In 1978 genetical variation for thiocyanate ion (SCN‐) was found in a kale population being selected for improved digestible organic‐matter yield. Four generations of half‐sib family selection for high and low SCN‐ content in young leaves resulted in subpopulations with almost a twofold difference in their means, no overlap in their distributions, no observable reduction in their genetical variation, and only small differences in other traits. The total observed divergence in SCN‐content was 82% of that expected for a random mating population in linkage equilibrium in the absence of genotype x environment interactions. Average heritability on a single plot basis was low, namely 30%, but was increased to 46% with two replicates. This maximised the predicted response to selection for a fixed number of families selected and a fixed total number of plots. Bulks of the high and low subpopulations were assessed in a trial in 1988. Averaged over September and December harvests of young and mature leaves, SCN‐ contents were 104.5 and 58.5 mg/100 g DM, respectively. Of eight controls, cv. Merlin had the lowest content, 63.5, and cv. Proteor had the highest content, 89.6. Cultivar x harvest date, cultivar x leaf age and cultivar x harvest date X leaf age interactions were removed by a logarithmic transformation of the data. Stem contents were very low and variable with an overall mean of 10.9 in September and > 5 in December, and no significant differences between cultivars or populations. The subpopulations will be valuable for research into possible effects of SCN content on the disease and pest resistance of kale and on the performance of animals fed kale. Copyright © 1990, Wiley Blackwell. All rights reserved