The performance of four insert earplugs was evaluated by determining the Noise Reduction Rating (NRR) and the Subject-Fit Noise Reduction Rating [NRR(SF)]. The NRR and NRR(SF) were calculated from real-ear attenuation at threshold (REAT) data collected using the experimenter-fit protocol described in the now-rescinded ANSI S3.19-1974 (American National Standards Institute, 1974) and the subject-fit protocol of the recently revised ANSI S12.6-1997 (American National Standards Institute, 1997) standards for REAT measurement. A comparison of the experimenter-fit and subject-fit REAT performance was conducted using four pools of subjects, one pool per protector. Each device was tested with at least 20 subjects, the minimum size necessary to estimate the NRR(SF) for an earplug. The REAT was measured with third-octave narrowband noise stimuli for center frequencies at 0.125, 0.25,0.5, 1, 2, 3.15, 4, 6.3, and 8 kHz. The REAT means and standard deviations were compared with the manufacturer data. This study showed that the NRR(SF) is typically lower than the NRR and that the NRR(SF) is not well-predicted by the NRR derating schemes recommended by the National Institute for Occupational Safety and Health and required by the Occupational Safety and Health Administration. The difference between the present NRR on hearing protector labels and the NRR(SF) is sufficiently large and unpredictable enough to render the application of derating schemes meaningless even though these schemes attempt to account for the difference between the laboratory and real-world outcomes. The only way to provide a protector noise rating that is predictive of a real-world outcome is to retest the protector according to the subject-fit method of ANSI S12.6-1997 (American National Standards Institute, 1997).