AbstractSix seed mixtures differing in number of species and their proportion of timothy (Phleum pratense L.) were tested during three/four production (ley) years in replicated field experiments at three climatically different sites in Norway; one a mountainous inland site at 61° N (Løken) and two in coastal environments, at 61° N (Fureneset) and 65° N (Tjøtta). There were significant differences in forage accumulation (FA) and digestible forage accumulation (DFA) between the three sites. There was a significant FA decline from the third to the fourth ley year for mixtures containing timothy, but not for mixtures without timothy. Estimated interannual FA‐ stability was higher for timothy‐based seed mixtures than for mixtures without timothy at the inland site, but FA‐stability was lower at the coastal sites. In the third‐year herbage of timothy‐based mixtures at the inland site consisted almost solely of timothy, whereas at the coastal sites meadow fescue (Festuca pratensis Huds.) and especially tall fescue (F. arundinacea Schreb.) dominated. In seed mixtures without timothy, cocksfoot (Dactylis glomerata L.) suppressed other species at the inland site, whereas at the coastal sites, tall fescue and ryegrasses (Lolium spp.) were the dominant species in the third‐year herbage. Length of growing season and site‐specific growing conditions were important drivers for the observed species changes. Timothy can thus be recommended for ley establishment at sites where the growing season is short (<4 months) and plant growth is intensive, but under conditions with a longer growing season it needs to be sown in mixtures with grass species that surpass the regrowth capacity of timothy.