Studies evaluated muff-type hearing protection devices (HPDs) in a mining environment. Noise reduction measurements were made using miniature microphones inside and outside the HPD cup. Laboratory tests conducted prior to field evaluations indicated that the results of this physical method were similar to the results of the American National Standards Institute (ANSI) real-ear-attenuation-at-threshold method. A total of 23 models of HPDs and 545 machines (20 different machine types) were evaluated in the field phase, resulting in 1265 HPD evaluations. The effectiveness of each HPD model in terms of dBA noise reduction is presented as a function of the metric (C-A), which characterizes the spectrum of the machine noise. The dBA reductions for various combinations of HPD models and machine types are also presented. The results show that the field performance of muff-type HPDs is significantly less than that specified by the Environmental Protection Agency Noise Reduction Rating (NRR), especially for low-frequency noise sources such as internal combustion engines. Thirty-two percent of the operators of such machines observed noise reductions of 10 dBA or less. Across all machine types, 20% of the workers realized an observed noise reduction of 10 dBA or less. It is concluded that the NRR grossly overestimates HPD performance. In addition, the NRR is not a good indicator for comparing HPD models since in many instances HPDs with lower NRRs out-performed those with higher NRRs. The laboratory-derived NRRs' failure to predict the HPD field performance was related primarily to the inaccuracy of the basic laboratory attenuation data from which the NRRs were computed, and to a lesser extent to shortcomings in the NRR computational procedure.