Exposure to crystalline silica ranks among the most frequent occupational exposures to an established human carcinogen. Health-based occupational exposure limits can only be derived from a reliable dose-response relationship. Although quartz dust seems to be a well-measurable agent, several uncertainties in the quantification of exposure to crystalline silica can bias the risk estimates in epidemiological studies. This review describes the silica-specific methodological issues in the assessment of exposure. The mineralogical forms of silica, the technologies applied to generate dust, protective measures, and co-existing carcinogens are important parameters to characterize the exposure condition of an occupational setting. Another methodological question concerns the measurement of the respirable dust fraction in the worker's breathing zone and the determination of the quartz content in that fraction. Personal devices have been increasingly employed over time, whereas norms for the measurement of respirable dust have been defined only recently. Several methods are available to analyse the content of crystalline silica in dust with limits of quantitation close to environmental exposure levels. For epidemiological studies, the quartz content has frequently not been measured but only calculated. To develop a silica-dust database for epidemiological purposes, historical dust concentrations sampled with different devices and measured as particle numbers have to be converted in a common exposure metric. For the development of a job-exposure matrix (JEM), missing historical data have to be estimated to complete the database over time. Unknown but frequently high-exposure levels of the past contribute largely to the cumulative exposure of a worker. Because the establishment of a JEM is crucial for risk estimates, sufficient information should be made accessible to allow an estimation of the uncertainties in the assessment of exposure to crystalline silica. The impressive number of silica dust measurements and the evaluation of methodological uncertainties allow recommendations for a best practice of exposure assessment for epidemiological studies.