Industrial exposure varies distinctly both between persons and for each person over time. It is often not possible to measure individual exposure repeatedly due to high costs. Therefore, a method for assessment of exposure is needed that accounts for inter- and intraindividual variability. We consider a strategy suggested by Preller et al. (1995, Scandinavian Journal of Work, Environment, and Health 21, 504-512), the idea of which is to predict exposure on several days via a linear model using additional variables as regressors. Those additional variables are easier to obtain than exposure measurements and are assumed to influence exposure. The paper gives a theoretical proof of the use of this method. An example is given using toluene exposure data from a study in a rotogravure printing plant.