Scientists have generally been more concerned to justify their theories than run the risk of falsifying them with counter-evidence. This is a criticism of science which has been argued most powerfully by Sir Karl Popper (1902–94). Unusually for a philosopher of science, Popper with his comments and criticisms actually changed the way science is carried out, establishing especially that hypotheses should be formulated in such a way as to be falsifiable. Popper based his ideas on examples in early 20th-century physics, especially on how Einstein's ideas challenged Newtonian physics, but his strictures are also applicable to biology and clinical hypotheses, as Sir Peter Medawar has discussed most cogently (1967, 1969).
One usually thinks of science as producing new truths by observation and experiment, and the more secure the better — so Popper's idea that it should be possible to falsify a hypothesis, if it is to be acceptable as a 'good' hypothesis, came as a considerable shock. The point is not that a hypothesis cannot be good unless it is likely to be false; rather it is that a hypothesis should be conceived and formulated in such terms that experiments or observations could deal it a mortal blow. The hypothesis thus attacked might die or might be merely wounded, to recover in a changed form, but Popper prefers sudden death to hanging on to life. For him a hypothesis that staggers on, mutilated, generally does so because it was not formulated with sufficient vulnerability to die cleanly and be forgotten.
As an example of a hypothesis which refuses to die because inadequately formulated, we may consider Freud's notion of trauma at birth having significant effects in adult life. Suppose to test this we looked at adults who were born by Caesarean section, and suppose we found that they had, as adults, the characteristics that Freud attributed to birth trauma. If the post-Freudians accepted this as falsifying evidence for the birth trauma hypothesis, then in Popper's terms well and good — for the hypothesis was evidently set up in such a way as to be falsifiable. And in showing it to be false something has been learned: that these characteristics are not due to birth trauma. But the post-Freudians would have lost their hypothesis, and losing hypotheses can itself be traumatic. Popper would object to the post-Freudians responding by redefining 'birth trauma' — so that, let's say, it is not the birth itself but the shock of coming into the world that produces the trauma — for then the hypothesis would have changed, rather than decently died, and it is hard to see how any evidence could count against it.
Popper would probably prefer that the whole idea be abandoned at this point, but it might be that such forced redefining is a way in which science can work well, cumbersome though it seems. For could not some hypotheses be refined and improved in this way? There is surely a danger of throwing out the baby with the bathwater.
Popper has compared falsification as a way of gaining knowledge (and for him it is the
only way, as he completely rejects
inductive generalizations) with the deaths of individuals leading to improved species in evolution by natural selection of the fittest to survive. There is now, based on this analogy, an active school of epistemology led by Donald Campbell (1974) and Stephen Toulmin (1972). It goes back at least to the American philosopher
Charles Sanders Peirce (1839–1914) with the view that hypotheses die by competition, much as species die out when they are inadequate, for the cumulative good of life and science. There are objections, however, to saying that it is
only by falsifying hypotheses that we gain new knowledge. It is very hard to believe that animals, including ourselves, with all the evidence of learning curves and so on, never learn by induction (Gregory 1981).
It seems clear that strictly speaking hypotheses cannot be falsified by observational or experimental evidence alone, for all evidence has to be interpreted according to subjective background knowledge or assumptions which may be wrong. Thus, we take it as obvious that the stars appear to move across the sky because the earth is rotating daily on its axis; but, before this was accepted, the stellar observation was differently described. Aristotle thought he had falsified the hypothesis of daily rotation of the earth by jumping up and then finding that he landed on the same spot, when according to his assumption of earthly rotation he should have come down west of where he took off, and there should have been a continuous easterly wind.
Is there a fundamental difference between
falsifying and
predicting as a means of testing hypotheses? Isn't a successful
surprising prediction excellent evidence for the hypothesis suggesting it? Perhaps there is no asymmetry between falsifying and predicting (logically, the failure of a prediction is no different from any other discordant evidence), except that dramatic true predictions may be rarer than equally surprising falsifications. Falsifications of the stationary and the flat earth hypotheses were extremely surprising — and so conveyed a very great deal of information. But there are comparably surprising predictions, such as the apparent shift of positions of the stars near the sun, photographed during its eclipse in 1919, which
supported Einstein's ideas against Newton.
(Published 1987)— Richard L. Gregory
Bibliography- Campbell, T. D. (1974). 'Evolutionary epistemology'. In Shilpp, P. A. (ed.), The Philosophy of Karl Popper.
- Gregory, R. L. (1981). Mind in Science.
- Medawar, P. (1967). The Art of the Soluble.
- — — (1969). Induction and Intuition in Scientific Thought.
- Popper, K. R. (1959). The Logic of Scientific Discovery.
- — — (1970). 'Logic of discovery or psychology of research?' In Lakatos, I., and Musgrave, A. (eds.), Criticism and the Growth of Knowledge.
- — — (1972). Objective Knowledge: An Evolutionary Approach.
- Toulmin, S. E. (1972). Human Understanding.
