's approach has been criticized on several grounds. The notion of "generalizing" is vague: many non-explanatory proofs generalize, and some explanatory proofs are quite specific. Moreover, the criterion seems to conflate explanatory power with generality, but mathematicians often find a proof explanatory precisely because it is tailored to the specific structure at hand. Despite these criticisms, Steiner's work established the central question: can we specify, in terms of proof structure, what makes a proof explanatory?
A different proof-centered approach, associated with Paolo Mancosu and others, focuses on the distinction between proofs that merely verify and proofs that "explain" in the sense of showing how the result arises from the definitions and prior theorems. This approach often draws on the mathematical tradition of explanatory proofs in number theory and algebra, where a result is said to be explained when it is shown to follow from a deeper structure—say, from the properties of a group or a field—rather than from a clever but ad hoc calculation. The challenge for this approach is to say what "deeper structure" means without simply appealing to the mathematician's subjective sense of depth.
A second major approach shifts attention from proofs to concepts. On this view, a mathematical explanation is not primarily a property of a proof but of the conceptual framework in which the result is stated. A theorem is explained when it is seen to follow from the "right" definitions—definitions that capture the essential nature of the objects involved. This approach is often associated with the tradition of structuralism in philosophy of mathematics, which holds that mathematical objects are positions in structures, and that the goal of mathematics is to articulate the structures themselves.
The concept-centered approach has a natural answer to the question of why some proofs are explanatory: they are explanatory because they use the right concepts, and the right concepts are those that reveal the structural reasons for the result. The classic example is the use of group theory in geometry. The fact that there are exactly five Platonic solids can be proved by a tedious case analysis, but it can also be explained by noting that the rotational symmetries of each solid form a finite group, and the classification of finite subgroups of the rotation group shows why only five are possible. The group-theoretic proof is explanatory because it locates the result in a general structural framework.
The weakness of this approach is that it risks circularity: what makes a concept "right" is that it yields explanations, and what makes a proof explanatory is that it uses the right concepts. To break the circle, proponents must appeal to an independent notion of mathematical depth or naturalness, which is itself a contested notion. Nevertheless, the concept-centered approach captures something real about mathematical practice: mathematicians often say that a result was not understood until the right concepts were introduced, and the introduction of new concepts is frequently described as the key to explanation.
A third approach, drawing on the work of Bas van Fraassen and others on scientific explanation, holds that explanation is not a property of proofs or concepts but of the use of mathematics in a context. On this view, a proof is explanatory relative to a questioner's background knowledge and interests. A proof that explains a result to a novice may not explain it to an expert, and a proof that explains a result in one context (say, a lecture on number theory) may not explain it in another (say, a lecture on combinatorics). The pragmatic approach denies that there is a single, context-independent property of "explanatoriness" to be discovered.
This approach has the virtue of accounting for the diversity of mathematical practice: mathematicians give different proofs of the same theorem in different contexts, and they often disagree about which proof is explanatory. But it has the corresponding vice of making explanation subjective or relative, which many philosophers find unsatisfying. If explanation is merely in the eye of the beholder, then the philosophical question "What is mathematical explanation?" seems to dissolve into a sociological question about what mathematicians happen to find illuminating.
A fourth approach, inspired by Philip Kitcher's unificationist account of scientific explanation, holds that a proof is explanatory to the extent that it is part of a systematic, unified account of a domain. On this view, the goal of explanation is to reduce the number of independent assumptions needed to derive a wide range of results. A proof that uses a single general method to derive many theorems is more explanatory than a collection of ad hoc proofs, because it shows that the theorems are not independent facts but consequences of a common pattern.
The unification approach has been influential in philosophy of science, where it offers an alternative to causal accounts, and it has been applied to mathematics with some success. The development of algebraic topology, for instance, can be seen as a unification of many disparate results in geometry and analysis under a single conceptual framework, and this unification is often described as explanatory. The approach faces the challenge of specifying what counts as "unification" in a way that is not merely a matter of taste, and it has been criticized for conflating explanation with systematization.
The extra-mathematical question—whether mathematics can explain physical facts—has generated a separate set of approaches, often drawing on debates in philosophy of science.
The most skeptical position holds that mathematics never explains physical facts; it only describes them. On this view, when a physicist says that the period of a pendulum is explained by the differential equation, the equation is not doing explanatory work. The explanation is causal: the pendulum swings because of gravity, the length of the rod, and the initial conditions. The mathematics is a language for expressing the causal relations, but the mathematics itself is not a cause and does not explain. This view is often associated with a broadly nomological or causal account of explanation, according to which an explanation must cite the causes of the phenomenon. Since mathematical facts are not causes—they are necessary truths about abstract objects—they cannot explain.
The descriptive view has a strong intuitive appeal. It respects the intuition that explanation is about what makes something happen, and mathematics does not make anything happen. But it faces a serious challenge: in many cases, the mathematical structure seems to be doing more than describing. The hexagonal honeycomb example is a case in point. The fact that bees build hexagonal cells is explained, in part, by the fact that a hexagonal tiling is the most efficient way to partition a plane into equal areas. This is a mathematical fact, not a physical cause. The bees do not know the isoperimetric theorem, and natural selection did not "choose" hexagons because it calculated the theorem. Yet the mathematical fact seems to explain why the cells are hexagonal rather than square or triangular. If the descriptive view is correct, this explanation must be an illusion, and the real explanation must be entirely in terms of physical forces and evolutionary history.
A more permissive position holds that mathematics can explain physical facts, but only in a derivative sense. On this view, a mathematical explanation of a physical fact is an argument that uses a mathematical theorem to infer the physical fact from other physical facts. The mathematics is not itself explanatory; it is a tool for drawing out the consequences of physical assumptions. The honeycomb explanation, on this view, is really an explanation of why the cells are hexagonal given that the bees are building a partition of the plane into equal areas with minimal perimeter. The mathematical theorem shows that any such partition must use hexagons, but the explanation of why the bees build such a partition is a separate, biological question.
This view has the virtue of preserving the causal account of explanation while acknowledging the role of mathematics in scientific reasoning. But it faces the objection that it makes the mathematical contribution merely instrumental, and that in many cases the mathematical fact is not just a tool but the very content of the explanation. When a mathematician explains why a certain equation has no solution by appealing to a theorem of number theory, the explanation seems to be about the equation itself, not about some physical system that the equation describes.
The most ambitious position holds that mathematics can genuinely explain physical facts, and that such explanations are not reducible to causal explanations. This view, defended by Marc Lange and others, argues that some mathematical explanations are non-causal explanations: they explain by showing that the physical fact is necessary in a way that transcends physical law. Lange's central example is the honeycomb again, but he also discusses cases from topology and number theory applied to physics. On this view, the mathematical fact explains the physical fact because it shows that the physical fact could not have been otherwise, given the mathematical structure of the world.
The genuine explanation view faces the challenge of specifying what "could not have been otherwise" means in a physical context. Physical laws are themselves necessary in some sense, so the mathematical necessity must be of a different kind. Lange argues that mathematical explanations are explanations in terms of constraints rather than causes: the mathematical fact constrains what is physically possible, and this constraint is explanatory even though it is not causal. This view has been influential but remains controversial, and the debate between the descriptive, inference-based, and genuine explanation views is one of the liveliest in the subfield.
The intra-mathematical and extra-mathematical questions are often treated separately, but they are connected in at least two ways. First, the same philosophical tools are used in both. The unification approach, for instance, has been applied to both questions: a proof that unifies a mathematical domain is explanatory in the same way that a scientific theory that unifies a physical domain is explanatory. Second, the answers to the two questions may be linked. If one holds that mathematical explanations are always non-causal, then the intra-mathematical question becomes a special case of the extra-mathematical question, and the same account of non-causal explanation should apply to both. Conversely, if one holds that explanation is always causal, then both mathematical and physical explanations must be causal, which seems to force a rethinking of what counts as a cause.
Some philosophers have argued that the two questions are not merely connected but identical. On this view, a proof that explains a mathematical theorem is explaining the same kind of thing as a mathematical explanation of a physical fact: both are explanations of why something must be the case, given the structure of the relevant domain. This view has the virtue of unifying the subfield, but it risks collapsing the distinction between mathematics and science, which many philosophers want to preserve.
The contemporary subfield is characterized by a diversity of approaches and a growing body of case studies. Philosophers of mathematical practice have produced detailed studies of explanatory proofs in specific areas—number theory, geometry, algebra, analysis—and these studies have enriched the abstract debate with concrete examples. The concept of mathematical depth has emerged as a related but distinct topic, and the relationship between depth and explanation is itself a subject of investigation. Some philosophers have argued that depth is a more fundamental notion than explanation, and that explanatory proofs are those that reveal deep structure.
The extra-mathematical debate has been sharpened by the development of non-causal explanation in philosophy of science. The recognition that some scientific explanations are not causal—for instance, explanations in terms of symmetry principles or conservation laws—has made it easier to see how mathematical explanations might fit into a broader taxonomy of explanation types. The question of whether mathematical explanations are a species of non-causal explanation, or whether they are sui generis, remains open.
A notable feature of the present landscape is the increased attention to explanatory practice in applied mathematics. The use of mathematics in physics, biology, economics, and engineering raises questions that are not captured by the traditional intra-mathematical/extra-mathematical distinction. When a mathematical model explains a phenomenon, is the explanation mathematical, physical, or some hybrid? The growing field of philosophy of modeling has begun to address these questions, and it has drawn on the literature on mathematical explanation while also challenging some of its assumptions.
The subfield remains methodologically pluralistic. There is no consensus on whether explanatory proofs exist, on what makes a proof explanatory, or on whether mathematics can explain physical facts. This pluralism is not a sign of immaturity but a reflection of the difficulty of the questions. Mathematical explanation sits at the intersection of philosophy of mathematics, philosophy of science, and the study of mathematical practice, and it draws on all three without being reducible to any of them. The questions it asks are among the oldest in philosophy—what is explanation, and what is mathematics—and the answers it offers are necessarily tentative. What the subfield has achieved is a set of well-defined problems, a rich body of examples, and a range of approaches that can be tested against those examples. That is a substantial achievement for a field that did not exist in its current form a few decades ago.