Loading field map
What makes a proof explanatory rather than merely correct? This question has driven a half-century of debate in the philosophy of mathematics. The puzzle is that two proofs of the same theorem can differ in explanatory power: one may reveal why the theorem holds, while the other only shows that it does. Understanding this difference has turned out to require rethinking what explanation itself means, both inside pure mathematics and in the applications of mathematics to science.
The earliest systematic framework for mathematical explanation was borrowed from the philosophy of science: the Covering-Law Model (often called the deductive-nomological model). According to this view, an explanation is an argument that subsumes a phenomenon under a general law. A proof would be explanatory if it deduces a theorem from more fundamental mathematical laws. This model dominated from the 1940s through the 1960s, but it faced a severe problem: many proofs that were clearly non-explanatory still fit the covering-law schema, while some proofs that mathematicians found explanatory seemed to violate it. The model could not distinguish between a proof that merely demonstrates a result and one that provides understanding. By the 1970s, philosophers recognized that the covering-law model was too coarse to capture the normative judgments mathematicians make about explanations.
In the late 1970s and early 1980s, the field divided into two research programs, each focusing on a different sense of 'mathematical explanation'. Mark Steiner's 1978 work launched the Intra-Mathematical Explanation framework, which sought to characterize explanation within pure mathematics. Steiner proposed that an explanatory proof is one that depends on a 'characterizing property' of the entities involved—a property that is 'deeper' and more general than the theorem itself. For example, the standard proof of the irrationality of √2 uses the Euclidean property of primes, which is more fundamental than the result. Steiner's account replaced the covering-law model by emphasizing mathematical content rather than logical form. Yet it also faced challenges: it seemed too permissive (some non-explanatory proofs also rely on characterizing properties) and too narrow (it excluded combinatorial proofs that mathematicians find explanatory).
Concurrently, Hartry Field opened a different direction with the Extra-Mathematical Explanation framework. Field asked how mathematics can explain empirical phenomena, such as why cicadas emerge in prime-numbered cycles. He argued that mathematics itself does not provide causal explanations; rather, it is a tool for representing and reasoning about the physical world. For Field, genuine explanations in science are causal, and mathematics plays a purely descriptive role. This view starkly contrasted with Steiner's. Where Steiner sought the source of explanatory power inside mathematical structures, Field located it in the causal structure of the physical world, with mathematics serving as an instrument. Field’s work also motivated a broader anti-realist position known as Fictionalism: since mathematics is not explanatory on its own, we can treat mathematical statements as useful fictions without ontological commitment. This position was further developed by other philosophers under the heading of Nominalism, which denies the existence of abstract mathematical objects. The key difference between Fictionalism and Nominalism is subtle: while both reject Platonism, Fictionalism emphasizes that mathematical discourse is literally false but useful, whereas Nominalism focuses on constructing nominalistic alternatives to standard mathematics. Both, however, were direct responses to the Indispensability Argument.
In 1981, Philip Kitcher proposed a rival to both Steiner and the covering-law model: the Unificationist Account. Kitcher argued that explanation is a matter of reducing the number of independent argument patterns needed to derive a wide range of conclusions. An explanatory proof is one that instantiates a pattern that also applies to many other cases, thereby unifying disparate phenomena. For example, the proof of the quadratic formula unifies many specific equations. The Unificationist Account directly challenged Steiner's focus on properties: Kitcher saw explanatory power in economy of patterns, not in depth of mathematical content. It also avoided the covering-law model's reliance on laws, which are rare in pure mathematics. However, critics noted that some highly unified proofs feel mechanical rather than illuminating, and that the notion of 'argument pattern' is vague. The Unificationist Account coexisted with the intra- and extra-mathematical programs, but it did not replace them; all three remained live options.
The question of explanation became central to a deeper ontological debate. The Indispensability Argument, rooted in Quine and Putnam, held that we should be committed to the existence of mathematical objects because they are indispensable to our best scientific theories. This argument assumed that mathematics plays an explanatory role in science. Field’s Fictionalism and Nominalism challenged that assumption: if mathematics is merely a convenient tool, its explanatory role is not genuine, and ontological commitment can be avoided. Field attempted to show that science can be reformulated without mathematics, though this project proved difficult.
In 2000, Alan Baker refined the Indispensability Argument into the Explanatory Indispensability Argument. Baker argued that mathematics sometimes provides the best explanation of empirical phenomena—the cicada case being a prime example. The mathematical explanation of prime-numbered cycles is not just a reformulation of a causal story; it is genuinely explanatory. Therefore, we have reason to believe in the mathematical entities invoked. The Explanatory Indispensability Argument directly countered Field’s Fictionalism and Nominalism. While the earlier Indispensability Argument had focused on the overall utility of mathematics in science, Baker’s version targeted specific explanations, making it harder to dismiss mathematics as a mere calculating device. The debate between these frameworks remains active: Indispensabilists (realists) argue that the mathematical explanations in science are autonomous, while Fictionalists and Nominalists maintain that the explanation can be paraphrased or that mathematical entities are not required for the causal structure.
By the 2000s, many philosophers concluded that no single framework captures all cases of mathematical explanation. This led to Pluralism, the view that there are multiple legitimate kinds of mathematical explanation, each suited to different contexts. Pluralism is not a rejection of earlier frameworks but a methodological shift: instead of trying to find the one true account, it takes the diversity of mathematical practice as its starting point. Proponents argue that both intra-mathematical and extra-mathematical explanations have distinct norms, and that unificationist, causal, and property-based models can all be valuable in appropriate domains. Pluralism draws on insights from the philosophy of mathematical practice, emphasizing how mathematicians actually evaluate proofs.
What do the leading frameworks agree on today? First, almost all reject the covering-law model as insufficient. Second, there is broad consensus that explanation in mathematics is non-causal in a way that distinguishes it from typical scientific explanation. Third, participants accept that any account must be sensitive to the evaluative judgments of working mathematicians. The major disagreements remain: (1) whether extra-mathematical explanation is genuinely mathematical or merely causal; (2) whether the Explanatory Indispensability Argument establishes realism or whether anti-realist responses (Fictionalism, Nominalism) prevail; and (3) whether pluralism is a stable position or a sign that the field has not yet found the right unified account. The intra-mathematical framework continues to be developed, with recent work connecting it to the notion of mathematical depth. Extra-mathematical explanation, especially through the lens of Baker's argument, remains at the center of the realism debate. Pluralism has become the default methodological stance for many, but it is not without critics who seek a more systematic theory. The philosophy of mathematical explanation is thus a lively subfield, animated by competing accounts of what it means to 'understand' a mathematical truth.