Philosophy of Science
Why the Same Evidence Can Prove Two Opposite Theories
Every scientific theory you have ever trusted is haunted by an infinite number of rival theories that fit the data just as well.
The Idea
Science is supposed to work like a filter: gather enough evidence and the wrong theories get eliminated, leaving only the true one standing. Underdetermination is the uncomfortable discovery that this filter has holes in it. The term refers to the gap between what evidence can show and what theory claims — specifically, that any finite set of observations is logically compatible with more than one theoretical explanation, sometimes with mutually contradictory ones. This isn't a failure of scientific practice. It's a structural feature of how evidence works. An X-ray of a broken bone tells you the bone is broken, but cannot alone tell you whether you fell or were pushed. Scale that up to cosmology or quantum mechanics, where we can observe effects but never directly inspect the underlying reality, and the problem deepens considerably. Philosophers distinguish two flavours. Contrastive underdetermination is the modest claim: for any theory, you can always cook up a logical alternative that saves the same phenomena. Holist underdetermination, associated with Pierre Duhem and later W.V.O. Quine, is sharper: when an experiment contradicts a prediction, logic alone cannot tell you which part of your theoretical web is wrong — you can always protect your core theory by adjusting an auxiliary assumption instead. Scientists resolve this in practice through judgment, elegance, and consensus. But judgment is not logic. That gap is exactly where the interesting questions live.
In the World
In the early twentieth century, Einstein's general relativity and a rival framework — Hendrik Lorentz's ether theory — made virtually identical predictions for every experiment then available. Lorentz's theory required a physical ether that was conveniently impossible to detect, but it was not strictly refuted by any observation. Einstein won not because his theory passed a test Lorentz's failed, but because it was structurally simpler, made fewer ad hoc assumptions, and felt, to the physicists who mattered most, like it was carved from something deeper. The same dynamic plays out in modern cosmology. Dark matter — an invisible substance that outweighs ordinary matter roughly five to one — is the dominant explanation for why galaxies rotate the way they do. But a competing framework called MOND (Modified Newtonian Dynamics) adjusts the laws of gravity at low accelerations instead of adding invisible stuff, and it fits a surprising range of galactic data comparably well. The two approaches differ radically in what they say the universe contains, yet they remain stubbornly difficult to fully separate on observational grounds alone. Neither dark matter nor MOND has been decisively eliminated. This isn't a scandal — it's science doing its honest, difficult work. But it is a reminder that the story 'we follow the evidence' leaves out the question of who decides when evidence is enough, and by what standard.
Why It Matters
Understanding underdetermination doesn't make you a skeptic about science — it makes you a more honest one. The productive response isn't to conclude that evidence is meaningless, but to notice what is doing the real work when scientists choose between theories: simplicity, coherence with other well-established ideas, predictive fertility, and something harder to name — a sense that a theory is pointing at something real. This has a practical edge. When you encounter scientific debates — about nutrition, economics, climate attribution, drug efficacy — underdetermination explains why genuinely smart, data-literate people can look at the same studies and reach different conclusions. They aren't necessarily cherry-picking or lying. They may be making different but defensible choices about which auxiliary assumptions to preserve. It also inoculates you against two opposite errors: the naïve view that science just reads truth off data, and the cynical view that because evidence doesn't determine theory uniquely, all theories are equally credible. Neither is right. Evidence constrains without dictating. That's not a weakness. It's precisely what makes science a human activity rather than a algorithm.
A Question to Ponder
When you trust a scientific consensus, what exactly are you trusting — the evidence, the judgment of the people who evaluated it, or the social process that produced the agreement?
Get a new one of these every morning.
Start learning with Thinkable