Mack's Principle and the Laws of Logic

In this paper I wish to raise a philosophical question about logic, namely, the question whether its laws can consistently be thought of as analogous to those of the empirical sciences, i.e., as subject in some sense or other to test and confirmation, or whether, as is more often maintained, they must be thought of as analytic and a priori if not as coventional. In order to float the question, some general idea of what kind of activity logic is must be presupposed. The problem of logic I take to be as follows: Given the truth (or probability) of sentences {P}, what can we say (with what degree of confidence, etc.) about the truth (or probability) of sentences {Q}? The method of logic I take to consist in performing operations on the sentences (P) or on supplementary sentences introduced for the purpose and in performing further operations on the sentences so generated, and so on until the sentences {Q} or some obviously related sentences are generated. According to the rules employed in these operations we may then say that the sentences {Q} are true or have a certain degree of probability in relation to the sentences {P}. We thus arrive at a degree of confidence in the sentences {Q}. But what of our confidence in the whole procedure by which this degree of confidence is arrived at? Well, we can construct a second-order scheme for that and talk about the sentences {P} and the rules by which we operated on them. We thus arrive at a degree of confidence in the procedure. But what of our confidence in this second-order scheme? And so on. It is tacitly agreed by almost everyone except Quine that this regressive problem presents itself in two distinct cases. The first covers deductive inference and gives us absolute confidence in the conclusion on the object level as well as in the rules at all subsequent levels. The second covers everything else; we can't even think of pursuing the regression