1. Cascading effects in ecological systems acting across three or more trophic levels can be either of a resource-based (bottom-up) or natural enemy-based (top-down) nature. But, due to their complexity these effects are often considered separately and their relative strength, acting simultaneously, remains unknown. 2. In a semi-natural field experiment using tansy (Tanacetum vulgare L.) and the specialised tansy aphid Metopeurum fuscoviride Stroyan as a model system, we compared the effects of four distinct plant chemotypes (i.e. bottom-up), defined by the bouquet of their volatile terpenoids, on aphid population dynamics by manipulating the presence/absence of mutualistic ants and presence/absence of naturally-occurring predators (i.e. top-down). 3. Predators reduced aphid abundance and colony survival but did not reduce initial growth rate due to a time lag until predators arrived on the plants. Ants directly benefited initial aphid growth rates and abundance, even in the absence of predators, but not the number of days an aphid colony persisted on the plant. 4. Plant chemotype directly affected aphid growth rate and final abundances across the different plants and indirectly affected the abundances of tending ants and predators through effects on aphids. We found that tending ants were more abundant on one plant chemotype. Although ant abundance did not affect aphid population development, it became clear that ants had a preference towards aphids on certain chemotypes. However, a higher number of predators led to a lower number of aphids. 5. The results confirm the importance of plant chemical variation, acting through multiple effects on many species in arthropod communities, and support results from field studies. In a natural population, with a diverse selection of host-plant variants, aphid populations and their interacting species can therefore be structured at the level of an individual plant. Specialist aphids on patchily-distributed host plants can exhibit metacommunity dynamics at very local scales. Plant within-species variation within a local population is often ignored in metacommunity ecology, yet our work shows that this can have strong effects on insect-ant-natural enemy dynamics and therefore future research should incorporate this into current theory and experimental studies.