Conventional approaches for assessing dietary exposure to contaminants and additives in food are deterministic, using point estimates for consumption and contamination. In reality, both consumption and contamination are variable. Furthermore our knowledge of them is uncertain, e.g. due to measurement uncertainty. Conventional approaches attempt to allow for this by using worst-case assumptions or safety factors, but these are often subjective and may result either in overestimation or underestimation of the true range of exposures. Probabilistic approaches take account of variability and uncertainty by using distributions rather than point estimates for consumption and contamination. The outputs are distributions for exposure, which provide a more complete and balanced description of risk for the decision-maker. These approaches also facilitate the use of sensitivity analysis to identify those factors that impact most on exposure, and to identify areas of uncertainty where additional data will improve exposure estimates. This paper reviews examples of the application of these methods to the assessment of dietary exposure to food contaminants, including dioxins in seafood, where it was found that the greatest uncertainties relate to toxicity rather than exposure. Further work required to implement probabilistic approaches for dietary exposure assessment is discussed.