Growth index (GI) is a measure of vegetative growth rate for pearl millet (Pennisetum glaucum (L.) R. Br.) and can be determined either at 10 d after flowering (i.e., GI1) or at maturity (GI2). Its improvement may result in greater grain yields of pearl millet. This dissertation research was conducted to determine (a) the inheritance of GI and related traits in pearl millet and (b) the direct and indirect effects after one cycle of recurrent selection for GI1 in Higrop and Senpop, two pearl millet gene pools that were formed for this study at the International Crops Research Institute for the Semi-Arid Tropics (ICRISAT). The inheritance of GI was studied via a generation means analysis and showed that the sum of additive and additive x additive epistatic genetic effects were more important for the traits than the sum of dominance and heterotic genetic effects, although dominance effects occasionally were significant. A breeding method which selects for additive genetic effects should be useful for improving GI of pearl millet;Selection in Higrop and Senpop was practiced for (a) GI1 only and (b) GI1 with culling of lines with low grain yield (GY) and/or late bloom date (BD). Direct response from selection was assessed via means for BD, GY and GI1, computed from C0 and C1 cycle populations and from 46 random S[subscript]1 lines from C0 and C1 of Higrop and Senpop. Indirect responses from selection were assessed via means for 11 unselected traits, computed from the 46 random S[subscript]1 lines. Cycle populations from the culling procedure were superior to those when only GI1 was selected. When averaged across both gene pools, selection solely for GI1 resulted in a 2.9% lower BD, 1.6% lower GY and 7.6% greater GI1 than the C0 mean, whereas selection that included culling resulted in a 6.6% lower BD, 11.5% greater GY and 8.2% greater GI1 than the C0 mean. Simultaneous improvement also occurred for BD, GY and GI1 when S[subscript]1 lines were evaluated. Significant correlated responses in Higrop were decreased panicle length (2 cm), and in Senpop for increased panicle length, panicle girth, and panicles per plant (2 cm, 1.7 mm, and 0.3 panicles plant[superscript]-1, respectively), and decreased plant height (10.2 cm). Grain yield in both gene pools was phenotypically correlated (P < 0.01) with panicle yield, harvest index, and threshing percent and genotypically with panicle yield, panicles per plant, and harvest index. Growth index in both gene pools was phenotypically correlated (P < 0.01) only with plant height, and genotypically GI1 was negatively correlated with panicle girth, panicles per plant, and harvest index. Future selection in Higrop and Senpop for GI1 which includes culling of low GY and late BD lines should be successful and result in few, if any, undesirable correlated changes.