Abstract
Quantifying costs and benefits of ostensibly mutualistic interactions is an important step toward understanding their evolutionary trajectories. In food‐for‐protection interactions between ants and extrafloral nectar (EFN)‐bearing plants, tending by aggressive ants may deter herbivores, but it may also deter pollinators. The fitness costs of pollinator deterrence are not straightforward for long‐lived iteroparous plants, because reproductive vital rates often contribute weakly to fitness relative to growth and survival (vital rates that may be enhanced by ant defense). We used field manipulations of ant and pollinator activity and demographic modeling to examine how pollination costs of ant defense translate to plant fitness, given the benefits of ant defense elsewhere in the plant life cycle. We contrasted the net fitness effects of alternative ant partner species. Our work focused on the tree cholla cactus, Opuntia imbricata , an EFN‐bearing plant associated with two ant species ( Crematogaster opuntiae and Liometopum apiculatum ) that differ in quality of defense against insect herbivores. We found that ant defense imposed pollination costs, despite evidence for ant‐repellent floral volatiles and temporal partitioning of ant and pollinator activity. The two partner species similarly reduced pollinator visitation and seed mass, and one ( C. opuntiae ) additionally reduced seed number. We used the experimental data and other long‐term demographic data to parameterize an integral projection model that integrated costs and benefits of ant defense over the complete plant life cycle. Model results indicated that the pollination costs of L. apiculatum were balanced by beneficial effects on growth, leading to a net fitness effect that was neutral to positive. By contrast, pollination costs outweighed benefits for C. opuntiae , the weaker defender, rendering this species a reproductive parasite. Thus, we infer that pollination costs destabilize mutualism with one partner species, but are offset by strong defensive benefits provided by the other, leading to contrasting selective pressures imposed by alternative associations. Accounting for ontogenetic turnover in ant partner identity indicated that most plants avoid the parasitic effects of C. opuntiae by associating nonrandomly with L. apiculatum at reproductive life stages. Our results highlight the value of a demographic approach to quantifying the costs and benefits of mutualism.