The aim of this dissertation is to advance the philosophical and linguistic understanding of conversational implicatures, but also to emphasize the important role they have, or should have, in philosophical fields such as epistemology and ethics. The first line of argument I want to pursue is linked to the role of indirect pragmatic means in the epistemological theory of testimony. According to the traditional picture testimony is delivered by statements, that is, literal utterances devoid of an indirect informational layer. The idea I want to suggest is that the understanding of epistemic testimony should be broadened to encompass conversational implicatures. This inclusion is justified by their systematic nature, the rationality and conventionalisation that lead their use. These features allow us to consider language users who stop at the level of what is said even though relevant factors point to the existence of an indirect communicational layer as uncooperative, incompetent or as liars. In addition, I will position conversational implicatures in the context of epistemic luck, arguing for the existence of communicational luck with epistemic consequences. According to the second line of thought that I present it is possible to lie with intended false conversational implicatures. In accordance with that, I present a definition of lying that includes them. According to the traditional picture, those kinds of implicatures are in fact cases of misleading, a way of creating false beliefs in the hearer that is epistemically and morally less harmful than direct lies. The main argument for this view rests on the fact that the information conveyed directly is true. In the dissertation I deny such picture claiming that lies should not be evaluated by their proximity to what is said, but by the intention of the speaker and the consequences of his act. Indirect lies, those performed with implicatures and other pragmatic means, can not be qualitatively differentiated from traditional direct lies. Different levels of directness are not enough for the claim that there are two distinct phenomena to consider especially if we acknowledge the existence of border semantic elements.