Polygenic sex determination (PSD) may show variations in terms of genetic and environmental components between populations of fish species exposed/adapted to different environments. The European sea bass (Dicentrarchus labrax) is an interesting model, combining both a PSD system and a genetic subdivision into an Atlantic and a Mediterranean lineage, with genetic substructures within the Mediterranean Sea. Here, we produced experimental progeny crosses (N = 927) from broodstock sampled in four wild populations (North Atlantic, NAT; Western Mediterranean, WEM; NorthEastern Mediterranean, NEM; SouthEastern Mediterranean, SEM). We found less females than males in the progeny, both in the global dataset (32.5%) and within each paternal group (from 25.1% for NEM to 39.0% for WEM), with significant variation among populations, dams, and sires. Sex, body weight (BW), and body length (BL) showed moderate heritability (0.52 ± 0.17, 0.46 ± 0.17, 0.34 ± 0.15, respectively) and sex was genetically correlated with BW and BL, with rAsex/BW = 0.69 ± 0.12 and rA sex/BL = 0.66 ± 0.13. A weighted GWAS performed both on the global dataset and within each paternal group revealed a different genetic architecture of sex determination between Atlantic and Mediterranean populations (9 QTLs found in NAT, 7 in WEM, 5 in NEM, and 4 in SEM, with a cumulated variance explained of 27.04%, 21.87%, 15.89%, and 12.10%, respectively) and a more similar genetic architecture among geographically close populations compared to geographically distant populations, consistent with the hypothesis of a population-specific evolution of polygenic sex determination systems in different environments. environment generating biased sex-ratios, or towards ESD if some environments increase the fitness of a specific sex (Bulmer and Bull 1982). Modelling approaches have shown that the orientation of polygenic sex determination in one or the other direction depends on complex combinations of environmental variation between and within environmental patches, and on migration rates between patches (Van Dooren and Leimar 2003; Bateman and Anholt 2017). Therefore, starting with the same