Abstract We experimentally compared the effect of fertilization to the effect of interplanting two species of legume trees ( Inga edulis and Gliricidia sepium ) on growth of a native tropical tree, Terminalia amazonia (Combretaceae). The experiment was a randomized block design with trees planted in eroded cattle pasture in Southern Costa Rica. After 8 years, both block and treatment significantly explained variance in tree growth. Blocks differed in degree of erosion and initial soil nitrate and phosphate. All three factors significantly predicted tree growth, erosion negatively and N and P positively. However, erosion best explained block effect. Fertilizer (10N:30P:10K) had no effect on tree growth. In the treatments with legume trees, I. edulis survived and grew better than G. sepium . Concordantly, plots in which I. edulis was interplanted with T. amazonia showed a stronger positive effect on T. amazonia than plots in which G. sepium was the interplanted species. Possibly I. edulis increased nitrogen availability and-or aided growth by partial shading. A treatment that mixed all three species also improved growth of T. amazonia significantly over controls. This result suggests that fertilizing native trees outplanted for restoration or forestry may be wasted investment. However, intermixing legume trees may increase economic benefits to farmers interested in reforesting degraded land.