Testing Hypotheses by Experiment

TESTING HYPOTHESES BY EXPERIMENT JAMES K. FEIBLEMAN* Hypotheses are tested in three well-defined steps. First, attempts are made to discover factual evidence for or against the hypothesis by experiments . Second, mathematical calculations are made to try to discover evidenceforor against the consistencyofthe hypothesis. Third, predicting and/or controlling phenomena are attempted, either experimentally or mathematically or both. In the second step the hypothesis is regarded as a theory, since it is in a stronger position for having survived the first step; in the third step the theory is regarded as a law, since it is in a correspondingly stronger position still. The first step is my subject here. I. The Meaning of"Experiment" "Experiment" may bedefined as deliberate interference with phenomena in such a way as to aid a decision concerning a hypothesis. A generalization is confronted with the evidence under controlled conditions. An experiment is controlled observation of what happens when some segment ofthe natural world is forced to an alternative in which it will be as easy to obtain a negative as a positive answer. Experimental science could be described as "planned experience." On the grounds that neither reason alone nor experience alone nor the two without manipulation can cope with the intricacies ofthe natural world, an elaborate logical method was devised which could weave these three sets of resources together in a fashion which enables all three to function to their fullest. Experiments test hypotheses whether these are implicit or explicit, covert or overt. The investigator may be unaware that he is relying upon a hypothesis when he conducts an experiment, but without it, it would be * Professor and University Chairman ofPhilosophy, Tulane University, New Orleans, Louisiana. 91 difficultto explain howoneinstrumentaldevicerather thananother, or one problem rather than another, was chosen. Experiments are not mere observations at the common-sense level. Instruments are employed because either the materials to be tested lie beyond ordinary observational levels, as in the case of refined chemicals to be mixed, or the method is beyond common experience, as in the separation which may take place by means ofa centrifuge. The more advanced the science, the more complex the instruments. The contemporary instruments of scientific research are complex indeed: the electron microscope, the cyclotron, or the digital computer, for instance. Experiments discover evidenceagainstas wellas for a hypothesis. A negative result may be as important as a positive one, as in the case of the Michelson-Morley experiment to determine the drift through the ether. The failures as well as the successes must be scrupulously noted, for they may be equally significant. The selection ofinstances which are apt to be favorable to the outcome is illegitimate and certain to defeat the very purposes for which the experiment was performed. Care must be taken, then, to insure that the instances chosen constitute a fair sample, and a fañsample is a random sample. The degree of verification or falsification obtained must be evaluated through interpretation of the data obtained by experiment, which is a quantitative assay of a qualitative object—the measure of an amount of energy or the dimensions of a material structure, for instance. This is as difficult a step as the making of the experiment itself. The qualitative aspects here are the energy and the material: energy is energy and material is material; these do not change, even though the forms and quantities do. There are even laws recognizing such conditions. The laws ofthe conservation ofmass and the conservation ofenergy recognize the constant nature ofthese entities and processes. Water turns to ice (qualitative change) at 32o F. (quantitative) although the amount ofwater is preserved. II. The Design ofExperiments Which hypotheses, then, are to be tested? The number ofpossible hypotheses is theoretically limitless; they cannot all be tested, for there would never be enough scientists, equipment, or time. On the other hand, any arbitrary restriction upon the discovery ofhypotheses is potentially dangerous to the progress of science. We cannot even establish the criterion 92 James K. Feibleman · Testing Hypotheses by Experiment Perspectives in Biology and Medicine · Autumn i960 ofverifiability, as some authorities have suggested. Ifwe were to test only those hypotheses that are readily susceptible ofexperimental verification, we should be overlooking many...