Hearing protection devices (HPD) are commonly used to prevent occupational noise-induced hearing loss. There is a large body of research on hearing protection use in industry, and much of it relies on workers' self-reported use of hearing protection. Based on previous studies in fixed industry, worker self-report has been accepted as an adequate and reliable tool to measure this behavior among workers in many industrial sectors. However, recent research indicates self-reported hearing protection use may not accurately reflect subject behavior in industries with variable noise exposure. This study compares workers' self-reported use of hearing protection with their observed use in three workplaces with two types of noise environments: one construction site and one fixed industry facility with a variable noise environment, and one fixed industry facility with a steady noise environment. Subjects reported their use of hearing protection on self-administered surveys and activity cards, which were validated using researcher observations. The primary outcome of interest in the study was the difference between the self-reported use of hearing protection in high noise on the activity card and survey: (1) over one workday, and (2) over a 2-week period. The primary hypotheses for the study were that subjects in workplaces with variable noise environments would report their use of HPDs less accurately than subjects in the stable noise environment, and that reporting would be less accurate over 2 weeks than over 1 day. In addition to noise variability, other personal and workplace factors thought to affect the accuracy of self-reported hearing protection use were also analyzed. This study found good agreement between subjects' self-reported HPD use and researcher observations. Workers in the steady noise environment self-reported hearing protection use more accurately on the surveys than workers in variable noise environments. The findings demonstrate the potential importance of noise exposure variability as a factor influencing reporting accuracy.