The phenomenon of descriptive and metalinguistic negation has been debated for a long time from a theoretical perspective. On the one hand, there are defenders of the ambiguist approach to negation, in which the descriptive negation basically serves to deny an utterance’s propositional content, and that this takes place by default (Horn 1985; 1989; Burton-Roberts 1989), while the metalinguistic negation surfaces only when the descriptive negation cannot be applied, and targets the non-truth-conditional contents of the utterance (e.g. implicatures, its register, its morphology or its phonology). Only the former is truth-functional, and the latter is claimed to be non-truth-functional as it does not operate on propositions. On the other hand, there are proponents of the non-ambiguist approach, who maintain that both types of negation are truth-functional since, in the case of metalinguistic negation, the process of pragmatic enrichment guarantees that the full proposition on which negation can operate will be reached (Carston 1996; 2002; Noh 1998; 2000; Moeschler 2010; 2013; 2017). Regarding processing, the ambiguist account predicts that it will take more time to treat metalinguistic negation because it always occurs as the second of two steps; in contrast, the non-ambiguist account makes no such prediction, since the interpretation of negation is contextually driven and the right context will issue the correct interpretation from the start. This paper will be devoted to the presentation of two self-paced reading experiments and of one offline elicitation experiment we carried out on French descriptive and metalinguistic negation. Our findings provide evidence in favor of the non-ambiguist approach.