While observing six simulated construction tasks in the field, trained analysts recorded arm, trunk and leg postures categorically with two fixed-interval observational protocols. Observations were compared to measurements obtained with an electronic postural assessment system coupled with video analysis. The electronic postural assessment system consisted of electronic inclinometers to measure upper arm posture, knee flexion and trunk flexion, coveralls to house the inclinometer wiring, and an eletrogoniometric system to measure trunk lateral bending and twisting. Video analysis included frozen-frame analysis that corresponded to the moment of observation and simulated real-time analysis. Measurements were made on five male participants who each performed three tasks representative of construction laborers' work. Agreement among the observational and reference methods was generally high, although significant differences in measured frequency of exposure existed for knee flexion, trunk lateral bending and trunk twisting. The results suggest that, under appropriate conditions, discrete observations can be used to obtain reasonably accurate estimates of exposure frequency for broad categories of certain body postures.