Hand and arm posture while keying is frequently mentioned as a risk factor for upper extremity musculoskeletal disorders (UEMSDs) among video display terminal (VDT) operators. However, many epidemiologic studies have not included measures of posture of VDT operators, in part, because of the difficulty of assessing posture rapidly and reliably among large numbers of subjects. For a single measure of posture to be useful for estimating dose-response relationships between posture and risk of UEMSDs, the within-subject variability of the postural measure must be smaller than the between-subject variability of the postural measure. In addition, the measure must be stable over time. We estimate the ratio of between- to within-subject variability for manual goniometry by measuring six postural angles on six occasions among 19 subjects using VDTs. For each postural angle, between-subject variability was substantially and statistically significantly larger than within-subject variability. Stability of postural measures over time was sufficient to justify a single postural measurement in epidemiologic studies. We conclude that manual goniometry can provide useful information about upper extremity posture among VDT users for use in epidemiologic studies of UEMSDs.