Theories can no more be adjudged ‘true’ or ‘false’ than languages can. But a theory, like a language, might prove more or less useful, and it is the useful ones that survive. We accept, deploy and believe useful theories.
New data, and tests of new hypotheses, depend on observation and experiment. An experiment is a way of making precise observations, not of ‘unconstrained nature’
but of a situation deliberately constrained so that all relevant variables are known. The ability to design and conduct experiments is as essential for a scientist as a thorough and up-to-date understanding of relevant theories.
A properly designed experiment must be reproducible. That is to say, it must give substantially the same results when it is repeated at a different time and place and by other (trained and competent)
experimenters. It must also be valid: anyone trained in the appropriate field of science must agree that it does just what its designer claims it does. If new techniques or equipment are involved then the experimenter must show that these meet appropriate standards of reliability
, independent of times, places and persons.
The results of the experiment must be interpretable: they must be free of ‘interfering variables’
and of errors of extrapolation or interpolation. This can be ensured by running controls, in which all variables except the one under investigation are kept at the same values as in the experiment itself. It is also important to ensure that a newly-designed experiment is practicable, ethical and economic. Good experiment design is an art that requires practice.