A total of 62 perennial grasses were evaluated for herbage production under low-fertiliser conditions at eight sites in the temperate zone of southern Australia from 1999 to 2001. A brief assessment of relative preference ranking by sheep was also made at the end of the experimental period. Four sites were in the high rainfall areas of south-east Australia, two in the drier mixed farming areas of western NSW, and two sites in Mediterranean southern Australia. Seven standard cultivars were included in the comparisons. Plants were grown from seed in glasshouses and transplanted to the field as spaced plants at 6-8 weeks age. Plants were harvested at irregular intervals by clipping to determine herbage production. Relative preference was determined by enclosing sheep in the experimental area at high stocking rates and estimating the amount of herbage grazed after 1, 3 and 5 days. Herbage production data were anal ysed using a multi-environment trial approach in which the environments comprised all combinations of sites and sampling times over the 3-year period. Overall, 73.6% of the total genetic variation for herbage production was accounted for.Average seasonal comparisons revealed good herbage production from Eragrostis curvula cv. Consol in both winter and summer, but in spring, several Dactylis glomerata lines were the most productive. Consol was not generally preferred by sheep, while D. glomerata lines were. Apart from cv. Consol, C4 species, including Bothriochloa macra, Chloris truncata, Enteropogon acicularis and Dichanthium sericeum had low herbage production in winter and spring, and had low acceptability to grazing sheep. These particular grasses were also not highly ranked for production in summer, although other C4 lines, notably selections of Themeda australis, Paspalidium jubiflorum and P. constrictum were. Few native C3 grasses had superior herbage production, although selections of Elymus scaber, Austrodanthonia fulva and A. duttoniana showed good growth rates. When considering further evaluation, grasses with very high recruitment (e.g. A. caespitosa) may need to be included, and the acceptability of grasses to sheep should also be assessed. © CSIRO 2005.