A spaced plant experiment and a sward experiment were conducted at INIA La Estanzuela, Uruguay in 1996 with six red clover (Trifolium pratense L.) populations, to study the efficacy of selecting genotypes for sward performance in a spaced plant nursery. This was examined through: (a) the ranking of the best linear unbiased predictors (BLUP) of populations and (b) the ratio of the correlated genetic advance in swards by selecting under spaced plant conditions to the direct genetic advance of selecting under sward conditions. From the eight characters analysed, only the rankings of BLUPs for leaf size and flowering were significantly the same between populations. The other six characters performed differently according to the level of competition imposed. Post-grazing leafiness and the difference between pre- and post-grazing leafiness (the two most important characters to measure preferential grazing) achieved greater genetic advance when selection was done as spaced plants: while for pre- and post-grazing height genetic advance was greatest by selecting under sward conditions. For other characters, the best conditions to select depended on the selection intensity achievable.