WHAT MAKES A ‘GOOD’ EXPERIMENT?*

The factors which make an experiment 'good' would seem to be relatively easy to identify. These include the relationship to existing theory or theories, or their call for new theories, and the ability of the apparatus to measure the quantities of interest to sufficient accuracy and precision. The physics community has little difficulty in applying these criteria although there are occasions when their application is not straightforward. There is a distinction, however, between a 'sociologically good' experiment, one which is accepted as good by the physics community, and one which might be considered 'good' on methodological or philosophical grounds. It is to be hoped that these two sets of criteria overlap to a large extent, and this is, I believe, shown to be true in the examples discussed below. In the following discussion I shall assume that all experiments are 'theory laden' in that the terms and apparatus used are all dependent on existing theory. Before beginning the discussion I distinguish between 'technically good' experiments, those in which improvements in existing apparatus or the construction of an entirely new apparatus, result in a more accurate and precise measurement of a physical quantity and those which are 'conceptually important' by their relation and relevance to existing theory, to be discussed in detail below. An example of the first type is the recent measurement of the speed of light by independently measuring the frequency and the wavelength of the same radiation using laser techniques (Evenson [1972]). Some members of the second class are also examples of the first, as for example Millikan's oil drop experiment to measure e, the charge on the electron (Millikan [1913]). Techniques developed in the first class can also be used later in conceptually important experiments as for example the test of relativity by Brillet and Hall [1979].