It is increasingly common for philosophers to rely on the notion of an idealised listener when explaining how the semantic values of context-sensitive expressions are determined. Some have identified the semantic values of such expressions, as used on particular occasions, with whatever an appropriately idealised listener would take them to be. Others have argued that, for something to count as the semantic value, an appropriately idealised listener should be able to recover it. Our aim here is to explore the range of ways that such idealisation might be worked out, and then to argue that none of these results in a very plausible theory. We conclude by reflecting on what this negative result reveals about the nature of meaning and responsibility.