The validity of soil toxicity databases for predicting ecological impacts in the field is rarely explored. The present study was set up to test whether laboratory toxicity data and the combined concepts of metal availability and mixture toxicity can predict ecological impact in mining‐affected soils. Metal and As contamination gradients were sampled approximately 5 different mines in Mexico where plant cover and abundances exhibited clear dose‐related responses. Soils were analyzed for total and isotopically exchangeable (labile) concentrations of Ni, Cu, Cd, Pb, and As and for soil properties affecting the availability of these elements. Six different indices of toxic doses were compared to evaluate their accuracy in describing the field response expressed as relative abundance and cover. Each index was based on a different method to calculate the sum of toxic units (normalΣTUs) in soil, with 1 toxic unit equal to the concentration of the element in soil yielding 50% adverse effect on plants with median sensitivity as recorded in a recent database of salt‐spiked soils. Toxic concentrations in the mine‐impacted soils were dominated by Zn and As. In the field, 50% reduced cover or abundance was found at 10 to 13 normalΣTUs if these were based on total soil concentrations and thresholds derived from freshly spiked soils, indicating a largely overestimated toxic effect. If thresholds were corrected for differences in availability among freshly spiked soils and spiked and laboratory‐aged soils, the overestimation of field toxicity was 5‐ to 6‐fold, irrespective of the consideration of soil properties. Finally, the normalΣTU calculated only with labile metals and As overestimated the field toxicity by factors 1.1 to 1.6 (95% confidence interval 1–7; i.e., rather accurate and indicating some Zn–As antagonism as confirmed in experimental studies). That latter index of dose yielded a bell‐shaped response on species richness peaking at approximately 1.6 normalΣTU. Overall, the present study shows that the current toxicity databases of metals can predict the impact of metal contamination on plant communities within factor 2, expressing the dose as soil‐labile concentrations and using the concentration addition concept in these mixed polluted environments. Environ Toxicol Chem 2020;39:1826–1838. © 2020 SETAC