From the archive · June 28, 2026
A model that survives attempts to kill it
What makes an idea scientific rather than just convincing?
01 · Word
Theory
noun
In science, not a guess but a model that has survived hard testing: it explains what we have already seen, predicts what we have not, and can still be proven wrong
Examples
“It's just a theory” misuses the word: in science, a theory is the most tested kind of explanation we have.
A theory can be overturned by a single reliable observation it cannot accommodate.
Origin
From the Greek theōria, “contemplation, a looking at.” The casual sense means a mere speculation, the scientific sense a rigorously tested model. The two pull in opposite directions, which is the source of endless confusion about phrases like “the theory of evolution.”
02 · Idea
Falsifiability
Popper argued you can never prove a general theory true, since no number of confirming cases rules out a future exception. You can prove one false with a single solid counterexample. So the mark of a scientific theory is not how much evidence supports it, but whether it makes risky predictions that could fail. A theory that can absorb any outcome, contradiction included, explains nothing. Good science courts its own refutation.
Karl Popper developed falsifiability in the 1930s as a “demarcation criterion,” a line between science and non-science, most fully in The Logic of Scientific Discovery. He was reacting to theories he felt could explain any observation whatsoever, contrasting them with Einstein's relativity, which made a specific, checkable, and potentially fatal prediction.
You cannot prove a theory true by piling up confirmations, but one clean contradiction can prove it false.
Trust the model that dares to be wrong and isn't, not the one that can never be wrong.
Limits and context
Strict falsifiability is too clean for real science: theories are rarely abandoned on a single failed test, because any experiment involves auxiliary assumptions that might be blamed instead. Later thinkers like Kuhn and Lakatos showed that theories are held, patched, and replaced in messier, more social ways. Falsifiability is an ideal that clarifies, not a rulebook scientists literally follow.
The everyday version is useful anywhere: ask of any confident claim, “what would change your mind?” If the honest answer is “nothing,” you are not looking at a theory. You are looking at a faith.
03 · Moment
The eclipse that tested Einstein
The total solar eclipse of May 29, 1919, observed from Príncipe off West Africa and Sobral in Brazil
Einstein's general relativity predicted that the Sun's gravity would bend the path of light from distant stars by a specific amount, about twice what Newton's physics implied. A total eclipse lets astronomers see stars near the Sun's edge; during the eclipse of May 29, 1919, expeditions led under Arthur Eddington photographed such stars from Príncipe and Sobral and measured how far their apparent positions had shifted. The results matched Einstein's prediction more closely than Newton's.
- Date
- May 29, 1919
- Test
- Starlight bending by the Sun
- Verdict
- Einstein confirmed
The caveat
Historians have shown the 1919 measurements were noisier and the analysis more contested than the triumphant headlines implied; the confident verdict was somewhat ahead of the data. Later, cleaner experiments decisively confirmed the prediction, so the theory was right, even if the founding test was rougher than legend holds.
The announcement in November 1919 made Einstein an international celebrity overnight and stood for decades as a model of science done right: a bold theory made a precise, falsifiable prediction, and an experiment went looking for a way it could fail. It is the exact example Popper cited to explain what a genuinely scientific theory looks like: one willing to be killed by a measurement.
The bending of light by gravity is now called gravitational lensing, and it has become a routine tool of astronomy rather than a daring test.