§1. A Cool Experiment

On a January morning in 1986, cold even in Florida, the launch of space shuttle Challenger ended abruptly seventy-three seconds after liftoff when a fireball erupted and the shuttle disappeared in a plume of smoke nine miles above the Atlantic Ocean, killing all seven of the crew.

At NASA they told him that Challenger’s O-rings showed scorching. Apparently they had been concerned about these O-rings for some time. O­rings are gaskets more than ten meters in circumference, sealing the seams between booster sections against escaping gas. Scorching suggests they failed.

This was not the first or even an early shuttle launch. There had been many; it was almost routine. So why should O-rings fail on January 28? Feynman also learned that the contractor had been testing O-rings under conditions of cold, indicating concern. The 1986 launch was the first to occur when the temperature was below freezing. None of this information—scorching, contractor’s concern, different weather—was in the evidence before the commission.

The New York Times heard what Feynman was hearing and published a story suggesting problems with cold O-r ings. The commission called a public meeting to review the evidence in the presence of reporters and tele­vision cameras. At dinner the night before, Feynman got the idea of taking a sample of the O-ring, clamping it tightly, and dunking it in ice water. Would it spring back to form? The next day, before the meeting, he went to a hard­ware store to purchase tools and a C-clamp. Then Feynman together with the head of NASA privately tried the experiment in his office.

Empiricisms. Barry Allen, Oxford University Press (2021). © Oxford University Press.

DOI: 10.1093/oso/9780197508930.001.0001.

Feynman took his place when the meeting came to order, dripping pliers and C-clamp in his pocket. A NASA manager began explaining O-rings, passing a sample to the commissioners. It was when the sample got to him that Feynman produced his tools and performed his cool experiment. As he was preparing the material, he spoke to the NASA manager who had been testifying:

I took this stuff that I got out of your seal, and I put it in ice water, and I dis­covered that when you put some pressure on it for a while and then undo it, it maintains—it doesn't stretch back. It stays the same dimension. In other words, for a few seconds at least—and more seconds than that—there's no resilience in this particular material when it is at a temperature of thirty- two degrees. I believe that has some significance for our problem.¹

That evening, all the major television networks showed Feynman's experi­ment. The next day it was the front-page story in the New York Times and the Washington Post. Feynman became a national hero and public figure.

Some years before, he had been asked to explain what science is. He answered, “The separation of the true from the false by experiment or experi­ence, that principle and the resultant body of knowledge which is consistent with that principle, that is science.” To students in his introductory physics course he explained that “the fundamental hypothesis of science, the funda­mental philosophy,” is that “the sole test of validity of any idea is experiment.” He reiterates this elegant creed at the conclusion of lectures on the concept of physical law:

If it disagrees with experiment it is wrong. In that simple statement is the key to science. It does not make any difference how beautiful your guess is. It does not make any difference how smart you are, who made the guess, or what his name is—if it disagrees with experiment it is wrong.

Yes, of course, one might say. The trouble comes when we have to de­termine whether a conjecture disagrees with an experimental result. Is it really that obvious? We have to decide that the experiment was well done. Responsibility for the decision cannot be foisted on “inductive logic.” No ob­servation can be stated without recourse to a language, and there is never one uniquely right choice. So what experiments show is always to some degree a decision, the discretion logic leaves to experience. It is impossible to separate what the experiment shows from the interpretation of those who carried it out and claim the right to say what it means.³

Is what Feynman did even an experiment? He never squeezed warm O- rings—is it an experiment at all without controls? Certainly everybody calls it an experiment. Every one of his obituaries describes the occasion, calls what Feynman did an experiment, and says it demonstrated something about why Challenger exploded. Feynman himself called it a “little experiment,” though he acknowledged that it was not strictly up to snuff. Examining a witness at the inquiry later the same day, he said, “I did a little experiment here, and this is not the way to do such experiments, indicating that the stuff looked as if it was less resilient at low temperatures”⁴

Suppose it was an experiment—what did it show? Feynman said it indi­cated that the material lost resiliency at low temperature. No one remembers this modest conclusion. Instead, the experiment is depicted as the pivotal mo­ment of the inquiry. The New York Times obituary says that it “perfectly dem­onstrated” the vulnerability of the seal. An obituary in Scientific American says it identified “one of the proximate causes of the disaster” Feynman did not himself mention the experiment or what conclusion he drew from it in the appendix he wrote to the final report expressing his dissenting opinion. Perhaps the most acute epitome came from Hans Bethe, another Nobel lau­reate, and Feynman's erstwhile colleague at Los Alamos.

Feynman indicated irregularities with his “little experiment” They be­come more or less glaring depending on what we understand the experiment to have accomplished. For instance, suppose we say the experiment demon­strated that cold seals were a proximate cause of the fatal explosion. To enroll the experiment in such a claim requires some assumptions worth making explicit. It was not raining on the morning of the launch; the seals were cold but not soaking wet, as Feynman's sample was. The metal in his clamp was probably not the same alloy as the boosters, and the pressure in the hand- tightened clamp was uncontrolled. It seems likely that the seals were sub­ject to much greater pressure due to the massive booster rockets they joined. How do we know these things make no difference? And what about time? The Challenger seals were exposed to the cold for hours, Feynman's sample for a few seconds. It may seem obvious that more time would only amplify the effect Feynman demonstrated, but how do we know? Obvious assumptions can turn out to be wrong.

Feynman tested a tiny portion of the sealant material. How do we know that a bigger piece responds the same way? Robert Boyle, prince of seventeenth-century experimentalists, observed, “Divers experiments suc­ceed, when tried in small quantities of matter, which hold not in the great.” In his Experimental History of Cold (1665), Boyle recounts his investiga­tion of reports that iron hoops surrounding water barrels break in very cold weather. Boyle wondered at the cause. Did cold change the iron, or was there some other agent? The expansion of water on freezing was not then a fact and had even been refuted. One of the first things Boyle had to do was confirm that a volume of water expands on freezing.

Feynman believes in experiments, but he wants them carefully checked. He was notorious for checking students' calculations, and found mistakes often enough to appreciate that little things need to be scrutinized. As he says, the experiment has to be rubbed back and forth.

When I say if it disagrees with experiment it is wrong, I mean after the exper­iment has been checked, the calculations have been checked, and the thing has been rubbed back and forth a few times to make sure the consequences are logical consequences from the guess [the conjecture being tested], and that in fact it disagrees with a very carefully checked experiment.⁷

Feynman professed chagrin at having tried his experiment privately. “Although I knew it would be more dramatic and honest to do the experi­ment for the first time in the public meeting, I did something that I'm a little bit ashamed of. I cheated. I couldn't resist. I tried it.... I discovered that it worked before I did it in the open meeting.” How would it be more dramatic unrehearsed? The audience is seeing it for the first time in any case. It would be reckless to leave the experiment untried until he was live on television to the world. If nothing interesting happened, he would look like a buffoon.⁸

A different question is why, having done his experiment privately and knowing the result, Feynman gave his performance at all. He could have sub­mitted a report to his chair, or recommended that engineers do a full test, but he chose instead to give his performance. I like to think he wanted others to have the experience he did when he performed the test privately. The value of the experiment is to alter the trajectory of inquiry, but to exercise that power it has to be experienced by those who stand to learn from it.

Experiments can be viewed in two ways: as probative proofs, or as instruments of discovery and invention.

Freeman Dyson, Feynman's colleague and friend, interprets Feynman's ex­periment in the first, probative way. Feynman gave the public the truth; he made truth present before them with his own hands. “The public saw with their own eyes how science is done, how a great scientist thinks with his hands, how nature gives a clear answer when a scientist asks her a clear ques­tion.” Is that what Feynman did, ask a clear question? I'd like to know what it was! Dyson does not say, or even say what the answer was. Be that as it may, if we follow Dyson's theorematic interpretation of the experiment, as proving a theorem, then the quibbles I mentioned become more serious, and the ex­periment looks rushed and shabby.⁹

The mandate of the Challenger commission was to “establish the probable cause or causes of the accident” In that form, the question is infinite. Cause or causes? Why does anything happen? But if we introduce a new question— did lost resilience in cold O-rings contribute to the disaster?—that is some­thing people can look into. Feynman took a nebulous question with no answer in sight and enforced a detour. If you want to pursue that big question now, you have to pass through the little questions he successfully posed along the way. What do we know about the effect of cold on O-rings? How cold? How long? What other components may have been affected by the tempera­ture, or by failing O-rings?¹⁰

His experiment did not answer a question, it invented one. Managers and engineers thought they understood the O-ring. “Now they were forced to agree that they had not. It was not just temperature effects that they did not understand. The failure analysis showed that other factors, previously not taken into account, had contributed to the technical failure: the potential for ice in the joint, putty behavior, the effect of violent wind shear.” All of these were new questions, new problems, newly exigent thanks to Feynman's little experiment.¹¹

My suggestion that the value of experiments is their power to alter the tra­jectory of inquiry, especially with new problems, seems confirmed from an­other direction in the announcement of a Higgs boson finding at CERN in July 2012. The principal experimental apparatus, the Large Hadron Collider, was at the time the most ambitious piece of equipment in the history of ex­perimental physics: a twenty-seven-kilometer ring of superconducting magnets chilled to within one degree of absolute zero. Design and construc­tion required twenty years and cost US$10 billion.

Official reactions to the confirmed discovery of a Higgs boson were couched in the probative idiom—the experiment proves, establishes, confirms, discovers. Among scientists, however, a different attitude some­times prevailed. Some had hoped the experiment would cast light on dark matter, or produce evidence of postulated supersymmetry particles, or even that it would fail and drive physics “beyond the Standard Model.” Writing two years before the announcement, Steven Weinberg said, “Finding one neutral Higgs particle would not pull us out of the doldrums, it would put us into the doldrums. It would be just what we're expecting and it would give us no clue to anything new. Finding several kinds of Higgs, or even no Higgs at all, would be better.”¹²

When confirmation was announced, on July 4, 2012, the event, for all the hoopla and good feeling, was ironically disappointing for many, especially those who understood the experiment best. Despite the impressive confir­mation of the predictions of the Standard Model, “Nothing unexpected has leapt out from the data. There are no hints of supersymmetric particles; no dark matter candidates; no clues to guide the development of particle physics beyond the Standard Model.” No new problems, just old answers verified.¹³

I mentioned two ways of looking at experiments. Their difference cor­responds to a duality in European empiricisms which I describe in terms from Euclid, who distinguished theorematic and problematic sciences. Theorematic sciences are bodies of theorems demonstrating something that exists independently of our knowledge. A theorem is grasped in an attitude of disinterested theorem, in Latin, contemplatio; for example, con­templating how the sum of a triangle's internal angles must be the same as two right angles. Such theorems invent or construct nothing. They disclose timeless relations among eternal forms; all that is new is our knowledge of them. Problems and problematic science are different. Problems require the construction of something that did not previously exist. Given a straight line, describe a square; inscribe an equilateral triangle in a circle; double the volume of a cube—the famous problem of the Delos altar.¹⁴

History's empiricisms tend to one or the other, theorematic or problem­atic, in how they understand the relation between experience and know­ledge. Is experience expected to make a theorem evident or confirm a hypothesis? Those are theorematic empiricisms, and the value of experience is the evidence it extends to theory and theory's truth—“Experience is ulti­mate evidence” is the slogan. Problematic empiricism discovers methodical experience as a power of invention and the solution of problems. It values experience for the questions it discovers, the techniques it invents, and the surprises it engenders.

Ancient empiricism was already split on these lines. The empiricism of Hippocrates, Democritus, and Epicurus was problematically oriented; em­pirical knowledge was expected to produce something, for instance, health or tranquility. Aristotle's sort of empiricism favored contemplative values, sub­ordinating experience to a demonstration of theorematic truth. Problematic empiricism predominated in the seventeenth century; for instance, in the work of Galileo, Boyle, and Newton. A theorematic trend rises with posi­tivism in the nineteenth century, while the radical empiricisms of the twen­tieth century reclaim experience for problematic thought.

Philosophers would do well to give up equating experience with present sense awareness, as in the expression “my current experience,” or “my present sensory experience.” Aristotle explains experience better when he says that it is not mere perception but perception remembered, a mnemic synthesis. Experience is “much memory,” in the pithy epitome of Thomas Hobbes. The conscious present is perception, which becomes experience when it is recol­lected and allowed to enhance the efficacy of action. Experience is not just a moment of conscious perception; it is having learned from perception. Experience is not something had, it is something recalled, a deferred, belated perception, a quality of the remembered past that was never the quality of a conscious present. The passage of time adds something to the present and its presence. It is called experience.

Empiricisms begins with a historical argument. I trace the tangled lines of what has been assumed, affirmed, and disputed concerning the values of experience and experiment from antiquity to the twentieth century. In an­tiquity, empiricism was a practice of inquiry and natural philosophy, from Alcmaeon in the sixth century bce to Galen in the second century ce, and including Hippocrates, Democritus, and Epicurus. Their approach to natural philosophy was the principal ancient alternative to the rationalism of Plato, Aristotle, and their schools. This empirical philosophy disappeared in late antiquity when inquiry disappeared as a practice, replaced by school com­mentaries on the “classics” that Plato and Aristotle had already become. A modernizing trend reprises empirical philosophy in an experimental key, from the recovery of Greek medicine and mathematics in the sixteenth cen­tury, to the consolidation of experiments as the principal method of natural philosophy by the eighteenth century.

“Experience” and “experiment” (or their Latin cognates) were inter­changeable terms until the eighteenth century. Modern empiricism is con­cerned with experiments rather than sensory impressions. Galen in antiquity took a step toward experimental natural philosophy. Robert Grosseteste, Roger Bacon, William Ockham, Galileo Galilei, Francis Bacon, and Robert Boyle advanced the idea in modern empiricism, usually cognizant of med­ical tradition. A new trend emerges in the late nineteenth century, an effort to make empiricism more consistently empirical, eliminating assumptions inherited, by a tortuous route I will recount, from the nominalism inaugu­rated by Peter Abelard and William Ockham. We shall see that much of what makes this empiricism “radical” is antipathy to the legacy of nominalism, and a return to the problematic ontology of empiricism’s oldest sources.