This book explores the part of natural language meaning that is geared towards convincing, persuading, and more generally changing the belief of an audience. A core claim of the book is that speakers have an agenda when they engage in conversation, and that each of their utterances serves a purpose: its argumentative goal. Grégoire Winterstein offers an account of how an utterance relates to these goals and shows that the linguistic form of an utterance matters as much as its content in the evaluation of its argumentative possibilities.The book proposes a Bayesian reinterpretation of the theory of argumentation within language, situated within contemporary approaches in semantics and pragmatics. The model describes three major argumentative components and their interactions: general argumentative laws (framed in probabilistic terms), argumentative operators that directly or indirectly constrain the argumentative profiles of the elements that they bear on, and raw argumentative possibilities associated with lexical elements, that could be associated with argumentative topoi. Argumentative meaning is described as separate from descriptive, truth-conditional, meaning, and while these two types of meaning inform each other, neither can be fully reduced to the other. To make sense of these relations, a revised typology of linguistic meaning is proposed, which distinguishes between the conventionally marked status of a content and its contextual at-issueness.This account also sheds light on non-linguistic aspects of argumentation such as the social identity of the speaker, as well as on non-referential, distributional, approaches to meaning. In particular, it discusses the relevance of modern language models as proxies to investigate and characterize the argumentative profile of specific expressions.