By estimating relative costs and benefits, we explored the role of the homopteran partner in the protection mutualism between the myrmecophyte Leonardoxa africana T3, the ant Aphomomyrmex afer, and sap–sucking homopterans tended by ants in the tree's swollen hollow twigs. The ants obtain nest sites and food from their host–plant (food is obtained either directly by extrafloral nectar or indirectly via homopterans). Aphomomyrmex workers patrol the young leaves of L. africana T3 and protect them against phytophagous insects. Because ants tended, either solely or primarily, coccids in some trees and pseudococcids in others, we were able to study whether the nature of the interaction was dependent on the identity of the third partner. First, the type of homopteran affects the benefits to the tree of maintaining a large ant colony. Larger colony size (relative to tree size) confers greater protection against herbivory; this relationship is more pronounced for trees whose ants tend pseudococcids than for those in which ants tend coccids. Second, for trees (and associated ant colonies) of comparable size, homopteran biomass was much larger in trees harbouring coccids than in trees with pseudococcids. Thus, the cost to the tree of maintaining ants may be greater when ants are associated with coccids. The net benefits to the plant of maintaining ants appear to be much greater with pseudococcids as the third partner. To explore how the type of homopteran affects functioning of the system, we attempted to determine which of the resources (nest sites, extrafloral nectar, and homopterans) is likely to limit ant colony size. In trees where ants tended coccids, ant–colony biomass was strongly dependent on the number of extrafloral nectaries. In contrast, in trees whose ants tended only pseudococcids, colony biomass was not related to the number of nectaries and was most strongly determined by the volume of available nest sites. We present hypotheses to explain how the type of homopteran affects functioning of this symbiosis, and discuss the implications of our study for the evolutionary ecology of ant–plant–homopteran relationships.