The attenuation provided by a hearing protection device (HPD) in the field is usually estimated by applying a derating factor to the laboratory-determined noise reduction rating (NRR) of the HPD. However, attenuation is highly dependent on individual-specific HPD fit. Prediction of an individual's attenuation depends on the accuracy of the measurement system and the variability of attenuation over time (e.g. after HPD refitting). Variability in attenuation and attenuation test systems has not been adequately characterized to allow for such an assessment. This study compared attenuation measurements made with two systems, Real-Ear-at-Threshold (REAT) and Microphone-in-Real-Ear (MIRE), on 20 workers using two earplugs (foam and custom-molded). Workers' perceptions of the earplugs were also evaluated. Individuals' attenuation results were summarized as personal attenuation ratings (PARs, similar to NRRs). Variability in PARs from between-subject, within-subject and within-day (i.e. repeated tests on a subject without earplug refitting) differences was assessed and used to present the lower confidence limit, or uncertainty factor (UF), of an average individual's attenuation. The custom-molded earplug PARs achieved a higher mean percentage of labeled attenuation than did the foam earplug with both test systems. The custom-molded earplugs also had higher overall acceptance among workers. MIRE PAR levels were lower than REAT levels for both earplugs, but the relationship between the two test systems was highly variable. The MIRE system had lower within-day variability than the REAT system. One individual's MIRE results were highly influential; removal of these results greatly reduced the UF for the custom-molded earplug/MIRE combination. UFs ranged from 8.8 to 13.5 dB. These findings highlight the importance of evaluating variability in individual-specific protection results for personal protective equipment like HPDs, rather than relying on single measurements.