IS GEOMETRY ANALYTIC?

In the fourth chapter of Language, Truth and Logic, Ayer undertakes the task of showing how a priori knowledge of mathematics and logic is possible. In doing so, he argues that only if we understand mathematics and logic as analytic,1 by which he memorably meant “devoid of factual content”2, do we have a justified account of a priori knowledge of these disciplines.3 In this chapter, it is not clear whether Ayer gives an argument per se for the analyticity of mathematics and logic. For when one reads that chapter, one sees that Ayer is mainly criticizing the views held by Kant and Mill with respect to arithmetic.4 Nevertheless, I believe that the positive argument is present. Ayer’s discussion of geometry5 in this chapter shows that it is this discussion which constitutes his positive argument for the thesis that analytic sentences are true


INTRODUCTION
In the fourth chapter of Language, Truth and Logic, Ayer undertakes the task of showing how a priori knowledge of mathematics and logic is possible.In doing so, he argues that only if we understand mathematics and logic as analytic, 1 by which he memorably meant "devoid of factual content" 2 , do we have a justified account of a priori knowledge of these disciplines. 3In this chapter, it is not clear whether Ayer gives an argument per se for the analyticity of mathematics and logic.For when one reads that chapter, one sees that Ayer is mainly criticizing the views held by Kant and Mill with respect to arithmetic. 4 Nevertheless, I believe that the positive argument is present.Ayer's discussion of geometry 5 in this chapter shows that it is this discussion which constitutes his positive argument for the thesis that analytic sentences are true 1 Now, I am aware that 'analytic' was understood differently by Kant, Carnap, Ayer, Quine, and Putnam.It is not even clear whether there is even an agreed definition of 'analytic' today.For the purposes of my paper, the meaning of 'analytic' is Carnap's sense as described in: Michael, Friedman, Reconsidering Logical Positivism (New York: Cambridge University Press, 1999), in terms of the relativized a priori. 2 Alfred Jules, Ayer, Language, Truth and Logic (New York: Dover Publications, 1952), p. 79, 87.In these sections of the book, I think the reason Ayer chose to characterize analytic statements as "devoid of factual" content is in order to give an account of why analytic statements could not be shown to be false on the basis of observation.
In other words, the reason why the analytic statements were necessary truths, according to Ayer, is that they did not assert that which required further facts or observation in order to establish their truth.One just needed to know the "meanings" or definitions of the terms in those statements in order to "see" that they are true independent of further observation or empirical data.For example, it is a fact that 'Euclidean triangle' refers to a planar three-sided figure.And it is a fact that every Euclidean triangle has angles adding up to 180 degrees.But given these facts, it follows that every planar three-sided figure has angles adding up to 180 degrees without need of any further observation.So, 'every planar three-sided figure has angles adding up to 180 degrees,' is an analytic statement in Ayer's sense of 'analytic'.3 Ayer, Language, Truth and Logic, p. 73. 4 Kant believed arithmetic is synthetic a priori while Mill believed that we arrive at mathematical beliefs on the basis of scientific induction.See Ayer, Language, Truth and Logic, p. 74 -75; 77 -78. 5 Ayer, Language, Truth and Logic, p. 82.Issue IV F Spring 2017 Is Geometry Analytic in virtue of the definitions of the terms in them and are thus "devoid of factual content".
What follows is a summary of the argument that Ayer makes in his discussion of geometry.Suppose, as Kant believed in his Critique of Pure Reason, that Euclidean geometry is synthetic a priori.On the one hand, Euclidean geometry is synthetic because one could not, by conceptual analysis alone, arrive at the truths of Euclidean geometry.Moreover, since geometry is synthetic, unlike analytic judgments (which, according to Kant, do not amplify [or increase] our knowledge), geometry does indeed increase our knowledge.On the other hand, Euclidean geometry is a priori, for it is grounded in our a priori idea of space, which, for Kant, was the pure form of sensible intuition.The idea of space itself is a priori (in the transcendental sense) insofar as it is only if we have that idea, due to the unfathomable constitution of our mind, is any experience possible.But, since Kant's Critique of Pure Reason, other consistent (and practically useful) non-Euclidean geometries6 have been developed.Thus, it cannot be the case that Euclidean geometry is synthetic a priori in Kant's sense or that its a priori status is due to the constitution of our mind."We see now," Ayer says, "that the axioms of a geometry are simply definitions, and that the theorems of a geometry are simply the logical consequences of these definitions.A geometry is not in itself about physical space; in itself it cannot be said to be "about" anything.But we can use a geometry to reason about physical space." 7st-forward about 30 years later to Hilary Putnam.In a chapter entitled "Analytic and Synthetic," in his book Mind, Language and Reality, Putnam, (somewhat) following Quine, 8 argues that principles of geometry are not analytic if the paradigmatic analytic sentence is 'all bachelors are unmarried'.Apart from the separate reasons, of which I shall not speak here, that Putnam gives for the analyticity of 'all bachelors are unmarried,' there are other reasons that he uses to argue that principles of geometry are not analytic.First of all, he rejects the "linguistic convention" account of analyticity that some philosophers were using to argue that the principles of geometry, like physical definitions (e.g., e=1/2 mv^2), are true by (linguistic) convention or stipulation, and hence analytic. 9Putnam argues that these definitions introduced by stipulation lose their conventional character and acquire systematic import within our conceptual system in such a way that it would be a mistake to construe them as analytic if analyticity is understood to mean 'true by linguistic convention or stipulation.' 10 Secondly (and this is where Putnam's acceptance Duhemian and Quinean holism 11 is very evident), mathematics and principles of geometry are characterized by their centrality, as framework principles, within our web of beliefs -revisable but only after holistic considerations.Thus, principles of Euclidean geometry, once thought to be analytic (in the sense of 'immune from revision'), were abandoned because a rival theory was available. 12at is intriguing, especially in light of Ayer's argument, is Putnam's conclusion that Euclidean geometry is false (my emphasis): If the paradigm for an analytic sentence is "all bachelors are unmarried"and it is -then it is of course absurd to say that the principles of geometry are analytic.Indeed we cannot any longer say that the principles of geometry are analytic; because analytic sentences are true; and we no longer say that principles of Euclidean geometry are true. 13 this paper, I will attempt to answer the question: is geometry analytic and in what sense?In doing so, I will begin by critically evaluating Ayer and Putnam's arguments on the analyticity (or lack thereof ) of geometry on the basis of historico-philosophical work on the foundations of geometry by Roberto Torretti in Philosophy of Geometry from Riemann to Poincaré and Michael Friedman in Reconsidering Logical Positivism.My critical evaluations of Ayer and Putnam will show that in their arguments against Kant and Linguistic Conventionalism respectively, they fail to distinguish clearly between what Einstein called "pure axiomatic geometry" and "practical geometry." 14n the one hand, I will show that Ayer fails to notice that Kant could have been talking about 'pure geometry,' and not 'applied geometry,' when Kant argued that geometry is synthetic a priori.On the other hand, I will show that Putnam fails to distinguish between applied Euclidean geometry and pure Euclidean geometry in the quoted passage.After my critical evaluations of Ayer and Putnam's arguments, I will conclude by suggesting how someone could plausibly think that applied Euclidean geometry is analytic in Carnap's sense.I will be drawing from Friedman

PURE GEOMETRY AND APPLIED GEOMETRY
On the one hand, what I choose to call 'pure geometry' is geometry studied as a branch of pure mathematics.Specifically, I take pure geometry to include classical Euclidean geometry as say: that which was pursued in Classical Antiquity and through the Middle Ages in Euclid's Elements; the analytical coordinate geometry invented by René Descartes in the 17th century, the non-Euclidean geometries 16 developed independently by Bolyai and Lobachevsky in the 1820s; Gauss's intrinsic geometry of surfaces in his Disquisitions of 1827, which, together with the earlier non-Euclidean geometries of Bolyai and Lobachevsky influenced Riemann in his 1854 lecture: On the Hypotheses that Lie at the Foundations of Geometry to come up with a generalized conception of space as a n-fold extended quantity i.e. an n-dimensional differentiable manifold -a conception which could accommodate both the Euclidean and non-Euclidean geometries.I would also include under pure geometry, the work of Felix Klein and Sophus Lie on transformation groups; projective geometry, which was axiomatized by Moritz Pasch in 1882; and the axiomatic conception of Euclidean geometry and both Euclidean and non-Euclidean geometry found in David Hilbert's 1899 Foundations of Geometry and 1902 article "On The Foundations of Geometry" respectively. 17Topology can also be classed within the study of pure geometry.Pure geometry is a priori in both senses of the term: "known independent of experience" and "[demonstrated] from the grounds." 18But as we shall subsequently see in this paper, saying why pure geometry is a priori generates a lot of controversy.
On the other hand, what I call 'applied geometry' is geometry in the literal sense of the term, i.e., "earth measuring." 19Applied geometry arises whenever the formal terms in pure geometry receive a physical interpretation i.e. when their designata are specified for use in the exact sciences like physics.In classical mechanics, for example, Euclidean geometry was used in kinematics by "building bridge equations" 20 from pure Euclidean geometry to physics.These 'bridge equations' are such as those that grew out of Descartes's analytic geometry, where he introduced coordinate systems and algebra to geometry (e.g., the equation of a straight line [ax + by + c = 0]), which found fruitful application in analysis and subsequently in kinematics.I think that 16 Cf.Roberto, Torretti, Philosophy of Geometry from Riemann to Poincaré (Dordrecht-Holland; Boston: D. Reidel Pub., 1978) Chapter 1 and 2 and 3. See Torretti, Philosophy of Geometry from Riemann to Poincaré, p. 185 for Hilbert's works.17 Torretti also points out the work of Levi-Cività in the geometric meaning of curvature and that of Weyl with the idea of an affine structure.So, we may include their contributions as contributions to pure geometry.18 The "from the grounds" sense of the a priori comes from the theory of demonstration.physical geometry in classical mechanics was still a priori, for I do not think that 'straight line' had received a physical interpretation in optics.However, in General Relativity, the apriority of physical geometry gets lost as physical geometry becomes entangled with the physical structure of the universe, especially with the distribution of mass and energy in the universe.For in General Relativity, the metric of the underlying topology of space-time depends on the distribution of mass-energy across the universe. 21Specifically, if a geodesic (which is the non-Euclidean equivalent of the straight line in Euclidean geometry) in a 4-dimensional semi-Riemannian manifold of non-constant curvature is interpreted as the path of an unimpeded beam of light, then the physical structure of the universe is such that two unimpeded parallel beams of light can converge globally (if not locally or infinitesimally) as they go past a star.22

CRITICAL EVALUATION OF AYER
Armed with this distinction between pure geometry and applied geometry, I would now like to engage in critical evaluation of Ayer's argument in the fourth chapter of Language Truth and Logic.One way of critically evaluating Ayer's argument is to ask whether the argument works in refuting Kant's epistemology of geometry as Ayer intended.With respect to this aspect of critical evaluation, the first point I want to make is that Ayer is correct in saying that "a geometry is not in itself about physical space:23 in itself it cannot be said to be "about" anything."He is correct insofar as he is talking about pure geometry as I have distinguished it above; however, I do not believe that this particular argument succeeds in refuting Kant's thesis that geometry is synthetic a priori.
First, for Kant, the 'synthetic' had a primary sense and a secondary sense.The primary sense of 'synthetic' is well known.It is roughly the idea that a judgment is synthetic whenever the predicate 'B' of a judgment is not contained (covertly) in the subject 'A' of the judgment.So, one can think of 'A' without necessarily thinking of 'B.' Or, put another way, mere conceptual analysis does not reveal that the predicate 'B' was contained in subject 'A' all along.The secondary sense of the synthetic is that while analytic judgments are explicative, which means that they reveal the concepts already contained in the subject albeit confusedly or less clearly; synthetic judgments are ampliative, which means that synthetic judgments extend our knowledge. 24condly, the 'a priori' also had two distinct senses for Kant.Firstly, whatever was a priori was both necessary and had universal validity.Secondly, whatever was a priori was so in a transcendental sense, meaning that it arose out of the unfathomable constitution of our mind insofar as it made experience possible.The implication here being that whatever makes experience possible cannot itself be known through experience. 25 the basis of these clarifications, I want to argue that Ayer's argument that seeks to refute Kant's epistemology of geometry in terms of geometry not being "about" anything, does not succeed.To see why, recall that for Ayer a proposition has factual content if and only if it "provides information about matters of fact." 26Elsewhere, he says that propositions that have factual content are empirical hypotheses. 27So, when Ayer says (emphasis mine), "[It] is natural for us to think, as Kant thought, that geometry is the study of the properties of physical space, and consequently that its propositions have factual content," I am at a loss as to why.There is no evidence in Kant's Critique to suggest that Kant's view of geometry was such that the propositions of geometry expressed empirical hypotheses.In fact, Kant is very clear: Geometrical propositions are one and all apodeitic…such propositions cannot be empirical or, in other words, judgments of experience, nor can they be derived from any such judgments. 28 may be objected that in the context of the quoted passage Kant says that, "geometry is a science which determines the properties of space synthetically, and yet a priori."This objection may be put forward to suggest that the term 'science' and the clause 'determines the properties of space synthetically' together imply that, for Kant, geometry is empirical and that it is about something (that it has factual content and that this content is physical space).Couldn't this, after all, be what Ayer is objecting to?My response is yes: Ayer is objecting to this view.However, Ayer's objection does not work for the following reasons.First, the quotation from Kant does not establish that the propositions of geometry are empirical hypotheses.That geometry is a 'science' could just mean that it is a systematically organized body of knowledge, which is just what a science means.Pure mathematics is also sometimes viewed as a science in the sense of being a systematically organized body of knowledge.Secondly, the quoted passage does not imply that the space in question is actual physical space.
In fact, that the passage does not imply physical space is strongly suggested by the fact that part of what Kant has in mind throughout the Transcendental Aesthetic is arriving at our idea of space.This idea of space -as a form of our sensible intuition -is a priori in the transcendental sense.So, if geometry is about this a priori idea, then it cannot be the case that in the quoted passage the space in question is physical space.Lastly, 'synthetically,' for Kant, does not mean that synthetic judgments are known by experience; for, after all, there are synthetic a priori judgments.The right way of interpreting 'synthetically' is in the primary and secondary senses of synthetic judgments that I have described above. 29at the foregoing aspect of critical evaluation shows is that one cannot explain away Ayer's overthrow of Kant's epistemology of geometry by saying that Ayer is talking about pure geometry while Kant is talking about applied geometry.The passages from Kant I have quoted above strongly suggest that Kant possibly had pure geometry in mind.The way that Ayer's argument does succeed is in challenging the grounds of the apriority of geometry.While Kant grounds the apriority of geometrical propositions on the a priori idea of space that arises as a result of the constitution of our mind, Ayer believes that a geometry is a priori because it is analytic (though not in Kant's sense of the term).Moreover, since Kant's time, there have been other consistent theories of pure geometry other than the classical Euclidean one that was the only one known to him.For within these geometrical theories, the theorems are logical consequences of the axioms and so these theories are all a priori because they are analytic.It is on this basis that Ayer is able to endorse the conventionalism of Poincaré in the quoted passage: roughly, that the idea that the choice of a pure geometry to be applied in physical theory is a matter of convention based on expediency and the overall fruitfulness and simplicity of working with the said geometry. 30e conventionalism of Poincaré will be better appreciated in the context of critical evaluation of Putnam's essay.In that essay, Putnam targets the linguistic conventionalism.Although some logical positivists (especially Moritz Schlick) were inspired by Poincaré's conventionalism, the linguistic conventionalism advocated by Carnap is unique and has important differences from the conventionalism of Poincaré.

CRITICAL EVALUATION OF PUTNAM 4.1 WHY WAS EUCLIDEAN GEOMETRY ABANDONED?
I begin my critical evaluation of Putnam's essay by looking at where Putnam says that the laws of geometry were abandoned because there was a rival theory. 31Here we make use of our distinction between pure geometry and applied geometry.While it is true to say that when Euclidean geometry is applied to our physical space, it turns out to be incorrect; it does not follow that pure Euclidean geometry itself is false and that it was abandoned.Viewed historico-philosophically, what actually happened is that geometers doubted that Postulate 5 was self-evident, and hence they doubted that it ought to be included with the rest of Euclid's axioms.Torretti points out that mathematicians like Proclus, John Wallis, Girolamo Saccheri, John Heinrich Lambert and Adrien Legendre made attempts to prove Postulate 5 from the other axioms. 32Although Gauss thought that there were no mathematical reasons for preferring Euclidean geometry to the non-Euclidean one,33 it was in the 1820s that Bolyai and Lobachevsky, working independently of each other, developed non-Euclidean geometry, which was constructed by denying postulate 5 and using the rest of Euclid's axioms that do not depend on Postulate 5. 34 Lobachevsky indeed did not view his system as contrary to Euclidean geometry.He viewed both systems as equally consistent. 35Then Riemann, by building on Gauss's work on the intrinsic geometry of surfaces, produced a generalized geometry capable of accommodating both the Euclidean and non-Euclidean geometries -Euclidean space is a special case of the genus of the manifold. 36reover, where Putnam says, "before the development of non-Euclidean geometry by Riemann and Lobachevski, the best philosophic minds regarded the principles of geometry as virtually analytic.The human mind could not conceive their falsity."Putnam's remarks need to be corrected in light of what actually happened historicophilosophically.If, by the best philosophic minds he counts Kant, then Kant did not regard the principles of geometry as analytic.For Kant, principles of geometry were synthetic a priori.It is important to distinguish analyticity from apriority.Analyticity is one way of explaining the apriority of mathematics and geometry.In fact, given my discussion in section 3 above, Kant's epistemology of geometry is compatible with the existence of non-Euclidean geometries.Friedman points out that the only difference between Poincaré and Kant is that the former was familiar with non-Euclidean geometry, while the latter was not. 37

THE RATIONALE OF THE ANALYTIC-SYNTHETIC DISTINCTION
A second way we may critically evaluate Putnam's argument is with the preliminary remarks leading up to his discussion of the analyticity (or lack thereof ) of principles of geometry.One such remark he makes is intended to defend the Quinean insight that he thinks is underappreciated by the philosophers who undertake to challenge Quine's views.Putnam thinks that citing garden-variety examples of analyticity will not do as a reply to Quine.Instead, he insists that what is needed is a definition of the nature and rationale of the analytic-synthetic distinction: "what point is there to having a separate class of statements called analytic statements?" 38 So thus, we may begin our critical evaluation of Putnam's view here, by responding to this question historico-philosophically.
On the one hand, I believe that the nature and rationale of the analytic-synthetic distinction lie in the fact that the logical positivists in general, and Carnap in particular, wanted to respond to the Kantian problem: how is pure mathematics possible? 39In doing so, they also wanted to avoid the Kantian synthetic a priori doctrine of the transcendental aesthetic. 40Since Putnam directs some of his criticisms of the analyticity of geometry towards Reichenbach, 41 whose conventionalism, he says, grew out of the Viennese circle, seeing what Reichenbach actually thought at one time in his career will serve us well as we seek to understand the nature and rationale of the analytic-synthetic distinction.

THE RELATIVIZED A PRIORI
As Friedman points out, Reichenbach had a unique conception of the relationship between the a priori and empirical science that was neither strictly Kantian nor radically empiricist. 42Friedman argues in his 1920 book, General Relativity and A Priori Knowledge, that Reichenbach distinguished between axioms of coordination and axioms of connection. 43On the one hand, the axioms of coordination preserved part of the Kantian sense of the a priori, namely that they make science, through the physical theory, possible.He held that whatever the axioms of coordination were, they were a priori relative to the scientific theory that was employing them.Friedman notes that, for Reichenbach, "these nonempirical axioms of coordination -which include, paradigmatically, the principles of physical geometry -are 'constitutive of the object of knowledge.'" 44Axioms of connection, on the other hand, are scientific inductive generalizations.What is important to note, according to Friedman, is that for Reichenbach, the axioms of coordination, relative to a given scientific theory (Newtonian Mechanics, Special Relativity, or General Relativity), are a priori and, as such, are not subject to any empirical confirmation or disconfirmation.But whichever axioms of coordination are in fact used depends on the scientific theory.So, the axioms of coordination could, in principle, be revised as a result of scientific advances, e.g., physical Euclidean geometry is a priori in classical mechanics but is not a priori in General Relativity, where "topology (sufficient to admit a Riemannian structure)" is instead a priori. 45This is the relativized a priori -the idea that relative to a scientific theory, a geometry, for example, is a priori.relativized a priori was lost. 46Later, as Friedman points out, Reichenbach devoted his writing to reconciling post-general relativity science to the conventionalism of Schlick and Poincaré. 47hlick's conventionalism grew out of that of Poincaré and the work of David Hilbert.From the former, Schlick was persuaded that the question of which of the geometries is to be applied to space is a matter to be settled by experience and requirements of overall simplicity of our scientific conceptual system.Applying the Helmholtz-Lie theorem, which states that based on the experience (or idealization in the case of Poincaré) that rigid motion is possible in our space, our space must be either Euclidean or characterized by one of the manifolds of constant curvature, 48 Poincaré thought that the choice between the non-Euclidean and Euclidean geometries was conventional based on which was the most expedient to work with. 49From Hilbert, Schlick got the idea of 'implicit definitions', namely that the axioms of a geometry implicitly define the geometry's primitive terms.Different geometries differ in so far as they employ different implicit definitions of 'point,' 'line,' 'between,' and so on.
As Friedman points out, the conventionalism of Poincaré and Schlick is in no way different from Duhemian holism, which is the idea that the theoretical components of our conceptual system face the tribunal of experience not in isolation, but as a whole. 50So, Schlick's conventionalism does not do justice to the principles of geometry and differs from the linguistic conventionalism of Carnap that I shall now proceed to explain.

LINGUISTIC CONVENTIONALISM
According to Friedman, where Reichenbach had failed to give a clear explication of the difference between the axioms of coordination and the axioms of connection, Carnap, in the Logical Syntax of Language, had the machinery to do so: L-rules 51 (the analytic) and P-rules (the synthetic) of the physical language of science.The choice of L-rules and the interpretation that made them true is a matter of convention, for these rules are purely formal and one has a lot of leeway in selecting the L-rules in the formulation of a language.In Classical Mechanics, for example, the L-rules include the principles of Euclidean geometry, while the P-rules are the general principles and laws of physics.In General Relativity, the L-rules include a topology sufficient to admit the 4-dimensional semi-Riemannian manifold of non-constant curvature, while the P-rules now include applied geometry through optics and the other general laws of physics.What is important to note for Carnap, says Friedman, is that the L-rules (or analytic sentences) are distinguished from the P-rules at a given stage in the development of the scientific enterprise. 52That they are revisable in light of the development of the scientific enterprise is exemplified by applied or physical geometry in the move from classical mechanics to general relativity.However, Carnap can still say this and can also hold on to the idea that in a physical language there is a distinction between analytic sentences and synthetic sentences.For it is the analytic sentences, axioms of coordination (in the case of Reichenbach) and conventions (in the case of Poincaré and Schlick), that are constitutive of any endeavor to gain any objective scientific knowledge. 53In other words, only if we are armed with the axioms of coordination or L-rules can we begin to generate hypotheses and test them. 54hus, it is useful to have the distinction between the analytic and the synthetic even in science.
On the other hand, I would argue that holists, like Putnam, who oppose the distinction in science, have a hard time responding to the Kantian question of "how is pure mathematics possible?"That is, they have a hard time explaining why mathematics and logic seem to be necessarily true, and therefore a priori.While they would say that mathematics and logic are framework principles which are to be identified with centrality and priority within our conceptual scheme, such that if revisions become necessary due to advances in science they are the last to be considered for revision, 55 I would still argue that the holists would be hard pressed to explain away our belief that there is a marked difference between mathematics, on the one hand, and physical laws, on the other.The former cannot be otherwise, while the latter can.
theory was analytic (in the sense of stipulated true by convention) while the principles of geometry in General Relativity were synthetic. 58Putnam says that both Hume and Reichenbach are wrong: The principle that light travels in straight lines is a law of optics, nothing more or less serious than that.We test the conjunction of geometry and optics indeed, and if we get into trouble, then we can alter either the geometry or the optics, depending on the nature of the trouble. 59 is possible that Reichenbach was correct in Putnam's footnoted discussion. 60For in a very enlightening discussion on the empiricism with respect to geometry of Ernst Mach, Torretti points out that there are alternative physical interpretations of the straight line.In fact, interpreting straight line as the path of a light ray may not be the best approximation of a straight line.In this discussion from Mach on how planes are constructed in practice: Physically a plane is constructed by rubbing three bodies together until three surfaces A, B, C, are obtained, each of which exactly fits the othersa result which can be accomplished […] with neither convex nor concave surfaces, but with plane surfaces only. 61rretti observes, "If you construct two adjacent planes by this method, their common edge will provide a better approximation of the straight line than any taut string or light ray." 62

CONCLUSION
What this observation by Torretti on the passage from Mach suggests to me is that we have a choice of what physical interpretation to give our pure geometrical concepts.Therefore, Reichenbach could have been right in saying that the optical theory is analytic, hence a priori relative to pre-general relativity physics.Recall that, in presenting Carnap's view, I said that Friedman argues that, for Carnap, the choice of L-rules and the interpretation that made them true is a matter of convention. 63Now when we stipulate the interpretation of a straight line as the path of a ray of light, the stipulation is a matter of convention, since there are alternatives that are equally good physical interpretations.Now, given that in the discussion of kinetic energy, Putnam seems to grant that conventions are valid only if there are alternatives; 64 and, given that there are alternative interpretations of a straight line, it follows that statements in the optical theory that interpret geometrical concepts are valid conventions, and are

's Reconsidering Logical Positivism in my presentation of Carnap's sense of 'analytic.' Friedman argues in that book that Carnap can accept Duhemian holism while
I think 'pure geometry' is a better term than 'pure axiomatic geometry.'Pure axiomatic geometry implies that all the pure geometries are approached axiomatically which is not necessarily the case.Also, 'applied geometry' sounds more accurate than 'practical geometry'; for it suggests that what we are talking about is pure geometry as it is applied to the study of physical space or in physics, e.g., in optics.In this paper, I may sometimes use the terms 'physical geometry' and 'applied geometry' interchangeably.15 Cf.Friedman, Reconsidering Logical Positivism, p. 70 In Peter Koellner's seminar class: Topics in the Philosophy of Mathematics: The Concept of Apriority (Philosophy 243, Fall 2015), for which I wrote this paper, Koellner had shown that 'a priori' did not always mean what it means today, i.e., "known independent of experience".'A priori [demonstration]' was used earlier by William of Ockham (ca.13 Century) when he distinguishes: (1) Demonstration from what is prior, i.e., why it is so from explanatory grounds; from (2) Demonstration from what is posterior, i.e., that it is so.He cited Ockham's Summa Logicae Part III Tractate II Chapter 17.This sense of 'a priori', Peter argued in seminar, can serve to illuminate certain accounts of justifications in mathematics.19 See Einstein, Sidelights on Relativity, p. 31 20 This is a term that Peter Koellner explicitly in his presentation during this seminar class.Since I was submitting this paper for that class, I wanted to acknowledge that it is not my own.