Rice is an important crop ranking third after maize and wheat in Kenya. Its demand is growing at 12% per annum, while production has stagnated for quite a number of years. This situation has lead to consumption outstripping production by about 84%. There is therefore an urgent need to step up domestic production. To understand farmers’ preferences, the first survey on rice production was carried out. To supplement on this survey information, focus group discussions and key informants were incorporated to further shed more information. The farmers demonstrated their preferences for varieties that were high yielding, hybrid rice and with high nutrient use efficiency and these constituted 53.7% of their wishes. Among the crops grown by the selected farmers, rice was the most important followed by maize, however, the hectarage of maize was higher followed by rice. The farmers also identified labour costs as a factor limiting production especially for irrigated rice and thus they preferred upland rice due to its ease of production. Farmers identified land preparation, inputs, planting, weeding and harvesting as the most expensive activities. The existing varieties were viewed as being highly dependent on inputs which they could not afford. The main fertilizers used were Diammonium Phosphate (DAP), Sulphate of Ammonia (SA), Calcium Ammonium Nitrate (CAN) and NPK (23:23:0), the amounts used were low even though the farmers were aware that soil fertility was low. No studies or attempts had been carried out to determine performance of rice genotypes and their heritability parameters for adaptation to low soil nitrogen (N) and phosphorous (P) conditions identified by the farmers as a major constraint in Kenya. Since the existing genetic base was narrow, accessions were acquired to broaden the variability of the local cultivars. However, their adaptation to local conditions was important for them to be useful. Evaluation of 390 accessions showed that genotypes and soil environments were highly significant for all the ten traits studied. The degree of genetic determination (H(2)) ranged from 8.0% for 1000 grain weight to 27% for top biomass. The phenotypic coefficient of variation of genotypes ranged from 12% for days to maturity to 149% for top biomass, while, that of days to maturity ranged from 14% to 160% for top biomass. The genetic advance (GA) had values ranging from, 0.2 for phosphorous tolerance to 1081 for grain yield, while the genetic advance expressed as percent of the mean was 6% for days to maturity and 88% for top biomass. The mean values for the ten characters studied had wide variability under the four soil environments with days to maturity ranging from 188 for genotype ARCCU1Fa1-L4P3-HB under both N and P application (N+P+) to 177 for genotype CT16333(1)-CA-1-M under none N and P application (N-P-) condition. The highest yielding genotype was CT16328-CA-18-M under none N and P application with 5916 kg ha(-1). The germplasm revealed usable variability under low soil N and P adaptation and thus warrant rice improvement for traits of interest to farmers. In order to establish the genetic factors controlling upland rice adaptation to contrasting soil N and P a study was conducted to determine genotypes with better performance under the prevailing farmers’ production environments. The GCA and SCA mean squares were significant and their interactions with environment were highly significant. The GCA:SCA ratios were mostly less than 1.0 for the majority of the traits under most of the soil N and P environments, indicating preponderance of nonadditive genetic effects. The maternal and nonmaternal mean squares were significantly different from zero (P<0.05) for most of the ten traits under study, indicating influence of cytoplasm effects and cytoplasm by nuclear gene interactions, respectively. The GCA effects for the parents were significant under different experimental environments but they had both positive and negative signs indicating different directions of influencing the trait of interest. The genotypes had both specific and broad adaptation as exhibited by their diverse rankings under different environments. The relationship between leaf and grain nitrogen (N) and phosphorous (P) with actual tolerance to low and high soil N and P conditions indicated significant genotypes, environments and genotypes by environment interactions. The analysis of genetic components gave highly significant GCA and SCA mean squares for the days to heading, anthesis and maturity, phosphorous and nitrogen tolerance, top biomass, plant height, number of panicles, 1000 seed weight and grain yield under the four soil experimental environments. The general combining ability (GCA) effects for parents were significant for both F(2) and F(3) segregants for the above ten traits and under the experimental conditions. The leaf and seed N and P contents gave significant genotype mean squares values for the three mega-environments. The GCA and SCA mean squares were significant for leaf and seed N and P for the three locations. The same case was observed for GCA effects for the parents, with all the nine parents showing significant values for leaf N and P for the three locations. The GCA:SCA mean square ratios were generally larger than 1.0 values for leaf N and grain P for the three locations. This clearly manifested the predominant role of additive gene action over the dominance effects and that improvement for tolerance to low soil N and P condition was practical. The fragrance in rice is an important quality traits and an experiment was conducted to establish the association of organoleptic test and the variations in SSR marker (RM223) among the parents. Plant leaves of the parental genotypes were sampled for DNA analysis using SSR markers RM223 and RM284. The RM223 was polymorphic while, RM284 gave monomorphic results. Four alleles were detected as follows: 146, 155, 161, and 163. The GCA:SCA ratio was 1.48 indicating that the fixable additive effects were greater than the nonadditive effects in inheritance of fragrance. Test for independence using Chi-square indicated that there was no association between organoleptic and variations in SSR markers. The performance of the aromatic lines was generally lower than that of non-aromatic lines for the agronomic traits. There was negative relationship between fragrant score and other agronomic traits such as grain yield across the three soil N and P regimes. The NERICA1 was found to be a good donor for aroma based on sensory testing and can be exploited in a breeding programme. Analysis on genetic components governing grain yield was investigated using Hayman’s analysis to generate various components of variation and to elucidate their potential, utilization, and provide a thorough understanding of their role in grain yield development and improvement. The additive gene action was significant for all cases of soil N and P environments and the two mega-environments. They were also far above the dominant gene effects, indicating the need to use mass selection in the early generations of segregating populations. The narrow sense heritability (h2(ns)) for upland rice grain yield was 21.52% and 4.22% under soil P and N conditions, respectively. / Thesis (Ph.D.)-University of KwaZulu-Natal, Pietermaritzburg, 2010.