We use a vector space model (VSM) to simulate semantic relat-edness effects in sentence processing, and use this connectionto predict N400 amplitude in an ERP study by Federmeierand Kutas (1999). We find that the VSM-based model is ableto capture key elements of the authors’ manipulations and re-sults, accounting for aspects of the results that are unexplainedby cloze probability. This demonstration provides a proof ofconcept for use of VSMs in modeling the particular contextrepresentations and corresponding facilitation processes thatseem to influence non-cloze-like behavior in the N400.