Résumé
Dispersal is a central life-history trait for most animals and plants: it allows to colonize new habitats, escape from competition or avoid inbreeding. Yet, not all species are mobile enough to perform sufficient dispersal. Such passive dispersers may use more mobile animals as dispersal vectors. If multiple potential vectors are available, an active choice can allow to optimize the dispersal process and to determine the distribution of dispersal distances, i.e. an optimal dispersal kernel. We explore dispersal and vector choice in the neotropical flower mite Spadiseius calyptrogynae using a dual approach which combines experiments with an individual-based simulation model. Spadiseius calyptrogynae is found in lowland rainforests in Costa Rica. It inhabits inflorescences of the understorey palm Calyptrogyne ghiesbreghtiana and is phoretic on a number of flower visitors including bats, beetles and stingless bees. We hypothesised that the mites should optimise their dispersal kernel by actively choosing a specific mix of potential phoretic vectors. In a simple olfactometer setup we showed that the flower mites do indeed discriminate between potential vectors. Subsequently we used an individual-based model to analyse the evolutionary forces responsible for the observed patterns of vector choice. The mites combine vectors exhibiting long-distance dispersal with those allowing for more localized dispersal. This results in a fat-tailed dispersal kernel that guarantees the occasional colonization of new host plant patches (long distance) while optimizing the exploitation of clumped resources (local dispersal). Additionally, kin competition results in a preference for small vectors that transport only few individuals at a time. At the same time, these vectors lead to directed dispersal towards suitable habitat, which increases the stability of this very specialized interaction. Our findings can be applied to other phoretic systems but also to vector-based seed dispersal, for example.