Disposal of solid residues and application of pesticides and fertilizers can lead to an increase in the concentration of heavy metals in soils and groundwater. The legislation on heavy metals nearly always refers to the total contents of these elements; however, an evaluation of their hazard potential and toxicity requires an assessment of the metal fraction that is mobile and possibly bioavailable. Few numerical parameters have been used as references to take decisions and to focus on prevention or remediation strategies in contaminated areas. The solid-solution distribution coefficient (K_d), defined as the relation between adsorbed metal concentrations and those present in the solution (K_d=[M]_ads/[M]_sol), allows a comparison between the behavior of elements in different systems. Environmental protection agencies in USA and Europe use generic K_d values found in bibliographic references; these values are often obtained under distinct conditions from those found in the tropical region, and it may lead to an erroneous hazard estimates. Thus, it becomes clearly necessary to obtain K_d values that will either validate these contamination estimates or provide guidance for intervention strategies in already contaminated areas. The objectives of this study were to evaluate heavy metal retention in 30 representative soils of the State of São Paulo, Brazil, based on a quantification of K_d values and their relation with chemical, physical, and mineralogical soil attributes, such as pH, cation exchange capacity (CEC), clay, organic matter (OM), and crystalline and poorly crystallized Fe, Al, and Mn oxides contents. Ion retention was obtained after batch equilibration. Samples (2 g) of each soils were mixed with 20 ml aliquots of a 0.01 mol L ^-1 NaNO₃ stock solution containing 0.1; 0.5; 1.0; 2.5; and 5.0 mg L^-1 of cadmium (Cd), cobalt (Co), chromium (Cr), copper (Cu), nickel (Ni), lead (Pb) and zinc (Zn). Samples were shaken for 24 h at 120 osc min^-1, the solid and liquid fractions were separated by centrifugation at 15000 x g. An aliquot of the supernatant was sampled and analyzed by High Resolution Inductively Coupled Plasma Mass Spectrometry (HR ICP-MS). The metal adsorption was estimated by subtracting the metal determined in the equilibrium solution from the metal initially added. The K_d values were obtained from the slope of linear adsorption isotherms. The smallest variations in K_d value were recorded for Pb (one order of magnitude – 121 to 7,020 L kg^-1) and Ni (two orders of magnitude - 6 to 998 L kg^-1), while variations indicating four orders of magnitude were observed for Cd (7-14,339 L kg^-1), Co (2 - 34,473 L kg^-1), and Cr (1 - 21,267 L kg^-1). Zinc K_d values were between 5 and 123,849 L kg^-1, with a variation of five orders of magnitude.The following order of affinity was obtained: Pb>>>Cu>>Cd>Zn@Ni@Cr>Co. Soil OM content apparently had little influence on the distribution of metals. More than 55% of the variation in distribution coefficients for metallic cations of the groups IIB (Cd and Zn) and VIIIB (Co and Ni) were explained by pH, while Cu, Cr, and Pb exhibited small pH dependence. In the multiple regression, CEC and pH explained about 80% of the variation in K_d values for Cd, Co, and Ni, and showed that nonspecific adsorption mechanisms are involved in the retention of these elements. pH and clay contents accounted for about 63% of the variation in Kd values for Cu and Pb. Distribution coefficients for Cr had an inverse correlation with pH, especially in the joint analysis with clay contents, and 61% of the variation were explained by these variables. Although the best regression model was metal-specific, this study suggests that the most beneficial and universal site-specific K_d adjustment is pH.