Why falsificationism is false

Karl Popper famously defended the view, known as falsificationism, that what distinguishes science from non-science is falsifiability. On this view, a theory is scientific if and only if it’s falsifiable, at least in principle. What this means for a theory to be falsifiable is that one can think of a possible observation that would be inconsistent with the theory. For instance, since Newton’s law of universal gravitation implies that every particle exerts a force of attraction on every other particle, it would be falsified if we observed a particle that repels another particle. Since it’s at least conceivable that we could observe this, Newton’s law of universal gravitation is falsifiable and therefore scientific. Popper wants to contrast this with theories like psychoanalysis, which according to him can be reconciled with any conceivable observation, hence is not scientific.

I have often been struck, when talking to scientists, by the influence that Popper seems to have among them. I would go as far as saying that, in many scientific fields, falsificationism has become the official philosophy of science. It’s drummed into the heads of scientists when they’re in graduate school and, with a few exceptions, they never learn anything else about philosophy of science and spend the rest of their career thinking that Popper’s conception of science is still the gold standard. In fact, however, not only is falsificationism not the gold standard, but it never was. Indeed, I think it’s fair to say that, despite the incredible popularity it has achieved, falsificationism was always opposed by most philosophers of science and it’s certainly the case that nowadays virtually all of them reject it. Moreover, philosophers of science have excellent reasons to do so, because there are very strong arguments against falsificationism. In this post, I want to explain why I think falsificationism is false, because I keep running into people, especially scientists but not only, who are surprised to hear that and I figured it would be useful to be able to refer them to something accessible to non-philosophers that explains it.

Given this post’s intended audience, I will not discuss the more sophisticated versions of falsificationism that have been proposed, which often stem from remarks Popper himself made that showed he was aware of the difficulties his theory faced. (For instance, even though above I have formulated the criterion of falsifiability as a necessary and sufficient condition for a theory to be scientific, Popper sometimes claimed that it was only necessary but not sufficient.) Rather, I’m going to address the somewhat crude version of falsificationism that most people seem to have in mind when they talk about Popper’s conception of science, which I summarized at the beginning of this post. However, I want to say here that, as far as I can tell, I think it’s roughly what Popper himself believed. Again, he was a sophisticated thinker and knew there was problems with this view, but his goal was to find a purely logical criterion of demarcation between science and non-science and this is what the view I summarized above is supposed to do.

Once you modify it to address those issues, as Popper sometimes did in response to his critics or because he was anticipating their objections, it no longer delivers a purely logical criterion of demarcation and in my opinion the view loses much of its original appeal. But it’s okay if you disagree with this. I know there are philosophers of science who think that falsificationism can be salvaged and that a more sophisticated version of that view is correct, so they will predictably not be moved by this post, but I’m fine with that since it’s the crude version that most people, especially scientists, have in mind when they talk about falsificationism. I will be satisfied if all I have achieved with this post is to convince you that, at least in this crude version, falsificationism is false. Although I ultimately disagree, it’s okay with me if you think the lesson one should draw from this critique of naive falsificationism is that one should adopt a more sophisticated version of the view, instead of abandoning the project altogether.

Popper’s philosophy of science is a product of his view on the problem of induction, so let’s talk briefly about that. The problem of induction, made famous by David Hume in the 18th century (though it had been noticed before), is about how we can ever be justified in believing in the kind of empirical generalizations that figure prominently in science. For instance, take Newton’s law of universal gravitation, which says that every body exerts a force of attraction on every other body proportional to the product of their masses and inversely proportion to the square of the distance between them. Unlike the claim that every body either has a mass or does not have a mass, which is a truth of logic and can be known without resorting to observation, the law of universal gravitation cannot be deduced logically from self-evident principles and can only be established by observation. We look at the world and, every time we set out to check whether the law holds, it appears to do so, so we eventually come to believe that it’s true.

This type of inference, by which a generalization is established by repeated observation of instances of it, is known as induction. But it’s really not clear why one is justified in making that kind of inference. In the case of deduction, there is a sense in which the truth of the premises necessitates the truth of the conclusion, but that is not the case with induction. Even if the law of universal gravitation was observed to hold a thousand times in the past, it doesn’t mean that it will still hold the next time we check. It seems that, in order to make that inference, we need to assume some kind of uniformity in nature. But it also seems that we could only justify the belief in the uniformity of nature by induction, which is precisely the type of inference the principle of uniformity was supposed to justify, so we can’t do that on the pain of making a viciously circular argument. My favorite summary of the problem of induction was given by Bertrand Russell, the famous British philosopher and mathematician, in The Problems of Philosophy:

We know that all these rather crude expectations of uniformity are liable to be misleading. The man who has fed the chicken every day throughout its life at last wrings its neck instead, showing that more refined views as to the uniformity of nature would have been useful to the chicken.

Although we are more sophisticated than poultry, by assuming that nature is uniform in the relevant way, we may put ourselves in a similarly precarious epistemic position as the chicken in Russell’s story.

Most philosophers assume that the problem of induction must have a solution, and that empirical generalizations can be confirmed by observational evidence (where confirmation is a matter of degrees), although they disagree on what the correct account of confirmation is. On this view, while such generalizations can never be definitely proven, they can be more or less supported by the evidence and sometimes we can be reasonably certain they are true or at least approximately true.  But Popper thought that the problem of induction was insoluble and that we must accept this fact. This means that a scientific theory can never be confirmed, but as Popper quickly noted, it can be falsified. In other words, while you can’t prove that a theory is true, you can at least show that it’s false.

Let’s go back to the law of universal gravitation. The problem of induction means that, no matter how often it has been observed to hold, it will never be proven. But all it would take to show that it’s false is just one counter-example. That’s because in order to show that a generalization is false, you just need to show that one of its instances is false. If I tell you that every cat is black, in order to show that it’s false, all you have to do is find a cat that is not black. On the other hand, even if all the cats you have seen so far have been black, there could always be another cat that is not black, so you can’t conclude that every cat is black no matter how many cats you have seen which seemed to confirm this generalization.

The problem of induction and this asymmetry between confirmation and falsification lie at the heart of Popper’s philosophy of science. For him, you can never show that a theory is true, but you can show that it’s false. So the way science works, according to Popper, is by conjuring up hypotheses such as the law of universal gravitation that can explain a whole array of phenomena and subjecting them to tests, whose aim is not to confirm them but to falsify them. On this view, science is essentially a deductive enterprise, not an inductive one. You come up with a theory, use deduction to derive observational consequences from it and perform experiments or make observations to check whether the results correspond to what your theory predicted. If they don’t, it means that your theory is false and you must look for another one, which is how scientific progress happens. A theory that can’t possibly be falsified, because it’s impossible to derive any observational predictions from it, is not scientific since it could never be shown to be false. Such a theory would be compatible with any observation whatsoever, which may be fine in some contexts (such as theology), but not in science, at least according to Popper.

What I have just described is, in a nutshell, falsificationism. Hopefully, at this point, you are feeling the appeal of that view. Perhaps you are even ready to officially embrace Popper’s philosophy of science. If you are, however, I recommend that you wait a little and keep reading, because I’m about to explain why most philosophers of science think falsificationism, despite its appeal, is actually mistaken. In order to understand why, we must take a closer look at what people actually do when they test a theory. Let’s go back once again to Newton’s law of universal gravitation and think about how one would go to test it. On its own, the law does not have any observable consequences, because it just says something about the forces that exist in the world and forces are not observable. You can’t see the gravitational force one physical object exerts on another. What you can see is only the effect it has on the motion of that object.

So you need something that provides the link between the forces that act on a physical object and its motion. This is precisely what Newton’s second law of motion does. This law of motion, expressed mathematically as F = ma, says that the total force on a physical object is equal to the product of its mass and its acceleration. Since the acceleration is the second derivative of position with respect to time,  it connects something unobservable, i. e. the total force on the object, to something observable, i. e. the position of this object at different times. In order to use to test the law of universal gravitation with the help of Newton’s second law of motion, however, you also need to know something about the masses of the physical objects, as well as their position at different times. This requires using various instruments to measure position, mass and time, which in turn means that you have to make several assumptions about the way in which those instruments work, guaranteeing they are reliable.

Thus, in order to test even a theory as straightforward as Newton’s law of universal gravitation, you need to make a lot of auxiliary hypotheses. The law of universal gravitation by itself doesn’t make any observable prediction. This isn’t just true of the law of universal gravitation, it’s true of any theory whatsoever. Just take any theory you’d like and think about how you’d go about to test it and you’ll soon realize that, in order to so, you need to make a lot of hypotheses that are not part of the theory itself. In general, a theory is never testable on its own, but only with the help of various auxiliary hypotheses or background assumptions. This means that a theory is never falsifiable simpliciter, but only relative to a set of background assumptions. Therefore, if we say that a theory is only scientific if it’s falsifiable, then it follows that no theory, not even a theory as successful as Newton’s law of universal gravitation, is scientific. Of course, this is absurd, so falsificationism is false.

But this was a bit quick and perhaps you are not entirely convinced yet, so let’s continue to examine the implications of the fact that a theory is only falsifiable relative to a set of background assumptions. Suppose that you derive a prediction from the law of universal gravitation, plus a bunch of auxiliary hypotheses, but it doesn’t come true. From a purely logical point of view, the only conclusion you can draw from this is at least one of the hypotheses you used to derive the prediction is false, but logic doesn’t tell you which one. It could be the law of universal gravitation, but it could also be one or several of the various auxiliary hypotheses you had to make in order to derive the prediction, you just don’t know. Strictly speaking, since the failure of a prediction derived from a theory plus a bunch of background assumptions doesn’t logically imply the falsity of the theory, it can’t falsify it. The failure of the prediction only falsifies the theory if you assume that the background assumptions necessary to derive that prediction are true.

In practice, although they frequently pay lip service to falsificationism when they talk about methodology, scientists are perfectly aware of that and don’t behave at all like they should according to naive falsificationism. Indeed, when a prediction derived from a theory they take to be well-established fails, scientists don’t just throw away the theory and start looking for a replacement. Instead, they generally assume that one of the background assumptions necessary to derive the prediction,  which are often left implicit, was false and try to figure out which one it was. Moreover, it’s not just that, as a matter of fact, scientists don’t behave as naive falsificationism implies they should, they are typically right not to do so.

After Newton formulated it, it took more than a hundred years before someone tested the law of universal gravitation experimentally, which incidentally shows that scientists are also not as obsessed with trying to falsify theories as Popper claimed they should. (Instead, they spend a lot of time devising methods to apply them to new problems, which is usually not straightforward and requires much ingenuity.) In fact, the purpose of the experiment wasn’t even to test the law of universal gravitation, but rather to estimate the mass of the earth. Suppose that, contrary to what actually happened, Cavendish’s experiment had yielded a result that was qualitatively incompatible with the law of universal gravitation. The reaction of physicists would certainly not have been to reject Newton’s theory of gravitation and start looking for another one.

Indeed, by the time Cavendish performed his experiment, Newton’s theory of gravitation had already been immensely successful. It had been used to explain a great variety of phenomena, from the motion of planets to that of projectiles on earth, but also the tides and many other things. It would have been completely irrational to throw such a successful theory away just because the experiment didn’t go as planned. The right thing to do would have been to assume some of the background assumptions was false, find which one and revise the experimental setup accordingly. For instance, perhaps the spheres used in the experiment were electrically charged, which according to Coulomb’s law, the electrostatic analogue to Newton’s law of universal gravitation, would have introduced another force that could have interfered with gravitation.

In general, when a prediction derived using a well-established theory fails, the rational thing to do is not to abandon that theory and start looking for another one. Doing so would considerably slow down scientific progress and could even stop it altogether. One thing Thomas Kuhn got right is that scientists only reject a theory they take to be well-established in the face of contrary experimental evidence if they have a viable alternative. (Kuhn was basically right about the sociology of theory change, but he drew crazy metaphysical/epistemological conclusions from that. Unfortunately, it’s the metaphysical/epistemological conclusions he drew that became popular, at least among non-philosophers. But this is a story for another day, so back to Popper and falsificationism.) It’s ironic that, although many scientists accept falsificationism uncritically, what they do every day is radically at odds with it.

That is not to say, of course, that it’s always a good thing that scientists operate in that way. One could argue that such a conservative bias toward currently accepted theory sometimes impedes the development of more satisfactory alternatives. For instance, many people think that, if economists initially dismissed empirical research suggesting that increasing the minimum wage might not reduce employment, it’s because they were committed to standard economic theory which predicts that increasing the price of labor will reduce employment since it assumes that demand curves in general are downward-slopping. Now, regardless of whether this particular instance of resistance to apparent falsification really is irrational, it certainly could be. So the point here is not that it’s never bad to stick to a theory even in the face of contrary evidence, but only that, before you throw away a theory that has served you good up until now, it’s perfectly rational to first examine whether the problem may not lay with one of the auxiliary hypotheses you used to derive the prediction that didn’t come true.

From a purely logical point of view, what scientists do when they blame the failure of a prediction on the background assumptions instead of theory they take to be well-established is no different from what charlatans peddling pseudoscience do when they conjure up bizarre hypotheses to explain away evidence that seem to contradict their theories, as when creationists posit that God planted fossils that seem very ancient to test our faith. The difference is that, when pseudoscientists do that, the hypotheses they conjure up are purely ad hoc and meant to protect theories that were never well-established in the first place and lack the kind of theoretical virtues that good theories have, but there is no purely formal criterion that allows you to decide that in a simple way. This is why the search for such a criterion of demarcation between science and non-science, whether in terms of falsifiability or not, is probably hopeless. If you want to argue that something is pseudoscience, there is no shortcut that will save you the pain of having to engage with it and grapple with the arguments of its proponents.

EDIT: I wrote a follow-up to this post in which I clarify a few things and reply to some critics. Due to the unexpected popularity of this post, I have already spent way too much time discussing falsificationism in the past few days, so I don’t plan on replying to comments either here or elsewhere. But the comments are open and, of course, you should feel free to criticize it. I apologize if you asked me a question or criticized my post and I didn’t reply to you, but like I said the response has been pretty overwhelming and I simply don’t have time.

47 thoughts

  1. Thanks Philippe for taking the time and effort to write this. For someone who only had a very slight exposure to philosophy of science, this is hugely informative.

    Now that I’m done with the praise, I can start the vicious attacks 🙂

    It appears to me that there’s a very significant difference between saying “falsifiability is the necessary and sufficient (or only necessary) condition for a discipline being called an science” and saying “scientists are pure deducters and zero inducters; if any experiment doesn’t comply with theory, they should jettison the theory without a thought”.

    My perception is that (i) scientists mostly work by induction, (ii) if an experiment fails they will first think they made some mistake before abandoning the theory, (iii) if they can’t find a mistake, the “failed” experiment is repeated consistently by them and others, they will start wondering if the theory is wrong or incomplete, (iv) the very fact that the theory can be falsified makes it scientific.

    I understand that you mention in the beginning of your post that there’s a “crude” falsificationism and that some philosophers believe that more sophisticated approaches “work”. So maybe you should count me in this camp.

    But I believe my objection goes deeper. It relies on the fundamental confusion between a criteria and a modus operandi.

    Happy to pursue this discussion 🙂

    Xavier

    1. Hi Xavier, I think critics would say the problem with the points you’ve written is that you haven’t demarcated science from pseudoscience. An astrologer or creationist can agree with all your 4 points and remain confident that what they do is indeed science. Your 4th point “the very fact that the theory can be falsified makes it scientific.” Strongly implies the crude version of falsification where it is both necessary and sufficient for determining what is science.

      1. Hi Fred. Not sure an astrologer or creationist would follow my (iii)

        Anyway, my point is about the confusion between criteria and modus operandi. My 4 points are here to illustrate a modus operandi that would not strictly follow deduction/falsification at every step.

        I’ll make a (necessarily imperfect) analogy with business. The ultimate criteria is whether customers purchase or not. Yet most of product development time is spent NOT asking customers what they want. You make observations, then make assumptions, then design a prototype, then build it, then test it, then show it to potential customers, then make observations… As Henry Ford famously said, customers would have said they wanted faster horses. Or Steve Jobs : “it’s not the customer’s job to know what he wants”.

        Another analogy: in his books, Taleb says that the Romans had a method for ensuring that bridges were solid. When the bridge was finished, they would make the engineer sit below it while a troop marched upon it. If the construction was flawed, the bridge would collapse and the engineer die. Like falsifiability, troop marching is the reality check. You don’t build the bridge by having troops marching on every beam at every step. But you know that in the end you will be tested that way. And that influences the whole process.

  2. Good read and very well written! That view of the importance of the auxiliary hypotheses also accounts for why it is so difficult to change one’s mind on a political subject – because the opinion rests on a nest of auxiliary premises that have so far had (a perceived) explanatory value.

  3. Science is the sacrifice to make to see a change. Gravity is not scientific, because when I throw an object in the air, I have to wait to see it fall, I can not do anything else. The same goes for psychology. Conversely, Newton’s theory tells us the sacrifice to be made (calculate the trajectory) to know where the object will fall. While I can not make a sacrifice to know if such person is angry or greedy.

    ​​Popers was right. Change is a before (a cause) and an after (an effect), but the sacrifice can not completely fill the void between before and after. There are therefore inevitably cases where the theory (sacrifice) gives false results. This is in fact what describes the paradox of Achilles and the turtle (Zenon).

    You may not see the relationship, but what is wrong in what you say is that the universe is not a divine thing that would exist independently of men, but what we can do (regardless of the universe). Only a God could let us know things that we could not verify (read Aristotle), so the universe is only what we can verify. Moreover, we can not observe what we have never checked (see where the object will fall according to the calculation, before having made the calculation). A theory is therefore always wrong somewhere, because it confronts an imaginary relation between things and what we can verify. In the case of Achilles and the tortoise, we confront a mathematical model coming from the imagination, and what man is able to observe by himself or by tools (hence the uncertainty principle).

    Popers might be right, because if we can imagine what the theory can not fill, it is because it induces a sacrifice. Except, that to fill it we have to wait for someone to imagine something else, like Einstein who discovered the “falsification” of Newton’s theorem. It is therefore Einstein who has demonstrated (for Popers) that Newton’s theory is scientific.

    La science est le sacrifice à faire pour voir un changement. La pesanteur n’a rien de scientifique, car quand je lance un objet en l’air, je dois attendre de le voir retomber, je ne peux rien faire d’autre. Il en va de même pour la psychologie. A contrario, La théorie de Newton nous dit le sacrifice à faire (calculer la trajectoire) pour savoir où va tomber l’objet. Alors que je ne peux pas faire de sacrifice pour savoir si telle personne est colérique ou gourmande.

    Popers avait raison. Le changement est un avant (une cause) et un après (un effet), mais le sacrifice ne peut pas combler entièrement le vide entre l’avant et l’après. Il y a donc forcément des cas où la théorie (le sacrifice) donne des résultats faux. C’est en fait ce que décrit le paradoxe d’Achille et de la tortue.

    Vous ne voyez peut-être pas le rapport, mais ce qui est faux dans ce que vous dites, c’est que l’univers n’est pas une chose divine qui existerait indépendamment de l’homme, mais ce que nous pouvons faire (indépendamment de l’univers). Seul un Dieu pourrait nous faire connaître des choses que nous ne pourrions pas vérifier, donc l’univers n’est que ce que nous pouvons vérifier. Par ailleurs, nous ne pouvons pas observer ce que nous n’avons jamais vérifier (voir où va tomber l’objet selon le calcul, avant d’avoir fait le calcul). Une théorie est donc toujours fausse quelque part, car elle confronte une relation imaginaire entre des choses et ce que nous pouvons vérifier. Dans le cas d’Achille et de la tortue, nous confrontons un modèle mathématique venant de l’imagination, et ce que l’homme est capable d’observer par lui-même ou par des outils (d’où le principe d’incertitude).

    Popers pourrait avoir raison, car si nous pouvons imaginer ce que ne peut pas combler la théorie, c’est qu’elle induit un sacrifice. Sauf, que pour le combler il faut attendre que quelqu’un imagine autre chose, comme Einstein qui a découvert la “falsification” du théorème de Newton. C’est donc Einstein qui a démontré (pour Popers) que la théorie de Newton est scientifique.

  4. You take Popper at his word re: falsification and induction, but I completely disagree with the premise. First, for falsification to work, you have to expect that the falsifying experiment which worked today will also work always in the future. This is induction! Second, any subsequent application of the falsifiable theory is obviously inductive in nature. Popper wants falsification to magic away the problem of induction but it absolutely does not.

    At the same time I don’t think the issue of auxiliary hypotheses is necessarily fatal to falsification, because in practice scientists don’t just come up with ad hoc hypotheses. When Newton’s predictions for the orbit of Saturn failed, scientists predicted additional planets further out (and thus discovered Neptune and Uranus). You might say: this is an example of how auxiliary hypotheses undermine falsification! But in practice scientists proceed by eliminating more and more options. The theoretical possibility of adding on mountains of ad hoc auxiliary doesn’t seem particularly relevant. Popper anticipates your argument in his book:

    “It might be said that even if the asymmetry is admitted, it is still impossible, for various reasons, that any theoretical system should ever be conclusively falsified. For it is always possible to find some way of evading falsification, for example by introducing ad hoc an auxiliary hypothesis, or by changing ad hoc a definition. […] I must admit the justice of this criticism; but I need not therefore withdraw my proposal to adopt falsifiability as a criterion of demarcation. For I am going to propose (in sections 20 f.) that the empirical method shall be characterized as a method that excludes precisely those ways of evading falsification which, as my imaginary critic rightly insists, are logically possible. According to my proposal, what characterizes the empirical method is its manner of exposing to falsification, in every conceivable way, the system to be tested.”

    Also, Kuhn didn’t really draw the crazy conclusions of his (crazier) followers, and indeed explicitly warns _against_ them in the book.

    1. » Alvaro de Menard:
      “You take Popper at his word re: falsification and induction, but I completely disagree with the premise. First, for falsification to work, you have to expect that the falsifying experiment which worked today will also work always in the future. This is induction!”

      Why would you say that Philippe takes Popper at his word when a) he doesn’t quote a single thing Popper actually said and b) he explicitly says that what Popper said is not what he is talking about here but what he calls “crude falsificationism” (i.e. “not falsificationism”)?

      Also, “falsification” doesn’t mean what you think it means. (See especially section 22 of LoSD.) It has, of course, nothing to do with induction.

  5. This is a good essay. I agree with your premise that all experiments jointly test a theory and some ancillary assumptions, and that it is difficult to tell whether to reject the theory or the ancillary assumptions.

    But I arrived as a falsificationist, and I leave the same. It never occurred to me that anyone considered falsification to be a perfect arbiter of science vs. pseudoscience. This seems like a straw man. It is not that surprising that we cannot cleanly differentiate science from pseudoscience any more than we can differentiate easterners from westerners or fruits from vegetables. Perhaps Popper did hold such an extreme view, but I don’t think that is what drives how most scientists think about science.

    As a researcher, I am almost entirely exclusively interested in falsifiable hypotheses and developing the means to test them. This is a tremendously useful guide to thinking about empirical work and thinking about whether theories are seriously supported by empirical data or unfalsifiable ideologies.

    1. Actually, his argument is this:

      premise: Naive falsificationism (a description of how scientific progress occurs) is false.
      conclusion: falsifiability (a predicate applied to scientific claims) is useless.
      corollary: there is no clear demarcation between pseudo-science and science.

      It’s a Socratic bait-and-switch, and just as infuriating. I’ll give him points for noticing the similarity in spelling, but I won’t accept his conclusion based on it.

      Not only that, but it’s not clear that the personal beliefs of scientists are all that relevant to scientific progress, any more than a baseball player’s philosophy of baseball matters for anything – whether ball players are good or not is judged on batting average and RBIs, not their sometimes nutty ideas on how they achieved those accomplishments.

      1. You take on thing and you put it near another one. You can never know if you could do something with this new thing before you try. We are only monkeys with better hands, nothing more. So to understand something you need to have already put these things togeteher and try to do something with them. Then, I can tell you the difference bewteen science and non-science which depends of the things you will put together.

  6. » “his goal was to find a purely logical criterion of demarcation between science and non-science”

    Let’s contrast this with what Popper actually wrote, shall we? Section 9 of The Logic of Scientific Discovery: “If therefore we characterize empirical science merely by the formal or logical structure of its statements, we shall not be able to exclude from it that prevalent form of metaphysics which results from elevating an obsolete scientific theory into an incontrovertible truth.”

    Section 20 of LoDS: “Only with reference to the methods applied to a theoretical system is it at all possible to ask whether we are dealing with a conventionalist or an empirical theory.” (Italics in the original.)

    My first question would consequently be: Why would you, somebody who is active in academic philosophy, write about something being “false” that you have apparently not even read, let alone studied?

    My second question would be: Why would you think that what Popper actually wrote is so inconsequential to the topic that there is no need to take it seriously and at least read it? Actually, not just read it but quote it, so that a) your readers can make up their own minds about it and b) you might realize that you were passing something off as Popper’s philosophy that in fact has nothing to do with it.

    I do realise that these question will initially seem rather antagonistic, but they are actually honest questions. Obviously, they mean that I am highly critical of your post, which I think is false in almost every detail you ascribe to Popper and his philosophy; but what’s important, in my book, is only that this is an opportunity to learn something (for everybody involved).

    1. If you want to show that Popper was aware of the problem that Duhem-Quine thesis posed for his criterion of demarcation, you don’t even need to go as far into The logic of scientific discovery as section 9, here is what he says in section 6:

      A third objection may seem more serious. It might be said that even if the asymmetry is admitted, it is still impossible, for various reasons, that any theoretical system should ever be conclusively falsified. For it is always possible to find some way of evading falsification, for example by introducing ad hoc an auxiliary hypothesis, or by changing ad hoc a definition. It is even possible without logical inconsistency to adopt the position of simply refusing to acknowledge any falsifying experience whatsoever. Admittedly, scientists do not usually proceed in this way, but logically such procedure is possible; and this fact, it might be claimed, makes the logical value of my proposed criterion of demarcation dubious, to say the least.

      I must admit the justice of this criticism; but I need not therefore withdraw my proposal to adopt falsifiability as a criterion of demarca- tion. For I am going to propose (in sections 20 f.) that the empirical method shall be characterized as a method that excludes precisely those ways of evading falsification which, as my imaginary critic rightly insists, are logically possible. According to my proposal, what charac- terizes the empirical method is its manner of exposing to falsification, in every conceivable way, the system to be tested. Its aim is not to save the lives of untenable systems but, on the contrary, to select the one which is by comparison the fittest, by exposing them all to the fiercest struggle for survival.

      See, I have also read Popper, but as I say pretty clearly in the post, it’s not the more sophisticated versions of falsificationism proposed by Popper and his followers I set out to refute, but rather “the crude version that most people, especially scientists, have in mind when they talk about falsificationism”. Since it’s this crude version of the view that most people believe and not one of the more sophisticated versions, as the comments I got on Twitter and elsewhere have clearly demonstrated again, this is the version of the view I thought it was most important to criticize in a way that can be understood by everyone.

      As I also say in the post, I think more sophisticated versions of falsificationism should also be rejected, because ultimately Popper was just wrong about induction and the more sophisticated kind of falsificationism he defended was motivated by his view about induction. But this is not something I can reasonably be expected to argue in a blog post that I specifically pitched at non-specialists. This is why I explicitly noted that people who hold a more sophisticated version of falsificationism than the crude version I criticized in that post would “predictably not be moved” by my arguments and that I was fine with that.

      Now, one thing I will concede is that it was uncharitable of me to say that Popper’s goal was “to find a purely logical criterion of demarcation between science and non-science”, because you’re right that in many passages he also emphasized the role of people’s attitudes toward a system, even though he wasn’t always as clear on this as in The logic of scientific discovery and apparently even endorsed the crude version of the view in print on at least one occasion, according to Sven Ove Hansson’s entry in the SEP’s entry on Science and Pseudo-Science:

      However, in what seems to be his last statement of his position, Popper declared that falsifiability is a both necessary and a sufficient criterion. “A sentence (or a theory) is empirical-scientific if and only if it is falsifiable.” Furthermore, he emphasized that the falsifiability referred to here “only has to do with the logical structure of sentences and classes of sentences” (Popper [1989] 1994, 82). A (theoretical) sentence, he says, is falsifiable if and only if it logically contradicts some (empirical) sentence that describes a logically possible event that it would be logically possible to observe (Popper [1989] 1994, 83). A statement can be falsifiable in this sense although it is not in practice possible to falsify it. It would seem to follow from this interpretation that a statement’s status as scientific or non-scientific does not shift with time. On previous occasions he seems to have interpreted falsifiability differently, and maintained that “what was a metaphysical idea yesterday can become a testable scientific theory tomorrow; and this happens frequently” (Popper 1974, 981, cf. 984).

      So it may be partly his fault if what most people have in mind when they talk about falsificationism is the crude view I’m criticizing rather than the more sophisticated view Popper usually defended in print, although I agree with you that philosophers of science have often been unfair to him, even if I think ultimately they were right that his philosophy of science is not correct. Indeed, I have linked on Twitter in the same thread where I shared this post to a paper by someone who defends Popper against what he claims are misrepresentations of his actual views, so it’s not as if I didn’t want people to know about that. But again my target in this post was the popular version of falsificationism and not the more sophisticated views that have been defended by Popper and his followers.

      In any case, I have no desire to debate this with you, because I don’t have time and also because, despite what you claim, I don’t believe for a second that you are really interested in learning anything. Instead, the fact that you were unnecessarily aggressive from the outset, that you have a blog called “The Open Society” (named after Popper’s famous, though very controversial book) and that you obsessively tracked down every conversation I’ve had on Twitter about this post to “correct the record” and explain that everything I say in this post is completely wrong in rather unpleasant terms, despite the fact that I had clearly no interest in engaging with you since I never replied, strongly suggests to me that you’re more like a cult member who is only interested in shutting down heretics who dare criticize your hero and that a debate with you would be a waste of time. So this will be my last reply to you and I will let you have the last word.

      1. Anybody seriously interested in an intellectually fruitful discussion would, upon being confronted with the assertion that “pretty much all of that is plainly wrong”, be tempted to reply: “How so? Can you explain that?” Especially if the critic in the very next sentence offers to do just that.

        And since you didn’t link to it, let’s let your readers decide whether this is “unnecessarily aggressive”:

        Pretty much all of that is plainly wrong, to be frank. I’d be curious to know whether you’d be interested to find out where and why—a public discussion might be instructive to just the audience you wrote for. If you’re up for having your mind changed, that is. 🙂

        If you think that is “aggressive”, then you are obviously not used to receiving any serious criticism at all. Which is really rather unfortunate for someone in philosophy.

        Also, nobody who cares about intellectually fruitful discussion will hide behind a tone troll—the lamest dodge of criticism imaginable. And inferring from the name of my blog and my replying to different aspects of your tweets (and explicitly inviting a response) that I am “more like a cult member” who is “interested in shutting down heretics”? Dude, whatever you’re smoking really isn’t good for you. 😀

      2. » Philippe:
        “[Popper] apparently even endorsed the crude version of the view in print on at least one occasion, according to Sven Ove Hansson’s entry in the SEP’s entry on Science and Pseudo-Science”

        Incidentally, did you notice that Hansson doesn’t even cite the one work in which Popper first and in meticulous detail laid out his view of the criterion of falsifiability, namely The Logic of Scientific Discovery? That would make anyone familiar with the topic suspicious enough. Worse, for some ideas that are allegedly central to Popper’s views, Hansson offers only obscure secondary sources.

        Hansson then goes on to ignore every bit of context needed to properly understand the few snippets from Popper himself that he quotes as evidence for his silly misrepresentation of “falsifiability”—hardly surprising as he ignores LoSD, where all this was explained in detail. Thus, Hansson completely misses the rather important distinction between the logical criterion of demarcation and the additionally necessary methodological rules that need to be put in place to prevent the criterion’s circumvention (cf. especially LoSD, § 20).

        Consequently, Hansson’s quote that you take as an instance of Popper espousing your “crude” version of falsificationism appears in Seiffert/Radnitzky explicitly only in the context of the purely logical aspect of falsifiability. He cites himself, Agassi, and Laudan in support of a point that can easily be seen as a misrepresentation of Popper’s actual ideas which neglects both the logical/methodological distinction and additionally the idea of degrees of falsifiability—again, something that is explained at length in the book Hansson ignores (cf. LoSD, ch. 6).

        For an academic philosopher, that is embarrassingly shoddy work. It is, however, par for the course in Popper criticism. As a rule of thumb, one should never rely on secondary sources about Popper, unless they properly quote the relevant original works and explain the context in which they were meant to be read.

  7. It has always seemed to me that Popper’s falsificationism was a development against someone and that the provisional truth that it implies must lead us to inaction and accept the current state of things. Would it be a political motivation?

  8. Just a couple of questions:

    What is the most-widely held view by philosophers of science?

    Whatever the criteria for demarcating science from pseudo-science, is falsifiability among them? Is falsifiability a necessary albeit not sufficient criterion? I ask because some statements I read by Richard Dawid seem to question this.

  9. While naive falsificationism is clearly wrong, I’m not sure that one has to make it all that sophisticated for it to become much less susceptible to the kinds of criticisms you make. For instance, what if we require of a theory merely that its proponents can explain what kind of experimental evidence would cause them to think that it would need to be significantly modified? A single failed experiment would not be enough, for the reasons you explain, but a consistent pattern of experiments that were inconsistent with the theory unless one repeatedly questioned background assumptions (such as suggesting that a wide variety of measuring devices all gave the same wrong answer) would count as a good reason.

    If, for example, you ask a believer in the power of prayer what would cause them to abandon that belief, you will not normally get an answer of that kind. If a hundred people in a church all pray fervently for a member of the congregation to recover from an illness and that person dies a month later, that will typically not have any effect on people’s religious beliefs, and one would be strongly criticized for suggesting that it should. And the same goes for any experiment, or indeed series of experiments, that one might propose. So that’s a good indication that belief in the power of prayer is unscientific. (That does not necessarily mean it is a bad thing — merely that it is a genuinely different kind of belief from belief in a well-established scientific theory.)

    For a slightly more sophisticated version still, which isn’t exactly falsificationism but is quite close under many circumstances, one could go for something with a more Bayesian style. That is, one could challenge proponents of a theory not to say what would cause them to decide that it is unsatisfactory, but merely what would have the effect of lessening their degree of belief in it.

    1. Just to add, I know you talked about more sophisticated falsificationism in the post, so I should clarify that the point I am trying to make is that one doesn’t have to modify it too much to deal with your criticisms. So it’s arguable that Popper is getting at a genuine distinction between science and pseudoscience, even if articulating precisely what that distinction is is not as easy as many people think.

  10. “Thus, in order to test even a theory as straightforward as Newton’s law of universal gravitation, you need to make a lot of auxiliary hypotheses. The law of universal gravitation by itself doesn’t make any observable prediction. This isn’t just true of the law of universal gravitation, it’s true of any theory whatsoever. Just take any theory you’d like and think about how you’d go about to test it and you’ll soon realize that, in order to so, you need to make a lot of hypotheses that are not part of the theory itself. In general, a theory is never testable on its own, but only with the help of various auxiliary hypotheses or background assumptions. This means that a theory is never falsifiable simpliciter, but only relative to a set of background assumptions. Therefore, if we say that a theory is only scientific if it’s falsifiable, then it follows that no theory, not even a theory as successful as Newton’s law of universal gravitation, is scientific. Of course, this is absurd, so falsificationism is false.”

    Didn’t get that part. Why on earth would it be a problem to use a theory to falsify another theory, as long as the theory you are using hasn’t been falisified yet? The point you are making is very curious indeed: similarly, would you agree that, since we need words to define words, definition is shit? “A theory is never falsificable simpliciter” yep so what? It won’t alter the fact it is falsificable (even if not “simpliciter”, who cares?). If you think the problem is Popper’s relying on uncertain theories to falsify a theory, then why don’t you go through every single demonstration of every single theorem you use in math throughout your own demonstration?

    Funny thing is, you “refute” Popper in a most pompous way without quoting him. “As regards auxiliary hypotheses we propose to lay down the rule that only those are acceptable whose introduction does not diminish the degree of falsifiability or testability of the system in question, but, on the contrary, increases it.” (The Logic of Scientific Discovery, 20). I think it clearly states that you oversimplify Popper to so to speak “refute” him. That’s not how that works.

    Btw. I don’t see what Bryan Caplan has to do with the whole thing. I like him but from a methodological/epistemological point of view, his amateurism beats everything.

    1. » H³:
      “Funny thing is, you “refute” Popper in a most pompous way without quoting him.”

      But, you see, Philippe did say he didn’t refute Popper but only a “crude version of falsificationism” (which has nothing to do with Popper or, in fact, falsificationism).

      Just don’t pay any attention to all the instances where he talks about “Popper’s conception of science”, about “what Popper himself believed”, about “Popper’s philosophy of science”, about what “Popper thought”, about what “Popper quickly noted”, about “the heart of Popper’s philosophy of science”, about “the way science works, according to Popper”, about what is and isn’t fine “in science, at least according to Popper”, and about what “Popper claimed”. Because that’s not what the post is about at all.

  11. Interesting. However, you lost me here:

    “Thus, in order to test even a theory as straightforward as Newton’s law of universal gravitation, you need to make a lot of auxiliary hypotheses.”

    My field is climate science. Suppose someone says “I have a new theory about how climate works, and it predicts that temperatures next March will average over 80°F in a certain area, measured at these three weather stations.”

    Now, when March is past, either the temperature did or didn’t average over 80°. If it was not over 80° we can say that his theory is wrong. It doesn’t tell us where the theory is wrong, but it does let us say that his hypothesis has been falsified.

    Where are the “lot of auxiliary hypotheses” that I “need to make” in order to test his prediction?

    Next, you say that in response to something being falsified, we often don’t just throw it out, viz:

    “When a prediction derived from a theory they take to be well-established fails, scientists don’t just throw away the theory and start looking for a replacement.” So instead, they might modify the theory.

    … but IT IS STILL FALSIFIED. You seem to think that if we don’t “throw away the theory”, that somehow it’s not falsified and Popper is wrong … but in fact, you’re just discussing the response to the falsification. And modifying a theory is really just throwing out one falsified theory and replacing it with a new theory.

    For example, we have the rule about English spelling that says ” ‘I’ before ‘E’ “.

    However, we can test that. We look around. We see the word “deceive”. At this point OUR RULE IS FALSIFIED. And I see no “auxiliary hypothesis” involved.

    So we do what some folks call “modifying” the rule, but in fact we create a brand new rule—”I before E, except after C”.

    The same is science. The theory gets falsified. So we “modify” it, meaning in fact that we throw it out because it was falsified and we propose a new, similar rule that prevents the falsification.

    Next, you talk about the falsification of “a theory they take to be well-established” … but usually, we’re applying falsification to new, untried theories, not Newton’s Laws …

    Finally, for me, the trick is to design your experiment or state your prediction so that it is NOT dependent on a “lot of auxiliary hypotheses”. Not easy, to be sure, but absolutely doable.

    Thanks for posting on an interesting subject.

    w.

    PS—The title of this post shows that you don’t really believe what you are saying, even if the headline is in jest … because according to you, there is no way for anyone to ever show that anything is false. Seems unnecessarily solipsistic, but hey, that’s just me …

  12. It’s really nice to see material about philosophy of science get attention on social media – well done! I think us researchers would benefit from knowing more phil of science.

    That said, Popper himself wrote pretty extensively about the problem of auxiliary hypotheses and how it might be resolved; there’s a lot more to his theory than you’ve presented, as you know. Do *some* scientists believe in a particularly naive version of Popper’s demarcation criterion? Sure, they probably do. But should they ditch Popper, or just develop a more sophisticated understanding of his ideas? I don’t think you can really answer that question if your focus is only on outgunning a strawman version of the theory. But maybe you’ll address this in subsequent posts! 🙂

  13. As a scientist myself (physicist and astrophysicist), I have always regarded this famous “falsifiability criterion” (attributed to Popper and religiously referred to by most scientists, to my surprise), as terribly naive, and terribly wrong!

    However, I must say that I was disappointed by this article. With due respect to the author, I find the arguments proposed here rather weak. The example with the law of gravitation does not demonstrate much, I think. The author says that one cannot falsify a theory such as Newton’s law of gravitation because, in order to test it experimentally, one needs to make a whole set of other assumptions about the second law of Newton, the way you define and measure masses, etc. Well, that’s pretty obvious! But then fine, you don’t test that particular theory on its own, isolated, you test it within a more general context. But for the advocates of falsifiability, what difference does it make? You can still falsify the theory in a given context, i.e. with respect to that context, and that’s good enough at that level. In any case, even to *formulate* the theory, you need a context! Be it only, ultimately, the context of langage!
    So this is really a lot of words for very trivial statements that do not demonstrate much, imho.

    But most of all, I think this article misses some more important arguments supporting the intended case.
    The author could have limited himself to simply say that, according to this falsifiability criterion, science itself, as an approach, is not scientific!
    How would it be so fundamentally crucial that one can refute a theory to label it as “science”, if one is not prepared to refute science altogether? Now, of course, when scientists do science, they are ready to abandon a theory if it is found not to give a fair or credible account of what we experience, but that will only be to search for a better theory that they will still accept as scientific. They will never say: “OK, science has failed on this particular point, so let’s move away from science”.
    When astrologers make a prediction based on the position of planets at a given time, or whatever, they are ready to accept that they were wrong if the prediction turns out to fail. But they will conclude that something was missing in their estimate, that they made a mistake in their calculation or reasoning, or that they did not take into account properly some other element, so they will formulate new hypothesis and perhaps enlarge their implicit theory, etc. This is a perfectly sound and “scientific” attitude. Why would they walk away from astrology just because they do not yet have the ultimate theory?

    In fact, this whole idea of trying to disqualify some approaches or world views based on such or such objective criterion seems to me almost ridiculous.

    The most important criterion, in my opinion, for one’s subscribing to a given corpus of knowledge or worldview, is whether a given line of explanation or description or, better, representation, of the world, satisfies us or not. Does it exhaust my questioning? At least partially. Does it appease my mind, heart and soul?
    That surely does not mean that all approaches are equally valid or defendable. Relativism as a moral value is, in my opinion, just as ridiculous.
    Especially since believing this or that can have some impact on the society, and imposing (or even just “respecting”) some views, doctrines and behaviours in accordance with them may have some undesirable effects or even terrible consequences from the human point of view, which I simply refuse to accept.
    But this is another issue altogether.
    From the general point of view, I think we should refrain from associating strongly connoted notions such as “values” to intellectual and emancipatory activities like science or philosophy.
    Unfortunately, the way (superficial) scientists or citizens use the falsifiability criterion to distribute ideas or actions between the “science” and “non-science” categories tends to be essentially moral. In the end, they use word “science” as a synonym of “Good”, as opposed to “Bad”. They pretend (and probably think sincerely) that they are referring to a method (science), rather than its productions, in order to get away from the slippery distinction between “truth” and “lie”, but I think they fail in that respect too.
    And anyway, who cares whether something is “scientific” or “not scientific”? Is that some sort of Imprimatur to be distributed by some magnanimous authority? Well, it should not be the case, in my opinion! That would be a terrible reduction of its actual value, as a human intellectual, cultural and spiritual activity.

    By the way, another reason why I never understood the worshipping of falsifiability, in addition to its being used essentially as a moral argument, is that it is actually very artificial and a posteriori. By this I mean that it is a false (or fake?) argument in the sense that nobody really cares about falsifiability when they do science! It is something which may acquire some value in the end, in retrospect, when the scientific activity has reached somewhere. But never during the activity itself!
    If we, scientists, were to consider falsifiability as a crucial criterion for science, then we should be somewhat preoccupied, because this is never what we do in practice. We don’t care about falsifiability when we develop a theory. We don’t ask ourselves whether it will be falsifiable or not in the end! We just try to make sense of something, in a way that we feel is satisfying (at least to us, at least partially).
    So it is certainly not a criterion for *doing* science. So again, how would it be so crucial for *being* science? Would it be that “doing science” is not “scientific”?

    Finally: in the last paragraph, the author writes, referring to pseudoscientists: “the hypotheses they conjure up are purely ad hoc and meant to protect theories that were never well-established in the first place and lack the kind of theoretical virtues that good theories have, but there is no purely formal criterion that allows you to decide that in a simple way.”

    First, it is funny to read here the term “pseudoscientist”, which the author does not define, as if it were self-evident what a “real scientist” and a “pseudoscientist” is. In an article which is precisely about the way (and possibility) to establish such distinction, that is somewhat peculiar, isn’t it? But most of all, I find it very presumptuous to judge that their hypothesis are “purely ad hoc”, that their theories “were never well-established in the first place”, and that they “lack the kind of theoretical virtues that good theories have”. Which kind? Isn’t that the whole point? And this tautology even sounds like a void statement, given the end of the sentence…

    Anyway, I hope the above will not be perceived as offensive by the author. I just wanted to share some thought in reaction to this blog, and actually support the main message of the author: stop worshipping “falsifiability”! Surely (well, I hope) anyone actually considering the question will have reached the same conclusion anyway. So let’s say that this message is essentially addressed to the scientists who keep using this mantra without thinking, as an element of pride, almost as a marker of identity (and superiority?), in a way that does not do credit to the actual value of science and philosophy.

    1. » Etienne Parizot:
      “I have always regarded this famous “falsifiability criterion” (attributed to Popper and religiously referred to by most scientists, to my surprise), as terribly naive, and terribly wrong!”

      Could it be that you have never actually read Popper either? What you take to be “falsifiability”, and which Philippe argued against in this post, is certainly terribly naive and terribly wrong. It also certainly has nothing to do with what Popper actually wrote. As a matter of fact, Popper explicitly dispelled all the terribly naive objections to his concept that Philippe’s post rehashes in the very book he wrote to promote “falsifiability”, The Logic of Scientific Discovery.

      1. >> Peter Monnerjahn:
        As you certainly understood, my message is to be read as a direct response to the initial post of Philippe Lemoine, with consideration of what most scientists believe Popper said regarding falsifiability as a criterion for scientificality (and _not_ as a comment of Popper’s actual philosophy and statements regarding falsifiability). The very sentence you extracted from my post makes it clear as well: I carefully wrote “*attributed* to Popper”, and “religiously referred to by most scientists”. So, no, this is not “what [I] take to be falsifiability”. The fact that this specific criterion that is referred to by most scientist is not what Popper ever wrote or thought is indeed an additional reason to be surprise of such a popularity. (This, I’m afraid, is not reflecting well on scientists! 🙁 )

  14. Thanks for an interesting read! I’m a Popper fan and I’m afraid my standpoint was not changed. If I understand right, the argument you are presenting can be summed up as: “but there are always other underlying assumptions” Perhaps I have missed something, but here are my thoughts:

    Yes, we do need to make the assumption that the instruments we use give accurate results. This can be counteracted by making sure multiple independent researchers observe the same result, preferably with different types of measurements.

    Yes, science builds upon itself and we assume previous theories hold when developing new ones based on them. Yes, if data is not in agreement with a proposed theory, it could be that actually it was something else (one of the underlying assumptions) which was wrong. For sure, we need to build science carefully.

    No, we do not discard an extremely well-established theory just because one experiment is not in agreement with it. This is exactly because of the reasons you state. Sometimes instruments give the wrong result or human error comes into play somehow in the experimentation. We must look at the total amount of evidence available and see if it is in general agreement with the theory or not.

    Overall I think you have raised valid concerns but they are not new and I don’t see how they are direct arguments against Popper’s idea of falsifiability. It sounds like a critique to science in general and the conclusion would be that science is not perfect. So I can totally agree with all of your points but I don’t see how they show that the specific ideas of Popper are wrong.

    Cheers,
    Andreas

    Some other minor points:

    I can only speak for what I have seen, but Popper’s philosophy is NOT taught to scientists in general and if you do take courses in the philosophy of science he is NOT at all presented as the foremost philosopher (rather the opposite).

    Your presentation on the idea of falsification focuses a lot on induction. While I sort of agree with what you say, I would say it is more about PREDICTABILITY of theories. A scientific theory makes predictions.

    The example with forces being impossible to measure and that only the resulting motion is observable is not correct. For instance, I can stand in strong wind and feel its force without being moved by it.

  15. Since I was so uncouth as to allege about this post’s claims about falsificationism that “pretty much all of that is plainly wrong”, it behoves me to back that allegation up with relevant facts and arguments—Philippe’s apparent lack of curiosity as to those reasons notwithstanding.

    First off, it is this post’s purported aim that is somewhat confusing. On the one hand, Philippe assures us that he is attacking “a crude version” of Popper’s philosophy that is not based on Popper’s actual words; on the other hand, on no less than ten occasions does he say that he is, after all, referring to something Popper said or thought. (One would have thought that an academic philosopher a) wouldn’t use a caricature of somebody’s position as a target of an intellectual disquisition and b) would use appropriate quotes on those latter occasions, but we’ll let that slide for the moment.) In any case, passing a “crude falsificationism” off as something that even remotely resembles what Popper said would be completely disingenuous—even if it’s done out of ignorance of Popper’s actual ideas.

    The confusion actually starts in the post’s title. It is, in fact, the criterion of demarcation (falsifiability) that Philippe thinks he can show to be false—not the methodology of falsificationism, which rests on falsifiability as one of its central building blocks, as well as on falsifiability’s corollaries and logical foundations.

    The first paragraph, at first blush, reads like a relatively accurate general description of Popper’s falsificationism. Unfortunately, it is more or less subtly wrong in every detail. First, the idea that “what distinguishes science from non-science is falsifiability” describes not falsificationism but a criterion of demarcation. Second, it is not potential observations that a theory can be inconsistent with statements of an empirical nature (its “potential falsifiers”). Third, a theory is not falsified because of any single observation; in fact, “a hypothesis is falsified by accepting a basic statement” that contradicts the theory (LoSD, § 22).

    The third point bears some further explaining. The logical “asymmetry between verifiability and falsifiability” (LoSD, § 6) means that however many empirical statements one accepts as true, no generalisation from those statements to a universal statement is logically valid (that would be induction); but as soon as one accepts as true a singular statement that contradicts a universal statement, one has to accept the falsity of the universal statement. This, it is crucial to notice, does not mean that we thereby find something to be certainly true or false; logic can at best (and only if it is deductive) force us to make a choice (cf. Notturno). It can force us, if we accept its rules, to “choose between the truth of some beliefs and the falsity of others ”. And that is how it is possible to learn. (It is not, contra Philippe, how it is possible to “show that [a theory is] false”.)

    And that is what Popper’s demarcation criterion of falsifiability actually means: “a theoretical system belongs to empirical science” (LoSD, § 19) if “the class of its potential falsifiers is not empty” (LoSD, § 21). It means that there have to be empirical statements whose falsity can be retransmitted to a theory for that theory to qualify as scientific. This in fact does, again contra Philippe, deliver “a purely logical criterion of demarcation”. Especially the whole discussion about auxiliary hypotheses is irrelevant in this regards, because the criterion only demands that the falsity of conclusions be able to be retransmitted to a theory—and the logic only demands that the falsity of conclusions be accepted, just as we can then (tentatively) accept certain initial conditions and auxiliary hypotheses as correct. Of course “logic doesn’t tell you which one” of the elements of a theoretical system is false; but logic can never do that, and falsifiability is not a criterion of truth but of demarcation.

    Philippe also thinks that his observation that “a theory is never falsifiable simpliciter”, i.e. without considering auxiliary hypotheses and initial conditions, is a simple refutation of Popper. He even references Newton’s theories in making this point more concrete: “The law of universal gravitation by itself doesn’t make any observable prediction.” Hence, Newtonian physics is unscientific according to Popper’s criterion, which would be “absurd, so falsificationism is false”. What makes this absurd, however, is that Popper himself gave the exact same example: “When writing this, I believed that it was plain enough that from Newton’s theory alone, without initial conditions, nothing of the nature of an observation statement can be deducible” (LoSD, § 28). And he goes on to explain that what is important is not that potential falsifiers be able to be deduced from the theory alone but that they exist: “if we are given a theory t and the initial conditions r, from which we deduce the prediction p, then the statement r·p̄ will be a falsifier of the theory”.

    How, then, does actual falsification work? Philippe thinks that “when a prediction derived from a theory they take to be well-established fails, scientists don’t just throw away the theory” is a relevant counter-argument to falsificationism. Well, of course they don’t, and Popper explicitly said they shouldn’t: “Once a hypothesis has been proposed and tested, and has proved its mettle, it may not be allowed to drop out without ‘good reason’.” (LoSD, § 11) Only a few years later, he added: “sticking to a theory as long as possible is of considerable significance” (“What is Dialectic?”, 1937).

    This refers back to the point we considered at the beginning of this comment: that it isn’t observations but our accepting a theory’s falsifier as true that actually falsifies a theory (because it is purely a matter of logic). “We do not take even our own observations quite seriously, or accept them as scientific observations, until we have repeated and tested them.” (LoSD, § 8) And even worse for the idea that anything about falsificationism can ever be “naive”, the falsifier has to be one whose truth would corroborate a rival theory: “If accepted basic statements contradict a theory, then we take them as providing sufficient grounds for its falsification only if they corroborate a falsifying hypothesis at the same time.” (LoSD, § 22)

    This is actual falsificationism: taking the criterion of falsifiability (and its inherent repudiation of induction) seriously and additionally putting methodological rules (cf. LoSD, § 20) in place to prevent its circumvention. Whenever somebody—be it Lakatos, Feyerabend, or Philippe Lemoine—talks about “naive falsificationism”, you can be certain that what they are talking about is actually simply not falsificationism.

  16. Thanks. I enjoyed reading this. I think it makes clear why falsification in the simple sense cannot work and why scientists don’t do it. I suspected this. At the LHC the standard theory predicted a particle in an energy range, and when a particle was found it was hailed as the Higgs Boson. But no one has ever seen how that particle behaves except that it produces certain decay products.

    The LHC experiments were not set up to falsify the standard model, they were set up to look for what was predicted (i.e. to verify the standard model). And the announcements suggested that they had verified it. When they failed to find other predicted particles, they did not announce the collapse of the standard model, they questioned the background assumptions.

    OTOH questioning the background assumptions is essential. When LIGO announced the detection of neutrinos moving faster than the speed of light, everyone assumed faulty equipment and that’s what it turned out to be. Because the speed of light is so well established and if it were to fall then so would the whole of relativity… and with it all of macro-scale physics.

    BTW I think there must be some synergy between this dynamic and how Hugo Mercier and Dan Sperber describe the workings of reason. I can’t quite put my finger on it, but their book The Enigma of Reason laid open a whole new way (for me at least) of thinking about how we make inferences.

  17. In short, I don’t think your main argument can be used to refute Popper’s concept of falsifiability.
    First, your distinction between theory and any kind of “background assumption” seems arbitrary and unnecessary. One can include any number of assumptions in the theory, thereby making them part of the theory (maybe calling them different components or aspects of the theory). Or one could treat the main theory just as another assumption in a larger collection of assumptions. Which specific part of a theory (or collection of assumptions) turned out to be false in a series of experiments is only relevant for the correction/amendment/modification/etc. of that theory.

    More importantly, your argument, that “[…], since the failure of a prediction derived from a theory plus a bunch of background assumptions doesn’t logically imply the falsity of the theory, it can’t falsify it.” does not refute Popper’s principle of falsification. It just means that, in cases with multiple possible explanations, you can’t be sure which one is the correct one without some additional information (i.e. more detailed observations or experiments are needed). Therefore, one would need to design an experiment that controls for other possibilities/assumptions to check whether the one in question holds.

    Here is an example: The observation of a broken down car on the side of the road doesn’t imply that there is something wrong with any specific component if the car (maybe it has a flat tire, maybe it’s the motor, or the electrical wiring or an empty tank, etc.). The fact that the car broke down only implies that some aspect about the car (some part in the collection of its components) caused it to stop. Analogously, if you have a collection of assumptions (or theory + assumptions) that doesn’t hold up under some experiment, that doesn’t mean that a specific assumption is false. Logically, it could be any one (or a combination) of them. It just means that “something” is wrong. To determine what that “something” is requires a more specific observation or experiment. (In the car example, this could mean, e.g., moving from the observation that the car has broken down to checking different parts of it for visible damage.)

    However, the feasibility of determining which part of a theory (which assumption in a collection of assumptions) is false, is not the point of Popper’s argument. The relevant question is whether a theory (or any aspect of it) *can* be tested at all, i.e. whether one can derive claims (predictions) that are indeed falsifiable through observations/experiments.

    To give another example: Assume someone drops an object of a certain mass from a certain height and the object can fall freely. One could come up with two possible theories:
    I) Newtonian mechanics, which can predict the velocity of the object at any point during its fall — this is a falsifiable claim. (One can take objects with different masses and drop them from different heights and see if the theory holds up or not. Btw., a change in background assumptions – e.g. accounting for air resistance or moving from Earth’s gravitational field to that of Mars – might require a modification of the theory but doesn’t necessarily change its falsifiability.)
    II) An alternative theory that postulates little, unobservable fairies who are pushing objects away from the heavens so as to not obstruct the path of an invisible intergalactic messenger — this is not falsifiable. There is no experiment, for example, that could test the existence of the little fairies if their only “effect” in the world is the same as that of what we call gravity.

    In principle, either one of these two theories could turn out to be true. They both provide an explanation for what we observe (i.e. gravity). However, we only attribute scientific merit to the first one because its claims can be tested and can therefore be falsified. Theory I) may well be false but at least we can come up with ways to check different aspects of it. This criterion for falsifiability does not hold for theory II).

    1. @Ben: I am not sure what you are trying to demonstrate here, but I suggest you consider the following change:

      “There is no experiment, for example, that could test the existence of the little fairies if their only “effect” in the world is the same as that of what we call gravity.”

      –>

      “There is no experiment, for example, that could test the existence of the gravitational force if its only “effect” in the world is the same as that of what we call little fairy pushing.”

      1. The difference is that the “fairy-aspect” of the second theory doesn’t provide any testable predictions. (For example, a modified version could state that it is actually Martians and not fairies that do the pushing — neither version is falsifiable because one cannot test their claim.) In contrast, for example, F=ma gives testable predictions.

        1. No. The little fairy could be hypothesised to be pushing in such a way that the resulting acceleration depends on the mass and distance to the nearby body in the appropriate way. The “mass” could become the “desirability” of the body, as judged by the fairies (so that the pushing activity would essentially be proportional to that desirability), and the inverse proportionality to the square of the distance would be attributed to increasing discouragement of the fairies in their wish to bring home the thing they are pushing on…
          We should not forget that giving the name “force” to something does not make it more scientific than giving it the name “fairy”.

        2. Ben: In this and your previous comment you suggest that the “scientific merit” of a theory depends on testability and repeatability (“testable predictions”). Unfortunately, this popular requirement can never apply to SINGULAR EVENTS (such as the Big bang or creation events in biology), since singular events are neither predictable nor repeatable. However, few (if any) scholars would be willing to call theories about the beginning of the cosmos or the emergence of life “non-scientific.” So the “merit” of a theory must rest on criteria that go beyond merely identifying it as “what scientists currently believe”—lest we beg the question regarding WHICH scientific opinions qualify as authoritative. Moreover, the truth of a matter is not determined by consensus.

          For example, biologists are struggling to salvage the Darwinian narrative of common descent (of all major phyla) via the mechanism of ‘natural selection’ operating on randiom (and hopefully favorable) mutations in pre-existing gene pools. However, the feasibility of that account has approached a probability of zero in light of current discoveries about DNA, the double helix, and how rigorously INFORMATION is encoded at the level of genes. Yet some macro-evolutionists (like Dawkins) still cling to the original gradualistic account by posing ad-hoc revisions to Darwin’s original theory.

          My point here is not to suggest (as has Etienne Parizot) the “empirical equivalence” of Newtonian mechanics and “little fairy pushing”, but to say that we must be cautious not to IDENTIFY one particular theory (about origins or whatever) as “THE scientific view”—and, by implication, all rival views as non-scientific—when even dominant theories are tentative and questionable.

          1. For example, biologists are struggling to salvage the Darwinian narrative of common descent (of all major phyla) via the mechanism of ‘natural selection’ operating on randiom (and hopefully favorable) mutations in pre-existing gene pools. However, the feasibility of that account has approached a probability of zero in light of current discoveries about DNA, the double helix, and how rigorously INFORMATION is encoded at the level of genes.

            That is a serious assertion there. How would those facts refute evolution? It seems obvious that imperfectly reproducing systems would generate similar systems with different reproductive ability, which would then, tautologically, reproduce at different rates. The ones that reproduce more would appear more. I don’t see how it is possible that evolution can be false.

  18. Came over because Sean Carroll retweeted it. I’ve been hearing (and suspecting) that Popper’s falsificationism is wrong for a while now and I think this is quite a good overview of how naïve falsificationism won’t work. The way modern physics works by placing constraints on various theories rather than discarding them outright seems to contradict it.

    The only replacement I’ve heard of is Thomas Kuhn’s paradigm shifts, which is an (I would say obviously) unsatisfying solution, due mainly to the vagueness of what constitutes a paradigm.

    Not exactly sure what replaced it though. Here’s hoping that I could get some pointers.

    1. » Ryan:
      “Came over because Sean Carroll retweeted it. I’ve been hearing (and suspecting) that Popper’s falsificationism is wrong for a while now…”

      I’m sure that any minute now Philippe will tell you, too, that in his post he was not at all talking about falsification as “proposed by Popper”. He—just like Sean Carroll, incidentally—prefers to attack an ignorant, silly, and conveniently-easily-knocked-down caricature of it, which he alleges “most people, especially scientists, have in mind” (without even quoting anybody, of course).

      If you are looking for “a replacement”, how about you try and find out what falsificationism really is: see this comment.

      That is what’s so pernicious about posts like Philippe’s: Many people will come away thinking precisely that it is about Popper’s falsificationism. Because there is no other version of it. “Naive falsificationism”, in any and all of its guises, is an ignorant caricature. To even call it “falsificationism” is a major disservice to philosophy and science literacy.

      Lastly, to not even care enough about the actual concept to take the time to understand it, let alone explain it to people—as Sean Carroll explicitly said: he wasn’t “even trying to carefully figure out what precisely [Popper] had in mind”—should be a source of intense embarrassment to anyone aspiring to be a scientist or a philosopher.

      1. Then why do physicists place bounds on theories rather than falsify them outright?

        Until falsificationism can explain this observation, I see no reason to believe falsificationism is an accurate account of scientific inquiry.

  19. Apparently the internet decided my comment was not worth posting.

    I came here because Sean Carroll retweeted it and I do think this explains why falsificationism is false nicely. The observation that made me doubt falsification in the first place was how physicists place constraints on theories rather than discard them outright (e.g. supersymmetry and the depressingly persistent MOND), and I don’t think any version of falsification could explain it.

    On the other hand, the only replacements I’ve heard of were Imre Lakatos’ research programs (which at least to me David Chalmers seems to endorse on Sean Carroll’s Mindscape), which no one has ever explained to me, and Thomas Kuhn’s paradigm shifts, which to me seems obviously false due to the lack of incommensurability between scientific paradigms (not counting those pre-Galileo), and unsatisfactory due to the vagueness of what is considered a paradigm.

    Here’s to hoping I could get some pointers as to what philosophers replace falsificationism with.

    1. » Ryan:
      Here’s to hoping I could get some pointers as to what philosophers replace falsificationism with.

      Here’s a genuinely honest question: Why, do you think, did you come away from reading Philippe’s post with the impression that he was talking about actual falsificationism (i.e. “Popper’s falsificationism”)? After all, he did say, “I will not discuss the more sophisticated versions of falsificationism that have been proposed, which often stem from remarks Popper himself made”.

      Here’s what self-respecting philosophers would do: Instead of replacing a straw-man caricature, they would actually read Popper and try to understand what falsificationism really is.

  20. I think we can consider that, from a first order logic perspective, “existence” is all that is required for satisfiability of a theory. And if there is only one observation of “existence”, this observation can not be both true or false at the same time, else contradictory. So if we propose to falsify the theory, we can not “unobserve” empirical
    evidence. We build the theory by concatenation (and operation) more observations such that it still remains true. This an empirical approach circumvents falsification–which is a modern machine learning approach. Sure we can say if there exists x such that y in the theory that this holds for all x, but really all we need is the empirical evidence for each x. e.g we enumerate the universe of blackbirds one by one, rather than search for a blue one to falsify it… this the nature of science is existential (for machine learning) rather than universal…

  21. “The law of universal gravitation by itself doesn’t make any observable prediction.”

    Philosophers make the silliest statements and get away with it. Newtonian mechanics is not just the law of gravity. So we send a spacecraft to Pluto without making any prediction where it would go and by some incredible accident it went to Pluto. Or we go by what all physicists would say, regardless of philosophers: if the spacecraft went somewhere else, Newtonian mechanics is falsified.

    If you can’t do science, do philosophy of science. Popper is poop, and so are you.

Comments are closed.