Abstract
In the first part of this paper I contend that Michael Dummett’s works contain a distinction between analytic and synthetic (logical) deductions. Dummett does not use Kant’s terminology. Nevertheless, based on general considerations about theories of meaning, Dummett shows how one can distinguish between deductions whose validity is recognised by merely grasping the concepts and deductions whose validity can be recognised only by “going beyond the concepts”, as Kant wrote. The latter are synthetic deductions, through which we can increase our knowledge. A special merit of Dummett’s distinction is its theoretical flexibility and generality. He does not propose a single clear-cut distinction between analytic and synthetic deductions; rather in his works we find an indication of how to draw a distinction once a theory of meaning is adopted. In the second part of the paper I consider possible developments within an inferentialist theory of meaning and draw some conclusions concerning the fallibility of deductions.