On the assumption that there are no objective normative facts, what is the best theory of normative language? I try to answer this question. Chapter 1 argues for a presumption against noncognitivism and explains why error-theories are of limited interest: they concern adverbs and adjectives like ‘moral’, but not words like ‘ought’, ‘good’ and ‘reason’. This narrows down the options: the best subjectivist theory of normative language is a truth conditional, non-error-theoretic account. Chapter 2 argues for contextualism about normative statements. Contextualists hold that their truth conditions (can) vary with the context of utterance. Chapter 3 starts the assessment of contextualist theories. It looks into Humean accounts. Problems are revealed with both Harman’s and Schroeder’s versions. Chapter 4 develops a form of indexical relativism according to which the truth of normative statements depends on contextually salient rules. I present imperative-based analyses of ‘ought’ and ‘reason’ and show how they can explain why ‘A ought to X’ entails that the balance of reasons favours that A X-es. Chapter 5 further develops the theory of chapter 4 and applies it to the words ‘good’ and ‘must’. It turns out to be hard to analyse ‘good’. It also emerges that ‘must’ and ‘ought’ cannot be given different truth conditions. Chapter 6 explains Stephen Finlay’s end-relational theory. On this account, normative statements concern the relation in which acts or objects stand to contextually salient ends. In the case of ‘ought’ and ‘good’, this relation is one of probability raising. Chapter 7 discusses and answers some familiar objections to Finlay’s view. Chapter 8 raises some new problems, related to the fact that normative judgments are often made in the light of several ends. Chapter 9 explains why the end-relational theory is nonetheless the best subjectivist theory of normative language.

Notre pratique morale ordinaire, l’éthique normative ainsi que l’éthique appliquée présupposent que nos questions morales sont décidables non arbitrairement. Autrement dit, ces activités présupposent qu’il existe des réponses non arbitraires à nos questions morales. Le présent travail de recherche vise à questionner ce présupposé en explorant les réponses des trois principales familles de théories métaéthiques, soient le réalisme moral, l’antiréalisme moral et le constructivisme métaéthique, à la question « Les questions morales sont-elles décidables de manière non arbitraire? ». Notre but n’est pas de déterminer quelle théorie métaéthique est la meilleure, mais plutôt d’évaluer la possibilité que les questions morales soient décidables non arbitrairement. Nous défendrons que le réalisme moral semble compatible avec la décidabilité des questions morales et qu’au contraire, l’antiréalisme ainsi que le constructivisme semblent plus difficilement compatibles avec la décidabilité morale. Nous défendrons également que l’indécidabilité des questions morales, un problème pratique engendré par ces cadres métaéthiques, implique une aporie bien gênante. Si a priori on admet que ces trois familles de théories métaéthiques sont équiprobables, on pourrait alors affirmer grossièrement que nous avons deux chances sur trois de faire face, en pratique, au problème de l’indécidabilité morale et donc à l’aporie qu’elle implique. Cela justifiera pour nous l’intérêt d’explorer la possibilité d’une solution à cette aporie. Nous proposerons donc l’hypothèse selon laquelle la pratique du questionnement moral de manière aporétique, considérée comme une activité non cognitive, implique une certaine manière d’être qui n’est pas arbitraire. / Our ordinary moral practice, normative ethics and applied ethics presuppose that our moral questions are decidable non-arbitrarily. In other words, these activities presuppose that there are nonarbitrary answers to our moral questions. This very research aims to question this presupposition by exploring the answers of the three main families of metaethical theories, namely moral realism, moral antirealism and metaethical constructivism, to the question "Are moral questions decidable non-arbitrarily?". Our goal is not to determine which metaethical theory is the best, but rather to assess the possibility that moral questions are decidable non-arbitrarily. We will defend that moral realism seems to be compatible with the decidability of moral questions and that, on the contrary, antirealism and constructivism seem less compatible with moral decidability. Also, we will argue that the undecidability of moral questions, a practical problem generated by these metaethical frameworks, would involve a troublesome aporia. If a priori we admit that these three families of metaethical theories are equiprobable, we could then roughly affirm that we have two out of three chances to face the problem of moral undecidability and therefore the aporia it implies. This will justify for us the interest of exploring the possibility of a solution to this aporia. We will therefore propose the hypothesis according to which the practice of moral questioning in an aporetic way, considered as a non-cognitive activity, implies a certain way of being that is not arbitrary.