Animal-mediated pollination is essential for the maintenance of plant reproduction, especially in tropical ecosystems, where pollination networks have been thought to have highly generalized structures. However, accumulating evidence suggests that not all floral visitors provide equally effective pollination services, potentially reducing the number of realized pollinators and increasing the cryptic specialization of pollination networks. Thus, there is a need to understand how different functional groups of pollinators influence pollination success. Here, we examined whether patterns of contemporary pollen-mediated gene flow in Heliconia tortuosa are consistent with the foraging strategy of its territorial or traplining hummingbird pollinators. Territorial hummingbirds defend clumps of flowers and are expected to transfer pollen locally. In contrast, traplining hummingbirds forage across longer distances, thereby increasing pollen flow among forest fragments, and are thought to repeatedly visit particular plants. If trapliners indeed visit the same plants repeatedly along their regular routes, this could lead to a situation where neighboring plants sample genetically distinct pollen pools. To test this hypothesis, we genotyped 720 seeds and 71 mother plants from 18 forest fragments at 11 microsatellite loci. We performed TwoGener analysis to test pollen pool differentiation within sites (among neighboring plants within the same forest fragment: Φ SC) and between sites (among forest fragments: Φ CT). We found strong, statistically significant pollen pool differentiation among neighboring mother plants (Φ SC = 0.0506), and weaker, statistically significant differentiation among sites (Φ CT = 0.0285). We interpret this pattern of hierarchical pollen pool differentiation as the landscape genetic signature of the foraging strategy of traplining hummingbirds, where repeatable, long-distance, and high-fidelity routes transfer pollen among particular plants. Although H. tortuosa is also visited by territorial hummingbirds, our results suggest that these pollinators do not contribute substantially to successful pollination, highlighting differences in realized pollination efficiency. This cryptic reduction in the number of realized pollinators potentially increases the vulnerability of pollination success to the decline of populations of traplining hummingbirds, which have been shown to be sensitive to forest fragmentation. We conclude that maintaining habitat connectivity to sustain the foraging routes of trapliners may be essential for the maintenance of pollen-mediated gene flow in human-modified landscapes.

Deforestation can impact the quality of pollen received by target plants (i.e. delivery of incompatible pollen, self‐pollen or pollen from closely related individuals). Such reductions in plant mating quality may be direct, when deforestation reduces plant population size and the availability of pollen donors, or indirect, when decreased mating quality results, for example, from shifts in the composition of the pollinator community. As most flowering plants depend on animal pollinators for reproduction, there is a need to understand the direct and indirect links between deforestation, pollinator community composition and plant mating quality. We quantified the direct, pollen donor‐mediated and indirect, pollinator‐mediated effects of deforestation on mating quality in Heliconia tortuosa, a tropical herb pollinated by low‐ and high‐mobility hummingbirds. We used a confirmatory path analysis to test the hypothesis that deforestation (amount of forest cover and forest patch size) influenced mating quality (haplotype diversity of pollen pools, outcrossing and biparental inbreeding) directly and indirectly through functional shifts in the composition of pollinator communities (proportion of high‐mobility hummingbirds). We found that deforestation triggered functional shifts in the composition of pollinator communities, as the proportion of high‐mobility hummingbirds increased significantly with the amount of forest cover and forest patch size. The composition of the pollinator community affected mating quality, as the haplotype diversity of pollen pools increased significantly with the proportion of high‐mobility hummingbirds, while biparental inbreeding decreased significantly. Although we did not detect any significant direct, pollen donor‐mediated effects of deforestation on mating quality, reductions in the amount of forest cover and forest patch size resulted in functional shifts that filtered out high‐mobility hummingbirds from the pollinator community, thereby reducing mating quality indirectly. Synthesis. Deforestation primarily influenced plant mating quality through a cascading effect mediated by functional shifts in the composition of the pollinator community. Our results indicate that plant mating quality strongly depends on the composition of local pollinator communities. Functional shifts that filter out highly mobile and effective pollinators may reduce the transfer of genetically diverse pollen loads from unrelated plants. Such shifts may have pronounced effects on plant population dynamics and disrupt genetic connectivity.

Summary Resolving the consequences of pollinator foraging behaviour for plant mating systems is a fundamental challenge in evolutionary ecology. Pollinators may adopt particular foraging tactics: complete trapline foraging (repeated movements along a fixed route), sample‐and‐shift trapline foraging (a variable route that incorporates information from previous experiences) and territorial foraging (stochastic movements within a restricted area). Studies that integrate these pollinator foraging tactics with plant mating systems are generally lacking. We investigate the consequences of particular pollinator foraging tactics for Heliconia tortuosa. We combine parentage and sibship inference analysis with simulation modelling to: estimate mating system parameters; infer the foraging tactic adopted by the pollinators; and quantify the impact of pollinator foraging tactics on mating system parameters. We found high outcrossing rates, ubiquitous multiple paternity and a pronounced departure from near‐neighbour mating. We also found that plants repeatedly receive pollen from a series of particular donors. We infer that the pollinators primarily adopt complete trapline foraging and occasionally engage in sample‐and‐shift trapline foraging. This enhances multiple paternity without a substantial increase in near‐neighbour mating. The particular pollinator foraging tactics have divergent consequences for multiple paternity and near‐neighbour mating. Thus, pollinator foraging behaviour is an important driver of the ecology and evolution of plant mating systems.