EVOPEDIA

← all systems · see it in The System

Mathematical proof: the practice of showing that a statement follows from axioms and earlier theorems, how it is written, checked (by people and machines) and accepted, and the derivations it works on

A mathematical proof is a deductive argument for a mathematical statement, showing that the stated assumptions logically guarantee the conclusion.

Properties published scienceour proposal

Sourced fields, and our own readings.

born at6th century BCE (Thales, 624-546 BCE)
born precisioncentury; origin contested
boundaryInside: the practice of stating, proving, checking and accepting mathematical statements (methods of proof, the standard of rigour, refereeing, formal derivation systems, proof assistants and their kernels, the community's acceptance) and the structure it works on (derivations from axioms by rules of inference). Outside: the mathematics proved (number, geometry, set theory as fields), the logic whose systems proofs are written in (studied as a field of its own), the computers and people that carry proofs, and empirical or statistical arguments, which are distinguished from proof.
nested levelA part of mathematics (the way its statements become theorems) and an object of mathematical logic (proof theory). Runs on people (mathematicians, referees, students), on writing (papers, books) and, since 1967, on proof checkers and computers.
scalesocial

Measures (28) published science

characteristic clock: four colour conjecture to proof (1852 to 1976)124 years (derived from two dates the source gives)
characteristic clock: time a false proof stood (Kempe 1879 to Heawood 1890; Tait 1880 to Petersen 1891)11 years
wait from Kepler proof (1998) to its publication (2006)8 years (derived from two dates the source gives)
Flyspeck formalisation (announced January 2003, completed 2014)about 11 years
CPU time for the hardest Flyspeck subclaimabout 5000 CPU hours
time to check the formal Kepler main statementabout 5 hours (scripts); about 40 minutes (recorded proof) on a 2 GHz CPU
nonlinear inequalities in the Kepler proofnearly a thousand inequalities
cases in the first four colour proof (proof by exhaustion)1,936 cases
reducible configurations after Robertson, Sanders, Seymour and Thomas (1997)633 configurations
referees' certainty in the Kepler proof99 percent (as stated by referees)
HOL Light kernel sizeabout 400 lines of code
HOL Light primitive inference rules10 rules, 3 axioms count
share of library theorems proved by fully automated procedures (Mizar)40 (Mizar); 47 (HOL Light/Flyspeck) percent
Logic Theorist theorems proved of Principia's first 5238 theorems (1956)
Odd Order proof length; quasi-thin classification length255; 1221 pages
mathlib size growth 2017-201915k to 140k lines of code
mathlib contributors (2019)73 contributors, 11 maintainers people
mathlib theorems (May 2025)over 210,000 theorems, 100,000 definitions count
Metamath set.mm libraryabout 23,000 theorems, 600k lines count
Mizar libraryover 3.1 million lines of code
size of the Boolean Pythagorean triples proofabout 200 TB
refereeing time for a machine-assisted proof7 years (not completely refereed); 3 years (another)
expansion from conversation to talk to paperminutes; an hour talk; a 15-20 page paper read in hours or days time and length (Thurston's estimate)
Thurston notes mailing listabout 1200 people
mathematicians scanned for beauty; equations rated15; 60 count
length of math demonstrations followed in EEG13-68 seconds
Fermat's Last Theorem Lean formalisation reported (2026)eleven days; about 13 million lines; roughly 30,000 intermediate theorems as reported by the encyclopedia
Proofs from THE BOOK (2003)32 proofs

Open contentions (6) open contention

Where sources or readings disagree. Both stay on the map.

Partners and parts published science

Built from / builds (11)

Sits inside / holds (3)

Trades with (valence) (24)

expressed by (26)

floor filled by (23)

represents (1)

Kinds (15)

Notes published science

Layers

The ways it takes in its world, where they meet, and how it acts back, each from L1 (where it meets the world) up.

a mathematical proof as mathematicians and logicians describe it: a statement shown to follow from axioms, definitions and earlier theorems by accepted steps, written or spoken, read and checked by people or machines, accepted by the community and built on (reader's view)

  1. L1: The argument as marks and sounds: proofs on paper, blackboard and screen in rigorous informal logic (symbols mixed with natural language), spoken in talks and conversation, and formal proof scripts in files.
  2. L2: Taking it in: reading symbols and figures, hearing a talk, a proof assistant parsing a script; readers 'translate back' from symbols to ideas.
  3. L3: The statement: definitions, axioms, the formal language, and the claim to be shown (a conjecture until proved).
  4. L4: Steps of inference: direct combination of earlier results, induction, contraposition, contradiction, construction, exhaustion, probabilistic and combinatorial arguments; in formal systems, each formula follows from those before by a rule.
  5. L5: Checking against a standard of rigour: readers, referees and colleagues looking for gaps; derivation systems that can be checked mechanically; proof-checker kernels; independent programs and verifiers.
  6. L6: Acceptance and understanding: the community convinced by communal standards, understanding gained, beauty felt; and the limits on what can be proved (undecidable statements, incompleteness).
  7. L7: Use: theorems cited without proof and built on, libraries of formal mathematics, teaching, verified hardware and software, and the advance of human understanding.

the boardroom (shared floors)

  1. L3: The statement: definitions, axioms and the claim, in a formal language or rigorous natural language; conjectures waiting for proof.
  2. L4: Steps of inference: methods of proof, intermediate results held in the head, cited theorems, computations and machine proposals, joined into a deductive pathway; in the brain, a fronto-parietal network with stronger delta synchrony in experts.
  3. L5: Checking: the standard of rigour applied by readers, referees and colleagues, and by mechanical checkers whose kernels admit only primitive steps; soundness of the derivation system.
  4. L6: Acceptance and understanding: the community convinced, understanding and beauty, the theorem taken into the shared body of knowledge; and the limits of provability (undecidable statements, incompleteness).
  5. L7: Use: theorems cited and built on, formal libraries, teaching, verified systems; for Thurston, the point is advancing human understanding.

way in: Automated provers and AI: proposed steps and whole proofs

  1. L1: Proposals from programs: the Logic Theorist (1956), sledgehammers, language models and agents writing Lean.
  2. L2: Search: automated procedures find proofs of 40-47% of library theorems; a language model drafts an outline for a sledgehammer to fill.
  3. L3: The proposal stated as a formal goal for the checker.

way in: Balance and posture (vestibular, proprioception) while following a proof

  1. L1: Gravity on the body, sitting or standing during a long demonstration.
  2. L2: Keeping balance when standing: a load that may compete with the concentration a complex proof needs, more for novices than experts.

way in: Computation: computer calculations as parts of a proof

  1. L1: Programs run on physical hardware (all devices can fail): case checks, linear programs, interval arithmetic.
  2. L2: Results: thousands of cases or inequalities settled (four colour: 1,936 cases; Kepler: nearly a thousand nonlinear inequalities, about 5000 CPU hours for one subclaim), or a certificate a checker can read.
  3. L3: The calculation stated precisely as a mathematical claim that the proof uses.

way in: Emotion (inner): beauty, joy, the wish to understand

  1. L1: Where the feeling comes from: contested (C-413); here met as the response to a formula or proof.
  2. L2: The felt response: a formula or proof judged beautiful, indifferent or ugly; joy in ways of thinking that explain and simplify.
  3. L3: Field A1 of the medial orbitofrontal cortex: activity rises with rated beauty, as for visual, musical and moral beauty.

way in: Hearing: talks, lectures, conversation, spoken statements

  1. L1: Speech: one-on-one conversation, seminar talks, colloquia; spoken statements judged true, false or meaningless in the scanner.
  2. L2: Listening: within a subfield a proof is passed in minutes and 'generally understood' in an hour's talk; outside it the audience is often lost.

way in: Machine input: a proof checker reads a formal proof script

  1. L1: The proof script: definitions and proofs in a formal language (HOL Light OCaml scripts, Coq, Lean files).
  2. L2: Parsing and elaboration: the script is turned into terms of the logic.
  3. L3: The formal statement: the claim fixed as a defined constant so its meaning is pinned down before checking.
  4. L4: The kernel: every theorem built only through a small secured kernel of primitive rules (about 400 lines in HOL Light).

way in: Sight and spatial sense: figures and diagrams

  1. L1: Figures and diagrams drawn in papers or at the board (Euclid's figures; the fragment P. Oxy. 29 shows one).
  2. L2: Vision and spatial sense take the figure in; there is no good 'inverse vision', so papers carry fewer and poorer figures than heads.

way in: Sight: written proofs, symbols and formulae

  1. L1: Marks: statements and proofs in symbols and natural language; formulae on a screen.
  2. L2: Reading: the expert reads 'between the lines', recognising idioms for concepts; formulae seen and rated (Zeki's 60 equations).

way in: The community: colleagues, referees, reviewers and chat

  1. L1: Messages from others: referee reports, letters, seminar questions, GitHub pull-request reviews and Zulip chat (mathlib).
  2. L2: Expert scrutiny: people find weaknesses better than they check formal correctness; referees may stop short (the Kepler referees at '99% certain').
  3. L3: Gaps filled on request: authors supply detail and corrections when called on.
  4. L4: A verdict: enough peers convinced, or an objection raised.

way in: The literature: theorems already accepted, cited without proof

  1. L1: Published papers and books, preprints, formal libraries (mathlib's 210,000 theorems, May 2025).
  2. L2: Citation: what others quote without proof and cite in their bibliography is taken up the same way; much is known with no written source.
  3. L3: The cited theorem enters as a statement to build on (a lemma, a library declaration).

way in: Thought (inner): logic, intuition, association, process

  1. L1: Where a proof idea comes from: contested (C-413); Thurston describes intuition as sensing without knowing the source.
  2. L2: Intuition, association and metaphor; 'listening' to intuitions takes a quieting and focusing of the mind.
  3. L3: Held logic: built-in reasoning (implication, contradiction); the proof's structure held in the head as a process of steps.

way in: Touch and kinesthetic sense: hands, gesture, Braille

  1. L1: Contact and motion: gestures, Braille under the fingers, things held in the hands.
  2. L2: Kinesthetic (motion) sense and the hands: thinking with objects held, scanned or moved through; access to the abstract content is the same whatever the input sense.

way out: Applying: verified software and hardware, cryptographic proofs

  1. L1: Running systems (a verified C compiler, an operating-system kernel) and protocols (zero-knowledge proofs).
  2. L7: Formal verification of designs: the same tools check hardware and software against specifications.

way out: Formalising: typing the proof into a proof assistant

  1. L1: Scripts and library entries (mathlib pull requests reviewed by maintainers).
  2. L5: The argument restated so a kernel can check every step; errors surface and are repaired.

way out: Speaking: talks, lectures, conversation, teaching

  1. L1: Speech with gestures, pictures and sound effects.
  2. L6: What the audience needs, two minutes for one subfield, an hour for another.

way out: Writing the proof: papers, preprints, notes, Q.E.D.

  1. L1: The paper, preprint or notes sent out (Thurston's notes to about 1200 people).
  2. L4: Steps written out in symbols and logic; the end marked.
  3. L6: The understood argument chosen for an audience.

Sources (32) published science

Mathematical proof (encyclopedia article, WME snapshot 2026-11-01)
Four color theorem (Wikipedia, via Wikimedia Enterprise snapshot 2026-11-01) (behind its links)
Avigad, Mathematics and the formal turn (arXiv 2311.00007v2, 2023) (behind its links)
Thurston, On proof and progress in mathematics, Bull. Amer. Math. Soc. 30(2) (1994) 161-177 (arXiv math/9404236) (behind its links)
Hilbert's program (Wikipedia, via Wikimedia Enterprise snapshot 2026-11-01) (behind its links)
Lean (proof assistant) (Wikipedia, via Wikimedia Enterprise snapshot 2026-11-01) (behind its links)
Hales, T. et al. (2017) A formal proof of the Kepler conjecture. Forum of Mathematics, Pi. arXiv:1501.02155 (behind its links)
Hales, T. C. (2014) Developments in formal proofs. Séminaire Bourbaki no. 1086. arXiv:1408.6474 (behind its links)
Zach R (2025) Sets, Logic, Computation: An Open Introduction to Metalogic (Open Logic Project, Fall 2025, revision 6178f32; CC BY 4.0). Chapters 1 (Sets), 3 (Functions), 4 (The Size of Sets), 14 (Turing Machine Computations) (behind its links)
The mathlib Community (2020) The Lean mathematical library. CPP 2020, doi:10.1145/3372885.3373824. arXiv:1910.09336 (behind its links)
Mathematics (encyclopedia article, WME snapshot 2026-11-01) (behind its links)
Kepler conjecture (Wikipedia, via Wikimedia Enterprise snapshot 2026-11-01) (behind its links)
Automated theorem proving (Wikipedia, via Wikimedia Enterprise snapshot 2026-11-01) (behind its links)
Benzmüller, C. (2019) What is a proof? What should it be? arXiv:1904.06332 (behind its links)
Davenport, J., Poonen, B., Maynard, J., Helfgott, H., Tiep, P. H., Cruz-Filipe, L. (2018) Machine-assisted proofs (ICM 2018 panel report). arXiv:1809.08062 (behind its links)
Zeki S., Romaya J.P., Benincasa D.M.T., Atiyah M.F. (2014) The experience of mathematical beauty and its neural correlates. Front. Hum. Neurosci. 8, 68, doi:10.3389/fnhum.2014.00068 (PMC3923150, CC BY) (behind its links)
Poikonen, H. et al. (2024) Math on cortex: enhanced delta phase synchrony in math experts during long and complex math demonstrations. Cereb. Cortex, doi:10.1093/cercor/bhae025 (behind its links)
Proof assistant (Wikipedia, via Wikimedia Enterprise snapshot 2026-11-01) (behind its links)
Amalric, Denghien, Dehaene 2018. On the role of visual experience in mathematical development: Evidence from blind mathematicians. Dev Cogn Neurosci. doi:10.1016/j.dcn.2017.09.007 (CC BY-NC-ND) (behind its links)
Curry–Howard correspondence (Wikipedia, via Wikimedia Enterprise snapshot 2026-11-01) (behind its links)
Dedekind R (1872) Continuity and irrational numbers (Stetigkeit und irrationale Zahlen), in Essays on the Theory of Numbers, tr. W. W. Beman (1901), Project Gutenberg #21016 (public domain) (behind its links)
Mathematical induction (Wikipedia, via Wikimedia Enterprise snapshot 2026-10-01) (behind its links)
Method of exhaustion (Wikipedia, via Wikimedia Enterprise snapshot 2026-11-01) (behind its links)
Dutilh Novaes, C. (2021, rev.) Argument and Argumentation. Stanford Encyclopedia of Philosophy (peer-reviewed reference work) (behind its links)
Logic (Wikipedia, via Wikimedia Enterprise snapshot 2026-11-01) (behind its links)
Proof theory (Wikipedia, via Wikimedia Enterprise snapshot 2026-11-01) (behind its links)
Euclid's theorem (encyclopedia article, WME snapshot 2026-11-01) (behind its links)
Gödel's incompleteness theorems (Wikipedia, via Wikimedia Enterprise snapshot 2026-11-01) (behind its links)
Zero-knowledge proof (Wikipedia, via Wikimedia Enterprise snapshot 2026-11-01) (behind its links)
Probabilistically checkable proof (Wikipedia, via Wikimedia Enterprise snapshot 2026-11-01) (behind its links)
Altschuler, B. M., Altschuler, E. L. (2016) Proof of the irrationality of the square root of two in Babylonian geometry problem tablets. arXiv:1603.06656 (behind its links)
Varshney, L. R. (2012) Toward a comparative cognitive history: Archimedes and D. H. J. Polymath. Collective Intelligence 2012. arXiv:1204.3534 (behind its links)