The present work assesses the effect of low-pass filtering on masking release in Mandarin sentence recognition. Mandarin sentences were masked by steady-state noise and two-talker masker, processed by low-pass filtering, and presented to normal-hearing listeners to recognize. Negative release was observed at low-pass filtering cutoff frequencies ≤750 Hz, with listeners showing better performance under the steady-state noise condition than under the two-talker masker condition. The negative release was affected by the input signal-to-noise ratio. Language characteristics might partially account for the negative release observed with Mandarin speech.