Rice (Oryza sativa L.) is the second most important cereal crop in Malawi. Rice productivity in the dominant (85%) rainfed ecosystem is very low, with mean yield of only 1.4 t ha-1. Farmers plant traditional landraces that have low yield potential and are susceptible to various stresses such as drought and diseases. Development and introduction of improved rice varieties, with stress tolerance traits from introduced varieties, such as the New Rice for Africa (NERICAs), could significantly increase productivity. Previous attempts to introduce high yielding irrigated varieties into the dominant rainfed ecosystem in Malawi have not been successful because farmers basically did not adopt the varieties, claiming that the varieties were lacking in grain traits that they preferred but that the traits were present in their landraces. The notable traits mentioned, through previous informal surveys, were long grains, medium to slender shape and aromatic grain with intermediate gelatinization temperature. No formal study has been conducted to ascertain the preferences, and the genetic control of the traits, including yield and yield related traits, have not been studied. The objectives of this study were to: 1) confirm farmers’ preferences for grain traits using participatory rural appraisal; 2) determine amount of genetic variability for yield and yield related traits in Malawi rice landraces, 3) determine the genetic control and correlations of grain length, grain shape and 1000-grain weight, 4) determine the inheritance of aroma and gelatinization temperature and, 5) determine the genetic control, correlations and path coefficients of yield and yield related traits, in F2 generations of Malawi rice landraces and NERICA varieties crosses. To confirm the farmers’ preferences for grain traits, a participatory rural appraisal was conducted in 2006 in two villages that were representative of rainfed rice growing areas in Malawi. The villages were Liundi and Nawanga in Machinga and Salima Districts, respectively. Qualitative and quantitative data were collected through questionnaires and discussions with 190 respondents, as well as through observations. To determine variability among Malawi rice landraces, 19 landraces were planted at Lifuwu in a Randomized Complete Block Design with three replicates in 2006. Data on plant height, days to 50% flowering, number of panicles per hill, panicle length, number of filled grains per panicle, 1000-grain weight, panicle weight, grain length and grain yield were collected and analyzed. Four Malawi rice landraces were crossed to four NERICA varieties in 2006 in a North Carolina Design II mating scheme to determine the genetic control of grain size. F1 plants were raised in 2007 and in 2008, 16 F2 populations together with their parents were planted in a Randomized Complete Block Design with three replicates at Lifuwu. Data on grain length, grain shape and 1000-grain weight were collected and analyzed. To determine the inheritance of aroma and gelatinization temperature, four Malawi rice landraces were crossed to four NERICA varieties in 2006 and F1 plants were raised in 2007. In 2008, 16 F2 populations together with their parents were planted in a Randomized Complete Block Design with three replicates at Lifuwu. Aroma and gelatinization temperature were evaluated. To determine the gene action of yield and yield related traits, four Malawi rice landraces were crossed to four NERICA varieties in 2006 in a North Carolina Design II mating scheme and F1 plants were raised in 2007. In 2008, 16 F2 populations together with their parents were planted in a Randomized Complete Block Design with three replicates at Lifuwu. Data on grain yield, the number of panicles per hill, days to 50% flowering, panicle length, panicle weight and 1000-grain weight were collected and analyzed The participatory rural appraisal confirmed that that long, slender or medium shape grains, with aroma and intermediate gelatinization temperature were the key traits preferred by farmers. Therefore the farmer preferred traits of long, slender grains, with aroma and medium gelatinization temperature, must be selected for in any high yielding varieties to be developed for the rainfed rice ecosystem. Results showed that differences were significant (P=0.05) for all the traits that were studied. Heritability estimates were low to moderate: 18.3% for panicle weight, 40.0% for panicles per hill and 56.3% for days to 50% flowering date. The high genetic variability among the landraces could be used in a breeding programme to develop improved varieties for various morphological traits. The number of panicles per hill and 1000-grain weight combined moderate heritabilities with relatively high genetic advance and therefore could be reliable traits for yield improvement. Genetic analysis of grain size showed that Malawi rice landraces were variable for all three grain size characteristics, namely grain length, grain shape and 1000-grain weight. NERICA varieties were variable for 1000-grain weight. Sixteen F2 progenies were variable for all three characteristics, and the variability was significant (P=0.05). Heritability estimates were high (45.4%) for grain length and low for grain shape (12.3%) and for 1000-grain weight (14.3%) suggesting that early generation selection would be effective for grain length. Predominance of additive gene action for grain length and grain shape suggested that early generation selection would be effective for these traits. Selection for 1000-grain weight would be more effective in later generations because of preponderance of non-additive gene action in the control of this trait. The correlation between grain length and grain shape was positive (r=0.769) and highly significant (P=0.01) suggesting that breeders would choose to select for both traits simultaneously, or they would choose one of the traits to develop varieties with long grains and medium shape. Crosses between aromatic and non-aromatic varieties had non-aromatic F1. The F2 progenies segregated into 3:1 ratio for non-aromatic: aromatic suggesting that, in the Malawi rice landraces, aroma was probably simply inherited through a single recessive gene. F1 progenies, between parents with high and intermediate gelatinization temperatures had intermediate gelatinization temperature. F2 progenies segregated into 1:3 ratios for high; intermediate gelatinization temperature in three out of four crosses suggesting control by one dominant gene. The segregation pattern in one cross was not significantly different from 3:13 ratio for high: intermediate suggesting that two dominant genes, one an inhibitor, were controlling the trait. Breeding and selecting for aroma and intermediate gelatinization temperature could be accomplished relatively easily because the traits are simply inherited. The genetic variability for yield and yield related traits was wide and significant (P=0.05) in the F2 populations of Malawi rice landraces and NERICA varieties crosses indicating that the populations would be valuable sources to develop varieties with improved yield. Panicle weight and the number of panicles per hill were positively correlated with, and had high direct effects on grain yield, therefore they could be used to indirectly select for high yield. Grain yield, the number of panicles per hill and plant height were predominantly controlled by additive gene action suggesting that bulk breeding methods would be adopted for these traits. The days to 50% flowering, panicle weight and 1000-grain weight were predominantly under the control of nonadditive gene action suggesting that hybrid development would be profitable for these traits. Faya Mpata, Faya Zidyana and NERICA 3 could be the best parents for improving yield and yield related traits because they had high general combining abilities for the traits. / Thesis (Ph.D.)-University of KwaZulu-Natal, Pietermaritzburg, 2010. / Thesis (Ph.D.)-University of KwaZulu-Natal, Pietermaritzburg, 2010.