Have You Seen One?

In 1897, at a meeting of the Imperial Academy of Sciences in Vienna, Ludwig Boltzmann finished a talk and sat down, and Ernst Mach stood up.

Boltzmann was the great theorist of heat. He had spent thirty years showing that the behavior of gases, the flow of heat from hot to cold, even the fact that a cup of coffee cools and never warms itself back up, could all be explained if you assumed one thing: that matter is made of enormous numbers of tiny particles bumping into each other. Mach was the most famous physicist-philosopher in the German-speaking world, the man whose name is still attached to the speed of sound. As Boltzmann’s biographer Carlo Cercignani tells it, Mach’s reply was short:

“I don’t believe that atoms exist.”

Mach is also remembered for a retort, often attributed to him in Viennese dialect, that he aimed at atom believers: Ham’s schon eins g’sehn? Have you seen one?

Nobody had. Nobody could. An atom, if there was such a thing, was far too small for any microscope of the day, and it would stay that way for another half century. Boltzmann’s whole life’s work rested on things no one had ever observed and that, as far as anyone could tell in 1897, no one ever would. To Mach, that made them worse than unproven. It made them the wrong kind of claim altogether. Not physics. Metaphysics.

Whose side would you have taken?

Two Kinds of Questions

Most of us carry around a simple map of what can be known.

On one side are questions an experiment can settle. What temperature does water boil at on top of a mountain? Does this drug lower blood pressure better than a sugar pill? How far away is the moon? You can go and check. When two people disagree, they can run the test, and one of them has to give ground.

On the other side are questions no experiment will ever touch. Does the world keep existing when nobody is looking at it? Is there something real underneath what our instruments report, or are the readings all there is? Why is there anything at all, rather than nothing? You can argue about these forever, and people have. Nothing you could weigh or measure would move either side an inch.

The first kind of question belongs to science. The second kind is metaphysics. The line between them isn’t about how hard a question is, or how important. It’s about what kind of question it is. And it’s permanent. A question lands on one side or the other because of its nature, the way a number is odd or even.

This map has a long pedigree. David Hume drew it with a match in his hand:

“If we take in our hand any volume; of divinity or school metaphysics, for instance; let us ask, Does it contain any abstract reasoning concerning quantity or number? No. Does it contain any experimental reasoning concerning matter of fact and existence? No. Commit it then to the flames; for it can contain nothing but sophistry and illusion.”

In plain words: real knowledge is either math or measurement. Everything else is talk. Auguste Comte built a whole “positive philosophy” on that rule in the 1800s, and by the twentieth century Sigmund Freud could say flatly that there is “no other source of knowledge of the universe, but the intellectual manipulation of carefully verified observations.” How the word “knowledge” came to mean only that is a story of its own. What matters here is the shape of the map it left behind: two territories, one border, and the border fixed.

It’s a good map, and the reasons are worth laying out. It explains the most striking fact in the history of thought, which is that science piles up agreed results century after century while philosophers argue about the same questions Plato argued about. It’s modest. It tells us to stop pretending to know what we can’t check. And it keeps science clean. Mach’s own career proved that. He went after Isaac Newton’s idea of “absolute space,” a fixed stage that exists apart from everything on it and that no measurement could ever detect. A physics that leaned on undetectable things, Mach said, was leaning on ghosts. A young Albert Einstein read Mach closely and later credited him with helping clear the way to relativity. “Don’t build on what you can’t test” is not narrow-mindedness. It’s one of the sharpest tools science has.

Working scientists mostly lived by it. In 1867 the chemist August Kekulé, whose famous work on the structure of benzene depended on picturing how atoms link together, still put it this way:

“The question whether atoms exist or not has but little significance from a chemical point of view: its discussion belongs rather to metaphysics.”

Kekulé used atoms every working day. He simply filed the question of whether they were real on the far side of the line, next to God and free will and the meaning of it all. Ask a room of good chemists in 1867 which kind of question “do atoms exist?” was, and most would have given Kekulé’s answer. You probably would have too. In that Vienna lecture hall, Mach was not the crank. He was the careful one.

Paris, 1908

Eleven years later, in a laboratory at the Sorbonne, Jean Perrin was watching yellow specks jiggle under a microscope.

The specks were grains of gamboge, a tree resin used to make yellow paint. Perrin had spent months sorting them by size until he had batches that were nearly identical, too small to sink quickly in water but big enough to see. They never sat still. They shuddered and wandered in random zigzags, a motion the botanist Robert Brown had noticed in 1827 and nobody had properly explained.

Three years earlier, in 1905, a patent clerk named Albert Einstein had published a paper saying what that jiggling was, and more important, what it should look like if atoms were real. If water is made of molecules, a floating speck gets kicked from every side, unevenly, moment to moment. Einstein worked out how far a speck should drift on average in a given time. His formula held one number nobody had ever pinned down: how many molecules sit in a fixed amount of a substance, what chemists call Avogadro’s number. Watch the specks, measure their wandering, and you could count the invisible.

Perrin did it. Then he did it other ways. He measured how the specks thinned out with height in a column of water, like a tiny atmosphere. He compared his results with completely different methods other physicists were using, based on the blue of the sky, on radioactive decay, on the electric charge of the electron. By the time he wrote his book Les Atomes in 1913, he could set out thirteen separate lines of evidence. They came from different corners of physics. They shared no equipment and no assumptions except the one being tested. And they all landed near the same number, roughly six to seven hundred billion trillion. (Today’s value is about 6.02 × 10²³.)

A match like that doesn’t happen by luck. Wilhelm Ostwald, the chemist who had fought atoms alongside Mach for years, gave in. In the preface to a 1909 edition of his chemistry textbook, he wrote that he was “now convinced that we have recently become possessed of experimental evidence of the discrete or grained nature of matter, which the atomic hypothesis sought in vain for hundreds and thousands of years.” Perrin ended his own book generously:

“The atomic theory has triumphed. Its opponents, who until recently were numerous, have been won over and one after another abandon the distrust that for a long time was legitimate and no doubt useful.”

Perrin is fair here. The skeptics weren’t fools. Their distrust was legitimate for a long time, and it did useful work, forcing the atom people to earn their case instead of assuming it.

But this is where the map breaks. Ever since Democritus said that in reality there is nothing but “atoms and the void,” the existence of atoms had been a textbook metaphysical question. Aristotle even explained why. The particular sciences, he wrote in the Metaphysics, “omit the question whether the genus with which they deal exists or does not exist.” Each science takes its subject for granted and studies how it behaves. Whether that kind of thing is out there at all is a different sort of question, and not one a working science settles for itself. Kekulé, in 1867, was following Aristotle’s rule to the letter. Then, between 1905 and 1909, the question moved. It didn’t move because philosophers finally won the argument. It moved because somebody found a way to make it answer to a measurement.

So the line between testable and untestable is real. At any given moment, some questions have an experiment waiting for them and some don’t. But it is not a line between two kinds of question. It’s the current edge of our instruments and our mathematics, and it moves. Calling a question “metaphysical” turns out to be less a verdict on the question than a date stamp: as of today, nobody has figured out how to make this one answer to a number. Some questions may never cross. Nothing here says they all will. What the atom story shows is that you can’t tell from your armchair which ones will, and that confidently filing a question under “just philosophy” has a way of aging badly.

A defender of the old map has a good reply ready. “Fine,” she says. “Then atoms were a scientific question all along. It just wasn’t ripe yet. The map is right; you only mislabeled one item.” The trouble is timing. She can only say that after 1908. Before Einstein’s paper, nobody, including the most careful people in the room, could tell the atom question apart from the questions that really are untestable. If you can only learn which side of the line a question belongs on after someone has carried it across, the line is not a fact about questions. It’s a fact about us.

And something did stay behind. What crossed in 1908 was a specific claim: matter comes in grains, and there are this many of them in a given amount. The deeper question under it did not cross. Should we believe that things we can never observe directly are actually real, or only that theories about them work? Perrin never saw an atom either. He saw specks move and reasoned his way down. In 1980 the philosopher Bas van Fraassen argued, in The Scientific Image, that a careful scientist may accept atomic theory as fitting all the evidence without believing the atoms themselves exist. That is roughly Mach’s position, now standing on the far side of Perrin, and it’s still defended today. The atom crossed. The question of what it means for something unseen to be real stayed on the bank.

The Bets Behind the Instruments

The easy way to tell the atom story is as a tale of patience. Metaphysics asks a question and waits, and eventually science shows up and answers it. On that telling, metaphysics is a waiting room. Questions sit there until their number is called.

That isn’t what happened. The question was pushed across, and it was pushed by people who had placed a metaphysical bet long before there was evidence to back it.

Boltzmann believed atoms were real, not a handy fiction, and that belief is why he spent decades building statistical mechanics, the math of huge crowds of particles. Einstein said plainly why he wrote the 1905 paper. Looking back in his Autobiographical Notes, he put it this way: “My principal aim in this was to find facts which would guarantee as much as possible the existence of atoms of definite finite size.” He didn’t stumble onto evidence for atoms while doing something else. He went hunting for it, because he already believed there was something there to find.

Max Planck, who spent years in public battle with Mach, made the point general. Science, he said, “cannot exist without some small portion of metaphysics,” and he explained why:

“Since one must assume the existence of that which one seeks, the scientist’s assumption of the actual existence of a ‘real world,’ in the absolute sense of the word, eventually grows into a firm conviction which nothing can shake any more.”

Put simply: you can’t search for something while believing there’s nothing to find. Before anyone builds an experiment, somebody has to believe there’s a world out there with a structure worth digging for. That belief isn’t a result of science. It’s the fuel.

Mach was betting too. His rule against unobservable things grew out of a picture of what the world is. For Mach, the world is made of sensations: colors, sounds, pressures, warmths. A “thing” is just a stable bundle of them. That isn’t a lab finding. It’s a metaphysics, and a bold one, every bit as untestable in 1897 as Boltzmann’s atoms. It did real work, too. It helped Einstein see that “at the same time” has to be defined by an actual procedure with clocks and light signals, which was the doorway into relativity. But by 1917 Einstein had seen its limits. In a letter to his friend Michele Besso, he wrote that Mach’s approach “cannot give birth to anything living, it can only exterminate harmful vermin.” A philosophy that only says no can clear ground. It can’t grow anything.

So the anti-metaphysical position was itself a metaphysical position. The fight in the Vienna lecture hall wasn’t physics against metaphysics. It was one metaphysics against another, and each shaped what its owner went looking for. Immanuel Kant saw this long before. Metaphysics exists, he wrote, “if not as a science, nevertheless as a natural disposition of the human mind,” and so “there has ever really existed in every man some system of metaphysics.” Nobody gets to choose whether to have one. You only get to choose whether you know which one you have.

John Stuart Mill found a good image for how this works inside a science. The first principles a science rests on, he wrote, “are really the last results of metaphysical analysis,” and their relation to the science “is not that of foundations to an edifice, but of roots to a tree, which may perform their office equally well though they be never dug down to and exposed to light.” A tree doesn’t need to see its roots to grow. But the roots are doing the work whether anyone digs them up or not. William James said the same thing about psychology and physics: the basic assumptions they start from “must sometime be overhauled,” and “the effort to overhaul them clearly and thoroughly is metaphysics.” Alfred North Whitehead got it down to four words: “Induction presupposes metaphysics.” You can’t learn anything from past experience unless you already trust that the future will resemble the past in some way, and experience alone can never prove that trust.

So metaphysics isn’t the waiting room. It’s closer to the steering wheel. Every question that has crossed was steered there by someone betting on an answer before it could be checked. The bet alone isn’t enough, though. Boltzmann bet for thirty years and couldn’t settle anything. Something else had to happen to turn his bet into a check.

Where the Mathematics Comes In

Nobody saw an atom in 1908. Perrin’s microscope showed him specks about the size of bacteria. What made the invisible countable was a chain of reasoning written in mathematics.

Einstein’s formula was a bridge. At one end sat a metaphysical claim: matter is made of separate particles of a definite size. At the other end sat something you could measure with a microscope, a stopwatch, and a great deal of patience: how far a resin speck drifts, on average, in thirty seconds. The formula tied the two ends together. If atoms were real and there were this many of them in a given amount of water, the speck should wander this far. If matter were smooth all the way down, there would be nothing to kick the specks, and their jiggling would need some other explanation that made its own, different predictions. The question “are there atoms?” became “what number do you get?” And a number can be wrong in public.

Then came the part that really settled it. Any single method might have fooled Perrin. A hidden error in one setup could produce a plausible number by accident. But for thirteen unrelated methods to land on the same enormous number by accident, the world would have to be conspiring against us. The only sensible explanation is that there really is something there to count.

G. H. Hardy, the Cambridge mathematician, described a physicist’s job in just these terms. The physicist, he wrote, “is trying to correlate the incoherent body of crude fact confronting him with some definite and orderly scheme of abstract relations, the kind of scheme which he can borrow only from mathematics.” A metaphysical claim, alone, just sits there. It becomes testable when someone builds a scheme that says: if this is true, then that number, over there, in a place you can measure, has to come out a certain way.

This cuts against an old worry. Francis Bacon wanted mathematics to “terminate natural philosophy rather than generate or create it,” to tidy up results after the fact, not lead the way. Aristotle warned that “the minute accuracy of mathematics is not to be demanded in all cases.” Both were protecting the study of nature from floating off into abstraction. But in the atom story, math did more than tidy up. It was the only road across. A question moves out of metaphysics when someone finds a calculation that makes each possible answer cost something: a prediction the other side can’t match.

There’s a strategy game that captures this well. In Europa Universalis, much of the world starts as terra incognita, blank gray map. The gray isn’t a kind of land. It’s land your country hasn’t explored yet. Send out an explorer with the right technology and the gray peels back into coastlines and provinces. Where the gray ends isn’t decided by the terrain. It’s decided by how far your ships can reach. On this picture, metaphysics is the gray part of the map, and mathematics plus instruments are the ships.

If that were the whole story, the gray would shrink a little every year, and science would be slowly eating metaphysics down to nothing.

Copenhagen and the Moon

The theory that finally explained how atoms behave, quantum mechanics, is often called the most accurate theory in the history of science. Some of its predictions have been checked to better than one part in a billion. It gave us transistors, lasers, and MRI machines. By the logic of the map, a success that big should have pushed metaphysics even further back.

It did the opposite. It dragged the oldest metaphysical question there is into the middle of a physics department: is nature there when nobody’s looking?

Philosophers had fought over this for centuries. George Berkeley argued that for ordinary objects, “their esse is percipi”: to be is to be perceived. Kant held that nature, as we can know it, is “the sum total of phenomena,” the world as it shows up in experience, while what things are in themselves stays out of reach. Planck held the opposite with equal force. There is a “real world, in the absolute sense,” he said, independent of any mind, and science is the long approach to it. For three hundred years this was a fight for philosophers.

Quantum mechanics made it a fight for physicists. In the theory’s standard form, a particle doesn’t have a definite position until it’s measured. Before that, the math describes only odds. Is that because we don’t know where the particle is? Or because it isn’t anywhere in particular? The theory doesn’t say.

Einstein thought the second option was absurd. The physicist Abraham Pais remembered a walk in Princeton when Einstein “suddenly stopped, turned to me and asked whether I really believed that the moon exists only when I look at it.” In 1935, with Boris Podolsky and Nathan Rosen, Einstein published an argument that quantum mechanics must be incomplete. If you can predict something about a particle with certainty without disturbing it, they said, there must be “an element of physical reality” behind that prediction, a fact already sitting there whether the theory mentions it or not. Niels Bohr answered that it makes no sense to talk about a particle’s properties apart from the whole experiment used to reveal them. The two men argued about it for decades and, as the story is usually told, agreed to disagree.

For most physicists, the argument went into the metaphysics pile. It seemed to make no difference to any prediction. Einstein’s view and Bohr’s view gave the same numbers for every experiment anyone could think of. The physicist David Mermin later summed up the working attitude in a phrase often wrongly credited to Richard Feynman: “Shut up and calculate!” By the rules of the map, that was exactly right. A question that makes no difference to any measurement is what “metaphysical” means.

So the most successful theory in physics had not shrunk the gray part of the map. It had created new gray, and it was closer to home than anything Hume worried about. The question was no longer about God or the soul. It was about the coffee cup on your desk.

Bell’s Number

In 1964, a Northern Irish physicist named John Stewart Bell, who worked at CERN on particle accelerators and thought about the foundations of quantum mechanics mostly on the side, found that the question everyone had filed under metaphysics could be made to answer to a number.

The idea fits in a paragraph. Take two particles created together and send them flying apart to two distant labs. In each lab, a scientist picks one of two ways to measure her particle and writes down the result. Now suppose Einstein’s picture is right in two ways. First, each particle carries its answers with it, settled from the start, like a sealed envelope. Second, nothing, no signal and no influence, travels faster than light, so what one lab chooses can’t affect what the other lab finds. Bell proved that under those assumptions, the results in the two labs can only line up with each other so much. Scored in the standard way worked out in 1969 by John Clauser and three colleagues, that agreement can go no higher than 2. Quantum mechanics predicts it can reach about 2.83.

A dispute that had been “just philosophy” for thirty years now came down to a number somewhere between 2 and 2.83. You could go and measure it.

People did, though it took nerve. Work on the foundations of quantum mechanics was widely seen as a poor career move, and several of the people who did it later recalled being told it wasn’t real physics. Clauser and Stuart Freedman ran the first test at Berkeley in 1972. Alain Aspect in Paris ran sharper ones in 1982 and got about 2.7, right where quantum mechanics said and well above Bell’s ceiling. Skeptics pointed to loopholes, ways an experiment might be fooled. Experimenters closed them one after another, and in 2015 three separate teams, in Delft, Vienna, and Boulder, closed the main ones at once. In 2022 Aspect, Clauser, and Anton Zeilinger shared the Nobel Prize in Physics for this work. The physicist and philosopher Abner Shimony gave the whole project a name that would have sounded like a contradiction to Hume: experimental metaphysics.

So this question crossed too. What came back across the line with it?

Not an answer to Einstein’s question on that walk. What Bell’s number rules out is a package: answers settled in advance, and no faster-than-light influence, and measurement choices made independently of the particles. At least one of those has to go. Which one is exactly what the experiments can’t tell you.

Give up the rule against faster-than-light influence, and you get something like David Bohm’s pilot-wave theory, where particles always have real positions and are steered by a wave that links distant places instantly. The philosopher Tim Maudlin argues that this is what Bell really proved: locality is what has to go. Keep locality but give up the idea that a measurement has one single outcome, and you get Hugh Everett’s many-worlds picture. Give up the independence of the experimenters’ choices, and you get what’s called superdeterminism, in which the “choices” were tied to the particles all along. Give up the idea that measurement reveals something already there, and you’re near Bohr, and near newer views like QBism, which treat the quantum state as a record of an agent’s information. Every one of these fits every experiment done so far.

The question “does the moon exist when no one looks?” was too vague to test. Bell sharpened it until part of it could be tested, and the test came back. What’s left over is sharper still: a forced choice among four strange options, each defended by serious physicists. The crossing didn’t shrink metaphysics. It made the gray smaller in one spot and more exact, and more unsettling, everywhere around it.

And the edge is still moving. Some physicists, including Roger Penrose, have proposed that the quantum wave collapses on its own once enough mass is involved, which would make a real, measurable difference. Those theories make predictions, so they can be tested, and they are being tested. In 2020 an experiment deep under the Gran Sasso mountain in Italy ruled out the simplest version of one of them. Another patch of gray is being charted right now.

This is the same shape the atom story had on a smaller scale. When atoms crossed, the question of whether to believe in unseen things stayed behind. When local realism crossed, a sharper question about which of four pillars to pull out stayed behind. Each time science takes a question from metaphysics, it pays for it with a new one, harder and more exact, that nobody could even have asked before. That’s why metaphysics hasn’t died despite three centuries of confident funerals. It isn’t the scraps science hasn’t gotten around to eating. It’s something science produces, every time it succeeds.

William James saw this coming before any of it happened. Every science, he wrote, runs “into queries which only a metaphysics alive to the weight of her task can hope successfully to deal with.” Then he added a line that sounds like an apology and is really a diagnosis: “the best mark of health that a science can show is this unfinished-seeming front.” Planck, from the other side of the argument, pictured the real world as “not the starting point, but the goal of all scientific endeavor, a beacon winking and showing the way from an inaccessibly remote distance.” You steer by a beacon. You never tie up at it. And the coastline you chart on the way keeps getting longer.

None of this means every metaphysical question is waiting for its Perrin. Some may never get an experiment. Gottfried Leibniz asked, “Why is there something rather than nothing?” and it’s hard to imagine the apparatus. Bertrand Russell warned that “very little can be proved a priori from considerations of what must be,” and that warning cuts both ways. It catches the metaphysician who proves from his chair that some feature of the world is impossible. It also catches the positivist who proves from his chair that some question can never be tested. The honest position is Kekulé’s without his confidence: this isn’t testable yet, and I don’t know whether it ever will be.

Duino

Boltzmann didn’t live to see his question cross. He had suffered for years from severe depression, and in September 1906, on a summer holiday at Duino, near Trieste, he took his own life. Historians warn against turning his death into a parable about atoms; his troubles were older and more tangled than any one quarrel. But the timing is hard to forget. Within three years, Ostwald recanted. Within twenty, Perrin had a Nobel Prize for proving the thing Boltzmann had been told to stop believing in. Mach, as far as anyone can tell, never fully came around.

Boltzmann is buried in Vienna’s Central Cemetery, and carved above his bust is a short equation: S = k log W. It says that the entropy of a system, the measure of disorder that grows as coffee cools and ice melts, depends on the number of ways its atoms can be arranged. In 1897 that equation was a metaphysical bet, a formula about the arrangements of things nobody had seen and that Mach said weren’t there. Today it sits in every physics textbook as plain physics. The stone didn’t change. The line moved past it.

When Mach stood up in that lecture hall, he thought he was defending physics against metaphysics. Boltzmann thought he was defending physics against a timid philosophy. Neither was standing safely inside science, looking out at the speculators. Both were standing at the edge of the map, making the bets that decide which way the next explorer sails.

The question carved on that gravestone is settled now. The one behind it isn’t, and somewhere, in some seminar room, someone is standing up to say they don’t believe it.