This paper shows that Bach's distinction between three levels of representation of utterances -'what is said', 'what is implicated' and 'impliciture'- generates semantic and pragmatic problems as the motivations about the literality of utterances/sentences underlying the distinction are in some cases overdetermined, whilst in others underspecified. Alternative analyses are offered, following Carston's proposal, which make it superfluous to consider more than one category of directly communicated content.