Slides showing nurses in different working postures were used to determine the reliability of OWAS observations. Each slide could be looked at for 3 seconds, while a new slide was shown every 30 seconds to resemble the normal practice of observation. Two observers twice scored a series of slides, some of them being identical at both viewings. To reduce effects of recall there was a time interval of 4 weeks or more between the two viewings and the slides were in a different order the second time. Different series were used to evaluate inter- and intra-observer reliability. The OWAS scores of corresponding slides were compared. In almost all comparisons percentages of agreement over 85% and kappa's over 0.6 were found, which is considered as good agreement. The procedure described seems to be a useful and simple technique to determine such reliability.