The Circe principle explains how resource-rich land can waylay pollinators in fragmented landscapes.

Global declines in pollinators, associated with land-use change [1-6] and fragmentation [7-10], constitute a serious threat to crop production and biodiversity [11]. Models investigating impacts of habitat fragmentation on pollen flow have categorized landscapes simply in terms of habitat and nonhabitat. We show that pollen flow depends strongly on types of land use between habitat fragments. We used paternity analysis of seeds and a combination of circuit and general linear models to analyze pollen flow for the endangered tree Gomortega keule (Gomortegaceae) [12] in the fragmented Central Chile Biodiversity Hotspot [13]. Pollination probability was highest over pine plantation, moderate over low-intensity agriculture and native forest, and lowest over clearfells. Changing the proportions of the land uses over one kilometer altered pollination probability up to 7-fold. We explain our results by the novel "Circe principle." In contrast to models where land uses similar to native habitat promote pollinator movement, pollinators may actually be waylaid in resource-rich areas between habitat patches. Moreover, pollinators may move with higher probability between habitat patches separated by some resource-poor land uses. Pollination research in fragmented landscapes requires explicit recognition of the nature of the nonhabitat matrix, rather than applying simple binary landscape models.