Seven tropically adapted, late maturity maize gene pools and populations (Populations 21, 22, 25, 29, 32 and 43, and Pool 24) developed and improved at CIMMYT were crossed in a 7 x 7 diallel mating system. The parent and 21 crosses were evaluated for grain yield, days to silk and plant height at five locations in Mexico, and one each in Colombia and Thailand. The objectives of the study were to determine the heterosis and combining ability among these materials and to identify appropriate germplasm suitable for hybrid development work. The combined analysis of variance for all three traits measured showed no significant genotype x environment (G x E) interaction. Genotypes themselves showed significant differences for all three traits, as did their partitions into parents and crosses. General combining ability (GCA) was significant for all traits whereas specific combining ability (SCA) was not significant for any of the traits. Mean grain yield for the trial was 6.98 t/ha. The highest yielding combination was Population 21 x Population 43 (7.83 t/ha) followed by crosses of Population 22 with Population 32 and 43 (7.55 t/ha). Population 29 x Population 32 (dent x flint) yielded 7.36 t/ha while exhibiting the maximum high-parent heterosis (12.7%) for yield. Population 43 (La Posta) was the tallest and latest parent and produced high yields in crosses with other populations. All crosses to Population 43 were shorter and earlier than Population 43 per se. Population 21 (0.24 t/ha), Population 22 (0.15 t/ha) and Population 43 (0.23 t/ha) possessed significant positive GCA for yield. Populations 21 and 43 also showed significant positive GCA for days to silk while only Population 43 showed positive GCA for plant height. Significant (P less-than-or-equal-to 0.05) positive SCA effects for yield were observed in two crosses involving Population 32 with Population 22 and 29 (flint x dent). Based on our study the best choices for initiating hybrid work are Populations 21, 22, 29 and 43. Many of the Tuxpeno based populations are ideal candidates for interpopulation improvement using Population 32 (ETO Blanco) as heterotic partner.