Starting in the 1960s, philosophers of science explored alternative approaches to scientific theories. Prominent among them was the so-called semantic conception, originally formulated by Patrick Suppes, according to which theories are viewed as collections of models together with hypotheses about how these models relate to parts of nature. Versions of the semantic conception differ in their views about the character of models, sometimes taking models to be abstract mathematical structures, susceptible to precise formal specifications, and sometimes taking them to be more concrete (as chemists do, for example, when they build models of particular molecules).

The semantic conception of theories has several attractive features. First, unlike the older approach, it provides a way of discussing aspects of science that are independent of the choice of a particular language. Second, it appears to do far more justice to areas of science in which theoretical achievements resist axiomatization. Darwinian evolutionary theory is a case in point. During the heyday of the axiomatic approach, a few philosophers attempted to show how the theory of evolution could be brought within the orthodox conception of theories, but their efforts tended to produce formal theories that bordered on triviality. The consequent debates about whether the theory of evolution was more than a tautology should have generated serious philosophical embarrassment. Philosophers deploying the semantic conception, by contrast, shed light on theoretical issues that arise in contemporary evolutionary biology.

Finally, the semantic conception is far better suited to an aspect of the sciences that was frequently neglected, the practice of idealization. Instead of thinking of scientists as aspiring to offer literally correct descriptions of general features of the world, the semantic conception supposes that they propose models accompanied by claims that particular parts of nature correspond to these models in specific respects and to specific degrees.

The historicist conception

The work of Thomas S. Kuhn (1922–96), to be discussed in more detail in the following section (see Scientific change), offered a third approach to scientific theories (although some supporters of the semantic conception tried to relate their own proposals to Kuhn’s). In his seminal monograph The Structure of Scientific Revolutions (1962), Kuhn displaced the term theory from its central position in philosophical discussions of the sciences, preferring instead to talk of “paradigms.” Although Kuhn’s terminology is now omnipresent in popular parlance, he came to regret the locution, partly because of criticism to the effect that his usage of paradigm was multiply ambiguous. In his description of everyday scientific work (so-called normal science), however, Kuhn had captured important aspects of theories that philosophers had previously overlooked. He had seen that scientists often draw inspiration from a concrete scientific achievement (the core meaning of paradigm) and that this achievement poses research questions for them and often furnishes styles of experimentation or explanation that they aim to emulate. He also saw that scientific work is often dominated by something larger and more enduring than a specific theory: to wit, a program for research that survives through a whole succession of theories. In the wake of Kuhn’s work, many philosophers attempted richer descriptions of the scientific background (the “body of theory”) on which researchers draw, talking variously of research programs, research traditions, and practices.

What, then, is a scientific theory? In recent decades there have been heated debates about this question. But there is no need to give an answer. In the course of their work, scientists do a wide variety of things. Philosophers of science try to understand aspects of the enterprise, offering reconstructions of scientific practice in the hope of addressing particular questions, and there is no reason to think that a particular style of reconstruction will be appropriate to every question. Just as carpenters decide which tools to use on the basis of the job at hand, philosophers might adopt different techniques of reconstruction for different purposes.

When the ways in which meaning accrued to theoretical vocabulary constituted a burning question for the philosophy of science, it was natural to adopt an axiomatic approach to scientific theories and to focus on the connections between theoretical terms and language that are more readily understood (and, to the extent that questions remain in the wake of Putnam’s insights about the theoretical-observational and observable-unobservable distinctions, the axiomatic approach can still be of value in this area). Similarly, when a philosopher (or scientist) wonders whether a specific assumption or a particular choice of a parameter value is necessary, the device of axiomatization helps to resolve the question; given an axiomatic presentation, one can explore whether every derivation using the assumption can be transformed into one without. However, when the topic under study is a science in which there are few generalizations, or when one is concerned to elucidate issues about idealization in science, the semantic conception seems much more illuminating. Finally, in probing the dynamics of large-scale change in science—reconstructing the ways in which Darwin won acceptance for his evolutionary theory, for example—the concepts introduced by Kuhn and those who reacted to his work seem more readily applicable. Insistence that there must be a unique answer to what scientific theories really are seems like misplaced dogmatism that obstructs philosophical inquiry.