Falsifiability
Falsifiability is Karl Popper’s proposed solution to the demarcation problem, the problem of distinguishing science from non-science (and from pseudoscience), set out in the SEP entry on Popper. A theory counts as scientific iff it forbids some observable state of affairs; i.e., it’s falsifiable.
Popper
If a theory is incompatible with possible empirical observations it is scientific; conversely, a theory which is compatible with all such observations… is unscientific.
Why is "explains everything" bad?
If “explains everything” is a vice rather than a virtue, then a lot of “scientific” theories are doing something other than science, and the asymmetry between confirmation and refutation is what tells you which.
Popper’s formative contrast was Einstein’s general relativity vs. Freudian psychoanalysis and Marxism. Einstein predicted that light would bend measurably near the sun during an eclipse. The prediction was risky: if the measurement came back wrong, the theory was sunk. (It didn’t.) Freud and Marx, by contrast, could explain every possible observation. Any behavior fit psychoanalysis; any historical turn fit dialectical materialism. That “explanatory power” Popper read as weakness: a theory that forbids nothing predicts nothing.
Popper accepts Hume’s verdict from the problem of induction: universal scientific generalizations cannot be verified by any number of observations. “All swans are white” is never proven by counting white swans. But it can be refuted by a single black swan. Asymmetric: confirmation is impossible, falsification is clean (in principle). So Popper flips the whole structure: science isn’t in the business of confirming theories, it’s in the business of trying to falsify them, and living with the ones that survive.
When a theory survives serious attempts to falsify it, Popper says it’s corroborated, provisionally the best we have. This is explicitly not confirmation: corroborated theories are still conjectural, still subject to future refutation. We use them because nothing better has survived yet.
Main objections:
- Duhem–Quine thesis: no theory is tested in isolation; it’s tested along with auxiliary hypotheses, measurement assumptions, instrument calibrations. When the data disagree with prediction, logic alone doesn’t tell you which assumption to blame. Clean falsification is never actually available
- Ad hoc rescues: scientists routinely modify auxiliary hypotheses to save favored theories (add an epicycle, invent a new particle, posit a correction). Popper has to distinguish legitimate modifications from illegitimate ad hoc ones, and that distinction is hard to make non-arbitrary
- Kuhn’s paradigms: normal science doesn’t abandon a paradigm on a single failed prediction; it accommodates anomalies until enough accumulate for a revolution. Falsificationism is a bad fit for how science actually proceeds
- Lakatos’s research programmes: falsification operates over whole programmes (a “hard core” plus a “protective belt” of auxiliaries) rather than individual theories. Programmes get judged progressive or degenerating over time, not falsified in a moment
My take
Popper is the attitude I want to have as a researcher, which is different from saying he’s correct. Falsificationism as an epistemic discipline (ask yourself what observation would prove you wrong, and if there isn’t one, you’re not doing science) is the most useful single heuristic in Popper’s book, and it’s how I try to frame my own research claims.
Falsificationism as an account of how science actually works is clearly wrong, and Kuhn + Lakatos + Duhem all land their punches. Nobody drops general relativity on one weird measurement; they check the equipment, the data, the auxiliaries. That’s rational, not irrational.
So I take Popper the way I take Hume on causation: the descriptive claim is shaky, the normative pressure is useful. “You can never prove your theory, only fail to refute it” is the right posture, even if the cleaner version of falsificationism doesn’t survive contact with the actual practice of science.
This is also why I keep pseudoscience and philosophy of science adjacent. The demarcation problem has no tidy solution, but Popper’s criterion still picks out most of the obvious cases (astrology, psychoanalysis, homeopathy) and that’s more than any rival criterion does.