The soil-water distribution (K(d)) of atrazine has been studied in 24 soil profiles, comprising 109 soil horizons derived from granodioritic materials; K(d) ranged from 0.01 L/kg to 64 L/kg with a mean value of (2.4 +/- 7.3) L/kg. The concentrations of organic carbon, cation exchange capacity, clay, and oxides of iron, aluminum, and manganese were determined. In a multiple linear regression analysis K(d) was strongly correlated with the organic carbon content (f(oc)) and weakly correlated with aluminum oxides. No other soil components were correlated with K(d). The organic carbon soil sorption coefficient (K(oc)) was estimated to be 216 +/- 9 L/kg (95% confidence limits) from a linear plot of K(d) versus the f(oc). In a logarithmic correlation between K(d) and f(oc) the non-linearity constant proved statistically significant from unity: logK(d) = (1.81 +/- 0.20) + (0.75 +/- 0.07) x logf(oc) which indicates that not only the concentration, but also the composition of the organic matter play important roles in soil sorption of atrazine. In 48 of the 109 soil horizons which contained more than 0.1% of organic carbon, the organic matter was fractionated into humic acids, fulvic acids, humin and free organic matter . A multiple linear regression analysis between K(d) and the various organic matter fractions revealed that humic acid (Ha) explained 71% of the variance compared to 26% for humin plus free organic matter (Hum + Fom). In addition, 3% of the variance was explained by manganese oxide: K(d)= (167 +/- 11) x Ha - (0.74 +/- 0.09) x (Hum + Fom) + (21 +/- 16) x MnO - (1.6 +/- 1.0) Previously published equations for predictions of K(oc) of a compound from its water solubility and its octanol-water partition coefficient (K(ow)) were evaluated. K(ow) proved to be the best predictor for atrazine.