Dummett's conception of deductive inference as indirect verification underlies his philosophy of logic. It is intended to serve as the core idea to make clear the complex relationships among the key notions such as legitimacy and utility of inference, applicability of deduction to empirical statements, the realist and semantic concept of truth, and so forth. In this paper, we focus on his constructivist idea of justification of inference and point out that it does not play the intended role in the entire picture. In particular, it faces a difficulty in explaining the modal notion of verifiability. Then an alternative model of justification is proposed. It is called bilateralism since it explains legitimacy of inference in terms of (in)coherence between verification and falsification, from which modality involved in justification process is derived.
