Rice production in Sub-Sahara Africa is hampered by drought stress, low farmer adoption of improved varieties, un-adapted germplasm, problems of crossability between species and problems of efficient emasculation. This study was initiated to: (1) determine farmer preferences for rice varieties and production practices, (2) identify drought tolerant parental lines for use in interspecific crosses, (3) determine crossability between different rice species, (4) identify drought tolerant interspecific segregants from crosses, and (5) study the efficacy of ethrel and gibberrilic acid (GA3) as chemical hybridising agents. Participatory rural appraisal (PRA) was conducted in Sikasso region in the month of September 2005 to determine farmer preferences for varieties and production constraints. A total of 125 rice farmers from 10 villages participated. This study found that farmer preferences for rice traits did not coincide with the breeders' objectives and trait preferences varied across ecologies. In upland and lowland rice ecologies, the most preferred characteristic was short duration. In the lowland tall plants was most preferred and short duration trait ranked second. Long duration ranked as the second most preferred trait after yield in the irrigated rice ecology. The estimated grain yield reduction was 60% due to drought stress from all ecologies. Sixteen genotypes including seven Oryza glaberrima Steud, six interspecific (0. saliva x 0. glaberrima) inbred genotypes and two improved 0. saliva L. subsp Gaponica) genotypes, with OS 6, as drought tolerant check were screened for drought tolerance at the vegetative stage during the dry seasons of 2004 and 2005 at Samanko research station, in Mali. Most of the genotypes could be regarded as exhibiting drought tolerance and avoidance mechanisms for most of the morpho-physiological characters under study. High selection index was observed for WAB 450-I-B-P-103-HB in the two season experiments. Drought stress tolerance (little leaf drying) was significantly correlated with leaf rolling, leaf area index, and plant height. Fast drought recovery was significantly correlated with drought tolerance, tiller number and leaf area index. Effects of genotypes, environment and their interactions were significant for tiller number, root dry weight and relative leaf water content. The most stable genotype was RAM 3 for tiller number and relative leaf water content. Thirteen genotypes were selected on the basis of these traits for rice hybridisation. A factorial arrangement involving three levels of gametocide concentrations and five application stages of the female rice parent was undertaken for each gametocide ethrel and GA3 for two seasons. Per cent empty spikelets and male sterility increased with concentration of ethrel. Male sterility induction of 41.5% and 42.8% at 2000 ppm was achieved in 2004 and 2005, respectively, while at 4000 ppm male sterility induction was 40.5 % and 46.1% in 2004 and 2005, respectively. The best application stage of ethrel that induced highest male sterility was at P4 (panicle primordium differentiation + spikelet primordium differentiation) of 46.6% in 2004 and 49.2% at PI (panicle primordium differentiation) in 2005. There was no measurement on GA3 for sterility due to lodging. Total male sterility was not achieved therefore, these gametocides could not be recommended for use as hybridising agents. Crossability was studied in four groups: group A (0. glaberrima x 0. saliva), group B (0. glaberrima x Interspecific), group C (Interspecific x 0. saliva) and group D (Interspecifics x Interspecific). High percent seed set was obtained in group D of 19.08% and group A had 9.29%, in group B the seed set was 11.26% and group Chad 6.62 %. The three best combiners as female parents based on percentage seed set were CG 14 (0.glaberrima), WAB 450-IBP-I05-HB and WAB450-I-B-P-I03-HB (interspecifics), and for male parents were NERICA 3, NERICA 2 (interspecifics) and WAB 375-B-9-H3-2 (0. sativa). Genetic components of variance and heritability estimates were studied in F3 and F4 generation progenies from North Carolina II design mating scheme. General combining ability (GCA), manly due to female and specific combining ability (SCA) were significant for tiller number, plant height and specific leaf dry weight indicating that additive and non-additive genetic effects controlled these traits. Heritability estimates for drought tolerance and leaf rolling were highly significant for all the populations examined, ranging from 0.32 to 0.72 and 0.22 to 0.56, respectively. Drought tolerance expressed as little leaf drying was significantly associated with less leaf rolling, larger leaf area index, and taller plants. Some genotypes and progenies such as WBK 39-B-B-B, WBK 30-B-B and WAB 450-I-B-P-I03-HB were identified as being drought tolerant with the ability to recover quickly from drought stress. There was a strong desire for tall plants and short duration rice in upland and lowland ecologies, while high yield was preferred in irrigated ecology. Therefore, the best breeding strategy would be ecosystem breeding targeting each of the ecologies. Screening for secondary traits that have been identified to be significantly associated with rice grain yield could be used for direct selection for grain yield under drought stressed conditions. Progenies identified, as drought tolerant would be considered for use as source germplasm in breeding that target Sikasso region and any other similar ecologies. This study therefore suggested that participatory research approaches that involve all stakeholders in rice production would be valuable in identifying and generating drought tolerant and adoptable rice cultivars in the tropics of Africa with considerable impact on food security. / Thesis (Ph.D.)-University of KwaZulu-Natal, Pietermaritzburg, 2006.