Abstract
A total of 262 noise spectra were divided empirically into three groups and the accuracy of the NRR estimate of hearing protector attenuation was evaluated for each group using 15 hearing protectors. The mean and the standard deviation of the error in the NRR estimate was most stable across the 15 protectors, when noise spectra were characterized by a simple monotonic band-level function yielding a dBC-dBA difference of 3.5 dB or less. Both the octave band attenuation function of the hearing protector and the octave band spectrum of the noise appeared to contribute to the error in the NRR estimate.