Sorghum (Sorghum bicolor (L.) Moench) is a staple food grain in many semi-arid and tropical areas of the world, notably in sub-Saharan Africa because of its adaptation to harsh environments. Among important biochemical components for sorghum for processors are the levels of starch (amylose and amylopectin) and total sugar contents. The aim of this study was to determine the genetic variation for total starch in the seed, its components and total sugar in the stalks of the sorghum accessions from Ethiopia and South Africa. Samples of 22 sorghum accessions were evaluated. Significant variations were observed in total starch (31.01 to 64.88 %), amylose (14.05 to 18.91 %), the amylose/amylopectin ratio (0.31 to 0.73) and total stalk sugar content (9.36 to 16.84 %). Multivariate analysis showed a wide genetic variation within and among germplasm accessions which could be used in the selection of parental lines for the improvement of traits of interest through breeding. The variation found among the sorghum accessions shows that an improved total starch and starch components and stalk sugar contents can be achieved through crossing these selected genotypes.

Qualitative morphological and amplified fragment length polymorphism (AFLP) markers were compared for assessment of genetic diversity. Nine qualitative morphological traits were recorded to compare genetic relationships among 17 sorghum accessions with information derived from six AFLP primer combinations analysis. The mean morphological genetic similarity was lower in comparison to similarity computed using AFLP markers. Genetic similarity measured by AFLP markers was similar within the Ethiopian and South African material, as well as between South African and Ethiopian material. Morphological similarity was much higher in the Ethiopian material than in the South African material, indicating that the genotypes were related. The two techniques described genetic variability in different ways. Dendrogram generated from the morphological data matrix separated accession 216737 as being genetically distinct from the rest of the accessions. Accessions M101 and 97MW6127 were the most dissimilar accessions based on AFLP data.