In this study we develop and test the hypothesis that spoken word recognition in context is facilitated by anticipatory assimilation but inhibited by perseveratory assimilation. Three experiments are presented. Experiment I shows that homorganic final nasals provide sufficient perceptual cues to reconstruct the place of articulation of the following stop which was electronically eliminated from the stimulus. In experiment II the target stops were not eliminated from the stimulus but presented - through digital cross-splicing - after homorganic versus heterorganic nasals. Removing anticipatory assimilation proved detrimental to the recognition of the target stops, thereby showing a positive effect of anticipatory assimilation on word recognition. In experiment III cross-splicing was used to simulate blocking of perseveratory assimilation of voicing. Here the results show that targets were recognised better when assimilation was blocked, showing that, indeed, perseveratory assimilation inhibits spoken word recognition.