Schlehofer et al.1 conducted a case-control study (1,178 cases of glioma, 1,987 community based controls) in 8 collaborating centres regarding occupational exposures. Occupational histories were obtained and data analysed by reference to several a priori occupational groups including electrical/electronic work, metal work, etc. An increased risk was found for men with low grade glioma and metal work and for women in the food processing industry. They concluded that “the results generally do not provide evidence of a strong association between occupational exposures and glioma development.” This conclusion should be questioned as it critically depends on the accuracy of the occupational histories. Histories were obtained by interview from cases and controls of every job held for more than 3 months, and exposure assessed by: (i) the kind of industry; (ii) “the precise characterisation of the occupational activity, including duration in years and the hours per week”; and (iii) a list of 78 substances of interest was presented to the interviewee. An exposure to a category was defined to occur if working lifetime exposure was >5 years. Also exposures were graded depending on the number of hours of exposure as none, low, medium or high. Exposures occurring over 25 years previous to and <5 years before diagnosis were excluded. Accuracy of recall of the hours worked and the exposures incurred up to 25 years ago depends on a lucid memory. For the cases memory is likely to be impaired by the brain tumor per se, the psychological distress of diagnosis of a fatal and painful illness, and the effects of treatment by deep X-ray therapy and surgery. No assessment of intactness of memory, such as application of a Mini-Mental State Examination, was conducted to ensure competence and consistency of accurate recall in subjects in the 8 collaborating centres. Both memory loss and recall bias will corrupt the data. There is no evidence presented of any attempt to validate the occupational histories and exposure data so obtained. Such variable data extraction will lead to considerable misclassification of exposures and increase the likelihood of a null finding. In a companion article Schlehofer et al.2 report that of the 1,178 glioma cases a proxy was interviewed in 315 (26%) cases because subjects were “not available.” The ability of a proxy (presumably a spouse or possibly numerous work mates) to recall the details of exposures by the case in hours up to 25 years ago with accuracy is highly doubtful. Thus a quarter of the exposure data is likely to be substantially inaccurate. The above comments are particularly pertinent to the electromagnetic field (EMF) exposures. Unfortunately the authors incorrectly refer to extremely low frequency fields (ELF, 1–300 Hz) as “low frequency” and to radio frequency radiation (RFR, 3 kHz–3 GHz) as “high frequency.” Standard terminology is helpful to the reader in scientific studies. The history-taking method used by the authors assumes workers will recognise when they are exposed to EMF. This is questionable. For example, it is doubtful many workers will realise that a powerful industrial electric motor (415 V, 50 Hz) enclosed behind the cowling of a machine with which they are working (i.e., when metal stamping) is emitting fields incident upon them (metal cowling is transparent to magnetic fields). Similarly how would ground crew at an airport know when they are being repeatedly exposed to intense radar from the nose cone of taxiing aircraft? Moreover, even when a source is recognised, because the intensity of fields decreases with the square of the distance, it invites inaccurate assessment and hence misclassification of exposure. The authors correctly state that direct measurement of 50 Hz magnetic field is recognised as the proper way to measure these exposures but do not alter their overall conclusion of no effects from these exposures even though recognising this major deficiency. The study has major flaws regarding the crucial occupational history taking from cases, particularly from the proxies, and hence the exposure analysis is vitiated. Because of this, no valid conclusions about the presence or absence of an association between these exposures and glioma, particularly regarding EMF, can be drawn. It is regrettable the abstract does not present more information on the data extraction problems, some of which the authors have recognised, to alert the reader to these shortcomings. Yours sincerely, Bruce Hocking.