1 Tachigali myrmecophila (Caesalpinaceae) is an Amazonian myrmecophytic canopy tree. The saplings are shade-tolerant, surviving in a suppressed state for several years in the rainforest understorey, the hollow leaf rachis and petiole being inhabited by the stinging ant Pseudomyrmex concolor (Pseudomyrmecinae). An experiment was designed (a) to evaluate the role of insect herbivory in the mutualism between T. myrmecophila and P. concolor and (b) to test a falsifiable hypothesis proposed by a current trade-off model of chemical and ant defences which predicts that myrmecophytic traits should be limited to plants with leaves of low longevity. 2 Plants from which the ants were experimentally removed had 4.3 times more herbivorous insects than plants with ants. All recorded orders of insects were attacked similarly by the ants. Rates of herbivory were: (a) ten times higher on experimental plants lacking ants, (b) about three times higher on immature than mature leaves, (c) about two and a half times higher in the wet than the dry season. After 18 months, the experimental plants presented an accumulated level of leaf herbivory which was about twice as high as for the plants with ants. 3 Observations on the phenology of control, experimental, and naturally unoccupied plants revealed that the leaf longevity of plants with active ant colonies was unexpectedly high for rainforest (81 months), and about 1.8 and 2.6 times as high as the experimental (45 months) and the naturally unoccupied (31 months) plants, respectively. The high leaf longevity of the myrmecophytic T. myrmecophila does not support the current trade-off model of chemical and ant defences. The rate of apical growth was 1.6 times higher for plants with ants than plants from the experimental group. The slow growth rate of T. myrmecophila, about 14 cm year(-1), is compatible with the resource availability hypothesis. 4 Phenological differences between experimental and naturally unoccupied plants suggest that descriptive-correlative studies would not always give a true picture and that the actual paradigm of ant-plant interactions, derived in part from this approach, should be carefully revised. 5 The results corroborate the hypothesis that the interaction between Tachigali myrmecophila and Pseudomyrmex concolor is mutualistic, and suggest that attack by phytophagous insects is the prime factor in the evolution of the myrmecophytism.824833842