The experiment reported here employed a word-onset gating technique to investigate the role of prosody in word recognition. Subjects were asked to identify words based on word onsets alone, word onsets followed by information about the word duration, or word onsets followed by information about full word prosody (i.e., both duration and stress). Results showed that words were correctly recognized with significantly less segmental onset information when word prosody was available to the subjects. Consistent with this finding, prerecognition error responses reflected correct length and prosody with less onset phonology when prosody information was provided in the stimulus than when only length information was provided. The findings of this experiment confirm the importance of word prosody for spoken word recognition.