Knowledge of the natural background concentrations of groundwater constituents is important for the management of groundwater resources, particularly for the assessment of groundwater contamination and the establishment of clean-up goals and regulatory target levels. In recent years, an increasing number of studies have assessed the natural background concentrations of dissolved constituents in groundwater using a variety of different methods, each with its own assumptions, advantages and limitations. The objective of this paper is to provide a methodological basis for improving the estimation of natural background concentrations of groundwater constituents. To this end, this paper critically reviews the different approaches used to determine natural background concentrations of dissolved constituents in groundwater. In addition, two regional case studies of fluoride in Canadian groundwater are presented to illustrate the estimation of background concentrations for natural groundwater constituents. The review of existing methods shows that the use of pristine groundwater samples is not possible in many cases, due to the widespread influence of human activities. The widely used pre-selection method can provide misleading results because of inadequate selection criteria and poor statistical significance associated with the reduction of the dataset. A variety of model-based methods have been developed, but these methods are all based on assumptions that cannot be verified. Relying on the user's experience and previous knowledge of the groundwater system, exploratory data analysis has many advantages and can be applied for both anthropogenic and natural constituents. The case studies show that the exploratory data analysis approach provides critical information to determine the sources of groundwater constituents and to properly delineate groundwater bodies for which background values will be established. Natural background concentrations should always be considered as theoretical values due to their spatio-temporal variability and scale dependence, and thresholds as concentration values above which further investigation is required.