Finger millet (Eleusine coracana) productivity in East Africa has remained low in all production agro-ecologies for decades owing to the low yielding potential of existing that are susceptible to the blast disease caused by the fungus Magnaporthe grisea (Hebert) Barr. and the limited research on the crop. The region holds large finger millet germpasm collections whose value is not yet been fully exploited. However, with the ongoing breeding efforts through hybridization, there is a need to comprehensively characterize the germplasm to identify valuable traits to address biotic and abiotic stresses that affect finger millet productivity. Studies on gene action and inheritance of key traits that contribute to yield improvement are also required to help formulate an effective breeding strategy for finger millet improvement. The objectives of this study were to (i) determine the genetic diversity in a set of germplasm from East Africa (ii) determine association between grain yield and its component traits (iii) identify genotypes for target production agro-ecologies (iv) identify blast resistant finger millet genotypes for use in breeding and production and (v) generate information on the inheritance of blast, grain yield and yield components for the development of an effective breeding strategy. A total of 340 finger millet accessions were collected from three countries in East Africa: Kenya, Tanzania and Uganda and 80 global minicore accessions sourced from ICRISAT-India. High phenotypic variability in the germplasm was recorded for 23 quantitative traits, blast reaction and five qualitative traits. Both morphological and molecular characterization (using SSR markers) of the 340 accessions revealed higher diversity within than among the countries Kenya, Tanzania and Uganda. Seven morphological clusters and three major genetic clusters were detected. Morphological diversity delineation was largely influenced by leaf sheath length, plant height, peduncle length, panicle exertion and grain yield. The mean polymorphic information content (PIC) of 19 polymorphic markers was 0.606 with mean alleles of 195 with sizes that ranged from 148-474 base pairs. The Kenyan and Tanzanian accessions had higher diversity than the Ugandan with the Kenyan and Ugandan, and the Kenyan and Tanzanian accessions being closely related than the Tanzanian and Ugandan. The low diversity in the Ugandan accessions could be attributed to higher research intervention in the country leading to the promotion and use of improved cultivars. Efforts have to be directed towards collection and conservation of valuable diversity before it is lost. The diversity in plant height, maturity, yield and blast reaction and the cluster groups detected in the germplasm should provide a basis for finger millet improvement through hybridization and selection. Higher genotypic than phenotypic correlations were recorded for most of the traits studied with grain yield having high positive correlations with finger width, grains per spikelet, threshing percent, peduncle length and panicle exertion. Both grain yield and days to flowering had negative correlations with all three blast types (leaf, neck and finger). Path coefficient analysis revealed that productive tillers per plant, 1000 grain mass, grains per spikelet and threshing percent had positive direct genetic effects on grain yield with strong indirect effects from several of the other traits which necessitates simultaneous selection for those traits with strong direct effects and those with strong indirect effects for grain yield improvement. High broad sense heritability estimates and high genetic advance as percent of mean were recorded in fingers per panicle, flag leaf sheath length, 1000 grain mass, finger length, peduncle length, panicle exertion, number of leaves per plant and leaf sheath length probably indicating the predominance of additive gene effects in controlling these traits hence the potential for improvement through selection. Adaptation and stability analysis using the GGE biplot model identified Lanet 2012 long rains, Serere 2012 long rains and Miwaleni 2012 long rains as the most discriminating environments for the low temperature, sub-humid mid altitude and dry lowland areas, respectively. Alupe 2012 long rains was the ideal environment for genotype discrimination for blast while Lanet 2012 long rains was best for grain yield. Genotypes G3, G5, G17, G25, G28, G36 and G71 were identified as being stable across environments and G1, G18, G19, G37, G54, G61, G74, G75, and G77 were found ideal for specific adaptation. Disease severity scores were highly negatively (P<0.01) correlated with days to flowering and grain yield suggesting that early lines suffered more disease damage leading to reduced yield. Resistant genotypes were slow blasting (probably associated with horizontal resistance) which may enable them to withstand blast pathogen variability for longer periods. Nine genotypes were identified with high resistance to blast and will be useful for breeding as blast resistance sources. Resistant genotypes had low AUDPC values and disease severity rating for the three blast types and vice-versa for susceptible genotypes. Further investigations need to be carried out to determine the possibility of the three blast types being controlled by the same genes. Early maturing blast susceptible genotypes with good yield potential could be utilized in areas with low blast prevalence. To understand the gene action for inheritance of the various traits 16 F2 families plus their four female and four male parents were evaluated at Alupe and Kakamega western Kenya under artificial blast inoculation. Significant additive genetic effects were recorded for all traits (except for finger width and grains per spikelet) meaning that improvement for these traits would be possible through the common selection methods for self pollinating crops. Parent lines KNE 392, and KNE 744 and IE 11 were found to be suitable for blast resistance breeding while Okhale 1 was found to be suitable for high grain yield and blast resistance improvement due to their high desirable GCA effects. Most of the F2 families showed transgressive segregation for the three blast types in either direction which gives hope for the development new pure lines with better blast resistance than the parents. Crosses IE 3104 x KNE 796, KAT FM 1 x Okhale 1, IE 11 x Okhale, IE 11 x P 224 and KNE 744 x KNE 392 have potential to generate lines with blast resistance due to their high desirable SCA effects. The F2 segregation distributions for blast indicated quantitative inheritance. However the one to four minimum number of genes (effective factors) detected for resistance control in all the three blast types was not in sync with the segregation patterns in the F2 families and further investigations are required. There were differences in segregation patterns between crosses which may suggest the presence of different resistance genes in the different parents used. This would call for gene pyramiding for durable resistance. These results confirm the potential of sourcing valuable parental stocks in the local germplasm for the development of genotypes to improve finger millet productivity in East Africa. Already some of the high yielding and blast resistant genotypes identified here have been incorporated in the regional cultivar trials. The diversity information generated will facilitate effective conservation and utilization of this germplasm. Results of gene action for inheritance of the various traits from this study will enable breeders to develop sound breeding strategies for finger millet improvement in the region. / Thesis (Ph.D.)-University of KwaZulu-Natal, Pietermaritzburg, 2013.