Risk assessments of existing substances, including metals, in the environment are being carried in the European Union, which require estimation of the predicted no-effect concentration (PNEC) derived from biological toxicity assays. Few toxicity data exist for Co and higher plants, and the influence of soil properties on Co toxicity also needs to be evaluated systematically. We investigated phytotoxicity and bioavailability of Co to barley (Hordeum vulgare L.), oilseed rape (OSR; Brassica napus L.) and tomato (Lycopersicon esculentum L.) shoot growth in ten soils varying widely in soil properties using a standardised shoot biomass assay. The effective concentration of added Co causing 50% inhibition (EC(50)) ranged from 40 to 1708 mg kg(-1), from 7 to 966 mg kg(-1) and from 7 to 733 mg kg(-1) for the barley, OSR and tomato test, respectively, representing 43-, 138-, and 105-fold variation among soils. The EC(50) based on the Co concentration in soil solution varied less among soils (4-15 fold) than that based on the total added Co, suggesting that solubility of Co is a key factor influencing its toxicity to plants. Regressions of soil Co toxicity threshold values with various soil properties, showed that exchangeable calcium (Ca) concentration was the most consistent single predictor. For risk assessment, we suggest that Co toxicity threshold values for plants be normalised using the soil exchangeable Ca.