Popper and Kuhn on Theory Change

Professor of Philosophy of Science John Worrall on the scientific revolutions, falsifiability and what are the main features of a scientific hypothesis

videos | October 8, 2020

There’s this nice modernist Enlightenment idea that the scientific revolution of the 16th and 17th centuries gave mankind the way to find out the truth about the Universe. As a famous couplet by Alexander Pope said, ‘Nature and nature’s laws lay hidden in the night; God said “Let Newton be” and all was light’. Many scientists actually complained that Newton had discovered the basic structure of the Universe, and all that was left was crossing t’s and dotting i’s, filling in the details. But then, of course, at the turn of the 20th century, there were two great scientific revolutions. It turned out that Newton’s theory certainly wasn’t the last word.

In particular, it was replaced by Einstein’s special and general theories of relativity, and that’s really quite a radical change at any rate at first glance. Newton assumes that gravity is an action of a distance force, there’s no mediation, there’s no gravitational waves or anything. In his view, it’s just straight: the movement of this planet affects that planet immediately, instantaneously. Absolute space, absolute time: there’s no concept of two events being simultaneous for one observer but not simultaneous for another.

Theoretical physicist Xavier Bekaert on Galilean relativity, the separation of time and space, and Einstein’s reconciliation of mechanics and electromagnetism
And yet, in Einstein, everything’s overturned. The world isn’t infinite, as it’s not part of Newton’s theory; there’s no action at a distance in Einstein. It is possible, famously, for two events to be simultaneous in one frame of reference and not in another, according to the theory of relativity.

So what’s going on? How can science be this great bastion of truth and rationality if theory changes as radical as that can occur? Well, the first philosopher of science who was really centrally influenced by scientific revolutions was called Popper, a predecessor of mine, very distinguished obviously at the London School of Economics where I teach: in fact, he set up the philosophy department. His whole view was that science is about falsifiability and is characterized by falsifiability. Even at the lowest levels, we can’t ever prove our theories, even silly little theories about all ravens being black, because we can’t observe every raven that there is. Famously, people in Europe assumed that all swans were white, and then it turned out that when Captain Cook went to Australia, there was a refutation.

Popper said science is not about induction, it’s not about establishing theories on the basis of evidence; it’s about falsifiability. So for him, the scientific revolution of the 20th century involving, in particular, Einstein’s theory was exactly meat and drink to his view, which was that science is characterized by its theories being falsifiable, that the only thing that you could say at any stage is that this theory, although falsifiable, has not yet been falsified. That’s the contrast being, according to Popper, with theories like Marx’s theory of history or Freud’s theory of psychoanalysis, which, according to him (and this is obviously controversial), could accommodate any observation. You could explain both the purpose of the behaviour in jumping into the lake to save the drowning child and in not doing that. Within a Freudian context, anything could be explained.

When it turned out that the Michelson–Morley experiment, a famous experiment that was done that supported Einstein’s theory, turned out the way that Einstein predicted and countered the way that Newton had predicted, that was the end of the story. Newton’s theory gets falsified, and Einstein’s theory is accepted in its place; it’s equally or more falsifiable than Newton’s theory, but unlike Newton’s theory, it evades eventual refutation because for many decades, Newton’s theory, according to Popper, was not falsified, and that’s why it continued to be accepted.

So the rationale for theory-changing science is very straightforward, according to Popper. A theory for a long time is accepted because despite the fact that it can clash with observation results it fails actually to do so, it gets all the observation results correct.

Then along comes another theory that says that certain experiments will turn out differently than the way that Newton predicted they would, and this leads to a falsification of Newton’s theory and confirmation or corroboration, in Popper’s terms, of Einstein’s theory.

Obviously, there’s more detail in Popper’s voluminous work starting in 1934, Logik der Forschung, but this is basically certainly the picture that people got from Popper, and it’s central to what he’s talking about. In 1962, Thomas Kuhn wrote a famous and very influential book called The Structure of Scientific Revolutions, which really, although not directly, addresses, if you like, a falsification of Popper’s own views.

What Kuhn pointed out was that scientists don’t operate the way that you would expect they do, according to Popper, looking for refutations, and once they get a refutation (that is, some observational result that clashes with their theories), they give up the theory. In fact, ironically, Popper emphasized one episode himself that shows that this view of falsification is incorrect. Kuhn talked about scientists treating apparently negative results as anomalies rather than refutations.

Physicist Martin Rees on Isaac Newton, theory of general relativity, and black holes
For example, in the 19th century, when Herschel discovered a purely observationally (and hitherto unsuspected) planet called Uranus, and Newtonian physics calculated what its orbit should be according to Newtonian theory, it turned out that the orbit was different, significantly different than it was predicted on the basis Newton’s theory. That looks like it should be a period falsification: you’ve deduced the consequence from the theory about an observable event that you’ve looked at, the observable event namely the motion of Uranus, and that event turns out to be different from what Newton’s theory said it would be.

So it should be a falsification, but it’s not how people operated at all. On the contrary, they said, no, Newton’s theory must be true because it’s been confirmed so often in the past, and so we must be making an incorrect assumption about an initial condition. Maybe there’s another planet that we didn’t see, a further planet. After all, Herschel discovered a planet that people didn’t know about before; perhaps there’s another planet that, when we add its gravitational interaction with Uranus into the interaction with the Sun and the other planets that we knew about already, would give us the right account exactly on Newtonian principles.

In fact, they worked backwards and said, okay, well, what would we have to assume about this unsuspected planet? You can’t just observe it; the planet is a very slowly moving spot of light, as far as we’re concerned, against the background of the fixed stars, fixed relative to one another. The planets move, but they move very slowly, so it’s not easy to spot planets. So it’s perfectly conceivable that there’s a hitherto unsuspected planet around. So they worked backwards from the refutation or, as Kuhn would have put it, anomaly with Uranus to make a prediction of a new planet. Where would it have to be? What would its mass have to be in order to account within Newtonian theory for what had hitherto been the anomalous or refuting behaviour of Uranus? This led to a prediction of the existence of Neptune, which was massively well confirmed.

And Kuhn catalogs lots and lots of cases where scientists treated events and observations that were inconsistent with accepted theory as not being refutations but as pointing to some auxiliary assumption, some initial condition, something more particular that we’re assuming in order to test the theory, and that was what was wrong rather than the theory itself.

So now we come back to the issue of why scientists change theories or science collectively changed its mind in going from classical physics from Newtonian physics to Einstein’s theory. It can’t be as straightforward as Popper was suggesting, that it’s a case of a refutation of Newton in an experiment that Einstein’s theory gets correct; it must be something else. Again, this is all subject to scholarly discussion and so on, but the basic message that Kuhn seems to be giving in the book is that there are no reasons why scientists change from what he called one paradigm to another. In science, it’s like a religious conversion; it just happens in a gestalt. there’s no reason, and there’s no set of universal principles of how evidence pinches on the theory that will tell you when it’s time to change a theory. We just have to wait for the older scientists who stick to the older view (there are always people who stuck to classical physics even though most of their colleagues switched to Einstein); we just have to wait for those people to die out. There aren’t any rules.

Imre Lakatos actually solved this problem. He accepted that Kuhn’s theory, Kuhn’s account, is much more accurate in terms of the way that scientists operate; in particular, he accepted that scientists standardly treat negative instances not as preparing falsifications but as Kuhnian anomalies. He accepted that but said that Kuhn’s theory was much too socially constructivist and science really does have genuine general rules.

The important thing to look at is not for falsification but for verification. You can always hold on to a theory when it clashes with the facts by modifying auxiliary assumptions in the way that Adams and Le Verrier did: these were the two people who independently postulated the existence of Neptune. You can always do that but it’s only scientific if in the process you make a new prediction that gets confirmed.

So that’s exactly what happened: it wasn’t just that you accommodated the negative data from Uranus within classical physics; you also predicted the existence of Neptune, the existence of a new planet. That’s a testable assumption, and that turned out to be correct. The contrast is with, for example, explaining the null results of the Michelson–Morley experiment, which was an important confirmation of the special theory of relativity by postulating in a completely ad hoc way that the rods are contracted in the direction of motion.

Let me give you a simpler example from the biology of the sort of thing that you mean by ad hotness. So there’s this problem: if you believe, like creationists believe, that contrary to Darwin, the Universe was created in 4004 BC, you’ve got all these fossils that seem to be massively older, and then you’ve got all these trees that seem to be much older, and you’ve got rates of decaying radioactive samples that seem to indicate the Universe has been around for millions and millions of years, not 6,000. But it’s still easy to explain: you just assume that God created the Universe with a lot of things looking already very old, so he created the fossils in the rocks. That’s the sort of typical, completely antiscientific reaction to a falsification that you’re just accommodating the data you can see. There’s no possible way that you can further test this hypothesis; it just accommodates the data.

Philosopher and sociologist of science Steve Fuller on the peer-review process, the relationship between scientists and policymakers, and public engagement in the scientific process
So, according to Lakatos (I think he was right; Kuhn missed that), there is a difference between treating an anomaly in a non-scientific way, just absorbing it within your framework because you’re committed to the theoretical framework, and getting an absorption that leads to a new prediction. So Lakatos’ account is that the Einsteinian revolution occurred not because its predecessor was refuted, as Popper would have said, while Einstein wasn’t, but that Einstein scored independent success, for example, with the Michelson–Morley experiment when it comes to the general theory of relativity, with the observations that two stars that have given distance apart during the night will be a different distance apart during the day. You can’t usually observe stars in the day, but of course, you can with an eclipse, and Einstein predicted exactly what the distance of the two consecutive stars would be.

So, in scientific progress, testability is the key. Lakatos gives you an account that corrects Popper but doesn’t go to the extreme of Kuhn. It’s all a question of what scientists had decided to do without there being any general rules for theory change.

Become a Patron!

Professor of Philosophy of Science, the London School of Economics and Political Science
Did you like it? Share it with your friends!
Published items
To be published soon