similar to scenarios of science advancing by paradigm replacement, the abandonment of degenerating research programs or problem-solving methodologies. In the present case, however, it depends on accepting interpretatively isolated theories as the fundamental units of explanation.

What does this mean for the truth and objectivity of scientific claims such as (8)-(10) (or less simplistic claims with more obviously theory-laden terms)? The thesis of the dispensability in principle of theory laden terms in observation reports would suggest rephrasing them along the lines: “I saw a needle pointing to a figure 2 on a scale and interpreted it as. . .” This is hardly plausible to anyone familiar with the controversies concerning protocol sentences, an observation language, and similar attempts to develop a radical empiricism. The alternative is to regard the bulk of such claims as theoretical claims, at least in a minimal sense. They may not be deduced from a theory, but their physical significance is theory dependent. As theoretical claims they are all replaceable, at least in principle, when presently accepted theories are replaced by better theories.

The radical skepticism this seems to entail is a bit like solipsism. Very few really hold such positions. Yet it is not easy to find a fatal flaw in the supporting arguments. I will separate rigorous theory reconstruction from Kuhnian relativism and focus on what I take to be the weak link in the chain of arguments just sketched. This is the claim that a rationally reconstructed theory and a functioning theory are the same theory in different forms37. This is eminently plausible if one thinks of a theory as a mathematical formulation that is to be accorded a physical interpretation. Why rely on a slipshod foundation when a competent craftsman can build a better one?

What should be recognized as the crucial factor here is that this is a reconstruction. Reconstruction begins with the primacy of pragmatics. A theory is selected as a target for reconstruction because of its established success, or its acceptance by the physics community. The theory chosen is never an empty formalism, but an already interpreted theory. A necessary, but not sufficient condition, for any reconstruction to be considered adequate is that it reproduces the essential features of this interpreted theory. Logically, this requires that there be a prior analysis of the functioning theory and its informal interpretation.

Here there is an illuminating contrast between axiomatic and semantic reconstructionists. The leading logical positivists recognized the need for an informal analysis of the language of functioning science prior to any analysis by reconstruction. Two outstanding examples of relating informal and formal analyses are Reichenbach's analysis of the qualitative properties of time terms (Reichenbach, 1956, chap. 2) and Carnap's informal analysis of the meaning of ‘probability’ in both ordinary language (Carnap, 1962, pp. 1-51) and in science (pp. 161-191). The need for such an informal analysis has been explained by Reichenbach's leading student. (See Salmon, 1979, pp. 3-15).

The key difference between an informal analysis of a functioning theory and a formal analysis of a reconstructed theory is in the semantic component. A functioning constructive theory relies on terms, such as 'mass', 'energy', 'time', 'space', and 'particle', whose meaning are set prior to and independent of the theory using them. These meanings depend on usage, presuppositions, and ultimately on the categorial core of

37 I owe the explicit form of this claim to a debate with R. I. G. Hughes.