Eragrostis curvula (Schrad.) Nees s. lat., a highly polymorphic polyploid c
omplex, can be considered as one of the most important warm season perennia
l grasses for the semi-arid regions of central Argentina. In apomictically
propagated and perennial plants such as weeping lovegrass, where successive
measurements can be done across time, repeatability estimates provide an i
ndication of the degree of influence of permanent effects on the phenotypic
variation and allow prediction of future performance from past records. An
alysis of variance of the experiment showed highly significant variation (P
< 0.01) for the main factors in all four traits. Although there was very h
igh variability between cuts, hybrids of E. curvula exhibited considerable
inter-entry variability, in particular for those traits determining forage
yield. Repeatability calculated in this experiment was highest for the crow
n diameter (0.86), leaf length (0.84) and dry matter (0.84), while in panic
le number (0.66) it was lowest. Repeatability estimates for the vegetative
characteristics indicate small effects of temporal environment. The four tr
aits studied, including panicle number with their moderate repeatability, d
o not require an essentially different number of observations to obtain mea
sures at the same level of accuracy. For vegetative characters two harvests
provided 98% of the accuracy of the total obtained with four cuts, and for
panicle number the same percentage was obtained for three harvests. This s
tability of performance is a desirable characteristic for grass cultivars.
Patterns of trait associations were also described. Because leaf length is
closely associated with dry matter and has high repeatability, to use leaf
length as an indirect evaluation criterion should be almost as efficient as
direct evaluation for aerial biomass yield. Reliable estimates of paramete
rs such as repeatability and phenotypic correlation are needed for predicti
on of production values and for the design of efficient improvement program
mes. For genotype evaluation additional research is required to quantify th
e extent of genotype x environment interaction across years and localities
of semi-arid regions.