Previous work has demonstrated that plant uptake of radiocaesium (RCs) is related to the activity concentration of RCs in soil solution, which is linked to the soil/soil solution distribution coefficient, K(D). The solid-liquid distribution of RCs is generally studied in soil suspensions in the laboratory and there are few reported measurements for in situ soil solutions. From a data set of 53 different soils (contaminated with either 134CsCl or 137CsCl) used in pot trials to investigate grass uptake of RCs, we analysed the variation of in situ K(D) with measured soil properties. The soils differed widely in % clay (0.5-58%), organic matter content (1.9-96%) and pH (2.4-7.0, CaCl2). The K(D) varied between 29 and 375,000 L kg-' (median 1460 L kg(-1)). Stepwise multiple regression analysis showed a significant correlation between the log K(D) and pH (p < 0.001), log %clay (p < 0.01) and log exchangeable K (p < 0.001) (overall R2 = 0.70). The in situ K(D) values were further compared to K(D)S predicted using an existing model, which assumes that RCs sorption occurs on specific sites and regular ion-exchange sites on the soil solid phase. Sorption of RCs on specific sites was quantified from the radiocaesium interception potential (RIP) measured for each soil and the soil solution concentrations of K+ and NH4+. The in situ log K(D) correlated well with the predicted K(D) (R2 = 0.85 before plant growth, R2 = 0.83 after plant growth). However, the observations were fivefold to eightfold higher than the predictions, particularly for the mineral soils. We attribute the under-prediction to the long contact times (minimum 4 weeks) between the RCs tracers and our experimental soils relative to the short (24 h) contact times used in RIP measurements. We conclude that our data confirmed the model but that ageing of RCs in soil is a factor that needs to be considered to better predict in situ KD values.