Karl Popper focused on these weaknesses of confirmation and concluded that additional ‘confirmations’ do not necessarily and substantially increase confidence in a hypothesis. In reaction, he created a philosophy for hypothesis testing known as falsificationism. Starting from the premise that the only compelling experiment is one that disproves a hypothesis, he argued that the task of science should be falsification, the rejection of false theories.
First proposed in 1920 and eloquently advocated by Popper, falsificationism had a substantial following among philosophers of science for several decades, and many aspects of it survive. Yet falsifiability has been virtually ignored by scientists. Popper’s vision of science is generation of a myriad of ideas followed by ruthless falsification and rejection of the majority. This vision does not correspond with the experience of scientists, but of course our subjective experience could be misleading.
Most scientists do agree that testability is a fundamental criterion for deciding which hypotheses are worthy of attention, but none agree with Popper’s assessment that falsifiability is supreme, nor that minor supporting roles are played by confirmation, discovery, insight, and subjective context-dependent evaluation. “An idea may be neither demonstrably true nor false, and yet be useful, interesting, and good exercise” [Trotter, 1941]. A concept may be embraced even without falsifiability, if it is capable of finding elegance of pattern among anomalous observations. Virtually the only mention of falsifiability that I have seen in my field (geology/geophysics) was Ken Hsü’s claim that Darwinian evolution is nonscientific because it is not falsifiable. Is the scientific method nonscientific, because its assumption of causality is neither provable nor disprovable?
Falsifiability is a tool, not a rule. The logical flaw in falsificationism is its deductive conclusion that a single inconsistent observation disproves a hypothesis. Scientists do not agree to follow this simple path for evaluating hypotheses, because the source of the inconsistency may be problems in the data, assumptions, or experimental conditions. Kuhn [1970], several other philosophers of science, and Wilson [1952] have cited numerous examples of theories surviving in spite of ‘falsifying observations’:
Newton’s laws exhibited incredible predictive value. Although they failed to account completely for planetary orbits, they were not rejected.
The chemical ‘law’ of Dulong and Petit is that the specific heat of each solid element multiplied by its atomic weight is approximately 2 calories per degree. This empirical relationship was used for many years, in spite of the early recognition that it did not work for either silicon or carbon. The exceptions were neither ignored nor used to reject the hypothesis. Ultimately they helped to guide establishment of a law more founded in theory. From the perspective of that new law, Dulong and Petit’s law was a special limiting case.
Copernicus’ 1543 proposal that the earth revolves around the sun initially conflicted with many observations. The ‘tower argument’ was particularly damning: if the earth really is spinning, then an object dropped from a tower should land west of the tower, not -- as observed -- at its foot. Fortunately the theory was not discarded.
Power of Evidence
The successful middle ground between avid justificationism and falsificationism is a concern with the power of evidence. Information is proportional to astonishment, or, in terms of informa -
