This paper describes a new analysis of data from the 1968-72 National Institute for Occupational Safety & Health (NIOSH) Occupational Noise and Hearing Survey (ONHS). The population consisted of 1172 (792 noise-exposed and 380 "controls") predominately white male workers from a cross section of industries within the United States. The analysis focused on how risk estimates vary according to various model assumptions, including shape of the dose-response curve and the amount of noise exposure among low-noise exposed workers (or controls). Logistic regression models were used to describe the risk of hearing handicap in relation to age, occupational noise exposure, and duration exposed. Excess risk estimates were generated for several definitions of hearing handicap. Hearing handicap is usually denoted as an average hearing threshold level (HTL) of greater than 25 dB for both ears at selected frequencies. The frequencies included in the biaural averages were (1) the articulation-weighted average over 1-4 kHz, (2) the unweighted average over 0.5, 1, and 2 kHz, and (3) the unweighted average over 1, 2, and 3 kHz. The results show that excess risk estimates for time-weighted average sound levels below 85 dB were sensitive to statistical model form and assumptions regarding the sound level to which the "control" group was exposed. The choice of frequencies used in the hearing handicap definition affected the magnitude of excess risk estimates, which depended on age and duration of exposure. Although data were limited below 85 dB, an age-stratified analysis provided evidence of excess risks at levels ranging from 80 to 84 dB, 85-89 dB, and 90-102 dB. Due to uncertainty in quantifying risks below 85 dB, new data collection efforts should focus on better characterization of dose-response and longitudinal hearing surveys that include workers exposed to 8-hour time-weighted noise levels below 85 dB. Results are compared to excess risk estimates generated using methods given by ANSI S3.44-1996.