We have reviewed the relevant issues in the exposure assessment of disinfection by-products (DBPs) of chlorination for epidemiological and health risk assessment. Various DBPs can be detected in drinking water and swimming pools, and the reported levels show a considerable range, but were generally below the current health standard for total trihalomethanes (TTHMs) (100 microg/l). Relatively little information is available on the correlation between the various DBPs in drinking water and in swimming pools. Chloroform was generally, but not always, the most predominant DBP. In epidemiological studies, TTHM levels have been used as an indicator for total DBP load, even though TTHM levels do not always correlate well with individual DPBs. Factors such as residence time, temperature, pH, organic content, including humic and fulvic acid and bromide levels affect the composition and levels of DBPs. Although there are biomarkers of DBPs, mainly for chloroform and more recently for the other volatile trihalomethanes (THMs) and the nonvolatile haloacetic acids (HAAs) such as trichloroacetic acid (TCAA) and dichloroacetic acid (DCAA), they have not been used in epidemiological studies. The THMs have been measured in exhaled breath and serum, while the HAAs have been measured in urine. These biomarkers have been useful to estimate the actual uptake of the DBPs and the relative contribution of various exposure routes. Physiologically based pharmacokinetic (PBPK) models exist for, e.g. chloroform, but their main target organs are the kidney and liver and they have not been used in epidemiological studies. Tap water ingestion, showering, bathing, swimming, boiling water and dishwashing are all activities that have been associated with the uptake of DBPs, and considerable variation in these activities has been observed between people. No studies have reported on the correlation between human uptake of DBPs and water-zone mean estimates, but various studies found a good correlation between THM concentrations in exhaled breath and THM concentrations in water during showering and swimming. In general exposure assessment in epidemiological studies has been limited which complicates the interpretation. These findings have implications for epidemiological studies, particularly with reference to Berkson and classical error type models, study power, attenuation and precision of health-risk estimates and study efficiency. Recommendations are made for further areas of study.