Evaluation of the agronomical triats of the 1981 Amazonian germplasm has continued in Cote dIvoire since 1985. These Amazonian genotypes are only foreseen as suitable for use in a mating programme with Wickham clones as the main partners because of their weak levels compared to Wickham clones. A new population, composed of 387 genotypes selected for grwth, yield and architecture, is studied in the paper and in particular the genetic structure and the expected genetic gain of the progenies. Analysis of yield and growth data shows that 20 to 25;of the total phenotypic variability is due to genetic sources. The level of the expected genetic gain seems to be higher when 10 to 30;of the whole population is integrated into the mating programme. The studies will be continued taking into account more characters of agronomical interest to precisely determine the size of the working population.