The objectives of this research were to evaluate the validity of real-time observational assessment of body postures and to provide information needed for improving ergonomic exposure assessment strategies in construction work. The first study involved the development and evaluation of an electronic direct postural measurement system designed to measure shoulder, knee and trunk postures. In a laboratory study on 5 subjects, the electronic inclinometers predicted shoulder, trunk and knee flexion to within 5 to 10 degrees of accuracy and the lumbar motion monitor (LMM) predicted trunk lateral bending and torsion to within 3 degrees of accuracy with no or very simple calibration procedures. The second study involved the evaluation of discrete-interval observations for the assessment of body postures during construction tasks, by comparing categories of body posture recorded with two work-sampling approaches (PATH and a simplified version of PATH) to measurements obtained with the electronic postural assessment system and video analysis. Five subjects were each observed performing 3 of 6 simulated construction tasks. Overall, agreement among the work-sampling and the electronic methods was high, although there were some notable differences in the measured frequency of some leg and trunk postures. The third study involved the evaluation of the efficacy of the task-based assessment approach through the examination of between-worker and within-worker components of exposure variability in a sample of construction tasks. In addition, bootstrapping (a computer resampling technique) was used to evaluate how the number of days of assessment affected the reliability of the exposure measure for groups of workers performing the same task. Exposures often varied greatly among tasks of a construction operation, indicating that the task-based approach is useful to improve precision in exposure assessment The between-worker component of exposure variability, for the most part, was much smaller than the within-worker exposure variability over several days supporting the hypothesis of high day-to-day exposure variability during construction work. Bootstrapping revealed that low frequency exposure variables may require as few as 1 or 2 days of assessment (particularly for low-variability exposures), while exposures with higher frequency may require 6 or more days (particularly for high-variability exposures) to allow a reliable assessment of exposure frequencies for construction tasks. This research demonstrates how work sampling approaches can be used to obtain valid and reliable measurements of ergonomic exposures. The results have implications for both the evaluation of control measures in the work place and for exposure assessment in epidemiologic studies.