The dissertation concerns Prawitz's theory of grounds, a constructive, proof-theoretic semantics the aim of which is that of explaining why and how deductively valid inferences are compelling. It is divided into three parts. In the first, we outline Prawitz's previous semantic approach in terms of valid arguments and/or proofs, and show the problems it suffers from when trying to account for the epistemic power of valid inferences. In the second, we illustrate Prawitz's own proposal, and evaluate the advantages it allows for with respect to its basic tasks. Finally, in the third part, we propose a formal development of the theory, in terms of (hierarchies of) languages of grounding, of (hierarchies of) systems of grounds, and with respect to correctness/completeness and decidability issues.