Exposure data from self-administered questionnaires on manual materials handling and work postures were validated in relation to direct measurements and systematic observations on 39 men and 58 women representing 45 different occupations. The agreement was tested at a dichotomous level and, when possible, with quantification of duration or frequency. At the dichotomous level the agreement was "acceptable" for nine variables concerning work postures and the handling of loads weighing > 5 kg. No variable showed "acceptable" agreement when the duration or the frequency was quantified in more detail (4- to 6-point scales). Musculoskeletal complaints seemed to introduce a differential bias for some lifting variables. Thus some variables for postures or the handling loads of > 5 kg may, under certain conditions, be acceptable for use in epidemiologic studies when the relative risks are high. However, self-reported exposure seems to be too crude if more-detailed information is required.