Predictive models were developed for Cordia alliodora branch and Theobroma cacao branch or leaf biomass,based on branch basal areas (r2 ≥ 0.79) but the model of C. alliodora leaf biomass, although significant, was of very low accuracy (r2 = 0.09) due to annual leaf fall. At age 10 years, shade tree stem biomass accounted for 80% of the total above-ground biomass of either tree. However, between the ages of 6 and 10 years, the biomass increment of T. cacao branches (3-4t.ha-1.a-1) was similar to that of the shade tree stems. During the same period, the net primary productivity was 35 and 28 t.ha-1.a-1, for the Erythrina poepigiana and and C. alliodora systems, respectively. Cocoa production under either of the shade trees C. alliodora or E. poeppigiana was 1000 kg.ha-1.a-1 (oven-dry; ages 6-10 yr). During the same period, C. alliodora timber production was 9 m3.ha-1.a-1 whilst the leguminous shade tree E. poeppigiana does not produce timber. Litterfall over the same 5 years, including crop and/or shade tree pruning residues, averages 11 and 23 t.ha-1.a-1, respectively. The main difference was due to E. poeppigiana pruning residues (10t.ha-1.a-1). Soil organic material reserves (0-45 cm) increased over 10 years from 198 to 240 t.ha-1 in the E. poeppigiana plots and from 168-184 t.ha-1 in the C. alliodora plots. These values, together with the productivity indices presented, provide evidence that the systems are sustainable. For economic reasons, the use of C. alliodora is recommended under the experimental conditions. however, on less fertile soils without fertilization, the greater biomass and hence nutrient return to the soil surface under E. poeppigiana, might make this the preferable shade tree. © 1990 Kluwer Academic Publishers.