Page:Sm all cc.pdf/17

 that it is correct, then determines its logical consequences. This inference may be deductive, a necessary consequence of the hypothesis, or inductive, a probable implication of the hypothesis. To be fruitful, this inference must generate a testable prediction of the hypothesis. An experiment is then undertaken to confirm or refute that prediction. The outcome affects whether the hypothesis is retained, modified, or refuted.

This method of hypothesis is the crux of scientific method, but scientific progress need not be quite as linear as shown. Hypotheses can be generated at any stage. Most die virtually immediately, because they are incompatible with some well-established observations or hypotheses. A single hypothesis may yield multiple predictions: some useless, many testable by a brief search of published experiments, some requiring an experiment that is infeasible, and few leading to actual experiments.

The insistence on verifiability, or its converse -- falsifiability, limits the scope of science to the pursuit of verifiably reliable knowledge. Reliability is, however, subjective (see Chapters 6 and 7), and hypothesis tests are seldom as conclusive as we wish. Though a hypothesis test cannot prove a hypothesis, some scientists (especially physicists) and many philosophers claim that it can at least disprove one. This argument, however, holds only for deductive predictions. More often, the test is not completely diagnostic, because assumptions buffer the hypothesis from refutation. Many hypotheses are abandoned without being refuted. Others are accepted as reliable without proof, if they have survived many tests; we cannot work effectively if we constantly doubt everything.

Is there a scientific method? The answer depends on whether one is a lumper or a splitter. Certainly the method of hypothesis is central to nearly all science, but scientific techniques and style depend both on the problem investigated and on individual taste.

For some, like Francis Bacon or Thomas Edison, experimentation is exploration; interpretations will inevitably follow. Trial and error, with many trials, is the method used by Edison, the medieval alchemists, and modern seekers of high-temperature superconductors. Others, like Aristotle, employ the opposite approach: develop an idea, then experimentally demonstrate its validity. A few, like René Descartes or Immanuel Kant, deduce the implications of premises. Many more, like Galileo, make predictions based on a hypothesis and empirically test those predictions. For most, each of these approaches is sometimes useful.

Research style is also fluid. At one extreme is Leonardo da Vinci, fascinated by everything he saw. Mohammad Ali, in describing himself, also described this research style: “Dance like a butterfly; sting like a bee.” At the other extreme is the Great Pyramid style -- systematically and possibly laboriously undertake multiple experiments in the same field, until the final foundation is unshakeable. Charles Darwin used this strategy for establishing his theory of evolution, except that he compiled evidence rather than experiments.

The scientific method is both very liberal and very conservative: any hypothesis is worthy of consideration, but only those that survive rigorous testing are incorporated into the framework of reliable knowledge. The scientific method is incredibly versatile, both in the range of knowledge amenable to its investigation and in the variety of personal scientific styles that it fosters and embraces. Invariably, however, it demands an intriguing and challenging combination: creativity plus skepticism.