Based on the results from a number of eye-tracking experiments, Frisson and Pickering (1999) and Pickering and Frisson (2001) proposed a model for the on-line processing of words with semantically related senses. According to this model, only the underspecified, schematic meaning of a word with multiple senses is activated initially. This underspecified meaning encompasses all related senses that are established in someone's lexicon. Once this underspecified meaning has been used to assign a semantic value, it can be followed by a homing-in stage in which context is used to arrive at the contextually appropriate sense. This article discusses the implications of the underspecification model with respect to the processing of figurative language and asserts that this model gives a better account of how incremental semantic processing can proceed smoothly and flexibly. In conclusion, the model is related to the linguistic distinctions of polysemy and monosemy, and it is argued that, at this moment in time, there is no compelling evidence to postulate more than one underspecified meaning.