THE DUHEM-QUINE PROBLEM PIERRE DUHEM: PHYSICAL THEORY AND EXPERIMENT I. Biographical A. lived in France, 1861-1916 B. noted for his work in theoretical physics, especially thermodynamics, as well as in the history and philosophy of science C. believed that a working scientist needs to understand the history of his subject D. also sought to defend a philosophy of science that made room for religious faith E. the selection in your anthology comes from a book first published in French in 1906 II. An Experiment in Physics Can Never Condemn an Isolated Hypothesis but Only a Whole Theoretical Group (59) A. A physicist who does an experiment assumes the truth of a whole group of theories B. Distinguishes two sorts of experiments: 1. experiments of application a. does not aim at discovering anything but only to draw on what one already knows b. example: measuring the electromotive force of a bunch of generators (59) 2. experiments of testing a. these are the ones he's interested in, as these are the ones through which science develops b. involve deriving a prediction of what should happen under certain experimental conditions (59-60) C. Example: Wiener's attempt to refute Neumann's hypothesis that in a ray of polarized light, the vibration is parallel to the plane of polarization (60) 1. Wiener deduced that a light beam reflected at 45 degrees allowed to interfere with incident polarized beam will produce an interference pattern 2. Since this pattern did not appear, concluded Neumann's hypothesis must be wrong D. in Duhem's opinion, this method is modeled on reduction to absurdity in mathematics, with experimental refutation playing the role of logical contradiction here E. however, the experimental method is not as rigorous as mathematical proof 1. in order to derive some prediction from an hypothesis and to interpret his experimental results, the scientist must make use of a whole group of theories 2. negative experimental results tell him merely that there is a mistake somewhere in this theory; they do not say where it is (60) F. to return to Wiener's experiment 1. in testing Neumann's hypothesis, he also had to assume: (61) a. that light is a wave b. that vibrations are normal to direction of light ray c. that the mean kinetic energy of the vibration measures light intensity d. that the photographic exposure indicates degree of intensity e. etc. 2. but, as Poincare has shown, we can rescue Neumann's theory by rejecting the assumption that mean kinetic energy measures intensity and substitute instead mean potential energy of the medium G. Another example: Foucault's experiment 1. designed to test Newton's particle theory of light a. which assumes, among other things, that the medium attracts the particles of light b. explains reflection and refraction c. implies 1.) that the index of refraction equals the velocity of light in the medium it just entered divided by that in the medium it just left 2.) in other words, that light travels faster in a denser medium (water/vacuum = 1.33) 2. Foucault, however, showed that light travels less rapidly in water than in air (62) a. but this shows only that the whole "system" or theory of emission is incompatible with the facts b. we don't know whether it is the hypothesis that light consists of particles that is false or whether it's the assumption that the medium acts on the light particles H. in sum, 1. an experiment does not test an isolated hypothesis 2. when the experiment shows his predictions are wrong, it does not show which hypothesis or assumption needs to be changed 3. physics is more like an organism than a machine a. the physicist is not like the watchmaker who can take apart the watch and isolate the defective piece b. he's more like the doctor who cannot dissect the patient but must guess the cause of the illness from what's affecting the whole body (62) III. A "Crucial Experiment" Is Impossible in Physics (63) A. The idea of a crucial experiment is based on the analogy between eliminative proofs in mathematics and experimental refutations in science 1. if we could enumerate all the possible hypotheses that explain some group of phenomena and experimentally eliminate all but one, that one would be shown to be certain 2. when there are only two alternative hypotheses, which predict different outcomes under the same experimental conditions, that is called an instance of the cross road, or a crucial experiment B. for example, Foucault's experiment could be considered a crucial experiment for deciding between the wave and the particle theories of light, as one implies that light travels faster in air than in water, the other that it travels faster in water C. of course, Duhem has already argued that experiments test whole systems of theories and not isolated hypotheses (64) D. but even if we could refute an isolated hypothesis, a crucial experiment is still not possible (64) 1. in mathematics there is no room for a third alternative: either X or not-X is true 2. but in physics, we can never be sure that we have enumerated all the possibilities 3. for instance, Foucault's experiment was designed to decide between the emission theory and the theory that light is a vibration of the medium. a. But Maxwell suggested that light is a vibration in a "dielectric medium" b. And of course since Duhem's day there have been further theories W. V. QUINE HOLISM, PART 1: TWO DOGMAS OF EMPIRICISM I. Introduction A. this paper is actually just the conclusion to a much longer paper B. in the longer paper, he argues against two "dogmas" 1. the distinction between analytic and synthetic statement -- what counts as a synonym depends on theory 2. that the meaning of a statement can be reduced to experience C. but what we are interested in here in his conclusion is the way in which he draws much more radical conclusions from the problem of the underdetermination of theory choice by experiment II. Empiricism without the dogmas (72) A. For Quine, the totality of human knowledge, including everything from math and logic to science to geography and history, is a fabric that impinges on experience only at the boundary 1. when it conflicts with experience, we have to make some changes in the interior 2. because things are logically connected, changes made in one place will entail the need to make changes elsewhere B. however, there is much freedom of choice as to which statements we will change: no particular statement is linked to any particular experience (q.v.) 1. if this is right, it is misleading to speak of the empirical content of any particular statement, especially as we get farther from the periphery of the fabric 2. note: Duhem does not draw such radical conclusions C. there is no difference between analytic and synthetic statements: (72) 1. any statement can be held true come what may 2. no statement is immune from revision 3. we could even decide to revise the logical law of the excluded middle, as we have done in quantum mechanics (72-3) 4. how is that any different than what has gone on in any other scientific revolution? (73, q.v.) D. to be sure, he has been speaking in terms of distance from the sensory periphery 1. but this reflects nothing more than the likelihood of choosing one statement rather than another to revise in the light of experience 2. for instance, we could imagine experiences that would lead us to revise the statement that there are brick houses on Elm Street a. but we could revise other statements instead b. that we choose to revise this statement reflects our natural conservatism E. for Quine, the conceptual scheme of science is just a tool for making predictions F. physical objects are merely posits within such schemes 1. epistemologically, they differ only in degree, not in kind, from the gods of Homer (73) (explain term) 2. the myth of physical objects has simply proven to be more effective than other myths in dealing with experience (q.v.) G. in addition to macroscopic physical objects, we posit: 1. atomic and subatomic entities 2. forces (74) 3. mathematical objects, such as classes H. metaphysical or ontological questions such as whether classes exist are on a par with scientific questions 1. such questions do not concern matters of fact 2. rather, they are about choosing the most convenient language or conceptual scheme for doing science I. although questions like whether classes exist seem more like questions about convenient conceptual schemes and questions like whether centaurs exist seem more like questions of fact, they differ only in degree J. conservatism and simplicity figure in the choices we make about what to change in the light of experience 1. other philosophers have taken a pragmatic line when it comes to choosing conceptual frameworks, but their pragmatism leaves off at the boundary between the analytic and synthetic 2. Quine is a more thoroughgoing pragmatist: there is no such boundary (74) HOLISM, PART 2: POSITS AND REALITY I. In this selection, Quine explores in a little more detail A. the question about how we choose conceptual schemes in science B. and the implications of his answer II. Subvisible particles (76) A. Quine's desk seems solid, but science tells him it is a "swarm of vibrating molecules" B. But the evidence scientists have for such a claim is only indirect 1. that is, all different sorts of indirect evidence, if we assume the molecular theory, can be brought under a set of laws 2. these laws continue to lead to successful new predictions C. According to Quine, there are thus five benefits that are credited to the molecular theory: 1. Simplicity: by bringing disparate phenomena under a unitary theory 2. familiarity: already familiar laws of motion are applied (76-7, q.v.) 3. scope: a unified theory has a wider range of testable consequences (77) 4. fecundity: theory can be extend to new applications 5. tests have turned out well so far D. Simplicity, however, is hard to define, and scientists may not always agree which is the simpler theory E. Not all theories have all these benefits: for instance, quantum and relativity lack familiarity F. And none of these benefits show that the theory is true 1. all these benefits concern the way the theory bears on observable reality 2. imagining the theory as a fabric again, if we cut out the center and preserve the ring, it still has all the same benefits 3. could the center perhaps even be false? III. Posits and Analogies (78) A. to say that the desk is just a swarm of vibrating molecules is just to use an analogy with things like bugs 1. the question of fact begins to dissolve into one of how good the analogy is 2. and analogies in science can be very limited 3. these analogies are at best aids to constructing theories (78) B. thus, molecules lose even their status of inferred or hypothetical entities for Quine: statements about molecules are simply useful devices for organizing statements about what we can observe (78-9, q.v.) C. unfortunately, however, it seems we can make a similar argument about things like desks: 1. they are merely useful posits for organizing our sensory experience 2. the only thing real is raw sense data 3. if we extend this to other people, we land in solipsism IV. Restitution (79) A. surely something has gone wrong 1. not only is the conclusion bizarre 2. but we cannot really say that we posit physical objects to explain past and present sense data, since we remember only physical bodies, not sense data B. from the fact that we have no evidence for the existence of physical bodies beyond that it helps us to organize our experience, perhaps what we should have concluded is that that's just what evidence is (80, q.v.) 1. the same applies to molecules and other posits in physics: the benefits that the theory provides are the best evidence we can ask for (80) 2. to say that something is a posit does not mean that it cannot be real 3. sense data, after all, are just posits, too, in psychology 4. things like molecules, physical bodies, and sense data are all part of our conceptual scheme V. Working from Within (80) A. Statements about bodies, whether common-sense or "recondite," make no empirical sense except within this conceptual scheme (81) 1. Any statement can be replaced by its negation, provided we make suitable adjustments elsewhere 2. Claims about observable bodies are underdetermined in the same way as claims in science B. but better than bemoaning this underdetermination, we should try to improve our physics by seeking to increase the sorts of benefits discussed earlier (82) VI. My comments: A. yes, but what about when people do not agree about the relative benefits of competing theories? B. I'll try to illustrate the problem through two historical examples: 1. the Copernican revolution 2. the Darwinian revolution