is failure an option? contingency and refutation

Stud. Hist. Phil. Sci. 39 (2008) 242–252

Contents lists available at ScienceDirect

Stud. Hist. Phil. Sci.

journal homepage: www.elsevier .com/ locate /shpsa

Is failure an option? Contingency and refutation

Allan FranklinDepartment of Physics, University of Colorado, CB 390, Boulder, CO 80309, USA

a r t i c l e i n f o

Keywords:
ContingencyRefutation17-keV neutrinoParity nonconservation
0039-3681/$ - see front matter � 2008 Elsevier Ltd. Adoi:10.1016/j.shpsa.2008.03.016

E-mail address: [email protected].

a b s t r a c t

In this paper I argue, using two case studies of episodes from recent physics (the discovery of parity non-conservation in the weak interaction and the decision that the proposed 17-keV neutrino did not exist)against the contingency view advocated by social constructionists. In this view, physics, or science in gen-eral, is, in Ian Hacking’s words, not determined by anything. Much of the previous discussion has centeredon examples of scientific success. In this paper I argue that experimental evidence and reasoned and crit-ical discussion played the crucial role in the refutation of a previously strongly believed hypothesis, andin the decision that a proposed new elementary particle did not exist, leaving no reasonable doubt. I sug-gest that the argument against contingency does not require the absence of all possible doubt, but ratherthe absence of reasonable doubt.

� 2008 Elsevier Ltd. All rights reserved.

When citing this paper, please use the full journal title Studies in History and Philosophy of Science

1. Introduction

Recently Ian Hacking (1999, Ch. 3) has provided an incisive andinteresting discussion of the issues that divide the constructionists(Harry Collins, Andrew Pickering), who believe that experimentalevidence plays a minimal role in the production of scientificknowledge, from the rationalists (perhaps objectivists would be abetter term), like myself, who believe that it is crucial. He setsout three sticking points between the two views: 1) contingency,2) nominalism, and 3) external explanations of stability. In this pa-per I will discuss only contingency, the idea that science is not pre-determined, that it could have developed in any one of severalsuccessful ways. This is the view adopted by constructionists.Hacking illustrates this with Pickering’s account of high-energyphysics during the 1970s during which the quark model came todominate (see Pickering, 1984).

The constructionist maintains a contingency thesis. In the case ofphysics, (a) physics (theoretical, experimental, material) couldhave developed in, for example, a nonquarky way, and, by thedetailed standards that would have evolved with this alterna-tive physics, could have been as successful as recent physicshas been by its detailed standards. Moreover, (b) there is no

ll rights reserved.

sense in which this imagined physics would be equivalent topresent physics. The physicist denies that. (Hacking, 1999, pp.78–79)To sum up Pickering’s doctrine: there could have been aresearch program as successful (‘progressive’) as that of high-energy physics in the 1970s, but with different theories, phe-nomenology, schematic descriptions of apparatus, and appara-tus, and with a different, and progressive, series of robust fitsbetween these ingredients. Moreover—and this is somethingbadly in need of clarification—the ‘different’ physics wouldnot have been equivalent to present physics. Not logicallyincompatible with, just different.The constructionist about (the idea) of quarks thus claims thatthe upshot of this process of accommodation and resistance isnot fully predetermined. Laboratory work requires that we geta robust fit between apparatus, beliefs about the apparatus,interpretations and analyses of data, and theories. Before arobust fit has been achieved, it is not determined what that fit willbe. Not determined by how the world is, not determined by tech-nology now in existence, not determined by the social practices ofscientists, not determined by interests or networks, not deter-mined by genius, not determined by anything. (Ibid., pp. 72–73;myemphasis).

mailto:[email protected]

http://www.sciencedirect.com/science/journal/00393681

http://www.elsevier.com/locate/shpsa

A. Franklin / Stud. Hist. Phil. Sci. 39 (2008) 242–252 243

One of the pervasive features of the constructionist argument istheir lack of belief in the efficacy of Nature, as revealed by experi-ment, to decide scientific issues. An extreme constructionist viewhas been given by Barry Barnes. He states, ‘Reality will toleratealternative descriptions without protest. We may say what we willof it, and it will not disagree. Sociologists of knowledge rightly re-ject epistemologies that empower reality’ (Barnes, 1991, p. 331).More extensively, Donald MacKenzie remarks.

Recent sociology of science, following sympathetic tendenciesin the history and philosophy of science, has shown that noexperiment, or set of experiments however large, can on itsown compel resolution of a point of controversy, or, more gen-erally acceptance of a particular fact. A sufficiently determinedcritic can always find reason to dispute any alleged ‘result.’ Ifthe point at issue is, say, the validity of a particular theoreticalclaim, those who wish to contest an experimental proof or dis-proof of the claim can always point to the multitude of auxiliaryhypotheses (for example about the operation of instruments)involved in drawing deductions from a given theoretical state-ment to a particular experimental situation or situations. Oneof these auxiliary hypotheses may be faulty, critics can argue,rather than the theoretical claim apparently being tested. Fur-ther the validity of the experimental procedure can also beattacked in many ways. (MacKenzie, 1989, p. 489)

Much depends here on what Hacking and MacKenzie mean by‘determined’ or ‘compel’. If they mean entailed then I agree. I doubtthat the world, or more properly, what we can learn of it, entails aunique theory. If not, as seems more plausible, they mean that theway the world is places no constraints on successful science, then Idisagree strongly. I would certainly wish to argue that the way theworld is constrains the kinds of theories that will fit the phenom-ena, the kinds of apparatus we can build, and the results we can ob-tain with such apparatuses. To think otherwise seems silly. I doubtwhether Kepler would have gotten very far had he suggested thatthe planets move in square orbits.

At the other extreme are the ‘inevitablists’, among whom Hack-ing classifies most scientists. He cites Sheldon Glashow, a NobelPrize winner in physics, ‘Any intelligent alien anywhere wouldhave come upon the same logical system as we have to explainthe structure of protons and the nature of supernovae’ (Glashow,1992, p. 28; cited in Hacking, 2000, p. 66). On a scale from 1 to5, where a score of 5 is a strong constructionist position and 1 isa strong objectivist position, Hacking and I both rate ourselves as2 on contingency.

One reason for leaning toward inevitability and away from con-tingency is the independent and simultaneous suggestion of theo-ries or hypotheses. This includes the suggestion of the V–A theoryof weak interactions by E. C. George Sudarshan and Robert Mar-shak and by Richard Feynman and Murray Gell-Mann; the proposalof quantum electrodynamics by Feynman, by Julian Schwinger, andby Sin-Itiro Tomonaga; the proposal of quantum mechanics by Er-win Schrödinger and by Werner Heisenberg; the suggestion of thetwo-component neutrino by Tsung-Dao Lee and Chen NingYang,by Lev Landau, and by Abdus Salam; and the independent sugges-tion of quarks by George Zweig and by Gell-Mann. There arenumerous other instances.

There are several reasons that make these simultaneous sugges-tions plausible. There are many factors which influence theory for-mation which, whereas they don’t entail a particular theory, doplace strong constraints on it. The most important of these is, I be-lieve, Nature. Pace Barry Barnes, it is not true that any theory can

1 In this case that suggestion failed and a new theory, namely, Einstein’s General Theor2 Collins (2004) makes this clear.

be proposed and that valid experimental results or observationswill be in agreement with it. Regardless of theory, objects denserthan air fall toward the center of the earth when dropped. A secondimportant factor is the existing state of experimental and theoret-ical knowledge. Scientists read the same literature. They also buildon what is already known, or at least on what is accepted as scien-tific knowledge, and often use hypotheses similar to those thathave previously proven successful. Thus, when faced with theanomalous advance of the perihelion of Mercury, some scientistsproposed that there was another planet in the solar system thatcaused the effect, a suggestion similar to the successful hypothesisof Neptune previously used to explain discrepancies in the orbit ofUranus.1 What are considered important problems at a given timewill also influence the future course of science. Thus, the need to ex-plain b decay and the problem of determining the correct mathemat-ical form of the weak interaction led to the extensive work, boththeoretical and experimental, that led to the simultaneous sugges-tion of the V–A theory of weak interactions. The mathematics avail-able also constrains the types of theories that can be offered.

The philosophical argument underlying the constructionist po-sition is known as the Duhem–Quine problem. In the modus tollensif a hypothesis h entails evidence e then the failure to observe e or:e (not e) entails :h. As both Duhem and Quine pointed out, it isreally h and b that entail e, where b is background knowledge orauxiliary hypotheses, and thus :e entails :h or :b, and one doesn’tknow where to place the ‘not’. They noted that one can maintainone’s belief in h if one is willing to make sufficient changes in one’sknowledge base.2 One might also suggest that among all the logi-cally possible alternatives to the hypothesis there is one that can alsoexplain the result.

Logically Duhem and Quine are correct, but in scientific practiceone is never confronted with the infinite number of possible alter-natives. There are usually only a finite number of physically inter-esting and plausible alternatives on offer. Nor are all modificationsto one’s knowledge base palatable. Sometimes the price one has topay is too high. Scientists typically start with what is accepted asscientific knowledge at a given time and some of the alternativesmay involve giving up well confirmed theories or well supportedexperimental results. An example of this occurred after the 1964observation of the decay of the Ko

L meson into two pions by Fitch,Cronin and their collaborators (Christenson et al., 1964). The com-bined particle–antiparticle and space–reflection (parity) symmetry(CP) allowed the decay of the Ko

S meson into two pions, but thesame decay for the Ko

L meson was forbidden. The most obviousinterpretation of the result was that CP symmetry was violated,and this was the interpretation given by most physicists, but nofewer than ten alternatives were offered. (For details see Franklin,1986, Ch. 3.) These included: 1) the local asymmetry between mat-ter and antimatter; 2) external fields; 3) the decay of a Ko

L mesoninto a Ko

S meson, with the subsequent decay of the KoS into two

pions; 4) the emission of a ‘paritino’, a particle that carries onlyspacetime properties, in the decay of the Ko

L meson; 5) a pion withangular momentum one, rather than zero, as believed by physi-cists; 6) another neutral particle produced coherently with the Ko

L

mesons; 7) a ‘shadow universe’, which interacts with our universeonly through the weak interactions, and that what was observedwas the decay of the Ko

L from the ‘shadow universe’; 8) the failureof the exponential decay law in the decay of the Ko

S meson; 9) thefailure of the superposition principle in quantum mechanics,needed for the theoretical derivation of the conclusion that Ko

L de-cay into two pions was forbidden; and 10) the possibility thatpions were not bosons. Eventually all of the suggested alternative

y of Relativity, was required to explain the observations.

244 A. Franklin / Stud. Hist. Phil. Sci. 39 (2008) 242–252

explanations were experimentally tested and rejected.3 This iswhat one might call a pragmatic solution to the Duhem–Quine prob-lem. Although not all of the logically possible alternative explana-tions were tested, a substantial number were tested. I note thatthe physics community had a rather liberal definition of plausiblealternative. Consider the shadow universe explanation offered byNishijima and Saffouri. Not only was it considered, it was both test-able and tested, and found wanting (see ibid., p. 96).

Most of the discussions have concentrated on examples of sci-entific success.4 There has been far less discussion of episodes inwhich a hypothesis or an experimental result has been shown tobe incorrect. There are also important differences between the twogroups on this issue. Harry Collins (1985), for example, has pre-sented such a case in his discussion of the early searches for gravitywaves. In that episode Joseph Weber (1969; Weber et al., 1973) re-ported positive results. Several other groups attempted to replicateWeber’s findings. None were successful and Weber’s results cameto be regarded by the physics community as incorrect. The reasonsoffered by different scientists for their rejection of Weber’s claimswere varied, and not all of the scientists engaged in the pursuitagreed about their importance. During the period 1972–1975 itwas discovered that Weber had made several serious errors in hisanalysis of his data. His computer program for analyzing the datacontained an error that generated spurious positive signals. Weberadmitted the error, although he claimed that it did not account forall of the signals observed. He did not, however, ever publish cor-rected results. In addition, his statistical analysis of residual peaksand background was questioned and thought to be inadequate. Inparticular, Weber was accused of varying his threshold for accep-tance of a signal to maximize his result. In his recent book, Gravity’sshadow (2004), Collins presents evidence that this criticism was jus-tified. He cites an unnamed scientist as follows:

Joe would come into the laboratory—he’d twist all the knobsuntil he finally got the signal. And then he would analyze thedata: he would define what he would call a threshold. Andhe’d try different values for the thresholds. He would have algo-rithms for a signal—maybe you square the amplitude, maybeyou multiply the things . . . he would have twelve different waysof creating something. And then thresholding it twenty differ-ent ways. And then go over the same dataset. And in the end,out of these thousands of combinations there would be a peakthat would appear, and he would say, ‘Aha—we’ve found some-thing.’ And [someone] knowing statistics from nuclear physicswould say, ‘Joe—this is not a Gaussian process—this is not nor-mal—when you say there’s a three-standard-deviation effect,that’s not right, because you’ve gone through the data so manytimes.’ And Joe would say, ‘But—What do you mean? When Iwas working to find a radar signal in the Second World War,anything was legal, we could try any trick so long as we couldgrab a signal.’ And [someone] said, ‘Yeah Joe, but somebodywas sending a signal.’ And Joe never understood that. (Collins,2004, pp. 394–395)

Weber also claimed to find coincidences between his detector andanother distant detector when, in fact, the tapes used to providethe coincidences were recorded more than four hours apart. Weberhad found a positive result where even he would not expect one.(Weber admitted the error and later stated that he had neverclaimed that this was strong evidence in favor of gravity waves.)Others cited the failure of Weber’s signal to noise ratio to improve,

3 Even before this was done Prentki, in a review talk, argued for CP violation by noting thbecause ‘the price one has to pay in order to save CP becomes extremely high’ and becausehere referring to the last three alternatives, which were well supported, crucial parts of q

4 See, for example, Franklin (Forthcoming).5 There was considerable cooperation among the early gravity wave researchers. They s

despite his ‘improvements’ to his apparatus. In addition, the side-real correlation, the observation of a signal when the gravity waveantenna pointed toward the galactic center, disappeared. This wasjust as well because Weber had found a twenty-four-hour periodic-ity, whereas a twelve-hour effect was expected.

There were other problems with Weber’s results. Other experi-mental groups calibrated their apparatuses by injecting a known,acoustic energy signal and demonstrating that the signal couldbe detected by their apparatus. Weber’s experiment failed this test.This was due to a difference in analysis procedures between Weberand his critics. Weber used a nonlinear analysis algorithm, whichwas sensitive only to changes in signal amplitude. His critics useda linear algorithm, which was sensitive to changes in both ampli-tude and phase. To be fair, Weber admitted that his critics’ analysisprocedures were twenty times better than his at detecting the cal-ibration pulses. He argued, however, that the calibration pulseswere short and that the real gravity wave signal pulses were longerand that his analysis procedure was better at detecting such pulses.If that were the case then a signal should have appeared when We-ber’s critics analyzed their data with Weber’s analysis program.5 Itdid not.

Perhaps most important were the uniformly negative resultsobtained by six other experimental groups. Collins points out thatonly one of these experimental apparatuses was not criticized byother groups, and that all of these experiments were regarded asinadequate by Weber. He states

Under these circumstances it is not obvious how the credibilityof the high flux case [Weber’s results] fell so low. In fact, it wasnot the single uncriticized experiment that was decisive . . .

Obviously the sheer weight of negative opinion was a factor,but given the tractability, as it were, of all the negative evi-dence, it did not have to add up so decisively. There was away of assembling the evidence, noting the flaws in each grain,such that outright rejection of the high flux claim was not thenecessary inference. (Collins, 1985, p. 91)

In Collins’s view neither experimental evidence nor critical discus-sion can resolve this, or any other such issue. I disagree. As dis-cussed in detail in Franklin (1994) this controversy was decidedon the basis of valid experimental evidence and on critical and rea-soned discussion.

In the next section I will discuss two episodes: the discovery ofparity nonconservation, the violation of space-reflection or mirrorsymmetry, and the how physicists came to reject the existence ofthe 17-keV neutrino, a proposed new elementary particle. Thesediscussions will support my view that evidence and critical discus-sion can be crucial in the refutation of hypotheses or in the rejec-tion of experimental results.

2. Case studies

2.1. The discovery of parity nonconservation

In 1956 Tsung Dao Lee and Chen Ning Yang (1956), who wouldwin the Nobel Prize for their suggestion, proposed that parity, ormirror-reflection symmetry or left–right symmetry, is not con-served in the weak interaction. (Physicists usually speak of fourinteractions. In decreasing order of strength they are: the strong,or nuclear, interaction, which holds the atomic nucleus together;the electromagnetic interaction, which holds atoms together; the

at although most of the alternatives had been ruled out by experimental observationsthe remaining alternatives ‘were even more unpleasant’ (Prentki, 1965). Prentki was

uantum mechanics.

hared both data tapes and analysis programs.

Fig. 1. Spin and momentum in real space and mirror space. As one can see in realspace the spin and momentum point in opposite directions. In mirror space theypoint in the same direction. This is an example of parity nonconservation.

Fig. 2. Relative counting rates for b particles from the decay of oriented 60Co nucleifor different nuclear orientations (field directions). More electrons are emitted o-pposite to the nuclear orientation than in the same direction. This demonstratesparity nonconservation (from Wu et al., 1957, p. 1414).


weak interaction, responsible for radioactive decay; and the gravi-tational interaction.)

Parity conservation was a well established and strongly be-lieved principle of physics. As students of introductory physicslearn, if we wish to determine the magnetic force between twocurrents we first determine the direction of the magnetic fielddue to the first current using a right-hand rule, and then determinethe force exerted on the second current by that field using a secondright-hand rule. We would get exactly the same answer, however,if we were to use two left-hand rules. This is left–right symmetry,or parity conservation, in electromagnetism and in classicalphysics.

Eugene Wigner (1927) proposed the concept of parity conserva-tion in quantum mechanics as a way of explaining some then re-cent results in atomic spectra. It quickly became an establishedprinciple. As Hans Frauenfelder and Ernest Henley stated, ‘Sinceinvariance under space reflection is so appealing (why should aleft- and right-handed system be different?), conservation of parityquickly became a sacred cow’ (Frauenfelder & Henley, 1975, p.359). An early indication of this came in 1933 when Wolfgang Paulirejected a theory proposed by Herman Weyl on the grounds that itwas not invariant under reflection, or because it did not conserveparity. ‘However, as the derivation shows, these wave equationsare not invariant under reflections (interchanging left and right)and thus are not applicable to physical reality’ (Pauli, 1933, p. 226).

In the early 1950s physicists were faced with a problem knownas the ‘s–h’ puzzle. Based on one set of criteria, that of mass andlifetime, two elementary particles (the s and the h) appeared tobe the same, whereas on another set of criteria, that of spin andintrinsic parity, they appeared to be different, a very unusual situ-ation in physics. (A good analogy to spin is the rotation of the earthon its axis. Intrinsic parity refers to the mirror symmetry propertiesof the wavefunction, a mathematical function used to describe theparticle.) In 1956 Lee and Yang realized that the problem would besolved, and that the two particles would just be different decaymodes of the same particle, if parity were not conserved in the de-cay of the s and h particles, a weak interaction. They examined theevidence for parity conservation and found, to their surprise, thatalthough there was strong evidence that parity was conserved inthe strong (nuclear) interaction and in the electromagnetic interac-tion, there was, in fact, no supporting evidence that it was con-served in the weak interaction. It had never been tested. It hadjust been assumed.

In general, the physics community, including many leadingphysicists, did not believe that the Lee and Yang suggestion wascorrect. Pauli skeptically remarked, ‘I do not believe the Lord is aweak left-hander, and I am ready to bet a very large sum thatthe experiments will give symmetric results’ (Bernstein, 1967, p.59). There were other bets between physicists. Richard Feynman,one of the leading theoretical physicists of the twentieth centuryand a Nobel Prize winner, bet Norman Ramsey, another winner,$50–$1 that parity would be conserved. Ramsey noted that Feyn-man believed that the real odds were one thousand to one, but saidthat he wouldn’t bet that much on anything. Felix Bloch, yet an-other Nobel Prize winner, offered to bet his hat with any othermember of the Stanford physics department that parity would beconserved.

Lee & Yang (1956) suggested several possible experimentaltests of parity conservation in the weak interaction. The first ofthe two most important tests was the b decay of oriented nuclei.(b decay is the transformation of one atomic nucleus into a differ-ent nucleus, with the emission of an electron and a neutrino.) Ori-ented nuclei are nuclei whose spins all point in the same direction.The second was the sequential decay p ? l ? e. This is the decayof a pi meson, an elementary particle, into a mu meson, anotherparticle, and a neutrino. The mu meson subsequently decays into

an electron and two neutrinos. These were the first experimentsdone and provided the crucial evidence for the physics community.

We can understand this if we look at Fig. 1. Suppose a radioac-tive nucleus, whose spin points upward, always emits an electronin the direction opposite to the spin. In the mirror the spin is re-versed, whereas the electron’s direction of motion is unchanged.Now the electron is emitted in the same direction as the spin.The mirror result is different from the real result. This violates mir-ror symmetry and shows the nonconservation of parity. Parity con-servation would also be violated if in a collection of oriented nucleimore electrons were emitted in the direction of the nuclear spinthan opposite to the spin, or vice versa. Only if we had a collectionof oriented nuclei and if equal numbers were emitted symmetri-cally with respect to the spin direction would parity be conserved.For p ? l ? e decay, parity nonconservation implies that the muon(the l) will be longitudinally polarized, which means that its spinwill point either parallel to or antiparallel to its direction of mo-tion. If the muon is stopped its polarization remains and its subse-quent decay will look just like that of an oriented nucleus. In thiscase we would look for an asymmetry in the distribution of themuon decay electrons emitted along the direction of the muon mo-tion and opposite to that direction.

The first experiment was performed by Chien-Shiung Wu andher collaborators (Wu et al., 1957). It consisted of a layer of ori-ented 60Co nuclei and a single, fixed, electron counter, which waslocated either along the direction of, or opposite to, the orientationof the nuclei. The direction of the orientation of the nuclei could bechanged and any difference in counting rate in the fixed electroncounter observed. Their results are shown in Fig. 2. With thecounter opposite to the nuclear orientation the ratio of the counts

Fig. 3. Relative counting rate as a function of precession field current (from Garwinet al., 1957, p. 1416).


observed when the nuclei were oriented to when they were notwas 1.20. With the counter parallel to the orientation the ratiowas 0.80. If parity were conserved the ratio would have been 1.(In statistical terms this was a 13 standard-deviation effect. Thismeant that it was extremely unlikely that the observed effectwas due to a statistical fluctuation in the number of counts). Thiswas a clear asymmetry. The experimenters concluded, ‘If an asym-metry between h and 180�—h (where h is the angle between theorientation of the parent nuclei and the momentum of the elec-trons) is observed, it provides unequivocal proof that parity isnot conserved in b decay. This asymmetry has been observed inthe case of oriented 60Co0 (ibid., p. 1413).

The second experiment, on the sequential decay p ? l ? e, wasperformed using two different experimental techniques by RichardGarwin, Leon Lederman, and Marcel Weinrich (1957) and by Jer-ome Friedman and Valentine Telegdi (1957). The Garwin experi-ment found a sinusoidal variation in counting rate, in contrast tothe symmetric distribution expected if parity were conserved(Fig. 3). Their statistically overwhelming result (22 standard devi-ations) led them to conclude that parity was not conserved. Moreinformally, Lederman called Lee at 7 am and announced ‘Parity isdead’. Friedman and Telegdi performed the same experiment usinga different technique. They found a forward-backward asymmetryof 0.091 ± 0.021 (a 4 standard-deviation effect). Both groups alsoconcluded that parity was not conserved.

The immediate reaction of the physics community was that par-ity nonconservation in the weak interaction had been clearly dem-onstrated. It is fair to say that as soon as any physicist saw theexperimental results they believed that parity wasn’t conserved.Even Pauli was convinced. He wrote:

Now, after the first shock is over, I begin to collect myself. Yes, itwas very dramatic. On Monday, the twenty-first, at 8 p.m. I wasto give a lecture on the neutrino theory. At 5 p.m. I received thethree experimental papers. I am shocked not so much by thefact that the Lord prefers the left hand, as by the fact that Hestill appears to be left–right symmetric when He expressesHimself strongly. In short, the actual problem now seems tobe the question: Why are strong interactions right and left sym-metric? (Bernstein, 1967, p. 60)

Pauli was fortunate that he had not wagered a very large sum ofmoney that parity would be conserved. Feynman paid Ramsey

6 I am referring here to merely tacking on an irrelevant statement such as ‘The moon is mcould exclude the possibility that at the same time as the experiments were being perforopposite effects. Such an alternative cannot, of course, be excluded, but suggest that it isscientists.

7 For details see Franklin (1995).

and Bloch remarked that it was lucky he didn’t own a hat. The NobelPrize in physics was awarded to Lee and Yang in 1957, less than ayear after their suggestion that parity was not conserved in theweak interaction.

This is, perhaps, the clearest example of a crucial experiment,one that decides unequivocally between two theories or, in thiscase, between two classes of theory, those that conserve parityand those that do not, in the history of physics. The evidencewas beyond a reasonable doubt. There were three different exper-iments with different types of experimental apparatus and involv-ing two different processes, the b decay of oriented nuclei andp ? l ? e decay. The statistical evidence was overwhelming.Friedman and Telegdi found a 4 standard-deviation effect, Wuand collaborators a 13 standard-deviation effect, and Garwin andcolleagues a 22 standard-deviation effect. (The probability of a 10standard-deviation effect is 1.5 � 10�23.) As my former student,Mark Corske remarked, ‘4 standard deviations is strong evidence,13 standard deviations is absolute truth, and 22 standard devia-tions is the Word of God’ (personal communication).

In addition, each of the experiments had been carefully cali-brated and checks were performed to ensure that the apparatusworked properly and that sources of background that might maskor mimic the effect to be observed were either absent or small (fordetails see Franklin, 1986, Chs. 6 and 7).

As far as the physics community was concerned, experimenthad shown that parity is not conserved in the weak interaction.The hypothesis had been refuted. No attempts to provide an alter-native explanation of the results appeared in the physics literature.I note here the importance of the fact that these experimentsdecided between classes of theories. There are, in fact, only twoclasses of theories, those that conserve parity and those that donot. These classes are both mutually exclusive and exhaustive. Itis this generality that allows a solution to a weak form of the Du-hem–Quine problem, one in which one accepts the experimentalresult, :e. Strong arguments were also provided for the correctnessof those experimental results.

I also note that it was quite difficult invent an alternative theoryof weak interactions that would both explain all of the previousexperimental results, that would also not conserve parity, andwhich was not a trivial extension of the V–A theory that wouldultimately explain parity nonconservation.6

The reader may have noted that the argument about classes oftheories would also apply to the episode of CP violation, in whichmany alternatives were offered. After all, there are only two classesof theories, those that conserve CP and those that do not. There is,however, a significant difference between the two episodes of par-ity and CP nonconservation. In the case of parity nonconservationone could see by inspection that the experiment violated parityconservation. In the case of CP violation what was observed wasthe decay of the K0

L meson into two pions. In order to relate thisto CP violation one has to assume the principle of superpositionin quantum mechanics, that pions were bosons, and that the expo-nential decay law held. It was these assumptions, along with oth-ers, that were questioned.

2.2. The appearance and disappearance of the 17-keV neutrino7

In this section I will discuss how the physics community dealtwith the claimed existence of a new elementary particle, the 17-keV neutrino and ultimately decided that such a particle did not

ade of green cheese’ to the V–A theory. A really determined might have asked how onemed on earth, similar experiments were being done on a distant planet, which gavenot plausible. As a Bayesian I would give it a prior probability of zero, as would all

Fig. 4. The dotted lines give the Kurie plots for both a heavy neutrino (lower) andthe ordinary light neutrino (upper). The heavy-neutrino line has a smaller energyintercept because the mass of the heavy neutrino reduces the maximum allowedenergy for the decay electron. (In this example the mass of the heavy neutrino istwo energy units.) The solid line is the sum of the two plots. A kink, a change inslope, is clearly at the maximum energy minus the mass of the heavy neutrino.

Fig. 5. Simpson’s results for DK/K (the fractional change in the Kurie plot) as afunction of the kinetic energy of the b particles. Eth is the threshold energy, thedifference between the endpoint energy and the mass of the heavy neutrino. A kinkis clearly seen at Eth = 1.5 keV, or at a mass of 17.1 keV (from Simpson, 1985, p.1893).

Fig. 6. The ratio of the measured 35S electron energy spectrum to the theoreticalspectrum. A three percent mixing of a 17-keV neutrino should distort the spectrumas indicated by the dashed curve (from Ohi et al., 1985, p. 324).


exist. A matter of fact was decided. This episode will include in-stances of discordant experimental results, alternative explana-tions of phenomena, and questions concerning selectivity in dataanalysis procedures.

2.2.1. The appearanceThe 17-keV neutrino was ‘discovered’ by John Simpson (1985).

He had searched for a heavy neutrino by looking for a kink in the b-decay energy spectrum of tritium, or in the Kurie plot, at an energyequal to the maximum allowed decay energy minus the mass ofthe heavy neutrino, in energy units. (The Kurie plot is a particularmathematical function of b-decay quantities plotted as a functionof the decay electron energy. It has the nice visual property thatthe Kurie plot for the correct theory is a straight line.) We mayunderstand this by looking at Fig. 4. The dotted lines give simu-lated Kurie plots for both b decay with a heavy neutrino (lowerline) and with the ordinary light neutrino (upper line). The hea-vy-neutrino line has a smaller energy intercept because the massof the heavy neutrino reduces the maximum allowed energy forthe decay electron. (In this example the mass of the heavy neutrinois two energy units.) If, as Simpson suspected, both types of neu-trino are emitted in b decay then the Kurie plot for the decay willbe the sum of the Kurie plots for each of the neutrinos (solid line).As one can see, there is a distinct kink, a change in the slope of theline, at an energy equal to the maximum allowed energy minus themass of the neutrino. Simpson, and others, expected that the heavyneutrino emission would be only a small fraction of that of the or-dinary neutrino. This would make the kink even more difficult toobserve than in the case illustrated in the figure.

Simpson actually examined the fractional change in the Kurieplot, DK/K, which is more sensitive to the presence of a second neu-trino. Simpson’s initial experimental result for the decay energyspectrum of tritium is shown in Fig. 5. A kink, a marked changein slope of the DK/K graph, is clearly seen at an electron energy,Tb, of 1.5 keV, corresponding to a 17-keV neutrino. (The maximumdecay energy for tritium is 18.6 keV. If there were no effect due tothe presence of a heavy neutrino this graph would be a horizontalstraight line.) ‘In summary, the b spectrum of tritium recorded inthe present experiment is consistent with the emission of a heavyneutrino of mass about 17.1 keV and a mixing probability [the frac-tion of heavy neutrinos] of about 3%’ (ibid., p. 1893). Later work,including some by Simpson, reduced the size of the positive effectto approximately 1%.

Within a year there were five attempted replications of Simp-son’s experiment (Altzitzoglou et al., 1985; Apalikov et al., 1985;Datar et al., 1985; Markey & Boehm, 1985; Ohi et al., 1985). Eachof them was negative. The experiments set limits of less than 1%for a 17-keV neutrino branch of the decay, in contrast to Simpson’svalue of 3%. A typical result, that of T. Ohi and others, is shown in

Fig. 6 and should be compared to Simpson’s result shown in Fig. 5.No kink of any kind is apparent.

Each of the subsequent experiments had examined the b-decayspectrum of 35S rather than that of tritium (3H1). Three of theexperiments used magnetic spectrometers. The other two used so-lid-state detectors, the same type used by Simpson. In the lattertwo cases, however, the source was not implanted in the detec-tor, as Simpson had done, but was separated from it. Such anarrangement would change the atomic physics corrections to the

Fig. 7. DK/K for the 35S spectra of Ohi and others as recalculated by Simpson. A kinkis clearly visible (from Simpson, 1986a, p. 114).


spectrum. (In order to compare the measured spectrum to the pre-dicted spectrum the experimenters needed to make atomic physicscorrections because of the interaction of the decay electrons withthe atomic electrons.) The 35S b-decay sources used in the experi-ments had a higher endpoint energy than did the tritium used bySimpson, which made the atomic physics corrections to the b-de-cay spectrum less important.

Simpson’s first report of the 17-keV neutrino was unexpected. Itwas not predicted, or even suggested, by any existing theory. Facedwith such an unexpected result the physics community took sev-eral approaches. The experiment was repeated, and alternativetheoretical explanations were offered. Some theoretical physiciststried to explain the result within the context of accepted theory.They argued that a plausible alternative explanation of the resulthad not been considered. This involved the question of whetheror not the theory used in the analysis of the data, and to comparethe experimental result with the theory of the phenomenon, wascorrect. This is an important point. An experimental result is notusually immediately given by an examination of the raw data,but requires considerable analysis. In this case the analysis in-cluded atomic physics corrections, needed for the comparison be-tween the theoretical spectrum and the experimental data.Everyone involved agreed that such corrections had to be made.The question was what were the proper corrections. The atomicphysics corrections used by Simpson in his analysis were ques-tioned by others. Several calculations indicated, at least qualita-tively, that Simpson’s result could be accommodated withinaccepted theory, and that there was no need for a new particle.‘A detailed account of the decay energy and Coulomb-screening ef-fects raises the theoretical curve in precisely this energy range sothat little, if any, of the excess remains’ (Lindhard & Hansen,1986, p. 965).

The combination of negative experimental searches combinedwith plausible theoretical explanations of Simpson’s result had achilling effect on the field. Although Simpson’s claim had been se-verely challenged, not everyone agreed that it had been refuted.Simpson, and others, continued working on the question. Simpson(1986a, b) presented further evidence in support of the 17-keVneutrino using a somewhat modified experimental apparatus. Healso took the criticism of his work seriously and presented an anal-ysis of his new data which incorporated the atomic physics correc-tions suggested by his critics. Although this reduced the size of hiseffect by approximately 20%, the effect was still clearly present.Simpson had shown that his result was robust under variationsin the atomic physics corrections to the decay spectrum. He alsoquestioned whether the analysis procedures used in the five nega-tive searches were adequate to set the upper limits they had re-ported. He argued that the wide energy range used by his criticsto fit the b-decay spectrum tended to minimize any possible effectof a heavy neutrino, which would appear primarily in a narrowenergy band near the maximum energy allowed for the heavyneutrino. Simpson further questioned whether or not the ‘shape-correction’ factor needed to fit the spectra in magnetic spectrome-ter experiments could mask a kink due the presence of a heavyneutrino. It is, in general, quite difficult to calculate the efficiencyfor magnetic spectrometer apparatuses. In order to get a good fitto the observed spectrum a ‘shape-correction’ factor is used. Thesequestions would have to be answered.

Simpson (1986a) also presented a reanalysis of Ohi’s data usinghis own preferred analysis procedure. He found a positive effect(Fig. 7). (Compare this with Ohi’s reported result shown inFig. 6.) The reader might well ask how the same data could bothsupport and refute the existence of the 17-keV neutrino? How thisconflict was resolved is discussed below.

Further negative evidence was provided by other experiments.Hetherington and others, for example, urged caution concerning

Simpson’s method of data analysis. They pointed out that ‘concen-trating on too narrow a region [of energy] can lead to misinterpre-tation of a local statistical anomaly as a more general trend’(Hetherington et al., 1987, p. 1512). The issue of the appropriateenergy range for the analysis of the data had not yet been decided.At the end of 1988 the situation was much as it was at the end of1985. Simpson had presented positive results on a 17-keV neu-trino. There were nine negative experimental reports as well asplausible theoretical explanations of his result. There was nostrong evidence for the existence of the 17-keV neutrino.

In 1989 Simpson and Andrew Hime presented two additionalpositive results, using both tritium and 35S, the spectrum used inthe original negative searches. In these reports the value of themixing fraction of 17-keV neutrinos had been reduced to approxi-mately 1% (Hime & Simpson, 1989; Simpson & Hime, 1989).

The situation changed dramatically in 1991. New positive re-sults were reported by groups at Oxford (Hime & Jelley, 1991)and at Berkeley (Sur et al., 1991). These results were quite persua-sive. Hime and Nicholas Jelley had improved the experimentalapparatus to eliminate the problem of the backscattering of elec-trons that had been found in earlier experiments. The Berkeleygroup took precautions to ensure that the decay electrons depos-ited their full energy in their detector. If the full energy wasn’tdeposited this would distort the decay energy spectrum and simu-late a kink that signaled the presence of a heavy neutrino. TheBerkeley group claimed that their result ‘supports the claim bySimpson that there is a 17-keV neutrino emitted with �1% proba-bility in b decay’ (ibid., p. 2447). These new results generated con-siderable new experimental and theoretical work. SheldonGlashow, a Nobel-Prize winning theorist, remarked, ‘Simpson’sextraordinary finding proves that Nature’s bag of tricks is notempty, and demonstrates the virtue of consulting her, not herprophets’ (Glashow, 1991, p. 257). Glashow was advocating theprimacy of experimental evidence over theoretical speculationand calculation. Unfortunately, as we shall see below, the positiveexperimental evidence was incorrect.

2.2.2. The disappearanceThe summer of 1991 marked the high point in the life of the 17-

keV neutrino. From this time forward only negative results wouldbe reported, and errors would be found in the most persuasive po-sitive results. Leo Piilonen and Alexander Abashian (1992) sug-gested that Hime and Jelley had overlooked a background effectthat might have simulated the effect of a 17-keV neutrino in theirexperiment. The appearance of several negative results (discussedbelow) encouraged Hime to consider the Piilonen–Abashian sug-gestion seriously and to reanalyze his own result. He found, usingan experimentally checked Monte Carlo calculation, that he could

Fig. 8. Residuals (the ratio of the measured result to the expected result minus one)from a fit to the 35S data assuming no massive neutrino. The solid curve representsthe residuals expected for decay with a 17-keV neutrino with a mixing probabilityof 0.85%. No evidence of a 17-keV-neutrino is visible (from Mortara et al., 1993, p.396).

Fig. 9. Residuals from fitting the beta spectrum of a mixed source of 14C and 35Swith a pure 35S shape. The solid curve indicates residuals expected including theknown 14C contamination. It is a very good fit to the data (from Mortara et al., 1993,p. 397).


explain his own Oxford result (with Jelley) without the need for a17-keV neutrino. ‘It will be shown that scattering effects are suffi-cient to describe the Oxford b-decay measurements and that themodel can be verified using existing calibration data’ (Hime,1993, p. 166). He also suggested that similar effects might explainhis earlier positive results obtained in collaboration with Simpson.

Hime briefly reviewed the evidential situation, noting that themajor evidence against the existence of the 17-keV neutrino hadcome from magnetic spectrometer experiments in which questionshad been raised concerning the shape corrections. He commentedthat Giovanni Bonvicini (1993)8 had shown that experimental diffi-culties combined with limited statistics could mask the presence of aheavy neutrino signature and still be described by a smooth ‘shape-correction’ factor. Bonvicini’s work was very important. By showingthat a smooth ‘shape-correction’ factor might either mask or en-hance a kink due to a 17-keV neutrino he cast considerable doubton the early negative results obtained with magnetic spectrometers.This work was influential in persuading scientists to perform the la-ter, more stringent, experimental tests. Hime remarked that

A measurement of the 63Ni spectrum [by the Tokyo group,Kawakami et al. (1992)] has circumvented this difficulty. Thesufficiently narrow energy interval studied, and the very highstatistics accumulated in the region of interest, makes it veryunlikely that a 17-keV threshold has been missed in this exper-iment. (Hime, 1993, p. 165)

He also cited a new result from a group at Argonne National Labo-ratory, discussed in detail below, that provided ‘convincing evi-dence against a 17-keV neutrino’ (ibid.). These negative resultsprovided the impetus for Hime’s reexamination of his result.

Some of the evidence that Hime cited against the 17-keV neu-trino was provided by the Argonne group (Mortara et al., 1993).This experiment used a solid-state, detector (the same type usedoriginally by Simpson) and an external 35S source. Their final resultfor the mixing probability of the 17-keV neutrino, shown in Fig. 8,was �0.0004 ± 0.0008 (statistical uncertainty) ±0.0008 (systematicuncertainty). This result, less than 0.2%, was far lower than the 1%reported by Simpson and others. They had found no evidence for a17-keV neutrino.

The experimenters also demonstrated the sensitivity of theirapparatus to a possible 17-keV neutrino. They mixed a small com-

8 This paper had been published earlier as a CERN report, CERN-PPE/92-54.

ponent of 14C in their 35S source and detected the resulting kink intheir composite spectrum (The two spectra had different endpointenergies. Combining them would result in a composite spectrumwith a kink. See Fig. 9). They had shown that their experimentwould have detected the 17-keV neutrino had it been present. (Acriticism of some of the earlier negative experiments was that theyhad not demonstrated this).

This exercise demonstrates that our method is sensitive to a dis-tortion at the level of the positive experiments. Indeed, thesmoother distortion with the composite source is more difficultto detect than the discontinuity expected from the massive neu-trino. In conclusion, we have performed a solid-state countersearch for a 17 keV neutrino with an apparatus with demon-strated sensitivity. We find no evidence for a heavy neutrino,in serious conflict with some previous experiments. (Ibid., p.396; myemphasis)

At this time the Berkeley group began to question their own posi-tive result. Further experimental runs had shown that their appara-tus seemed to generate a spurious 17-keV neutrino signal. Theysearched for the cause of the artifact. In 1993 they found that, de-spite their precautions, not all of the electrons were depositing theirfull energy in their detector, simulating the presence of a 17-keVneutrino. Their earlier result was incorrect.

The newer negative results were persuasive not only because oftheir improved statistical accuracy, but also because they wereable to demonstrate that their experimental apparatuses could de-tect a kink in the spectrum if one were present. This was a directexperimental check that there were no effects present that wouldmask the presence of a heavy neutrino. The Tokyo group had alsoshown that the ‘shape-correction’ factors used in their experimentdid not mask any possible 17-keV neutrino (Kawakami et al., 1992;Ohshima, 1993; Ohshima et al., 1993). This combination of almostoverwhelming and persuasive evidence against the existence of a17-keV neutrino, combined with the demonstrated and admittedproblems with the positive results, decided the issue. There wasno 17-keV neutrino.

It seems clear that this decision was based on experimental evi-dence, discussion, and criticism or, in other words, epistemologicalcriteria. It had been shown that the two most persuasive positiveresults had overlooked effects that mimicked the presence of a17-keV neutrino. In addition the new negative results had

Fig. 10. Morrison’s reanalysis of Simpson’s reanalysis of the result of Ohi et al.(from Morrison, 1992).


answered the criticisms made previously concerning the ‘shape-correction’ factor and had demonstrated that they could detect akink in the spectrum if one were present. What then of the ques-tion of selectivity that Simpson had raised? The reader may stillbe asking why Simpson’s reanalysis of Ohi’s data and Ohi et al.’sown analysis gave such different results.

Selectivity was illustrated in this episode in the choice of theenergy range used to fit the decay energy spectrum, so that theexperimental and theoretical spectra could be compared. Simpsonhad argued that because 45% of the effect expected occurred within2 keV of the neutrino threshold a narrow energy range around thatthreshold should be used

in trying to fit a very large portion of the b spectrum, the dangerthat slowly-varying distortions of a few percent could bury athreshold effect seems to have been disregarded. One cannotemphasize too strongly how delicate is the analysis whensearching for a small branch of a heavy neutrino, and how sen-sitive the result may be to apparently innocuous assumptions.(Simpson, 1986b, p. 576)

D. Hetherington and others, however, suggested caution.

It has been argued [by Simpson] that in order to avoid system-atic errors, only a narrow portion of the beta spectrum shouldbe employed in looking for the threshold effect produced byheavy neutrino mixing. If one accepts this argument, our datain the narrow scan region set an upper limit of 0.44% [muchlower than the 3% effect originally found by Simpson]. However,we feel that concentrating on a narrow region and excluding therest of the data is not warranted provided adequate care istaken to account for systematic errors. The rest of the spectrumplays an essential role in pinning down other parameters suchas the endpoint. Furthermore, concentrating on too narrow aregion can lead to misinterpretation of a local statistical anom-aly as a more general trend which, if extrapolated outside theregion, would diverge rapidly from the actual data. (Hethering-ton et al., 1987, p. 1512)

Hime, one of Simpson’s collaborators, agreed.

The difficulty remains, however, that an analysis using such anarrow region could mistake statistical fluctuations as a physi-cal effect. The claim of positive effects in these cases [by Simp-son] should be taken lightly without a more rigorous treatmentof the data. (Hime, 1992, p. 1303)

This issue was dramatically demonstrated by Simpson’s reanalysisof the data of Ohi et al. Recall from Fig. 6 that Ohi’s result showedno evidence for a 17-keV neutrino. Simpson’s reanalysis of thatsame data shows clear positive evidence (Fig. 7). How can the samedata provide both positive and negative evidence for the same ef-fect? The answer is that the analysis procedures were quite differ-ent. Ohi and collaborators had used a wide energy range for theiranalysis. As Douglas Morrison showed, the positive effect foundby Simpson was due to his use of a narrow energy range for hisreanalysis of Ohi’s data.

The question then is, How could the apparently negative evi-dence of Fig. 6 become the positive evidence of Fig. 7? Theexplanation is given in Fig. 10, where a part of the spectrumnear 150 keV is enlarged. Dr. Simpson only considered theregion 150 keV, ±4 keV (or more exactly +4.1 and �4.9 keV).The procedure was to fit a straight line, shown solid, throughthe points in the 4 keV interval above 150 keV, and then tomake this the base-line by rotating it down through about 20�to make it horizontal. This had the effect of making the pointsin the interval 4 keV below 150 keV appear above the extrapo-lated dotted line. This, however, creates some problems, as it

appears that a small statistical fluctuation between 151 and154 keV is being used: the neighboring points between 154and 167, and below 145 keV, are being neglected although theyare many standard deviations away from the fitted line. [Simp-son’s straight-line fit to the data just above 150 keV and itsextrapolation is the line going from lower left to upper right.Comparing the data points to this line generates the positiveeffect seen in Fig. 10. The dotted curve above the data is theeffect expected for a heavy neutrino.] Furthermore, it is impor-tant, when analyzing any data, to make sure that the fittedcurve passes through the end-point of about 167 keV, whichit clearly does not (Morrison, 1992, p. 600).

The caution urged by both Hetherington and collaborators and byHime was justified.

How was this issue dealt with? Several later experiments usedboth a narrow and a wide energy range in the analysis of their data.For example, Hetherington et al. concluded that their results fromboth the wide and narrow-energy range analyses agreed and thatboth argued against a 17-keV neutrino.

Ultimately, the decision that the 17-keV neutrino did not existwas based on finding errors in the two most persuasive positive re-sults and by the overwhelming negative evidence provided byexperiments which avoided explicitly the data analysis issuesposed by narrow vs. wide energy range. The first of these experi-ments was that of a Tokyo group. These experimenters noted someof the problems that plague experiments using wide energy re-gions and decided therefore to concentrate ‘on performing a mea-surement of high statistical accuracy, in a narrow energy region,using very fine energy steps’ (Kawakami et al., 1992, p. 45). Theysaw no evidence of a 17-keV neutrino and set an upper limit forthe mixing probability of 0.073%. This was the most stringent limityet and was certainly far lower than the approximately 1% effectfound by Simpson and others.

Bonvicini noted that this experiment, with its very high statis-tics, had answered essentially all of his criticism of spectrometerexperiments convincingly. ‘Thus, I conclude that this experimentcould not possibly have missed the kink and obtain[ed] a good[fit] at the same time, in the case of an unlucky misfit of the shapefactor’ (Bonvicini, 1993, p. 115). The Tokyo group had shown thattheir result was robust against changes in the energy range usedin their analysis and was not sensitive to their use of a ‘shape-cor-rection’ factor.


The second experiment that provided decisive evidence againstthe existence of the 17-keV neutrino was that of a group at the Ar-gonne National Laboratory. Not only did this experiment find noevidence for a 17-keV neutrino, but the experimenters had demon-strated that they would have observed such evidence had it beenpresent. The fact that an effect of both the right size and shapewas observed when the spectra of 35S and 14C were mixed demon-strated that their analysis procedure used did not mask a real ef-fect. The failure to observe any effect in the 35S spectrum aloneshowed that the analysis procedure did not create an artifact thatwould mimic the effect of a heavy neutrino.

There was no 17-keV neutrino. The kink was dead. It had died ofnatural or, should we say, evidential, causes.

3. Conclusion

The two case studies presented in this paper have shown thatcontra MacKenzie ‘experiment . . . can on its own compel resolutionof a point of controversy, or, more generally acceptance of a partic-ular fact’ (MacKenzie, 1989, p. 489). In the case of the discovery ofparity nonconservation a strongly-believed symmetry principlewas overthrown. The 17-keV neutrino episode showed the com-plex history of how an initial experimental claim, supporting theexistence of a new elementary particle, was shown to be incorrect.In both cases the issue was decided by experimental evidence andreasoned and critical discussion. Such a conclusion can also bedrawn from the brief discussions of the early attempts to detectgravity waves and the discovery of the violation of CP symmetry.9

These were reasonable conclusions. Without offering a full theory ofrationality I would suggest that it is rational to strengthen one’s be-lief in a hypothesis or theory when one observes evidence entailedby that theory. It is also rational to accept an experimental resultwhen strong arguments are offered for its correctness. It is also ra-tional to reject a hypothesis when credible experimental evidencerefutes it and to reject an experimental result when it has beenshown that there were errors in the experiment or in its analysis,or when an overwhelming preponderance of evidence is offeredagainst it.

In the case of parity nonconservation the evidence was so over-whelming that no alternative explanations of weak interactionsinvolving neutrinos appeared in the physics literature. This, asdiscussed earlier, was clearly not the case when CP violation wasdiscovered. In that episode, however, all of the alternative explana-tions were eliminated by experiment within a few years. Would ithave been possible to offer similar alternatives in the case of par-ity? Of course, the answer is yes (see note 6), but one might askwhether it was reasonable to do so. Parity nonconservation notonly explained, indeed it predicted, the results of the experimentsof Wu and collaborators (1957), of Garwin, Lederman, & Weinrich(1957), and of Friedman & Telegdi (1957), but also solved the s–hpuzzle that had started the investigation. It also made other predic-tions that were soon confirmed.

Could a ‘determined critic’ have maintained that there was, infact, a 17-keV neutrino. Again, the answer is yes, but at what price?In order to do so the critic would have needed to present argu-ments that the carefully checked negative experimental results ofthe Argonne and Tokyo groups were wrong, and that the experi-menters who thought that their own positive results were wrongwere mistaken. No such critic appeared.10

I suggest that the argument against contingency does not re-quire the absence of any possible doubt, but rather the absence

9 For details see Franklin (1986, 1994).10 The physics community need not be unanimous. Weber maintained that he had detect

mass electron–positron states did not exist (for details see Franklin, 1998). By refusing tocommunity such individuals become marginalized.

of reasonable doubt. Failure is not only an option, sometimes itseems to be required.

References

Altzitzoglou, T., Calaprice, F., Dewey, M., Lowry, M., Piilonen, L., Brorsen, J., Hagen, S.,& Loeser, F. (1985). Experimental search for a heavy neutrino in the betaspectrum of 35S. Physical Review Letters, 55, 799–802.

Apalikov, A. M., Boris, S. D., Golutvin, A. I., Laptin, L. P., Lyubimov, V. A., Myasoedov,N. F., Nagovitsyn, V. V., Novikov, E. G., Nozik, V. Z., Soloshchenko, V. A.,Tikhomirov, I. N., & Tret’yakov, E. F. (1985). Search for heavy neutrinos in bdecay. JETP Letters, 42, 289–293.

Barnes, B. (1991). How not to do the Sociology of Knowledge. Annals of Scholarship,8(3–4) (Special issue Rethinking objectivity, P. 1), 321–335.

Bernstein, J. (1967). A comprehensible world. New York: Random House.Bonvicini, G. (1993). Statistical issues in the 17-keV neutrino experiments.

Zeitschrift für Physik A, 345, 97–117.Christenson, J. H., Cronin, J. W., Fitch, V. L., & Turlay, R. (1964). Evidence for the 2p

decay of the Ko2 meson. Physical Review Letters, 13, 138–140.

Collins, H. (1985). Changing order: Replication and induction in scientific practice.London: Sage Publications.

Collins, H. (2004). Gravity’s shadow. Chicago: University of Chicago Press.Datar, V. M., Baba, C. V. K., Bhattacherjee, S. K., Bhuinya, C. R., & Roy, A. (1985).

Search for a heavy neutrino in the b-decay of 35S. Nature, 318, 547–548.Franklin, A. (1986). The neglect of experiment. Cambridge: Cambridge University

Press.Franklin, A. (1994). How to avoid the experimenters’ regress. Studies in History and

Philosophy of Modern Physics, 25, 97–121.Franklin, A. (1995). The appearance and disappearance of the 17-keV neutrino.

Reviews of Modern Physics, 67, 457–490.Franklin, A. (1998). Selectivity and the production of experimental Results. Archive

for the History of Exact Sciences, 53, 399–485.Franklin, A. (Forthcoming). Are the laws of physics inevitable. Physics in Perspective.Frauenfelder, H., & Henley, E. M. (1975). Nuclear and particle physics. Reading, MA:

W.A. Benjamin.Friedman, J. L., & Telegdi, V. L. (1957). Nuclear emulsion evidence for parity

nonconservation in the decay chain p+—l+—e+*. Physical Review, 105,1681–1682.

Garwin, R. L., Lederman, L. M., & Weinrich, M. (1957). Observation of the failure ofconservation of parity and charge conjugation in meson decays: The magneticmoment of the free muon. Physical Review, 105, 1415–1417.

Glashow, S. L. (1991). A novel neutrino mass hierarchy. Physics Letters, 256B,255–257.

Glashow, S. L. (1992). The death of science? In R. Q. Elvee (Ed.), The end of science?Attack and defense (pp. 23–32). Lanham, MD: University Press of America.

Hacking, I. (1999). The social construction of what? Cambridge, MA: HarvardUniversity Press.

Hacking, I. (2000). How inevitable are the results of successful science? Philosaphyof Science, 67, 58–71.

Hetherington, D. W., Graham, R. L., Lone, M. A., Geiger, J. S., & Lee-Whiting, G. E.(1987). Upper limits on the mixing of heavy neutrinos in the beta decay of 63Ni.Physical Review C, 36, 1504–1513.

Hime, A. (1992). Pursuing the 17 keV neutrino. Modern Physics Letters A, 7,1301–1314.

Hime, A. (1993). Do scattering effects resolve the 17-keV conundrum? PhysicsLetters, 299B, 165–173.

Hime, A., & Jelley, N. A. (1991). New evidence for the 17 keV neutrino. Physics Letters,257B, 441–449.

Hime, A., & Simpson, J. J. (1989). Evidence of the 17-keV neutrino in the b spectrumof 3H. Physical Review D, 39, 1837–1850.

Kawakami, H., Kato, S., Ohshima, T., Rosenfeld, C., Sakamoto, H., Sato, T., Shibata, S.,Shirai, J., Sugaya, Y., Suzuki, T., Takahashi, K., Tsukamoto, T., Ueno, K., Ukai, K.,Wilson, S., & Yonezawa, Y. (1992). High sensitivity search for a 17 keVneutrino. Negative indication with an upper limit of 0.095%. Physics Letters,287B, 45–50.

Lee, T. D., & Yang, C. N. (1956). Question of parity nonconservation in weakinteractions. Physical Review, 104, 254–258.

Lindhard, J., & Hansen, P. G. (1986). Atomic effects in low-energy beta decay: Thecase of tritium. Physical Review Letters, 57, 965–967.

MacKenzie, D. (1989). From Kwajelein to Armageddon? Testing and the socialconstruction of missile accuracy. In D. Gooding, T. Pinch, & S. Schaffer (Eds.), Theuses of experiment: Studies in the natural sciences (pp. 409–435). Cambridge:Cambridge University Press.

Markey, H., & Boehm, F. (1985). Search for admixture of heavy neutrinos withmasses between 5 and 55 keV. Physical Review C, 32, 2215–2216.

Morrison, D. (1992). Review of 17 keV neutrino experiments. In S. Hegarty, K. Potter,& E. Quercigh (Eds.), Proceedings of the Joint International Lepton–PhotonSymposium and Europhysics Conference on High Energy Physics, Geneva,

ed gravity waves until his death and Jack Greenberg never accepted the fact that low-accept both the arguments and evidence presented and the judgment of the physics


Switzerland, 25 July–1 August 1991 (2 vols.) (Vol. 1, pp. 599–605). Geneva: WorldScientific.

Mortara, J. L., Ahmad, I., Coulter, K. P., Greene, J. P., Schiffler, J. P., Trzaska, W. H., &Zeuli, A. R. (1993). Evidence against a 17 keV neutrino from 35S beta decay.Physical Review Letters, 70, 394–397.

Ohi, T., Nakajima, M., Tamura, H., Matsuzaki, T., Yamazaki, T., Hashimoto, O., &Hayano, R. S. (1985). Search for heavy neutrinos in the beta decay of 35S.evidence against the 17 keV heavy neutrino. Physics Letters, 160B, 322–324.

Ohshima, T. (1993). 0.073% (95% CL) upper limit on 17 keV neutrino admixture. InJ.R. Stanford (Ed.), Proceedings of the XXVI International Conference on High EnergyPhysics: August 6–12, 1992, Dallas, Texas (2 vols.) (Vol. 1, pp. 1128–1135). Dallas:American Institute of Physics.

Ohshima, T., Sakamoto, H., Sato, T., Shirai, J., Tsukamoto, T., Sugaya, Y., Takahashi, K.,Suzuki, T., Rosenfeld, C., Wilson, S., Ueno, K., Yonezawa, Y., Kawakami, H., Kato,S., Shibata, S., & Ukai, K. (1993). No 17 keV neutrino: Admixture < 0.073% (95%C.L.). Physical Review D, 47, 4840–4856.

Pauli, W. (1933). Die Allgemeinen Prinzipen der Wellenmechanik. Handbuch derPhysik, 24, 83–272.

Pickering, A. (1984). Constructing quarks: A sociological history of particle physics.Edinburgh: Edinburgh University Press.

Piilonen, L., & Abashian, A. (1992). On the strength of the evidence for the 17 keVneutrino. In G. Chardin, O. Fackler, & J. Tran Thanh Van (Eds.), Progress in atomicphysics, neutrinos and gravitation: Proceedings of the XXVIIth Rencontre deMoriond, Les Arcs, Savoie, France, January 25–February 1, 1992 (pp. 225–234).Gif sur Yvette: Editions Frontières.

Prentki, J. (1965). CP Violation. In Oxford International Conference on ElementaryParticles, 19/25 September 1965: Proceedings (pp. 47–58). [Chilton]: RutherfordHigh Energy Laboratory.

Simpson, J. J. (1985). Evidence of heavy-neutrino emission in beta decay. PhysicalReview Letters, 54, 1891–1893.

Simpson, J. J. (1986a). Is there evidence for a 17 keV neutrino in the 35S b spectrum?The case of Ohi et al. Physics Letters, 174B, 113–114.

Simpson, J.J. (1986b). Evidence for a 17-keV neutrino in 3H and 35S b spectra. In O.Fackler, & J. Tran Thanh Van (Eds.), ’86 massive neutrinos in astrophysics and inparticle physics: Proceedings of the Sixth Moriond Workshop, Tignes-Savoie-France,January 25–February 1st, 1986 (pp. 565–577). Gif sur Yvette: Editions Frontières.

Simpson, J. J., & Hime, A. (1989). Evidence of the 17-keV neutrino in the b spectrumof 35S. Physical Review D, 39, 1825–1836.

Sur, B., Norman, E. B., Lesko, K. T., Hindi, M. M., Larimer, R., Luke, P. N., Hansen, W. L.,& Haller, E. E. (1991). Evidence for the emission of a 17-keV neutrino in the bdecay of 14C. Physical Review Letters, 66, 2444–2447.

Weber, J. (1969). Evidence for discovery of gravitational radiation. Physical ReviewLetters, 22, 1320–1324.

Weber, J., Lee, M., Gretz, D. J., & Rydbeck, G. (1973). New gravitational radiationexperiments. Physical Review Letters, 31, 779–783.

Wigner, E. (1927). Einige Folgerungen aus der Schrödingerschen Theorie für dieTermstrukturen. Zeitschrift für Physik, 43, 624–652.

Wu, C. S., Ambler, E., Hayward, R. W., Hoppes, D. D., & Hudson, R. P. (1957).Experimental test of parity conservation in beta decay. Physical Review, 105,1413–1415.

is failure an option? contingency and refutation

Documents