I am developing a construction-based model of coercion, an interpretive adjustment that people perform when there is conflict between the meaning of a word and the meaning of the grammatical pattern in which that word appears. A simple example from the domain of verbs occurs when an event-selecting adverb like suddenly is combined with a state verb like know. When we hear a nonce production like Suddenly, she knew it, we are able to construct a 'compromise construal' in which knew it means 'came to know it'. My research on coercion focuses on how context can change the combinatoric requirements of verbs (argument structure) and what kind of event or state a verb denotes (aspect). In my 2001 monograph with Josef Ruppenhofer, I use this framework to model innovative uses of the German applicative (be-) construction.I have recently begun exploring coercion from a neuroscience perspective, learning from researchers in CU's Cognitive Neuroscience of Language Laboratory. I am working with former CU LING doctoral student Les Sikos, whom I co-advised with CU Psychology professor Al Kim, to design ERP-based studies of the neural processes used to resolve coercion effects. We are particularly interested in observing the temporally dynamic patterns of recruitment of left versus right hemispheres during imaginative-language processing. We plan to explore the twin hypotheses that (a) the right hemisphere is engaged during the processing of imaginative language and (b) the degree of right-hemisphere involvement is affected by word and/or construction frequency. The basic intuition behind these hypotheses is that coerced tokens like a beer and some newspaper are so entrenched as to be indistinguishable from noncoeced tokens like a glass and some water, while nonstandard tokens like a ketchup and some pillow, even when appropriately contextualized, occasion some pondering. We believe that the studies we are designing are the first to combine ERP techniques and corpus analysis in the exploration of imaginative-language processing.