Measurements of exposure to vibrations from hand-held tools are often conducted on a single occasion. However, repeated measurements may be crucial for estimating the actual dose with good precision. In addition, knowledge of determinants of exposure could be used to improve working conditions. The aim of this study was to assess hand-arm vibration (HAV) exposure during different grinding operations, in order to obtain estimates of the variance components and to evaluate the effect of work postures. Ten experienced operators used two compressed air-driven angle grinders of the same make in a simulated work task at a workplace. One part of the study consisted of using a grinder while assuming two different working postures: at a standard work bench (low) and on a wall with arms elevated and the work area adjusted to each operator's height (high). The workers repeated the task three times. In another part of the study, investigating the wheel wear, for each grinder, the operators used two new grinding wheels and with each wheel the operator performed two consecutive 1-min grinding tasks. Both grinding tasks were conducted on weld puddles of mild steel on a piece of mild steel. Measurements were taken according to ISO-standard 5349 [the equivalent hand-arm-weighted acceleration (m s(-2)) averaged over 1 min]. Mixed- and random-effects models were used to investigate the influence of the fixed variables and to estimate variance components. The equivalent hand-arm-weighted acceleration assessed when the task was performed on the bench and at the wall was 3.2 and 3.3 m s(-2), respectively. In the mixed-effects model, work posture was not a significant variable. The variables 'operator' and 'grinder' together explained only 12% of the exposure variability and 'grinding wheel' explained 47%; the residual variability of 41% remained unexplained. When the effect of grinding wheel wear was investigated in the random-effects model, 37% of the variability was associated with the wheel while minimal variability was associated with the operator or the grinder and 37% was unexplained. The interaction effect of grinder and operator explained 18% of the variability. In the wheel wear test, the equivalent hand-arm-weighted accelerations for Grinder 1 during the first and second grinding minutes were 3.4 and 2.9 m s(-2), respectively, and for Grinder 2, they were 3.1 and 2.9 m s(-2), respectively. For Grinder 1, the equivalent hand-arm-weighted acceleration during the first grinding minute was significantly higher (P = 0.04) than during the second minute. Work posture during grinding operations does not appear to affect the level of HAV. Grinding wheels explained much of the variability in this study, but almost 40% of the variance remained unexplained. The considerable variability in the equivalent hand-arm-weighted acceleration has an impact on the risk assessment at both the group and the individual level.