What follows is an excerpt from chapter 11 of Karl Popper’s lucid and, as far as I’m concerned, inspiring book Conjectures and Refutations. Unfortunately I have only had the opportunity to read this chapter, but mark my words, by the end of this summer I’ll have the rest of Popper in my brain. The title of the chapter is ‘The Demarcation between Science and Metaphysics’ and the excerpt was gleaned from section 2, titled ‘My own view of the problem’.
// It was in 1919 that I first faced the problem of drawing a line of demarcation between those statements and systems of statements which could be properly described as belonging to empirical science, and others which might, perhaps, be described as ‘pseudo-scientific’ or (in certain contexts) as ‘metaphysical’, or which belonged, perhaps, to pure logic or to pure mathematics.
This is a problem which has agitated many philosophers since the time of Bacon, although I have never found an explicit formulation of it. The most widely accepted view was that science was characterized by its observational basis, or by its inductive method, while pseudo-sciences and metaphysics were characterized by their speculative method, or as Bacon said, by the fact that they operated with ‘mental anticipations’—something very similar to hypotheses.
This view I have never been able to accept. The modern theories of physics, especially Einstein’s theory (widely discussed in the year 1919), were highly speculative and abstract, and very far removed from what might be called their ‘observational basis’. All attempts to show that they were more or less directly ‘based on observations’ were unconvincing. The same was true even of Newton’s theory. Bacon had raised objections against the Copernican system on the ground that it ‘needlessly did violence to our senses’; and in general the best physical theories always resembled what Bacon had dismissed as ‘mental anticipations’.
On the other hand, many superstitious beliefs and many rule-of-thumb procedures (for planting, etc.) to be found in popular almanacs and dream books, have had much more to do with observations, and have no doubt often been based on something like induction. Astrologers, more especially, have always claimed that their ‘science’ was based upon a great wealth of inductive material. This claim is, perhaps, unfounded; but I have never heard of any attempt to discredit astrology by a critical investigation of its alleged inductive material. Nevertheless, astrology was rejected by modern science because it did not fit accepted theories and methods.
Thus there clearly was a need for a different criterion of demarcation; and I proposed (though years elapsed before I published this proposal) that the refutability or falsifiability of a theoretical system should be taken as the criterion of its demarcation. According to this view, which I still uphold, a system is to be considered as scientific only if it makes assertions which may clash with observations; and a system is, in fact, tested by attempts to produce such clashes, that is to say by attempts to refute it. Thus testability is the same as refutability, and can therefore likewise be taken as a criterion of demarcation.
This is a view of science which takes its critical approach to be its most important characteristic. Thus a scientist should look upon a theory from the point of view of whether it can be critically discussed: whether it exposes itself to criticism of all kinds; and—if it does—whether it is able to stand up to it. Newton’s theory, for example, predicted deviations from Kepler’s laws (due to the interactions of planets) which had not been observed at the time. It exposed itself thereby to attempted empirical refutations whose failure meant the success of the theory. Einstein’s theory was tested in a similar way. And indeed, all real tests are attempted refutations. Only if a theory successfully withstands the pressure of these attempted refutations can we claim that it is confirmed or corroborated by experience.
There are, moreover (as I found later), degrees of testability: some theories expose themselves to possible refutations more boldly than others. For example, a theory from which we can deduce precise numerical predictions about the splitting up of the spectral lines of light emitted by atoms in magnetic fields of varying strength will be more exposed to experimental refutation than one which merely predicts that a magnetic field influences the emission of light. A theory which is more precise and more easily refutable than another will also be the more interesting one. Since it is the more daring one, it will be the one which is less probable, But it is better testable, for we can make our tests more precise and more severe. And if it stands up to severe tests it will be better confirmed, or better attested, by these tests. Thus confirmability (or attestability or corroborability) must increase with testability.
This indicates that the criterion of demarcation cannot be an absolutely sharp one but will itself have degrees. There will be well-testable theories, hardly testable theories, and non-testable theories. Those which are non-testable are of no interest to empirical scientists. They may be described as metaphysical.
Here I must again stress a point which has often been misunderstood. Perhaps I can avoid these misunderstandings if I put my point now in this way. Take a square to represent the class of all statements of a language in which we intend to formulate a science; draw a broad horizontal line, dividing it into an upper and lower half; write ‘science’ and ‘testable’ into the upper half, and ‘metaphysics’ and ‘non-testable’ into the lower: then, I hope, you will realize that I do not propose to draw the line of demarcation in such a way that it coincides with the limits of a language, leaving science inside, and banning metaphysics by excluding it from the class of meaningful statements. On the contrary: beginning with my first publication on this subject, I stressed the fact that it would be inadequate to draw the line of demarcation between science and metaphysics so as to exclude metaphysics as nonsensical from a meaningful language.
I have indicated one of the reasons for this by saying that we must not try to draw the line too sharply. This becomes clear if we remember that most of our scientific theories originate in myths. The Copernican system, for example, was inspired by a Neo-Platonic worship of the light of the Sun who had to occupy the ‘centre’ because of his nobility. This indicates how myths may develop testable components. They may, in the course of discussion, become fruitful and important for science. In my Logic of Scientific Discovery I gave several examples of myths which have become most important for science, among them atomism and the corpuscular theory of light. It would hardly contribute to clarity if we were to say that these theories are nonsensical gibberish in one stage of their development, and then suddenly become good sense in another.
Another argument is the following. It may happen—and it turns out to be an important case—that a certain statement belongs to science since it is testable, while its negation turns out not to be testable, so that it must be placed below the line of demarcation. This is indeed the case with most important and most severely testable statements—the universal laws of science. I recommended, in my Logic of Scientific Discovery, that they should be expressed, for certain purposes, in a form like “There does not exist any perpetual motion machine’ (this is sometimes called ‘Planck’s formulation of the First Law of Thermodynamics’); that is to say, in the form of a negation of an existential statement. The corresponding existential statement—‘There exists a perpetual motion machine’—would belong, I suggested, together with ‘There exists a sea-serpent’ to those below the line of demarcation, as opposed to ‘There is a sea-serpent now on view in the British Museum’ which is well above the line since it can easily be tested. But we do not know how to test an isolated purely existential assertion.
I cannot in this place argue for the adequacy of the view that isolated purely existential statements should be classed as untestable and as falling outside the scientist’s range of interest. I only wish to make clear that if this view is accepted, then it would be strange to call metaphysical statements meaningless, or to exclude them from our language. For if we accept the negation of an existential statement as meaningful, then we must accept the existential statement itself also as meaningful.
I have been forced to stress this point because my position has repeatedly been described as a proposal to take falsifiability or refutability as the criterion of meaning (rather than demarcation), or as a proposal to exclude existential statements from our language, or perhaps from the language of science. Even Carnap, who discusses my position in considerable detail and reports it correctly, feels himself compelled to interpret it as a proposal to exclude metaphysical statements from some language or other.
But it is a fact that beginning with my first publication on this subject (‘Ein Kriterion des empirischen Charakters theoretischer Systeme’, Erkenntnis, 3, 1933, pp. 426 ff., now in The Logic of Scientific Discovery, pp. 312-14, see also Sections 4 and 10), I always dismissed the problem of meaninglessness as a pseudo-problem; and I was always opposed to the idea that it may be identified with the problem of demarcation. This is my view still.
