Coefficient of parentage (CP) and restriction fragment length polymorp
hism-based genetic similarity estimates (RFLP-GS) have been proposed a
s measures of genetic distance in crop species. If these measures are
to have application in practical breeding, it is important to validate
their utility in predicting genetic traits of interest such as hetero
sis. The objectives of this paper were to (i) estimate heterosis for y
ield in soybean [Glycine max (L.) Merr.] adapted to the southern USA,
and (ii) predict heterosis by mi:ans of CP and RFLP-GS as genetic dist
ance measures. Twenty-four F2 populations were developed by crossing t
hree testers ('Young', 'Centennial', and 'Tracy') to eight contrasting
parents, representing a wide range of CP and RFLP-GS. The experimenta
l material was divided into three sets representing the testers above,
and was evaluated in eight replications at Clayton and Plymouth, NC,
in 1994. Midparent heterosis for yield was 7.9, 4.5, and 7.9% for Sets
1, 2, and 3, respectively. Heterosis was 3.5, 1.6, and 3.0% for 100-s
eed weight, and 4.1, 5.4, and 13.2% for plant height. The CP and RFLP-
GS were highly correlated (r = 0.80, 0.92 and 0.95 for Sets 1, 2, and
3, respectively, P = 0.01), but neither predicted heterosis well for y
ield averaged across locations because of a large genotype x environme
nt (G x E) interaction. In contrast, CP and RFLP-GS predicted heterosi
s well for 100-seed weight and plant height in two of the three sets a
veraged over locations. Our estimates of high parent heterosis for yie
ld (as high as 11% over locations), may justify soybean hybrids as a b
reeding objective. However, the limited predictive value of CP and RFL
P-GS in our study indicates that the identification of favorable heter
otic combinations may require extensive field testing.