In this article, I argue against a widely accepted model of utterance
interpretation, namely the LS model (literality-based serial model), a
ccording to which the literal interpretation of an utterance (the prop
osition literally expressed by that utterance) must be computed before
nonliteral interpretations can be entertained. Alleged arguments in f
avor of this model are shown to be fallacious, counter-examples are pr
ovided, and alternative models are sketched.