How we know

Foundations

“Nobody was there”

“The Big Bang is just a theory. Nobody was there. It is a guess.” You will hear this, and it contains a real misunderstanding worth taking apart, because the answer is the whole point of science.

In everyday speech a theory is a hunch. In science it is the opposite: the highest rank an idea can hold. A theory is a framework that makes predictions about measurements nobody has made yet, predictions that could come out wrong. If they come out right, again and again, the theory earns its name. The theory of gravity has never been observed either. What has been observed is that everything it predicts happens.

The Big Bang has been on trial for seventy-five years, and for the first twenty of them it had a serious rival.

The rival

In 1948 Hermann Bondi, Thomas Gold and Fred Hoyle proposed the steady-state theory. The universe expands, as Hubble had shown, but it has no beginning and no end. As the galaxies drift apart, new matter is created in the gaps, about one hydrogen atom in each cubic metre every ten billion years or so, far too little to ever catch in the act, and just enough to keep the universe looking the same for ever.

It was a beautiful idea, and it was a scientific one, because it made predictions that differed from the Big Bang’s. If the universe has always looked the same, then the distant past, which we see by looking far away, should look like the present. And a universe that was never hot should have no leftover glow. The Big Bang predicted the opposite on both counts. So the two could be told apart.

Test one: does the universe change?

The first blow came from radio astronomy. In the late 1950s Martin Ryle’s group at Cambridge counted radio sources by brightness and found far more faint ones than a steady universe allowed. Faint means far, and far means long ago. The universe had contained more powerful radio galaxies in the past than it does now. It had changed. Quasars, discovered soon after, exist only at great distances, only in the past. Today the Hubble and Webb telescopes show galaxies at every lookback time, and the further back they look, the smaller, bluer and more chaotic the galaxies are. The universe has a history.

Test two: the afterglow

The second blow was the one in lesson 1.4. A hot early universe should have left a glow, cooled by expansion to a few degrees above absolute zero. Ralph Alpher and Robert Herman said so in 1948, the same year the steady state was born. In 1965 the glow was found, and by 1990 its spectrum had been measured as a perfect thermal glow. The steady state had no way to make it. Hoyle tried, ingeniously, and failed.

The afterglow also gives a test that the steady state could never have passed. Expansion says the glow was hotter in the past, by exactly the factor the universe has stretched since. Astronomers have measured its temperature in gas clouds seen as they were ten and eleven billion years ago, by the way the glow excites molecules in them. It was three to four times hotter than today, exactly as predicted.

Test three: the recipe

The third blow was the helium of lesson 1.2. A quarter of the universe’s ordinary matter, by mass, is helium, and stars cannot have made anything like that much in the time available. Only a hot early universe could. In 1964 Hoyle himself, with Roger Tayler, published a paper admitting it. Half a century later the argument is quantitative: the amount of deuterium in pristine gas matches the Big Bang’s prediction, made using the density of ordinary matter read from the afterglow. Two different measurements, of two different epochs, agree to a few percent.

Tests that could have failed, and did not

A good theory keeps offering ways to prove it wrong. Here are four the Big Bang has offered and survived.

The ages. A universe cannot be younger than its stars. In the 1990s it briefly seemed to be: estimates of the expansion rate implied a universe of ten or eleven billion years, while the oldest star clusters seemed to be fourteen or fifteen billion years old. Better distances brought the cluster ages down to twelve or thirteen billion, and the discovery of dark energy pushed the universe’s age up to 13.8 billion. The stars are younger than the universe, by a comfortable but not enormous margin.

Time dilation. If distant galaxies are receding because space is stretching, their clocks should appear slow. A supernova at redshift 0.5 should take one and a half times as long to brighten and fade. It does, and by exactly that factor. The rival idea that light merely loses energy on its journey predicted no such stretch.

The peaks. In 1970 Jim Peebles and Jer Yu predicted that the afterglow should carry a pattern of ripples at particular sizes, from sound waves in the early universe. The pattern was first clearly seen in 2000, and its largest ripple came out at exactly the angle a flat universe predicts.

The ruler in the galaxies. Those same sound waves should have left a faint preference for galaxies to sit about 150 megaparsecs apart. In 2005 a survey of tens of thousands of galaxies found it, at the predicted scale.

One model, six numbers

The current picture, called the concordance model, describes the universe with six adjustable numbers: the amounts of ordinary matter, dark matter and dark energy, the expansion rate, and two numbers describing the primordial ripples. Those six fit the afterglow’s thousands of data points, the galaxy surveys, the supernovae, gravitational lensing and the counts of galaxy clusters, all at once. No rival comes close.

What “how we know” does not mean

It does not mean everything is settled. The model works, and two of its ingredients are things nobody has identified: dark matter and dark energy make up ninety-five percent of it. An early phase called inflation is well supported and unconfirmed. Two ways of measuring the expansion rate disagree. Knowing how we know includes knowing where the map ends, which is the next two lessons.

Deeper

Prediction versus accommodation

Any theory can be adjusted to fit data it already knows about. The evidence that counts is a prediction made before the measurement. The Big Bang’s record here is unusually strong. The afterglow’s existence and rough temperature: predicted 1948, found 1965. Its blackbody spectrum: predicted, measured 1990. The acoustic peaks: predicted 1970, seen 2000. The baryon acoustic scale in galaxies: predicted from the afterglow, found 2005. The deuterium abundance: predicted from the afterglow’s baryon density, matched in quasar spectra. Each of these could have come out otherwise.

The tests, side by side

The six numbers

From the 2018 Planck analysis: the expansion rate, 67.4 ± 0.5 km/s per megaparsec; ordinary matter, 4.9 percent of the total; dark matter, 26.5 percent; dark energy, 68.5 percent; the tilt of the primordial ripple spectrum, 0.965 ± 0.004; and the fraction of afterglow photons scattered during reionisation, about 5 percent. Everything else follows, including the age, 13.787 ± 0.020 billion years.

The age crisis and its resolution

In the mid-1990s the Hubble Space Telescope’s first distance measurements suggested an expansion rate near 80 km/s per megaparsec. In a universe of only matter, that implies an age of about eight billion years, younger than the globular clusters. Three things changed. Cluster ages fell as the distance scale was corrected. The expansion rate settled near 70. And the 1998 supernova results showed the expansion is accelerating, which makes the universe older for a given expansion rate today, because it expanded more slowly in the past. The crisis became a confirmation.

Why several lines matter

Each pillar depends on different physics: the expansion on gravity and geometry, the afterglow on thermodynamics, the light elements on nuclear physics. Each depends on different instruments and different people. A mistake in one would not propagate into the others. That independence, more than any single measurement, is why the hot early universe is treated as established.

Frontier

The tension

The one place the model creaks is the expansion rate. Extrapolating from the afterglow gives 67.4 ± 0.5. Measuring directly, with Cepheid variables and supernovae in nearby galaxies, gives 73.0 ± 1.0. The gap is five standard deviations. Either one side has an error no one has found, despite a decade of looking, or something happened between the afterglow and today that the six numbers do not capture. The James Webb Space Telescope is re-measuring the nearby distances; so far the gap has not closed. Active research.

Galaxies too early

Webb has found luminous galaxies less than three hundred million years after the Big Bang, more and brighter than most models expected. This is not a problem for the hot early universe; it is a problem for how quickly the first stars formed, which the next lesson takes up.

The ingredients

The model’s success is also its embarrassment. It works only if most of the universe is made of things we cannot see and have never identified. Lesson 1.7 is about that.

Transcript and shot list

Target runtime about 7 minutes at a measured narration pace, with room for the visual beats to breathe. Shot types: ANIM is rendered from code with Manim, AI is a generated shot from Higgsfield with the shared style reference, REAL is agency imagery with credit, CAM is the narrator. Timecodes are targets and will move to match the recorded narration.

[00:00] S01 · AI · A classroom; a hand goes up; on the board, the words “just a theory”.

“The Big Bang is just a theory. Nobody was there. It’s a guess.” You will hear this, and it deserves a proper answer, because the answer is the whole point of science. In everyday life a theory is a hunch. In science it’s the opposite: the highest rank an idea can hold. A theory makes predictions that could fail. If they keep not failing, it earns the name. So here’s the Big Bang on trial, with a real rival, and every way it could have lost.

[00:40] S02 · ANIM · Cosmic calendar; every test in this lesson sits in the last quarter-second before midnight on 31 December.

Everything in this lesson happened in the last hundred years: the final quarter of a second of the cosmic calendar.

[00:55] S03 · ANIM · Two timelines side by side: Big Bang, hot and dense at the start, galaxies changing; steady state, identical frames for ever with new atoms appearing in the gaps.

In 1948 three physicists, Bondi, Gold and Hoyle, proposed the rival. The steady state. The universe expands, but it has no beginning. As galaxies drift apart, new matter appears in the gaps, about one hydrogen atom per cubic metre every ten billion years, just enough to keep the universe looking the same for ever. A beautiful idea. And a scientific one, because it predicted things the Big Bang didn’t. No leftover glow, because the universe was never hot. And a distant past that looks exactly like today. Both of those could be checked.

[01:50] S04 · ANIM · A radio-source count rising steeply at the faint end; galaxies at increasing lookback times get smaller and bluer.

First check: does the universe change? In the 1950s Martin Ryle counted radio sources and found far too many faint ones. Faint means far, and far means long ago. The universe used to have more powerful radio galaxies than it does now. It changes. Look far enough away today and the galaxies are smaller, bluer, messier. The universe has a history. The steady state said it didn’t. Strike one.

[02:35] S05 · ANIM · The afterglow’s blackbody curve; then a gas cloud at high redshift with a thermometer reading three to four times 2.7 K.

Second check: the afterglow. A hot early universe should leave a glow, cooled to a few degrees above absolute zero. Predicted in 1948, found in 1965, a perfect thermal spectrum by 1990. The steady state had no way to make it. And here’s the check the steady state could never have passed: the glow should have been hotter in the past. Astronomers measured its temperature in gas clouds seen as they were ten billion years ago. Three to four times hotter than today, exactly as expansion predicts. Strike two.

[03:25] S06 · ANIM · A bar of ordinary matter, one quarter helium; a star icon with a tiny helium output; then a deuterium match between two epochs.

Third check: the recipe. A quarter of ordinary matter is helium, and stars can’t have made that much. Only a hot early universe could. In 1964 Hoyle himself published the calculation and admitted it. Today the argument is precise: the deuterium in pristine gas clouds matches the prediction made from the density of matter read off the afterglow. Two epochs, two kinds of physics, one answer. Strike three.

[04:10] S07 · ANIM · Four quick tests: a ruler of ages, star clusters just inside the universe’s age; a supernova light curve stretching by 1.5; the acoustic peaks appearing; galaxies pairing at 150 megaparsecs.

A good theory keeps offering ways to prove it wrong. Four it survived. A universe can’t be younger than its stars; in the 1990s it briefly seemed to be, and better measurements put the oldest clusters just inside the universe’s 13.8 billion years. If space is stretching, distant clocks should run slow; a supernova halfway across the universe fades one and a half times slower, exactly. The ripples in the afterglow were predicted in 1970 and seen in 2000, at the angle a flat universe demands. And those same ripples should show up as a faint preference for galaxies to sit 150 megaparsecs apart. In 2005, they did.

[05:15] S08 · ANIM · Six dials; behind them, the afterglow spectrum, a galaxy map, supernova points and a lensing arc all fit at once.

The whole picture now runs on six numbers: how much ordinary matter, dark matter and dark energy, how fast the expansion, and two numbers for the primordial ripples. Those six fit the afterglow, the galaxy surveys, the supernovae, gravitational lensing and the counts of galaxy clusters, all at once. Nothing else comes close.

[05:50] S09 · ANIM · Three labels: ESTABLISHED for the hot early universe; ACTIVE RESEARCH for the expansion-rate tension and the dark ingredients; a map with its edge drawn in.

The scorecard. That the universe was once hot, dense and small, and has expanded and cooled for 13.8 billion years: established, by several independent lines that could each have failed. Where the map ends: two ways of measuring the expansion rate disagree, and the model only works if most of the universe is made of things nobody has identified. Active research. Knowing how we know includes knowing where the edges are.

[06:35] S10 · AI into ANIM · The classroom board; “just a” is wiped away, leaving “theory”; the calendar returns.

So, nobody was there. True. Nobody was there for the formation of the Himalaya either, or the extinction of the dinosaurs, or any event before the invention of writing. We know those things the same way: by predictions that could have failed, tested against evidence. That is not a guess. That is what the word theory means. Next lesson: the hundred million years when nothing shone, and the first light.

Shot list

Two AI shots, eight code-rendered shots. Estimated Higgsfield use: 3 to 6 generations including retries.