Poincaré’s Conventionalism
1) About Henri Poincaré
Henri Poincaré (1854–1912) was a French mathematician, physicist, philosopher of science, and one of the most influential thinkers in the development of modern mathematics and the philosophy of science. His work ranged across number theory, differential equations, topology, celestial mechanics, mathematical physics, and the foundations of geometry. Although he is principally remembered as a mathematician, Poincaré also developed a distinctive philosophical account of scientific knowledge, particularly concerning the status of mathematical principles and physical theories. His philosophical writings, including La Science et l’Hypothèse (Science and Hypothesis, 1902), La Valeur de la Science (The Value of Science, 1905), and Science et Méthode (Science and Method, 1908), examined how scientific reasoning operates and what can legitimately be claimed about reality.
Poincaré’s philosophy emerged during a period in which traditional assumptions about mathematics and physics were being seriously challenged. The discovery of non-Euclidean geometries had demonstrated that Euclidean geometry was not the only logically coherent geometrical system. At the same time, developments in physics were calling into question older conceptions of space, time, matter, and motion. Poincaré was therefore confronted with a fundamental philosophical problem: if several mathematical descriptions are internally consistent, and if experience alone does not uniquely determine which one should be adopted, what explains the choice of one framework rather than another? His response was neither a straightforward realism nor a simple rejection of objective knowledge. Instead, he emphasised the role of conventions in organising scientific thought.
This position is commonly described as conventionalism. Poincaré argued that certain fundamental principles used in science cannot be regarded simply as empirical discoveries forced upon us by observation. They function partly as conventions that scientists adopt in order to organise experience effectively. Such conventions are not arbitrary in the sense of being chosen without reason. Their selection is influenced by considerations such as simplicity, convenience, coherence, and compatibility with established scientific practice. Scientific systems therefore involve both empirical elements, which are constrained by observation, and conventional elements, which provide the conceptual frameworks through which observations become intelligible.
Poincaré’s conventionalism is especially important in relation to geometry. He rejected the idea that experience straightforwardly demonstrates the truth of Euclidean geometry as a description of physical space. Physical measurements always involve instruments, physical assumptions, and interpretations. Consequently, when observations appear to support a geometrical proposition, it is difficult to separate the geometrical assumptions from the physical principles used in carrying out the measurement. Poincaré maintained that alternative combinations of geometry and physical laws could sometimes account for the same observable phenomena. The choice between such alternatives could therefore depend upon broader methodological considerations rather than upon empirical evidence alone.
His position should not, however, be confused with the claim that scientists are free to choose any theory they please. Poincaré did not believe that scientific knowledge was merely a matter of personal preference. Experience imposes significant constraints upon scientific theories, and a scientific framework must remain capable of organising and predicting phenomena successfully. What conventions determine is not the total content of science but the underlying conceptual arrangement within which empirical knowledge is expressed. This distinction allows Poincaré to occupy a position between naïve realism and radical relativism.
Another important feature of Poincaré’s thought is his emphasis on the practical activity of scientific reasoning. Scientists do not merely receive facts from the world and record them mechanically. They select concepts, formulate hypotheses, construct mathematical representations, and determine which principles will provide the most effective organisation of phenomena. Scientific knowledge is consequently shaped by human intellectual practices. Yet these practices remain answerable to the behaviour of the natural world. A successful scientific system must strike a balance between conceptual economy and empirical adequacy.
Poincaré’s conventionalism also reflects his broader view of mathematics. He regarded mathematical structures as indispensable instruments for scientific reasoning, but he resisted reducing mathematics entirely to empirical observation. Mathematical principles provide frameworks through which physical relations can be represented and calculated. Their usefulness in science does not necessarily establish that they correspond to independently existing structures in nature. This insight became particularly significant as mathematical physics developed increasingly sophisticated theoretical descriptions that could be formulated through different mathematical frameworks.
The lasting significance of Poincaré’s conventionalism lies in the way it draws attention to the active role of conceptual choice in scientific knowledge. His philosophy does not imply that scientific theories are invented without constraint; rather, it shows that evidence alone may leave room for alternative theoretical organisations. Scientific rationality therefore involves more than collecting observations. It also involves decisions concerning definitions, mathematical structures, explanatory principles, and the overall economy of a theory. Poincaré’s account consequently became an important precursor to later discussions about underdetermination, theory choice, the nature of scientific representation, and the relationship between mathematical structures and physical reality.
2) Undetermination of Theory by Evidence
A central feature of Poincaré’s conventionalism is the idea that empirical evidence does not always determine a single theoretical interpretation. Scientific observations may be perfectly compatible with more than one conceptual framework, meaning that the evidence itself cannot necessarily decide between competing theories. This position is often described as the underdetermination of theory by evidence. For Poincaré, the problem becomes particularly clear when considering scientific laws that involve both empirical observations and mathematical or conceptual assumptions. What appears to be a direct discovery about nature may partly depend upon the framework through which the observations are interpreted.
The significance of underdetermination can be seen in the relationship between geometry and physical science. Suppose measurements appear to confirm that physical space behaves according to Euclidean geometry. The measurements themselves, however, do not occur independently of physical assumptions. Measuring the length of a line requires instruments, and determining whether an instrument is accurate requires assumptions about the behaviour of physical objects and processes. Thus, an apparent test of geometry may simultaneously involve assumptions about physics. The resulting evidence may therefore be interpreted in more than one way, depending upon which elements of the theoretical system are held fixed.
Poincaré used this insight to challenge the idea that experience simply reveals which geometry nature has chosen. If observations conflict with the predictions of a geometrical framework, scientists have several possible responses. They may modify the geometry, alter the physical laws governing the relevant phenomena, or revise assumptions concerning the behaviour of measuring instruments. Consequently, the same observational results can sometimes be incorporated into different theoretical systems. One system might retain Euclidean geometry while introducing more complicated physical laws, whereas another might employ a non-Euclidean geometry alongside simpler physical principles.
This does not mean that all theories are equally supported by evidence. Empirical observations continue to place genuine restrictions on scientific reasoning. A theory that repeatedly fails to account for observable phenomena cannot simply be protected by declaring every inconvenient feature a convention. Poincaré’s point is subtler: evidence constrains the overall system, but it may not uniquely determine how the individual components of that system should be divided between empirical laws, geometrical principles, definitions, and conventions. Scientific testing therefore operates on interconnected theoretical structures rather than on isolated propositions considered independently.
The issue becomes especially important because scientific theories normally contain many mutually supporting assumptions. An experiment does not test a single statement in complete isolation. It may depend upon assumptions concerning the operation of instruments, mathematical calculations, background laws, measurement standards, and initial conditions. When an unexpected result occurs, scientists must decide which component of the wider system requires revision. Poincaré’s conventionalism highlights precisely this flexibility within scientific reasoning. The evidence may tell us that something within the system needs to change without telling us uniquely what that something must be.
Poincaré therefore distinguished between the empirical content of science and the organisational principles used to structure that content. Some statements are closely tied to observation and can be directly confronted with experience, whereas others function as principles through which observations are represented and connected. The latter may possess a degree of conventionality without being meaningless or arbitrary. Their value lies partly in the way they enable scientists to construct coherent and workable descriptions of phenomena. Scientific knowledge consequently involves an interaction between what experience constrains and how human reason organises what experience provides.
The underdetermination of theory by evidence also explains why scientific progress cannot be understood merely as the accumulation of observations. New evidence does not automatically produce a unique theoretical conclusion. Scientists must interpret that evidence within an existing conceptual framework and determine which modifications produce the most satisfactory result. Two competing frameworks might initially explain the same body of evidence, yet differ in simplicity, mathematical elegance, practical usefulness, or compatibility with other areas of science. Theory choice therefore involves methodological judgement in addition to empirical testing.
Poincaré’s argument ultimately gives scientific rationality a more complex character than a simple model in which facts dictate theories. Observation remains indispensable, but observations acquire scientific significance through theoretical organisation. Different theoretical structures can sometimes accommodate the same empirical material, and choosing between them requires considerations extending beyond direct observational confirmation. This insight became particularly influential in later philosophy of science, where questions about the relationship between evidence and theory developed into broader debates concerning scientific realism, theoretical equivalence, model choice, and the limits of empirical determination.
3) Disk World Thought Experiment
The Disk World thought experiment is one of the clearest ways of illustrating Poincaré’s argument about the relationship between geometry, physical laws, and the description of the world. Poincaré imagined a two-dimensional world consisting of a vast disk in which the physical properties of objects change systematically as one moves towards the boundary. The inhabitants of this world would attempt to understand the geometry of their environment through measurements made with physical instruments. Their situation demonstrates that what appears to be a geometrical property may alternatively be interpreted as a consequence of physical laws affecting measuring instruments and objects.
In the imagined world, the inhabitants live on a large circular surface whose physical conditions vary according to location. Objects become progressively smaller as they approach the outer boundary of the disk. More precisely, the scale of their measuring instruments changes in such a way that the inhabitants themselves do not notice the alteration. Their rulers, bodies, and other physical objects contract according to the same systematic pattern. Because everything involved in ordinary measurement changes together, the inhabitants have no straightforward way of detecting the contraction simply by comparing one object with another.
This creates a striking problem concerning the geometry of their world. Suppose the inhabitants construct what they regard as a straight line and measure the circumference of a circle. They may discover relationships that appear inconsistent with Euclidean geometry. Their measurements could suggest that the ratio between a circle’s circumference and its diameter differs from the familiar Euclidean value of π. From one perspective, they might conclude that the geometry of their world is non-Euclidean. The geometrical structure itself could therefore be regarded as different from the geometry familiar to ordinary Euclidean intuition.
Poincaré’s thought experiment shows, however, that another interpretation is possible. Instead of saying that the geometry of the Disk World is non-Euclidean, the inhabitants could retain Euclidean geometry and argue that physical forces cause objects and measuring instruments to change size according to their position. Under this interpretation, the unusual measurements would not indicate that space itself possesses non-Euclidean geometry. They would instead result from the physical behaviour of matter. The same observational situation could therefore be represented through two different theoretical arrangements.
The first arrangement would attribute the observed effects primarily to geometry. The inhabitants could adopt a non-Euclidean geometry and regard their physical laws as comparatively simple. The second arrangement could preserve Euclidean geometry but introduce more complicated physical laws governing the contraction of bodies and instruments. Both approaches could, in principle, reproduce the same observable measurements. The difference would lie in where the theoretical system places the complexity: in its geometry or in its physics.
This example is important because the inhabitants cannot simply settle the question by measuring more carefully. Their measurements are themselves affected by the physical conditions of their environment. Every ruler they use undergoes the same relevant transformation as the objects being measured. Consequently, there is no purely observational procedure that automatically identifies one interpretation as the uniquely correct description. What counts as the geometry of their physical world depends partly upon the conventions and theoretical principles adopted for interpreting their measurements.
The Disk World therefore illustrates Poincaré’s broader distinction between a description of physical reality and the conceptual framework used to organise that description. Geometry is not treated as a completely isolated science whose propositions can be read directly from physical measurements. Instead, geometrical assumptions interact with physical hypotheses. A change in one part of the theoretical framework can compensate for a change in another. The inhabitants may preserve the same empirical predictions while reorganising the relationship between geometry and physics.
The thought experiment consequently provides a vivid demonstration of why Poincaré regarded the choice of geometry as partly conventional. The issue is not that one can freely declare any geometry to be true regardless of experience. Rather, different geometrical frameworks can sometimes be connected with different physical assumptions in ways that preserve the same observable consequences. The most appropriate framework is therefore selected according to broader scientific considerations, such as simplicity, coherence, and convenience. The Disk World makes this philosophical point concrete by showing how the very same measurements can support different ways of conceptualising the structure of space.
4) Role of Convenience and Simplicity
For Poincaré, the selection of a scientific framework cannot always be settled by observation alone. When more than one theoretical arrangement is compatible with the available evidence, scientists must rely upon additional methodological considerations. Among the most important of these are convenience and simplicity. These considerations help scientists decide how best to organise their knowledge when empirical evidence leaves more than one viable possibility. Poincaré therefore regarded scientific reasoning as involving not only the discovery of facts but also the practical organisation of those facts into an intelligible system.
Convenience, in this context, does not mean mere personal preference. A convention is convenient when it allows scientists to describe, calculate, compare, and predict phenomena efficiently. Scientific concepts are valuable partly because they make complex relationships manageable. A framework that enables many different observations to be expressed through a relatively small number of principles will generally be preferable to one that requires a large collection of special qualifications. Convenience is consequently connected with the practical effectiveness of a scientific language or system.
Simplicity plays a closely related role. When two theoretical systems account for the same observations, scientists are often inclined to favour the one that has a simpler structure. Poincaré recognised that simplicity can arise in different ways. A theory may have simpler mathematical principles, fewer independent assumptions, more economical laws, or a more unified description of apparently unrelated phenomena. The preference for simplicity therefore helps determine how scientific knowledge is organised even when observation does not uniquely dictate the arrangement.
The relationship between geometry and physics provides a particularly important example. A scientist might choose Euclidean geometry together with complicated physical laws, or a non-Euclidean geometry together with simpler physical laws. If both combinations successfully reproduce the same observations, there may be no isolated empirical fact that forces one choice. The decision can instead depend upon which complete system offers the greater economy. Poincaré’s conventionalism therefore does not make scientific choice irrational; it explains why rational choice may involve criteria other than direct empirical confirmation.
Poincaré also connected simplicity with the structure of scientific language and calculation. Scientists need systems that allow them to communicate results and derive consequences without unnecessary complication. A complicated framework can technically describe the same phenomena as a simpler one, but it may be less useful because it requires more effort to operate and offers fewer opportunities for generalisation. The preference for simplicity thus has a practical dimension: scientific theories are tools for reasoning, prediction, and coordination among researchers.
At the same time, Poincaré did not regard simplicity as an absolute measure of truth. A simpler theory is not automatically a more accurate representation of reality. Simplicity is a methodological virtue that becomes particularly significant when several systems have comparable empirical success. Its importance arises from the need to choose among alternative ways of organising knowledge, rather than from the assumption that nature itself must possess the simplest possible structure. This distinction prevents conventionalism from being reduced to the claim that whatever is simplest must necessarily be true.
The value of convenience also extends to the establishment of scientific conventions. Scientists adopt standards for measurement, terminology, mathematical representation, and classification partly because common conventions make collective inquiry possible. Once established, these conventions can become deeply integrated into scientific practice. They provide stable reference points that allow researchers to compare observations and formulate laws consistently. Their conventional character does not make them useless; on the contrary, their usefulness often depends upon widespread and sustained agreement within scientific practice.
Poincaré’s emphasis on convenience and simplicity therefore reveals an important dimension of scientific rationality. Scientific frameworks are judged not solely by whether they correspond directly to isolated observations, but also by how effectively they organise the total body of knowledge. A successful system should be empirically adequate while remaining coherent, economical, and workable. This approach explains why scientists can rationally adopt conventions without believing that those conventions are arbitrary creations. They are selected because they provide effective structures for thought, allowing scientific inquiry to proceed with greater unity and efficiency.
5) DIstinction Between Geometric Space and Representative Space
Poincaré distinguished between geometric space and what may be called representative space in order to clarify the difference between space as an abstract mathematical structure and the way spatial experience is organised through perception and scientific representation. This distinction is important because human beings do not encounter geometrical space in a completely direct or theory-free manner. We experience objects, distances, directions, and movements through our sensory and intellectual capacities, and these experiences are subsequently organised through mathematical concepts. The spatial world of immediate experience is therefore not identical with an abstract geometrical system.
For Poincaré, geometric space possesses certain characteristics that distinguish it from ordinary sensory experience. Mathematical geometry treats space as continuous, homogeneous, and structured according to definite relations. It provides idealised concepts such as points, lines, planes, and distances, none of which are encountered in their perfectly mathematical form through ordinary perception. A mathematical point, for instance, has no physical extension, while every physical mark that might represent a point possesses some size. Geometry therefore operates at a level of abstraction that cannot simply be identified with the raw contents of sensation.
Representative space, by contrast, concerns the way spatial information is presented to us through our senses and organised by the mind. Visual perception, for example, gives us impressions of shape, position, depth, movement, and distance. These impressions are influenced by the structure of our sensory apparatus and by the ways in which we have learned to coordinate different experiences. What we perceive as a spatial arrangement is consequently already an organised representation rather than an unmediated apprehension of mathematical space. Poincaré’s distinction helps explain why the geometry used in science cannot simply be equated with immediate perception.
This becomes particularly significant when considering the role of the senses in the formation of geometrical knowledge. If geometry were simply copied from sensory experience, there would be little reason to distinguish between different geometrical systems. Yet the development of non-Euclidean geometry demonstrated that several internally coherent geometrical structures could be constructed. Human perception by itself does not provide a complete logical proof that one of these systems must describe physical space. Geometry therefore involves intellectual organisation beyond what is directly supplied by sensation.
The distinction also illuminates Poincaré’s understanding of measurement. When people measure a physical distance, they do not compare an object with an abstract mathematical line in isolation. They use physical instruments and interpret the results through established mathematical and physical assumptions. A ruler is itself a material object whose behaviour can change under different physical conditions. The numerical result of a measurement therefore belongs to a broader representational system. What is being measured is not simply an immediately given geometrical magnitude but a quantity defined and interpreted through scientific procedures.
Poincaré’s position also allows him to explain why different geometrical descriptions can be applied to the same physical phenomena. The physical world does not present scientists with labels saying which mathematical structure they must employ. Scientists construct representations that organise observed relationships in systematic ways. One representation may employ Euclidean geometry, while another may use a non-Euclidean framework together with corresponding physical assumptions. The important question is not whether experience contains a geometrical theory ready-made, but which representational system provides the most effective organisation of experience.
This distinction should not be interpreted as a denial of the external world. Poincaré was not claiming that space is merely a private mental invention. Rather, he was analysing the means through which human beings represent and understand spatial relations. Physical objects and their behaviour impose constraints upon our representations, but the mathematical structure used to describe those relations is not simply identical with the objects themselves. Scientific knowledge consequently involves a process of conceptual mediation between physical phenomena and their mathematical representation.
The distinction between geometric and representative space ultimately supports Poincaré’s broader conventionalism by showing that spatial knowledge involves both experience and intellectual construction. Sensory experience provides the material with which scientific inquiry begins, while mathematical structures organise that material into a systematic representation. Because several forms of organisation may sometimes be compatible with the same physical observations, the choice of geometrical framework cannot always be settled by perception alone. Poincaré’s distinction thus helps establish why geometry occupies a special position between empirical description and conceptual convention.
6) Axioms as DIsguised Definitions
Poincaré’s conventionalism gives a distinctive interpretation of the axioms of geometry. Rather than treating geometrical axioms as straightforward empirical discoveries or as necessary truths imposed upon the physical world, he regarded some of them as functioning like definitions in disguise. An axiom can establish the meaning and rules of a geometrical system without necessarily making a factual claim about the physical universe. This interpretation helps explain how different geometrical systems can be developed without requiring scientists to believe that one of them has been directly revealed by experience.
The idea can be understood by considering how mathematical concepts acquire their significance within a formal system. Terms such as ‘point’, ‘line’, and ‘parallel’ do not need to correspond perfectly to individual physical objects. Instead, their meaning is partly determined by the relations established among them. An axiom specifies how these concepts are to function within the system. Consequently, adopting a particular set of axioms is similar to establishing the rules according to which a certain mathematical language will operate. The axioms determine the structure of the system rather than simply reporting independently observed facts.
This interpretation was especially relevant to the status of Euclid’s parallel postulate. For centuries, the parallel postulate had been regarded as a candidate for a fundamental truth about the nature of space. The development of non-Euclidean geometries demonstrated, however, that coherent geometrical systems could be constructed by rejecting or modifying it. Poincaré took this development as evidence against the idea that Euclidean axioms possess an unquestionable empirical status. If several geometries can be constructed consistently, then the choice of geometrical axioms cannot simply be explained by saying that one set is logically unavoidable.
Calling axioms ‘disguised definitions’ does not mean that they are arbitrary stipulations. A useful geometrical system must possess internal consistency and must enable mathematicians and scientists to reason effectively. Moreover, when geometry is applied to physical phenomena, the resulting system must remain compatible with experience. The conventional element therefore concerns the way the basic concepts and relations are established, while empirical investigation remains important when determining how those structures are connected with the physical world. Poincaré’s view combines freedom of conceptual organisation with significant practical and empirical constraints.
The notion of disguised definition also clarifies why Poincaré resisted treating geometry as an experimental science in exactly the same way as physics. Physical laws make substantive claims about how objects and processes behave and can therefore be confronted with observations in a more direct fashion. Geometrical axioms, by contrast, help establish the framework within which physical relations are represented. If a physical prediction based upon a particular geometrical framework fails, scientists can potentially respond by changing the physical laws rather than changing the geometry itself. The geometrical axioms consequently occupy a different methodological position from ordinary empirical hypotheses.
Poincaré’s account also avoids the opposite conclusion that geometry is entirely subjective. A definition is not simply a private opinion, and a mathematical system cannot be altered without consequences. Once a set of axioms has been adopted, its logical consequences follow independently of individual preference. Furthermore, different systems can be compared according to their coherence, usefulness, simplicity, and suitability for scientific applications. The conventional origin of an axiom therefore does not undermine the objectivity of the mathematical reasoning carried out within the resulting framework.
The concept of disguised definitions is particularly important because it reveals the flexibility available within scientific description. Scientists may preserve an established geometrical framework while modifying physical laws, or they may adopt a different geometry and retain a simpler account of physical behaviour. In either case, the axioms determine the structural language in which the description is expressed. What changes is not necessarily the observational content but the conceptual organisation used to represent it. This is one reason Poincaré considered scientific frameworks to involve both factual and conventional components.
Poincaré’s interpretation of axioms challenges the traditional assumption that the foundations of geometry must either be self-evident truths or empirical facts about physical space. Axioms can instead function as principles that establish a conceptual framework, much as definitions establish the meaning and use of terms. Their justification therefore depends upon more than correspondence with isolated observations: consistency, convenience, simplicity, and scientific usefulness also matter. Through this conception of axioms, Poincaré developed a sophisticated form of conventionalism in which mathematics provides structured systems for organising knowledge without requiring every foundational principle to be understood as a direct statement about external reality.
