Data-logging noise dosimetry was used to assess the exposure levels of electricians working for a major electrical subcontractor in Washington State at five sites using four types of construction methods. Subjects documented activities and work environment information throughout their work shift, resulting in an activity/exposure record for each of the 174 full-shift samples collected over the 4-month duration of the study. Over 24% of the TWA samples exceeded 85 dBA; 5.2% exceeded the federal Occupational Safety and Health Administration permissible exposure limit of 90 dBA. The National Institute for Occupational Safety and Health exposure metric, which specifies a 3-dB ER, was also utilized; using this metric, 67.8% of the samples exceeded 85 dBA and 27% exceeded 90 dBA. Subjects were directly observed for a subset of 4469 min during which more detailed activity and environmental information was recorded. Linear and logistic regression models using this subset were used to identify the determinants of average exposure, and exposure exceedences, respectively. These models demonstrated the importance of multiple variable modeling in interpreting exposure assessments, and the feasibility and utility of modeling exposure exceedences using logistic regression. The results further showed that presumably quiet trades such as electrician are at risk of exposure to potentially harmful noise exposures, and that other workers' activities and the general environment contribute substantially to that risk. These results indicate that noise control strategies will have to address the construction work environment as an integrated system.