QUANTITATIVE INHERITANCE AND CORRELATIONS FOR AGRONOMIC AND QUALITY TRAITS, INCLUDING KERNEL HARDNESS, IN A RANDOM-MATED POPULATION OF SORGHUM

Efficient utilization of any breeding material for specific traits requires an understanding of the inheritance and the correlations of the traits in question. In this research, genetic parameters were estimated for agronomic and quality traits, including kernel hardness (percentage floater) by testing 236 S(,2) lines derived from a Purdue random-mated population (PP18) of sorghum, (Sorghum bicolor, L. Moench). Additive variances were estimated using both unweighted and weighted least-squares procedures. All tested traits showed wide range and significant variation, which was primarily additive. Unweighted (3.4 to 30%) and weighted (2.9 to 25.5%) additive genetic coefficients of variation were lowest for days to bloom and highest for lodging. Unweighted narrow-sense heritabilities for S(,0) (0.46 to 0.74) and S(,1) (0.47 to 0.94) family selection were highest for days to bloom and lowest for plant height. Both additive genetic (-0.444 to 0.774) and phenotypic (-0.277 to 0.710) correlations were significant and positive for days to bloom with plant height, panicle compactness, and panicle weight; for plant height with lodging, panicle weight, and kernel weight; for panicle weight with lodging, panicle compactness, and kernel weight; and for the kernel traits, hardness, weathering, light vitreosity, and endosperm section vitreosity with each other. Both correlations were significant, but negative, for days to bloom with lodging and weathering; and for each of lodging and panicle weight with weathering. Direct responses from S(,0) (5.1 to 41.3%) and S(,1) (7.1 to 56.9%) family selection (% of mean) were lowest for days to bloom and highest for kernel hardness and consistently exceeded correlated responses. However, the difference between the two responses was not appreciable for some pairs of traits. The results suggested good potential for conventional breeding methods, early generation selection, and bidirectional selection for all traits tested. More improvement would be expected from S(,1) than S(,0) family selection, and from direct rather than indirect selection. However, substantial simultaneous improvement would be expected for some pairs