AbstractIn recent years we have seen major advances in the variety and quality of ground water sampling instrumentation, in almost every case involving increases in sensitivity of one or more orders of magnitude. Software number‐crunching ability brings us the capability of running sophisticated mathematical and statistical programs, while modern analytical chemistry routinely makes determinations of metals, organics and radioactive elements accurate in the parts per billion range.The construction of sample points rarely supports these levels of precision. The well is the basic sampling machine for the ground water industry and its ability to produce a truly representative sample underpins all data and conclusions. Regardless of the type of instrumentation employed, a sample point that cannot be shown to be truly representative and that fails to provide a series of reliable samples over the long term–let us say 25 years minimum–is a negative investment.In the authors’professional sampling work, it has been observed that, wherever it is possible to improve the quality of the ground water sample point, a reduction in the measured level of pollutant elements in the sample is found and the confidence level improves. Volatile organic values are the sole exception to this observed reduction in concentration, the reason being that sample point and method improvements capture the fully representative water sample, thus reducing volatile losses.It is not possible to address all hydrogeologic environments and sampling methods in this review. Instead, features were selected that had been found to be important in commercial, high‐precision, water quality sampling in Western U.S. environments. In this area, water movement is slow and little or no exchange of water occurs between wells and the surrounding aquifer without artificial stimulation. Water quality stabilization requires multiple casing volume evacuations in wells ranging to 1,000 feet deep. High‐volume pumping or mechanized bailing is an economic requirement. Dedicated installations cannot supply the volume of water required for comprehensive analytical programs.It is the nature of these sources, as well as practical recommendations for corrective action, that will be addressed in this article.Any attempted improvement in utilization of data by mathematical processing or statistical treatment will be inadequate because:• Samplings are infrequent and often accumulate at rates of only two or four sample suites per year. Seasonal variations may obscure evidence of passage of a low‐intensity pollution front and some years may elapse before the true situation becomes apparent from statistical analyses. By such time, the problem pollutant may have escaped from the treatment area.• Powerful biases occur due to the failure of the sample point to eliminate false contributions to the sample from sources other than the natural environment or a pollution source, most frequently originating in casing materials and