Occupational exposure limits (OELs) for hazardous substances are usually defined as exposures averaged over 8 h or 15 min which must not be exceeded. In the European Union (EU), examples are the limits in the Chemical Agents Directive (EU, 1998) and the Carcinogens Directive (EU, 2004). But if we define the limits as exposures which must never be exceeded, we disregard the complex patterns of real exposure. It is common experience that exposure distributions often include results which are several times the median value, not because of a failure of control but because there is a statistical chance that the many factors which determine exposure combine in a way which produces an outlying result. Do these outliers constitute non-compliance or are they freak happenings which can be disregarded, and if so what is the criterion for distinguishing ‘freak’ measurements from significant results which must be taken into account? This has been a problem ever since OELs moved from being ‘guidelines to be used by professional industrial hygienists’ (ACGIH, 2009) to being sharp cutoffs of legally allowable exposure, and many attempts have been made to write guidance on how to demonstrate compliance with the limits. Probably, the most famous is of Leidel et al. (1977), published by the US National Institute of Occupational Safety and Health (NIOSH). In a commentary last year, Ramachandran (2008) described this guidance and its limitations and NIOSH’s plans to revise it. This work continues. In Europe, the most prominent guidance is the European Standard EN689 (CEN, 1995). A couple of years before that was published, the British Occupational Hygiene Society (BOHS) had produced a Technical Guide on the problem (BOHS, 1993). These documents all try to deal with the underlying difficulty that an OEL defined as a sharp threshold is out of step with the laws of physics that determine how exposure varies in the workplace. A second problem is that individuals apparently doing the same job in the same way can differ substantially in exposure, which is not fully taken into account in EN689 and the earlier documents. The two problems of variability from shift to shift and from worker to worker mean that a statistically valid test of compliance with the OEL requires far more measurements than would normally be considered practicable. Two years ago, a joint Working Party on the problem was appointed by the BOHS and Nederlandse Vereniging voor Arbeidshygiene, the Dutch Occupational Hygiene Society (NVvA). This has produced draft guidance which is open for comment on the societies’ websites until the end of 2009 (http:// www.bohs.org and http://www.arbeidshygiene.nl/ index.php). The guidance is not finished, but the Working Party felt that it would be useful to expose it to public comment at this stage. Its approach to the fundamental problem is to emphasize the importance of the principles and methods of control, not just compliance with OELs. In this, it follows the European Chemical Agents Directive (EU, 1998), which in Articles 5 and 6 requires application of good control practice alongside but independently of compliance with exposure limits. This dual approach is also emphasized in the British COSHH regulations (HSE, 2005). However, when it comes to measuring compliance, the statistical problem has to be faced. Like most other guidance, the British–Dutch draft uses a criterion that 95% of exposures should be below the OEL, arguing that this represents good professional practice even if it does not strictly guarantee compliance with the law. The main test involves division of *Author to whom the correspondence should be addressed. Tel: þ44-1332-298101; fax: þ44-1332-298099; e-mail: editor@ogs.org.uk