A prototype agroforestry system, which combined the low-input features and environmental benefits of the traditional Indonesian multi-species 'jungle rubber' system with high yielding rubber clones developed in monoculture plantations, was tested under on-farm conditions. Secondary forest was allowed to regenerate between weeded rows of clonal rubber. Clonal rubber establishment was studied, and the interactions between it, secondary forest species and farmer management were investigated using a combination of researcher- and farmermanaged weeding trials. In one trial, alteration of below-ground resources (using root barriers and trenches to create three soil volumes) did not affect above or below-ground growth of clonal rubber, although weeding significantly increased stem diameter and volume. It was concluded that secondary forest regrowth interfered with resource capture at the level of individual roots; interference was not due to depletion of total available resources. Shoot:root ratios and ratios of horizontally- to vertically-oriented proximal roots were not affected by weeding. Growth of clonal rubber in N-fertilised plots, in the presence of weeds, was significantly greater than in corresponding unfertilised plots, indicating that N-addition may overcome some negative effects of competition in the system. However, a bioassay of nutrient limitation showed no significant differences in root biomass or root-length density, for either rubber or weed rootingrowth into soil cores enriched with various nutrients. The second researcher-managed trial, on steep slopes, showed that the survival rate of clonal rubber was 33% higher than that of the 'seedling' rubber variety traditionally used, and that mean stem height and diameter of clonal rubber trees were significantly greater than those of seedling rubber, 21 months after planting. Damage to trees by banded leaf monkeys (Presbytis melalophos nobilis) and feral pigs (Sus barbatus) was severe, unexpected, and greater for seedling than for clonal rubber. For undamaged trees, weeding frequency within the rubber-tree row had no significant effect, indicating that the major influence on rubber tree growth was interference from secondary forest regrowth between rows, operating both aboveand below-ground. In a farmer-managed, trial, vertebrate pest damage was the major influence on clonal rubber establishment, explaining almost 70% of the variation in rubber growth. The amount of labour invested in weeding was positively correlated with rubber growth. However, fanners generally decided to completely cut back the secondary forest regrowth between rows of rubber trees, including potentially valuable trees, rather than weeding within the rows and selectively pruning inter-row trees. Farmers considered that the inter-row vegetation may harbour vertebrate pests and compete with the clonal rubber, and they had access to fruits, firewood and non-timber forest products on other land. Thus, contrary to expectations, when offered clonal germplasm, these 'progressive' farmers opted to use plantation methods to protect what they considered a valuable asset suited to monoculture, rather than maintain the traditional multispecies strategy they use with local germplasm. Thus, although clonal rubber can technically be established in a 'jungle rubber'-like system (albeit with lower growth rates than achieved in plantations), not all farmers may be prepared to adopt this type of system.