Abstract

It is usually held that what distinguishes a good inference from a bad one is that the former is but the latter is not truth-preserving. What is behind this view is the basic assumption that whether a certain inference is truth-preserving or not is a genuine issue, an issue the answer to which is determined objectively. This view is called the realist view of logic. ;In this dissertation the realist view of logic is criticized and an alternative view is presented. The alternative is called the fictionalist view of logic. On this view, what makes a logical system good is not that its inferential rules are truth-preserving, but rather that it has a certain proof-theoretic property, called deductive conservativeness. More specifically, it is argued that if the inferential rules for the logical operators involved in the system are conservative over the deductive system which is devoid of logical operators, then the conservative system will have all the virtues we expect in a good logical system. It is shown however that insofar as the system is conservative, there is no way to determine whether its inferential rules are truth-preserving or not; thus the latter is simply irrelevant to the goodness of the logical system. To support this claim, the view that the meaning of a logical operator is determined not by its semantic properties such as truth but by its inferential role, is advocated. A significant implication of this view is that there is no answer to the question which of the conservative logical systems is the correct logic for the world. ;Furthermore, it is maintained that metalogical claims of implication and consistency, namely claims of the forms 'A implies B', and 'A is consistent', can be taken as modal claims which involve modal operators as logical operators, and thus the above approach is applicable to such metalogical claims, too. This idea is used to avoid our ontological commitment to mathematical entities such as models and proofs employed in metalogic.