Does the feature [sonorant] influence access to lexical entries? We created two nonwords by replacing a fricative in a real word with either another fricative (match condition) or a resonant (mismatch condition). Participants heard the nonwords and were asked to recover the real word. Errors were greater in the mismatch condition, suggesting that [-sonorant] is represented in the target fricative. Similar effects appeared when we used stops instead of fricatives. Resonant targets showed no difference in match and mismatch conditions.