Traditional stability-performance tradeoffs pose limitations on active noise reduction (ANR) using feedback control, which are evident in circumaural communication headsets. In these systems, the cavity resonant behavior necessitates low feedback gains, redcing performance. Feedforward ANR using a Lyapunov-tuned least-mean-square filter dramatically enhances noise reduction performance compared with feedback ANR [Cartes et al., J. Acoust. Soc. Am. 111, 1758–1771 (2002); Collier et al., NOISECON (2003)]. However, feedforward performance is sensitive to the noise source stationarity, and the frequency-dependent forward path gain reduces stability margins. This paper presents experimental results for a hybrid feedforward-feedback ANR system, which enhances performance and gain margins for both stationary and nonstationary noise. Algorithms are optimized and measurements are performed with Thayer’s rapid prototyping system and associated low-frequency acoustic test cell using a circumaural hearing protector. In the frequency range 50–800 Hz, the hybrid system provides an average of 27 dB active noise reduction (40 dB total) for tonal noise and 17 dB reduction (32 dB total) for nonstationary noise. Performance below 100 Hz improves by as much as 15 dB over that of individual control components, and gain margin of the hybrid system improves substantially over individual feedforward or feedback components.