This paper describes the further development of a read-across approach applicable to the toxicological assessment of structurally-related xenobiotic metabolites. The approach, which can be applied in the absence of definitive identification of all the individual metabolites, draws on the use of chemical descriptors and multi-variate statistical analysis to define a composite “chemical space” and to classify and characterize closely-related subgroups within this. In this example, consideration of the descriptors driving grouping, combined with empirical evidence for lack of significant further biotransformation of metabolites, leads to the conclusion that, in the absence of any specific structural alerts, the relative toxicity of metabolites within a single grouping will be determined by their relative systemic exposure as described by their ADME characteristics. The in vivo testing of a smaller number of exemplars, selected to have representative ADME properties for each grouping, is sufficient, therefore, to evaluate the toxicity of the remainder. The approach is exemplified using the metabolites of the herbicide S-metolachlor, detected in the leachate of a soil lysimeter.