Exposure to wood dust was measured among 26 carpenters at 13 building projects. Sampling days were chosen randomly. Individual tasks, based on technology applied and material used during a working day, were sampled separately. From these task-based measurements, 8 h time-weighted average concentrations were calculated. Sampling was performed in accordance with a protocol that was developed by the carpentry and furniture industry and which was especially designed for sampling of wood dust. Eight hours time-weighted average exposure to wood dust ranged from 0.8 to 11.6 mg m(-3) with a geometric mean (GM) of 3.3 mg m(-3) and a geometric standard deviation (GSD) of 2.1. The probability of exceedance of the OEL, when comparing the estimated concentrations against the Dutch OEL of 2 mg m(-3), was 75%. The highest exposures were measured during sawing of Cempanel sheets. Task-based measurements showed lowest exposures when working outdoors (n = 11, AM = 2.2 mg m(-3)), but even then 5 out of 11 task-based exposures exceeded 2 mg m(-3). Indoors the exposure was 5.2 mg m(-3) (AM, n = 29) and when working both indoors and outdoors exposure was 16.2 mg m(-3) (AM, n = 4). In conclusion, long-term average exposure to wood dust among carpenters at construction sites is more than 1.5 times the present occupational exposure limit of 2 mg m(-3). The estimated probability of exceedance of the OEL was 75% and a reduction of exposure with a factor 5 is needed to bring the probability of exceedance below 5%. It is intended to lower the exposure limit to 1 mg m(-3) by 1 January 2007. In that case the probability of exceedance of the OEL is 95% and a reduction of exposure with a factor 10 is needed to bring the probability of exceedance below 5%. This can be achieved by using alternative materials, preparation of building material in workshops equipped with exposure controls, alternative equipment and improved ventilation and good housekeeping.