The relevant exposure to environmental pollutants or relevant confounders can virtually never be measured directly in sufficient detail in a sufficient number of people included in an epidemiological study. Instead, surrogate indicators of exposure are used that are implicitly or explicitly linked by (conceptual) models to the 'relevant exposure'. Using specified indices (e.g., the coefficient of alienation and forecasting efficiency), the models of different forms can be compared and tested in special investigations. This can evaluate the criterion validity of the model through analysis of the agreement between the model estimates, based on measured surrogate indicators of the exposure, and the actual exposure. The perspective on the construct under study may further change the validity of the exposure indicator. Therefore, sub-studies evaluating the quality of exposure indicators should be considered as integral part of any study in environmental epidemiology, preferably as a pre-study or pilot study.