Three winter-active populations of phalaris (Phalaris aquatica L.), selected over two generations for improved persistence under grazing, were evaluated with commercial cultivars of phalaris and other temperate perennial grasses from 1999–2003 in three environments of south-eastern Australia as part of a program to develop a cultivar for more sustainable pastures and to assess genotype × environment interaction. Grazed sites were located at Bulart in western Victoria, and Rye Park on the Southern Tablelands and Tamworth on the North West Slopes of New South Wales. At the conclusion of the experiment, the frequency of live plant base was highest at Rye Park despite soil acidity and drought. Significant variance among half-sib families in each population was also observed most frequently at this site. Frequency was intermediate at Bulart but lower than expected considering high soil fertility, probably because of high grazing pressure. Frequency was lowest at Tamworth where severe drought occurred from 2001 onwards. There was significant genotype × environment interaction for frequency among half-sib families. Significant common family variance for frequency across the Bulart and Rye Park sites was demonstrated, but not between Tamworth and either of the other sites in later years. The relationship between winter herbage mass potential and persistence differed with population and site, and was negative for one population at Bulart but positive for another population at Tamworth. Mean persistence of all families was 30% higher than winter-active controls at Rye Park and at least 40% higher at Bulart. Phalaris generally persisted better than cultivars of tall fescue (Festuca arundinacea Schreb.), cocksfoot (Dactylis glomerata L.) and perennial ryegrass (Lolium perenne L.) with some exceptions, particularly at Bulart. Development of a winter-active phalaris cultivar with improved persistence under grazing was considered possible for the Southern Tablelands and western Victorian environments with these populations but a separate program using additional germplasm will be needed for the North West Slopes environment.