A common method for compensating for grain-size differences in suites of sediment samples is to normalize potential contaminants by regression with a particular grain-size fraction, the <63 microm fraction being most often selected. However, this fraction is unlikely to represent accurately the clay content, which represents a major factor in the ability of sediments to adsorb contaminants. Moreover, no reliable estimation of clay content can be made from a coarser grain-size fraction. As a result, regression with coarser-grained fractions can produce spurious interpretations of background values and contamination. Normalization with the clay content or by an alternative grain-size proxy is recommended.