Inter-generation correlations between F2-F3, F2-F4 and F3-F4 in six crosses of chickpea were calculated by using individual plant/progeny means. In general, correlation values in case of plant height, seeds per pod and 100-seed weight were higher between F3 and F4 than those between F2 and F3 and F2 and F4, However, inter-generation correlations were mostly nonsignificant in case of pods per plant and grain yield. These results clearly show that selection in F3 can be fruitful for seeds per pod and 100-seed weight. However, selection for pods per plant and grain yield in early generations may not show any relationship with later generation performance for these traits.To achieve maximum efficiency and progress in breeding for any character, effective selection carried out in early generations would be advantageous. SHEBESKI (1967) suggested yield testing in early generations when most of the variability is present, assuming that there is a positive relationship of the yield of the genotypes in the early generations with the yielding ability of the later generation derivatives of such genotypes. MAHMUD and KRAMER (1951) observed in soybean that F3 lines provided good estimates of the average yield potentials of F4 segregates if genotype-environment interaction can be kept under control.Results using the yields of F2 derived lines to predict the lines of later generations have been inconsistent (FREY 1954, MCKENZIE and LAMBERT 1961, BRIGGS and SHEBESKI 1971and KNOTT and KUMAR 1975 and often poor correlations were obtained. In some of these experiments, the generations were grown in different years, so the interaction between geriotype and environment could have affected the correlations. Often heterozygosity and variability within an F2 derived line is the cause of poor relationship between it and the ') Part of Ph, D, thesis submitted by the senior author to IARI,