Organic matter is the soil constituent most often associated with herbicide adsorption. Structural diversity makes humified organic material an ideal substrate for the adsorption of many pesticides, but variability in composition and distribution in situ complicates interpretation of its quantitative effect on adsorption. Variability in the adsorption distribution coefficient (KD) of a herbicide among soils often is due to differences in organic matter content and can be reduced by adjusting KD for soil organic carbon content and computing the organic carbon partition coefficient (Koc). Koc can be estimated from the octanol-water partition coefficient (Kow) of organic compounds, but the correlation weakens as compound polarity increases. Koc also can be correlated with aqueous solubility if a correction is made for the melting point of compounds that are solids at 25 C. Relative adsorption can be estimated from parachor and molecular connectivity indices; but corrections are needed for polar compounds, and correlations with KD or Koc have been variable. Such predictive methods may be useful for broad classification purposes, but accurate extrapolation generally requires site-specific adsorption measurements. Empirical models which accommodate the multiple regression of organic matter content and other soil properties such as clay content, pH, and cation exchange capacity on herbicide adsorption can increase accuracy, but interpretation may be restricted to a small number of sites.