Scientific Methodology

In the 1950s, Carl Hempel offered what has become a widely-accepted description of how theories are constructed and tested in the natural sciences, drawing on arguments from the philosophy of science in the first half of this century. Compared to the simpler idea of direct induction from data to theory as proposed by Bacon and Mill in the seventeenth century (see Figure 1-A), Hempel portrayed scientific methodology in terms of a “hypothetical-deductive” path. One moves from data indirectly to the level of theory in a process which involves imagination, analogy and models as well as logical inference. Then, as Karl Popper had shown earlier, theories are then open to falsification against the data by the predictions one deduces (see Figure 1-B).

In the 1950s-60s, Thomas Kuhn,Norwood Hanson,Stephen Toulmin,Imre Lakatosand others supplemented Hempel’s account in a broader view which stresses the historical and contextual dimensions of scientific research. Barbour provides a particularly helpful overview of their work (see Figure 1-C). According to Barbour, these philosophers showed that metaphysical concepts and assumptions pervade scientific theories and underlie scientific methodology. Data are ‘theory-laden,’ and theories influence the decisions as to which data are relevant. The testing of scientific theories is complicated, too, by the fact that ad hoc auxiliary hypotheses can always be constructed to ward off potential falsifiers. Networks of theories, and not just isolated concepts or equations, are tested as a whole. Finally, the criteria for choosing between rival theories goes far beyond predictive success to include coherence with other, accepted theories, explanatory scope, fertility in suggesting new domains for application, conceptual simplicity (“Occam’s Razor”), aesthetic qualities like beauty, and the avoidance of ad hoc moves. Since such criteria transcend the details of the particular theories being considered, they provide a framework for a rational choice between rivals.

Figure 1: Three versions of scientific methodology.
(adapted from Barbour, 1990)

Contributed by: Dr. Robert Russell