Kepler's Pragmatic Ellipse

Belief, Simplicity, and the Shape of the Sky

Science runs on a tension between what we believe nature should look like and what it actually does. We come to problems with priors: ideas about the right kind of answer, the right level of complexity, the kind of order the world ought to have. Sometimes those priors are right. Often they are not. The interesting cases are the ones where a scientist’s beliefs and the evidence pull in different directions, and the scientist has to decide which to follow.

Johannes Kepler’s career is one of the best illustrations of that tension in the history of science. He began with strong, specific beliefs about what the solar system should look like. He ended up discovering laws far simpler than he expected, read directly from data rather than deduced from any deeper principle. And the path between those 2 points runs through one of the most instructive acts of pragmatism in scientific history: a moment when he set aside the answer he believed in for a simpler tool he could actually use, and found, to his own surprise, that the simpler tool was right.

Kepler was born in 1571, a generation before Newton, 2 generations before the full machinery of calculus and analytical mechanics. He was trained as a theologian at Tübingen, and spent a large part of his career as the official astrologer to the Habsburg court, casting horoscopes not out of belief but out of financial need, since aristocratic patronage required it of him. There is a quiet irony in this: the man who was helping to replace mysticism with empirical law was being paid to maintain one of its most persistent superstitions. He was, in many ways, a man of the late Renaissance, steeped in mysticism, preoccupied with divine harmony, convinced that the structure of the cosmos reflected the mind of God.

And yet Kepler was also something genuinely new: arguably one of the first practitioners of what we would now recognize as modern empirical science. For him, data and observation were not just inputs to a preconceived framework but final judges that could overrule theory, intuition, and deeply held belief. When the numbers disagreed with him, he changed his mind. When the data pointed somewhere he did not want to go, he followed it. This habit of letting evidence lead over personal conviction was not common in his time, and it is not universal in ours. In Kepler it was fundamental.

His first great theory shows what he was up against in himself. In 1596, at 25, he published the Mysterium Cosmographicum, proposing that the spacing of the 6 known planets around the Sun was set by the 5 Platonic solids, nested one inside another like a set of geometric shells. The octahedron inscribed in Mercury’s sphere, the icosahedron in Venus’s, the dodecahedron in Earth’s, the tetrahedron in Mars’s, the cube in Jupiter’s. It was a beautiful idea, mathematically elegant, theologically resonant. Kepler loved it. He returned to it for the rest of his life, even as his other work moved in completely different directions.

The model was wrong. But it reveals exactly what Kepler believed nature should be: geometrically harmonious, structured by objects of perfect mathematical symmetry, divinely ordered. This was his prior. It was rich, elaborate, and specific. The universe, in his view, should be complex in the right way.

What his career demonstrates is that strong priors are not disqualifying. What matters is not whether a scientist starts with beliefs, but whether they are willing to let those beliefs be overruled. Theories need to be consistent, not theorists. A scientist who abandons a wrong model when the evidence demands it is not being inconsistent; they are doing their job. The prior is a starting point, a necessary one, but it has no special claim on loyalty once the data speaks against it.

What happened over the following decades is the story of that belief being educated out of him by reality. Kepler’s 3 laws of planetary motion, that orbits are ellipses with the Sun at one focus, that a planet sweeps out equal areas in equal times, that the square of a planet’s orbital period is proportional to the cube of its average distance from the Sun, are often taught today as if they were theorems derived from deeper principles. They are not. They were phenomenological laws, read directly from observation, with no theory behind them. Kepler did not deduce them. He found them in Tycho Brahe’s data by following the numbers wherever they led.

His story divides into 2 parts: what he believed, and what he did when belief ran into a wall. They are different lessons, and both are worth understanding.

Belief

In the winter of 1600, Johannes Kepler traveled to Prague to work with Tycho Brahe, still convinced that the solar system had a real physical structure that a good enough set of observations could reveal.

Tycho Brahe was not a man who was easy to work with or easy to like. He was a Danish nobleman who had built the most sophisticated observatories in Europe and spent decades recording planetary positions with a precision no one had achieved before. He was also jealous of his data, protective of his legacy, and suspicious of the younger man’s ambitions. Their collaboration was brief and fraught. Brahe died in October 1601, and Kepler inherited 2 things: the position of Imperial Mathematician at the court of Rudolf II in Prague, and Tycho’s archive.

The archive was the prize. What Brahe had built was the most accurate account of planetary positions that pre-telescopic astronomy had ever produced, aimed at arcminute precision, about 1/60th of a degree. When the telescope arrived in the early 1600s and astronomers looked back at Tycho’s records, they found them consistently reliable in a way that was remarkable for manual observation.

Kepler trusted this data enough to let it override his theoretical preferences, which is a harder thing than it sounds. Small errors can always be blamed on instrument imprecision or measurement noise. But when Kepler worked through the Mars data, the best available circular model fit some of Tycho’s observations very well and failed elsewhere. At its worst, the residual error was 8 arcminutes.

8 arcminutes is a tiny angle, invisible in ordinary life. But Kepler understood what it meant. This distinction matters: random errors scatter, and you average over them and move on. Systematic errors have structure, and that structure is information. When a residual appears consistently across observations and cannot be blamed on measurement variability, it means your model is missing something real. Kepler recognized the 8-minute discrepancy for what it was, and he refused to let it go. He wrote, famously, that those 8 minutes “have led the way to the reformation of all of astronomy.” He followed them, into the labyrinth that would consume the next several years of his life.

The natural question, once you have decided that circular orbits are wrong, is: what shape is right?

Kepler’s answer was not the ellipse. It was an oval, an egg-shaped curve, broader near aphelion, the planet’s farthest point from the Sun, and narrower near perihelion, the closest point. This came directly from his physical picture of why planets move at all. He believed the Sun was the active driver of planetary motion, exerting an influence that diminished with distance, which meant planets closer to the Sun moved faster and planets farther away moved more slowly. This picture is, remarkably, correct in structure — it anticipates what Newton would later formalize. But its geometric consequence looked asymmetric: broader on the aphelion side, tighter on the perihelion side. Not a circle, and not an ellipse, which is symmetric about both its axes. Something egg-like, with a different local curvature at each end. Historians have called it the “puffy-cheek” orbit.

And here came the wall.

In a 1603 letter to David Fabricius, a Lutheran pastor in East Frisia who did serious observational astronomy alongside his clerical work, Kepler put the difficulty plainly. Fabricius was not a famous theorist, but Kepler regarded him as the finest observer in Europe after Tycho, and their letters are one of the best records we have of science in mid-investigation: tentative, iterative, positions held and then revised. Fabricius had also, in 1596, become the first person to identify a periodic variable star, Mira in the constellation Cetus, a discovery that speaks to the caliber of astronomer he was. In 1617 he was killed by a farmer he had denounced from the pulpit.

In the letter, Kepler wrote that he lacked “the geometric generation of the oval or face-shaped curve.” He added that if the figure were a perfect ellipse, then Archimedes and Apollonius would suffice. To grasp what this meant, consider that Cartesian coordinate geometry did not yet exist. Descartes published La Géométrie in 1637, nearly 30 years after Kepler’s Astronomia nova. In the pre-Cartesian world, curves were geometric objects defined by classical constructions, and the ellipse had been exhaustively studied since Apollonius of Perga in the 3rd century BCE. Kepler’s oval had no such theory behind it. He could not find its foci, compute its areas, or use it to make precise predictions. It was, in the mathematical sense of his time, an unusable object: visible in outline, but not available as a working tool.

He was stuck. He knew, or thought he knew, what shape Mars traced in the sky. He simply had no way to do astronomy with it.

Action

Kepler did not stop.

This is worth pausing on, because stopping would have been the reasonable choice. The frontier of scientific knowledge is enormous, and one of the hard-won lessons of research is knowing when a problem simply isn’t ready to be solved. The mathematics Kepler needed did not exist. The physical framework that would eventually explain why orbits take the shapes they do would not exist for another 80 years. A scientist who looked at this situation clearly could have concluded, not unreasonably, that the shape of planetary orbits was a question for a future generation better equipped to answer it.

There is also a career logic to setting hard problems aside. The frontier is full of tractable questions, problems where the tools exist and progress is possible. A scientist who concentrates there can publish, build a reputation, train students, and contribute steadily to the accumulation of knowledge. The person who locks onto a problem that the current state of mathematics cannot handle risks years of frustration with nothing to show for it. It is often the wiser professional move to leave the hard problem for later, and many fundamental questions in the history of science have sat unanswered for generations precisely because each successive researcher made this calculation and moved on to something more accessible.

Kepler did not make that calculation. He did what only a certain kind of scientist does: he found the best available tool that would let him keep moving, and he used it, even knowing it wasn’t the right tool in any final sense.

The ellipse was that tool. By 1603 it was one of the most thoroughly understood geometric objects in the mathematical tradition, with a fully developed classical theory going back nearly 2,000 years. Kepler knew how to construct one, compute its area, find its foci. He could use it as a quantitative tool in a way that his oval completely resisted.

So he used it, not as the answer. He was explicit in the 1603 letter to Fabricius that he did not believe the orbit was an ellipse. The relevant passage runs close to: if the figure were a perfect ellipse, then Archimedes and Apollonius would suffice. He was stating a counterfactual — if we were lucky enough to be dealing with an ellipse, we would already have the tools. The implication was that he did not think they were that lucky.

Instead, he was using the ellipse as an approximation to his oval, a workable stand-in that allowed calculation to proceed while he continued searching for the correct curve. You are not claiming the approximation is exact. You are claiming it is useful, which is a different and more modest and more honest thing to claim. There is a version of this pragmatic move in every serious scientific investigation: use the linear model to build intuition about the nonlinear one, keep the investigation alive when the direct path is blocked.

This tends to get erased in retrospect because retrospect knows how the story ends. From the vantage point of knowing the orbit really is an ellipse, Kepler’s use of the ellipse as a calculation tool looks like unconscious recognition. That reading is almost certainly wrong. The more accurate reading is simpler: the ellipse was workable, the oval was not, and Kepler was the kind of scientist who preferred a good tool he could work with to a correct tool he could not.

Through 1604 and into 1605, the oval models continued to fail. The constructions assumed, in one form or another, the very quantity they were supposed to derive. The physical account was also fraying: his “magnetic hypothesis” did not produce the asymmetric orbit he expected when he worked out its consequences carefully. He was in what he himself described as a “veritable labyrinth of calculation.”

At some point in this period, something changed. The exact sequence isn’t clear from the surviving letters. Kepler had been comparing the ellipse’s predictions to Tycho’s observations, and he began to see that the ellipse was not merely close. It was right. When he set up the comparison carefully, with the Sun at 1 focus and the orbit sweeping out equal areas in equal times, the numbers matched within the measurement error.

His first response was not acceptance. It was suspicion. He had physical reasons for thinking the orbit should be asymmetric in a way an ellipse is not. When the ellipse started fitting, his reaction was not triumph but doubt. The historical record indicates that by early 1605 he had tentatively identified the Martian orbit as elliptical but still considered it “inexplicable” — physically unaccounted for. The empirical success was there. The physical understanding was not.

What closed the gap was the area law. Working through how a planet’s speed should vary under his physical picture, Kepler arrived at what we now call his second law: a line from the Sun to the planet sweeps out equal areas in equal times. The area law, it turned out, was not independent of the ellipse. When you combine equal-area sweeping with his force picture, the resulting curve is an ellipse. The ellipse was not an arbitrary fit. It was a geometric consequence of his physical scheme.

He had not been wrong to use the ellipse as a scaffold. He had been wrong to think the scaffold wasn’t the building. His conviction that the orbit should be asymmetric, egg-like, broader at aphelion, had simply been mistaken. The symmetry of the ellipse, which had seemed like the wrong kind of shape for the physics he envisioned, turned out to be exactly what the physics required.

What Science Actually Does

When Astronomia nova appeared in 1609, Kepler did not have a proof that planetary orbits are ellipses. Finite data with finite precision cannot prove the exact form of a physical law. This is basic to how science works. Tycho’s observations were the most accurate pre-telescopic planetary positions ever recorded, and they were still a finite collection of measurements covering a finite arc of Mars’s path. In principle, infinitely many curves fit within any dataset’s error bounds. No amount of data, however rich, can prove that nature follows a specific mathematical form.

What Kepler had was something more modest and more honest: the curve that fit Mars’s orbit better than the circular models, better than the oval models, better than any of the alternatives he had seriously explored. It organized the data into a stable, predictive, quantitative law that worked across the full range of Tycho’s observations. The 8-arcminute discrepancy that had broken the circular model disappeared. That is not proof, but it is not nothing. It is what philosophers of science call comparative empirical support, and it is the actual currency of scientific progress.

Newton, publishing the Principia in 1687, supplied something structurally different: a derivation. Starting from the inverse-square law of gravitation and the laws of motion, he showed that a body under a central inverse-square force must follow a conic section, an ellipse for a bound orbit, a parabola or hyperbola for an unbound one. The ellipse was no longer an empirical discovery. It was a theorem. Kepler discovered that planetary orbits are ellipses. Newton explained why they must be. Both achievements were real. They were different in kind.

Consider the simple pendulum: a weight on a string, swinging back and forth under gravity. The setup is about as elementary as classical mechanics gets. And yet the exact period at large amplitude is not a simple expression. It involves a complete elliptic integral, a function so complex it resisted closed-form treatment for generations. An elementary mechanical system with a complicated answer. Physical systems are under no obligation to yield simple answers just because they are simply stated.

A modern scientist confronting Kepler’s problem would reach for a different tool: Occam’s razor. The principle, named for the 14th-century English friar William of Ockham, holds that among competing explanations consistent with the evidence, the simplest should be preferred. It has become something close to a reflex in modern science. A researcher today, faced with a choice between the ellipse and some more complex family of curves, would apply it almost without thinking: use the simpler model until the data forces you to something more complicated. The ellipse fits? Use the ellipse. Don’t go looking for an egg unless the residuals demand it.

Kepler did not think this way, and could not have, because Occam’s razor had not yet become the methodological standard it later became. He started from physical intuition, and the ellipse violated that intuition. It was too symmetric, too simple, too clean. He reached for it not out of parsimony but out of pragmatic need: it was workable when nothing else was.

The deeper irony is that Kepler’s story has become one of the clearest justifications for Occam’s razor. The ellipse was the simplest curve consistent with the data, and it turned out to be correct. The more complex oval he believed in was wrong. Not because simplicity is a law of nature, but because in this case the underlying physics happened to be simple, and the simplest model faithful to the data happened to track that simplicity. Kepler stumbled into the right answer without using a principle that his stumble helped justify. He did not choose the ellipse because it was parsimonious. He chose it because it was all he had. And it was enough.

That is the full arc of the story: a belief about what nature should look like, overruled by data; and an action taken under constraint, vindicated by discovery. The belief and the action pull in different directions, and yet both were necessary. The empirical discipline made Kepler willing to follow the ellipse when it fit, even though he hadn’t expected it to. The pragmatic stubbornness kept him at the problem long enough to find out. A scientist with only one of those two qualities would have ended up somewhere else entirely.