High starch content is an important component of root quantity and quality for almost all uses of cassava (flour, chips, and industrial raw material). However, there is scanty information on genetic variability for dry matter and starch contents and relatively little attention has been paid to genetic improvement of root dry matter content and starch content in Tanzania. The major objective of this research was to develop improved cassava varieties that are high yielding, with high dry matter and starch content for Tanzania and specifically to: i) identify farmers’ preferences and selection criteria for cassava storage root quality characteristics and other traits of agronomic relevance for research intervention through a participatory rural appraisal; ii) determine the genotypic variability for starch quantity and dry matter content evaluated for three harvesting times in four sites; iii) determine the inheritance of dry matter and starch content in cassava genotypes; and iv) develop and evaluate clones for high storage root yield, high dry matter content and starch. Attributes desired by farmers were yield, earliness, tolerance to pests and diseases. The complementing attributes associated with culinary qualities were sweetness, good cookability, high dry matter content or mealyness and marketability. The preliminary study conducted to evaluate the variability in root dry matter content (RDMC) and starch quantity and yield of ten cassava cultivars indicated that RDMC ranged from 29 to 40% with the mean of 34.3%. The RDMC at 7 months after planting (MAP) was higher than at 11 and 14 MAP. Starch content (StC) ranged from 20.3% to 24.9% with the mean of 22.8%. The StC differed significantly between cultivars, harvesting time and sites. An increase in StC was observed between 0 and 7 MAP, followed by a decline between 7 and 11 MAP, and finally an increase again noted between 11 and 14 MAP. However, for most of the cultivars at Kibaha an increase in StC between 11 and 14 MAP could not surpass values recorded at 7 MAP. At Kizimbani, cultivar Kalolo and Vumbi could not increase in StC after 11 MAP. At Chambezi and Hombolo, a dramatic gain in StC was observed for most of the cultivars between 11 and 14 MAP. Starch yield ranged from 0.54 to 4.09 t ha-1. Both StC and fresh storage root yield are important traits when selecting for commercial cultivars for starch production. Generation of the F1 population was done using a 10 x 10 half diallel design, followed by evaluation of genotypes using a 4 x 10 á-lattice. Results from the diallel analysis indicated that significant differences in fresh storage root yield (FSRY), fresh biomass (FBM), storage root number (SRN), RDMC, starch content (StC), and starch yield (StY), and cassava brown streak disease root necrosis (CBSRN) were observed between families and progeny. The FSRY for the families ranged from 15.0 to 36.3 t ha-1; StC ranged from 23.0 to 29.9%; RDMC ranged from 31.4 to 40.1%; and StY ranged from 3.3 to 8.3 t ha-1. The cassava mosaic disease (CMD) severity ranged from 1.7 to 2.7, while cassava brown streak disease (CBSD) severity for above ground symptoms ranged from 1.0 to 1.9. Additive genetic effects were predominant over non-additive genetic effects for RDMC, StC, and CBSRN, while for FSRY, FBM, SRN, and StY non-additive genetic effects predominated. Negative and non-significant correlation between RDMC and FSRY was observed at the seedling stage (r=-0.018), while at clonal stage the correlation was positive but not significant (0.01). The RDMC and StC were positive and significantly correlated (r=0.55***) at clonal stage. However, the StC negatively and non-significantly correlated with FSRY (r=- 0.01). High, positive and significant correlation (r=0.94; p.0.001) was observed between the StY and FSRY at clonal stage. High, positive and significant correlations between the seedling and clonal stage in FSRM (r=0.50; p.0.01), RDMC (r=0.67; p.0.001), HI (r=0.69; p.0.001), and SRN (r=0.52; p.0.01) were observed, suggesting that indirect selection could start at seedling stage for FSRM, RDMC, HI, and SRN. The best overall genotype for StC was 6256 (40.9%) from family Kiroba x Namikonga followed by genotype 6731 (40.6%; Vumbi x Namikonga). Among the parents, Kiroba and Namikonga were identified as the best combiners in terms of GCA effects for StC. Genotype 6879 from family Vumbi x AR 42-3 had the highest StY value of 34.8 t ha-1 followed by genotype 6086 (30.4 t ha-1; Kalolo x AR 40-6). Among the parents, Kalolo and AR 42-3 were identified as good combiners for the trait. Mid-parent heterosis for StC ranged from 41.6 to 134.1%, while best parent heterosis ranged from 30.4 to 119.6%. Genotype KBH/08/6807 from family Vumbi x TMS 30001 had the highest mid-and best parent heterosis percentage for StC. For StY, mid-parent and best parent heterosis ranged from 168.0 to 1391.0%, and from 140.4 to 1079.0%, respectively, with the genotype 6879 (Vumbi x AR 42-3) exhibiting the highest mid- and best parent heterosis percentage for StY. Improvement for StC, RDMC, and CBSRN may be realized by selecting parents with the highest GCA effects for the traits and hybridize with those that combine well to maximize the positive SCA effects for the StC, RDMC and CBSRN. Selected genotypes from the clonal stage will be evaluated in preliminary yield trial and advanced further to multi-locational trials while implementing participatory approaches involving farmers and processors in selection. New promising lines should be tested at different sites and the best harvesting dates should be established. / Thesis (Ph.D.)-University of KwaZulu-Natal, Pietermaritzburg, 2009.