1. To evaluate selection criteria in grass breeding, a series of investigations was conducted measuring plant performance in a wide range of field environments.
2. This paper reports on the behaviour of one variety of Lolium perenne, S. 24, grown in a logarithmically related series of densities and frequently defoliated for 2 years. Half the plots received only natural rainfall, the rest were irrigated so that the calculated soil moisture deficit did not exceed 2 in.
3. In 1958, with little effect of irrigation, swards yielded most per unit area, widest spacing (24 in.) least; there were no significant differences in yields from the other densities (3, 6 and 12 in.). In 1959, when yields were lower, irrigation increased production in all plots, the higher the density the greater the increase. In contrast to 1958, over the range of spaced densities, the higher the density the greater the yield. Individual harvests did not always conform to the annual pattern.
4. Lowest mean production per plant was from broadcast plots in both 1958 and 1959, yields of plants from 3, 6 and 12 in. plots showed a linear increase with a reduction in density, those from 24 in., though heaviest, yielded less than linear expectation. In 1959 irrigation increased yields per plant.
5. Of the components of yield, number of sterile tillers gave the best estimate of total fresh weight (95·3% reliability). However, total number of tillers (sterile plus fertile) was almost as efficient (94%). Tiller weight was a much less reliable index.
6. The results and their value in establishing efficient criteria for selection in grass breeding are discussed.