Exposure assessment is one of the key parts of the risk assessment process. Only intake of toxicologically significant amounts can lead to adverse health effects even for a relatively toxic substance. In the case of chemicals in foods this is based on three major aspects: (i) how to determine quantitatively the presence of a chemical in individual foods and diets, including its fate during the processes within the food production chain; (ii) how to determine the consumption patterns of the individual foods containing the relevant chemicals; (iii) how to integrate both the likelihood of consumers eating large amounts of the given foods and of the relevant chemical being present in these foods at high levels. The techniques used for the evaluation of these three aspects have been critically reviewed in this paper to determine those areas where the current approaches provide a solid basis for assessments and those areas where improvements are needed or desirable. For those latter areas, options for improvements are being suggested, including, for example, the development of a pan-European food composition database, activities to understand better effects of processing on individual food chemicals, harmonisation of food consumption survey methods with the option of a regular pan-European survey, evaluation of probabilistic models and the development of models to assess exposure to food allergens. In all three areas, the limitations of the approaches currently used lead to uncertainties which can either cause an over- or underestimation of real intakes and thus risks. Given these imprecisions, risk assessors tend to build in additional uncertainty factors to avoid health-relevant underestimates. This is partly done by using screening methods designed to look for "worst case" situations. Such worse case assumptions lead to intake estimates that are higher than reality. These screening methods are used to screen all those chemicals with a safe intake distribution. For chemicals with a potential risk, more information is needed to allow more refined screening or even the most accurate estimation. More information and more refined methods however, require more resources. The ultimate aims are: (1) to obtain appropriate estimations for the presence and quantity of a given chemical in a food and in the diet in general; (2) to assess the consumption patterns for the foods containing these substances, including especially those parts of the population with high consumption and thus potentially high intakes; and (3) to develop and apply tools to predict reliably the likelihood of high end consumption with the presence of high levels of the relevant substances. It has thus been demonstrated that a tiered approach at all three steps can be helpful to optimise the use of the available resources: if relatively crude tools - designed to provide a "worst case" estimate - do not suggest a toxicologically significant exposure (or a relevant deficit of a particular nutrient) it may not be necessary to use more sophisticated tools. These will be needed if initially high intakes are indicated for at least parts of the population. Existing pragmatic approaches are a first crude step to model food chemical intake. It is recommended to extend, refine and validate this approach in the near future. This has to result in a cost-effective exposure assessment system to be used for existing and potential categories of chemicals. This system of knowledge (with information on sensitivities, accuracy, etc.) will guide future data collection.