The African Rice Gall Midge (AfRGM), Orseolia oryzivora Harris and Gagné (Diptera: Cecidomyiidae), is an endemic rice pest found throughout Africa. The failure of most other control methods imposes the need to use crop resistance. This study was initiated: (1) to develop an accurate method for assessing damage caused by AfRGM; (2) to determine AfRGM resistance genes’ modes of action, the heritability estimates of their resistance to AfRGM and the behavioural pattern of progenies with resistance to AfRGM attack; (3) to reveal convergent evolution of same or similar resistance gene(s) in geographically distinct landraces, or divergent evolution of genotypes carrying the same gene, by analysing the genetic diversity among five AfRGM parental lines; (4) to build a core sample of progenies to be used as a reduced mapping population, largely reflecting the entire genome of the whole population, after an estimate of the heritability of 15 agro-morphological descriptors and; (5) determine Simple Sequence Repeat (SSR) markers flanking genes or quantitative trait loci (QTLs) linked to resistance to AfRGM. A method of accurately assessing damage caused by AfRGM was determined by comparing four methods of assessment including the International Rice Research Institute’s (IRRI) Standard Evaluation System (SES) for rice and three methods based on resistance index (RI) assessments differing in the computing of the percentage of tillers with galls on a resistant check variety. The RI-based assessment (RI-BA) methods consistently provided a better evaluation of AfRGM damage than the SES, regardless of the trial size. Within RI-BA methods, RI-BA2 was always more accurate than RI-BA1 and RI-BA3 when the plot was large. RI-BA2 and RI-BA3 were equally accurate when the plot size was small, and they provided better estimates than RI-BA1. When the plot was of medium size, RI-BA2 was more accurate than RI-BA3; RI-BA3 also surpassed RI-BA1. Overall, the best method of assessing AfRGM damage was RI-BA2, regardless of the plot size. Five rice populations including F1, F2 and F3 generations involving ITA306, a susceptible variety of Oryza sativa subsp. indica, and four varieties having different reactions against AfRGM were used to determine the genetic basis of resistance and estimate the heritability of resistance to AfRGM. All the F1s were susceptible, suggesting recessive gene inheritance. The F2 generations’ segregation pattern of 1R:15S in both ITA306-TOS14519 and ITA306-TOG7106 crosses as well as the segregation of 1R:8Seg:7S in ITA306-TOS7106 F3 families indicated that the AfRGM resistance expression being studied is governed by two genes. The deviation of the segregation patterns of crosses involving ITA306 and the tolerant parental lines from Mendelian segregation ratios suggests that the tolerance to AfRGM shown by BW348-1 and Cisadane is under complex mechanisms of control rather than under simple genetic control. The narrow-sense heritability estimates of resistance to AfRGM were low in populations involving tolerant varieties and were high in populations involving resistant varieties. They ranged from 0.086 in the ITA306-Cisadane population, to 0.4 in the ITA306-TOG7106 population. Conversely, the broad-sense heritability estimates ranged from 0.23 (ITA306-Cisadane) to 0.63 (ITA306-TOS14519). The behavioural patterns of progenies against AfRGM attack were evaluated for 532, 413 and 479 F2 progenies from ITA306-BW348-1, ITA306-Cisadane and ITA306-TOS14519 crosses, respectively, in addition to 90 BC1F2 progenies from the ITA306 and TOG7106 cross. One F3 generation of 649 families from a cross between ITA306 and TOS14519 was also tested. Four types of behavioural pattern categories were observed: (1) progenies were more resistant than the resistant check entry at 45 DAT and 70 DAT; (2) progenies were more resistant at 45 DAT and became susceptible at 70 DAT; (3) progenies were susceptible at both 45 DAT and 70 DAT; (4) progenies were susceptible at 45 DAT but reverted to resistant at 70 DAT. The first three categories were the most frequently observed and occurred in all cross combinations. The last category was observed only for a few progenies from the ITA306-TOS14519 F2 and F3 generations and, surprisingly, many from the ITA306 and BW348-1 cross. Heritability estimates were calculated for 15 major traits in an F3 population in order to predict the genetic gain associated with each trait, together with the resistance to AfRGM and to estimate the influence of the environment on phenotypic values. Broad-sense heritability (H2) estimates were high for the penultimate leaf length (PLL) - 0.99, penultimate leaf width (PLW) – 1.0, flag leaf length (FLL) - 0.99, flag leaf width (FLW) – 1.0, ligule length (LigL) - 0.99, tillering ability (Til) - 0.99, number of days to booting (DB) - 0.95, number of days to first heading (DFH) - 0.96, number of days to heading (DH) - 0.89, number of days to maturity (DM) - 0.98, culm length (CL) - 0.99, plant height (PH) - 0.99, panicle length (PanL) - 0.95, secondary branching (SB) - 0.95 and the thousand grains weight (TGW) - 0.71. Conversely, narrow-sense heritability estimates were very low (nearly 0) in PLL, FLL, Lig, DB, DFH, DM and SB or low (at most 0.267) in PLW, FLW, DH and PH, with a high value of 0.727 for TGW. Inheritance of the traits studied was therefore under non-additive gene effects rather than additive genetic effects and can therefore be improved using pedigree breeding schemes along with breeding for AfRGM resistance. Fine genetic evaluation of five AfRGM parental lines was studied in terms of polymorphisms using 303 SSR primers covering the rice genome. Of the 178 polymorphic primers identified, 60 were highly polymorphic and informative. The number of alleles amplified by these primers ranged from one to five for a total of 1,041 alleles. The polymorphism rate was globally high, ranging from 45.2% to 66.8%. The mean of the polymorphism information content (PIC) was 0.553. Factorial analysis, based on the allelic diversity, demarcated the parental lines into Oryza glaberrima Steud, Oryza sativa subsp. japonica and O. sativa subsp. indica groups, while a cluster analysis distinguished them into four groups: AfRGM resistant, susceptible, moderately resistant and tolerant. BW348-1 and Cisadane showed the least diversity, despite their distant geographical origins. TOS14519 and TOG7106 showed more divergence to ITA306 despite their common West African origin. This variability amongst the genotypes tested is the result of farmer-based selection for AfRGM resistance rather than direct breeding efforts through breeder intervention. A method of selecting individuals for a mapping population, based on a core sample, was developed in order to speed up the mapping procedure. A diversity study amongst F2 and F3 generations involving 15 quantitative and 26 qualitative agro-morphological characters was carried out and led to the dropping of seven non-discriminant descriptors. The diversity index (H) was calculated for each remaining character and the discriminant descriptors were selected based on a diversity index threshold value above 0.4. Four descriptors of H values less than 0.35 were therefore dropped. The sizing of the core collection of 64 individuals and the selection of these individuals were done using MSTRAT version 4.1 package in redundancy mode, a construction run of 100 times with an iteration number of 500. The core sample was similar to the whole population for clustering pattern, minimum and maximum quantitative values and diversity index, while mean values and coefficient of variation distinguished them. The core sample, which represents 10% of the whole population, also revealed the same phenotypic variation and the same genotypic segregation according to two SSR markers. It can therefore efficiently reflect the whole population as a mapping population. Finally, a study was undertaken to identify flanking markers to the gene/QTL involved in the resistance against AfRGM using bulked segregant analysis (BSA). A polymorphism study between ITA306 and TOS14519 displayed 145 polymorphic SSR markers, which were used to screen the bulks that originated from the two tails, and depicted only two SSRs as candidate markers linked to gall midge resistance. These markers included RM317 and RM17303 which displayed strong significance after an analysis of variance using an F test, meaning that they were segregating with the resistant alleles. / Thesis (Ph.D.)-University of KwaZulu-Natal, Pietermaritzburg, 2012.