<p>This study focused on whispered speech in which a key feature, the voicing, is missing and examined whether variations in the duration of whispered consonants influence the identification of the voicing feature. We also examined lexical influences on the identification of the voicing feature during whispered speech perception by comparing words (/ʒaʁ/ <i>jars</i> “gander” vs. /ʃaʁ/ <i>char</i> “tank”) and nonwords (/ʒœʁ/ <i>jeur</i> vs. /ʃœʁ/ <i>cheur)</i>. We used an identification task on an 11-step continuum going from the fricatives /ʒ/ to /ʃ/. Participants were tested either on the minimal pair of words /ʒaʁ/-/ʃaʁ/ or on the minimal pair of nonwords /ʒœʁ/-/ʃœʁ/. For both words and nonwords, the percentage of identification of the voiced /ʒ/ consonant decreased as the duration of the consonant increased. Inversely, the percentage of identification of the voiceless /ʃ/ consonant increased as the duration of the consonant increased. The effect of consonant duration was found stronger for words than for nonwords. Together, our results strengthen the claim that durational cues of voicing are exploited during whispered speech perception, and influence consonant identification. Moreover, the stronger effect of consonant duration for words than for nonwords indicates that activation of lexical representations exerts some influence on the way whispered consonants are identified. The implications of these results regarding spoken word recognition models are discussed.</p>