Metal hyperaccumulation in plants is an ecological trait whose biological significance remains debated, in particular because the selective pressures that govern its evolutionary dynamics are complex. One of the possible causes of quantitative variation in hyperaccumulation may be local adaptation to metalliferous soils. Here we explored the population genetic structure of Arabidopsis halleri at fourteen metalliferous and non-metalliferous sampling sites in Southern Poland. The results were integrated with a quantitative assessment of variation in zinc hyperaccumulation to trace local adaptation. We identified a clear hierarchical structure with two distinct genetic groups at the upper level of clustering. Interestingly, these groups corresponded to different geographic sub-regions, rather than to ecological types (i.e. metallicolous vs non-metallicolous). Also, approximate Bayesian computation analyses suggested that the current distribution of A. halleri in Southern Poland could be relictual as a result of habitat fragmentation caused by climatic shifts during the Holocene, rather than due to recent colonization of industrially polluted sites. In addition, we find evidence that some non-metallicolous lowland populations may have actually derived from metallicolous populations. Meanwhile, the distribution of quantitative variation in zinc hyperaccumulation did separate metallicolous and non-metallicolous accessions, indicating more recent adaptive evolution and diversifying selection between metalliferous and non-metalliferous habitats. This suggests that zinc hyperaccumulation evolves both ways – towards higher levels at non-metalliferous sites and lower levels at metalliferous sites. Our results open a new perspective on possible evolutionary relationships between A. halleri edaphic types that may inspire future genetic studies of quantitative variation in metal hyperaccumulation.,SSR_Ahalleri_data-fileFor 14 Arabidopsis halleri populations there is i) a worksheet with the raw SSR data and ii) a list of primers used for these markers. The data-file consists of 343 genotypes and 10 SSR markers.,