Résumé
Leonardoxa africana is an understorey tree of the coastal rainforests of Cameroon that provides food (extrafloral nectar) and nest sites (swollen, hollowed internodes) for ants. Only two host-specific species of ants inhabit this tree. The present study showed that their distributions were mutually exclusive: the formicine ant Petalomyrmex phylax inhabited three-fourths of all occupied trees, while the myrmicine Cataulacus mckeyi occupied the remaining trees. In order to determine whether the presence of each of these ants was beneficial to the trees, we compared rates of damage on young leaves patrolled by ants and on young leaves from which ants were experimentally excluded. We also studied activity rhythms of the two ants and their responses to phytophagous insects they encountered. Young leaves patrolled by Petalomyrmex suffered significantly less damage than those from which ants were excluded (2% versus 24%). In contrast, young leaves patrolled by Cataulacus suffered much greater herbivory (31%) than those patrolled by Petalomyrmex, and herbivory on ant-patrolled leaves was not significantly different from that on ant-excluded leaves (46%). Behavioural observations help to explain the difference between the effects of the two ant species. Number of workers active on plant surfaces was much greater on Petalomyrmex-occupied trees than on trees occupied by Cataulacus. Petalomyrmex workers patrolled young leaves constantly (day and night) and chased out or killed microlepidopteran larvae placed on young leaves. The patrolling activity of Cataulacus was only diurnal (rather well correlated with the activity of nectar production), and Cataulacus workers failed to consistently attack herbivores. These results confirm that Petalomyrmex is a mutualist of Leonardoxa and demonstrate that Cataulacus, exploiting the resources of its host, providing no protection and excluding the mutualist from trees it occupies, is a parasite of the mutualism. We discuss origin, ecological persistence and evolutionary stability of such a parasitic strategy in plant-ant symbioses, and we ask whether mutualisms may evolve from such parasitic relationships.