Hypothesis and Rebuttal: The Growth of Scientific Knowledge

Author: Karl Popper
Publisher:
Publish Date: 2003-06-01
Features:
II I might take the theories mentioned earlier as examples. Einstein's theory of gravity clearly meets the criterion of falsifiability. Even if our measuring instruments at the time did not allow us to draw firm conclusions about the results of tests, it is clearly possible to falsify this theory. Astrology cannot pass this test. Astrologers place extreme emphasis and superstition on the evidence they believe in, so they are completely indifferent to any adverse evidence. Moreover, they make their interpretations and predictions quite vague, so that any possible falsification of their theory (if the theory and predictions were stated more clearly) can be explained away. To avoid falsification, they undermine the testability of their theory. Making predictions extremely vague makes it almost impossible for them to fail, which is the typical trick of astrologers; making predictions unrefutable. Marxism's historical doctrine, despite the serious efforts of some of its founders and followers, eventually adopted this astrological approach. In some of its early formulations (for example, Marx's analysis of the nature of "future social revolutions"), their predictions were testable, and in fact, they have been falsified. However, Marx's followers not only did not accept these falsifications but instead reinterpreted the doctrine and the evidence to make them compatible. In this way, they saved the doctrine from refutation; but the price for doing so was adopting a trick that made the theory unrefutable. As a result, they gave the theory a "conventionalist twist"; and through this trick, they undermined the scientific status that the theory so proudly proclaimed. The two psychoanalytic theories, however, belong to a different type. They are simply untestable and unfalsifiable. No imaginable human behavior can overturn them. This does not mean that Freud and Adler did not get certain things right: personally, I do not doubt that many of their observations are quite important and will one day play a role in a testable psychology. But those "clinical observations" that the psychoanalysts naively believe confirm their theories are no better than the frequent proofs found by astrologers in their practice. As for Freud's grandiose trilogy of the self, the super-ego, and the id [Id], it is like the stories Homer 38 [Homer] collected from Olympus—none of them qualify as scientific. These theories describe certain facts, but in a mythological form. They contain very interesting psychological insights, but they lack a testable form. At the same time, I recognize that this myth can be developed to make it testable; historically, all (or almost all) scientific theories originated from myths [myths], and a myth may contain important predictions for a scientific theory. Examples include Empedocles's trial-and-error evolution [evolution] theory, or Parmenides's never-happened block-universe myth; if we add another dimension, it becomes Einstein's block universe (in which nothing ever happened, because according to the four-dimensional view, everything was predetermined and arranged from the beginning). Therefore, I feel that discovering a theory to be unscientific or "metaphysical" (we might say this) does not thereby make it unimportant, irrelevant, "meaningless," or "absurd." But it cannot be considered to be supported by empirical evidence in the scientific sense, although in some sense of occurrence, it is very likely "the result of observation." (There are many such pre-scientific or pseudo-scientific theories, some of which, unfortunately, are also quite influential [influence], like Marxism's historical interpretation; for example, the historical interpretation of racism is another influential theory that resonates with the unthinking people like a revelation.) Therefore, the problem that the falsifiability criterion seeks to address is neither the question of whether something is meaningful 39 nor the question of truth or acceptability. It is the problem of drawing a line between empirical scientific statements or systems of statements and all other statements (whether religious, metaphysical, or simply pseudo-scientific), as far as this can be done. Many years later—this would be around 1928 or 1929—I called this problem the "problem of demarcation." The criterion of falsifiability is one way to solve this demarcation problem, because it states that a statement or system of statements must be capable of contradicting possible or imaginable observations to qualify as scientific. III Of course, I now know that this demarcation criterion—whether testability, falsifiability, or refutability—is far from obvious; because until now, its significance has rarely been understood. In 1920, I felt it was utterly insignificant, although it solved an intellectual problem that deeply troubled me and had clear practical consequences (such as some political issues). But I had not yet grasped its full implications and its philosophical importance. When I explained this problem to a classmate in mathematics (who is now a famous British mathematician), he suggested I publish it. At the time, I thought it was utterly absurd; because I firmly believed that since this problem seemed so important to me, it must have struck many scientists and philosophers and that they must have already found my obvious solution. Later, through Wittgenstein's writings and the acclaim for his work, I learned that this was not the case; therefore, I published my research results 13 years later in the form of a critique of Wittgenstein's meaning criterion [criterion of meaningfulness]. As you all know, Wittgenstein [Wittgenstein] attempted to show in his Tractatus (e.g., see propositions 6.53, 6.54, and 5) that all so-called philosophical or metaphysical propositions are actually non-propositions or pseudo-propositions; they are meaningless. All true (or meaningful) propositions are basic propositions or atomic propositions—the truth-function of these propositions that describe simple statements of possible states of affairs and can, in principle, be affirmed or denied through observation. In other words, meaningful propositions can be completely reduced to basic propositions or atomic propositions, which are simple statements describing possible states of affairs and can, in principle, be affirmed or denied through observation. If we call a statement an "observation statement" not only because it states an actual observation but also because it states anything observable, we must say (according to the Tractatus, 5 and 4.52) that any true proposition must be a truth-function of observation statements, and thus can be derived from observation statements. All other apparent propositions would be meaningless pseudo-propositions, in fact just nonsense. This idea was used by Wittgenstein to contrast science with philosophy. We read (as in proposition 4.11, where he regards natural science as opposed to philosophy): "All true propositions are the whole of natural science (or all of natural science)." This means that all propositions belonging to science are those that can be derived from true observation statements; they are those that can be affirmed by true observation statements. If we knew all true observation statements, we would also know all that natural science can affirm. This is equivalent to a rough criterion of verifiability for demarcation. To make it less rough, it could be rewritten as: "Propositions that can potentially be verified by observation statements are those that are consistent with all true or meaningful propositions." Thus, according to this method, verifiability, meaningfulness, and scientificity all align. Personally, I have never been interested in the so-called meaning [meaning] problem; on the contrary, I regard it as a verbal issue, a typical pseudo-problem. I am only interested in the demarcation problem, which is to find a standard for the scientificity of theories. It is precisely this interest that made me immediately see that Wittgenstein's criterion of verifiability of meaning also attempts to serve as a demarcation criterion; this made me realize that according to his view, this criterion is entirely inappropriate, even if we set aside all doubts about the vague concept of meaning. Because, in my own terms, Wittgenstein's demarcation criterion here is verifiability or deducibility from observation statements. However, this criterion is too narrow (and too broad): it almost excludes everything that is actually typical of science (although it actually excludes astrology). No scientific statement can ever be deduced from observation statements, nor can it be described as the truth-function of observation statements.