Explanatory concepts can be given different labels, depending on our confidence in their reliability. A law is an explanation in which we have the greatest confidence, based on a long track record of confirmations. A theory, for most scientists, denotes an explanation that has been confirmed sufficiently to be generally accepted, but which is less firmly established than a law. An axiom is a concept that is accepted without proof, perhaps because it is obvious or universally accepted (e.g., time, causality) or perhaps to investigate its implications. A hypothesis is an idea that is still in the process of active testing; it may or may not be correct. Models are mathematical or conceptual hypotheses that provide useful perspectives in spite of recognized oversimplification. Whereas laws and theories are relatively static, hypothesis formation, testing, and evaluation are the dynamic life of science.
Laws, theories, and hypotheses also differ in generality and scope. Theories tend to be broadest in scope (e.g., the theories of relativity and of natural selection); most provide a unified perspective or logical framework for a variety of more specific and more limited laws and hypotheses. All three are generalizations; rarely do they claim to predict the behavior of every particular case, because they cannot encompass all variables that could be relevant. Most laws are expected to be universal in their applicability to a specified subset of variables, but some are permitted exceptions. For example, the geological law of original horizontality states that nearly all sediments are initially deposited almost horizontally. Hypotheses have not fully bridged the gap between the particular and the universal; most are allied closely with the observations that they attempt to explain.
Researchers do not take these terms too seriously, however. The boundaries between these three categories are flexible, and the terms may be used interchangeably.
Hypotheses, theories, and laws are explanations of nature, and explanations can be qualitative or quantitative, descriptive or causal (Chapter 3). Most explanations involve variables -- characteristics that exhibit detectable and quantifiable changes (Chapter 2) -- and many explanations attempt to identify relationships among variables (Chapter 3).
All scientific concepts must be testable -- capable of confirmation or refutation by systematic reality checking. Uncertainty is inherent not only to explanatory concepts, but also in the terms describing concept testing: confirmation, verification, validity, reliability, and significance. Scientific confirmation does not establish that an idea must be correct, or even that it is probably correct. Confirmation is merely the demonstration that a hypothesis is consistent with observations, thereby increasing confidence that the hypothesis is correct.
Some concepts can be tested directly against other, more established concepts by simple logical deduction (Chapter 4). More often, we need to investigate the hypothesis more indirectly, by identifying and empirically testing predictions made by the concept. Data are the experimental observations, or measurements, that provide these tests.
The interplay between hypothesis and data, between speculation and reality checking, is the
heart of scientific method. Philosophers of science have devoted much analysis to the question of
how hypothesis and data interact to create scientific progress. In the latter half of the twentieth century, the leading conceptualization of the scientific process has been the hypothetico-deductive
method. Popper [1959, 1963], Medawer [1969], Achinstein [1985], and many others have provided
perspectives on what this method is and what it should be. Most suggest that the gist of this
method of hypothesis is the following:
Scientific progress is achieved by interplay between imagination and criticism. Hypotheses are the key, and hypothesis formation is a creative act, not a compelling product of observations. After developing a hypothesis, the scientist temporarily as -
