This thesis is about mental and linguistic representation. It addresses the question of how people understand words, how they are used to convey meaning and how this meaning is acquired. I take a theoretical approach to the issue, and try to develop a constructive account on the basis of insights from Relevance Theory and Jerry Fodor s construal of informational semantics. I investigate the theoretical viability of two hypothesis underlying work in relevance theoretic pragmatics: 1) the input to pragmatic processing is in most cases an impoverished semantics in the form of encoded concepts (=radical pragmatics) 2) these concepts are mental items which have their content in virtue of standing in a constitutive relation to something in the world (=semantic externalism) Against these hypotheses the thesis raises two main challenges: one of which is explaining how the notion of the words-concepts relation being one of encoding squares with data from cross-linguistic lexical variation. The second challenge is how concepts representing abstract (non-perceivable) entities are acquired. As a positive contribution, the thesis makes two main claims. The first is that the relationship between words and concepts should be thought of as one of potential activation. By this I mean that words can on a given occasion be used to express any of a number of candidate concepts, but none of these are seen as the word s encoded or default meaning. A consequence of this view is that lexical items are not plausibly seen as having any meaning inherently. Rather, words get their meaning in context, by way of a semantically constrained pragmatic process that picks out the most relevant of a range of corresponding concepts from occasion to occasion. The second main claim is that the meaning of many words are epistemically incomplete, in that the concepts they potentially activate have their meaning as a result of speakers relying on each other to entities that are not or cannot be perceived. The idea here is that language speakers to a great extent defer to other people for the content of a concept.