Abstract 1. Spatial habitat structure can influence the likelihood of patch colonisation by dispersing individuals, and this likelihood may differ according to trophic position potentially leading to a refuge from parasitism for hosts. 2. Using a tri-trophic thistle-herbivore-parasitoid system we tested whether habitat patch size, isolation and host-plant heterogeneity differentially affected host and parasitoid abundance, and parasitism rates. 3. Cirsium palustre thistles (n = 240) were transplanted in 24 blocks replicated in two sites creating a range of habitat patch sizes at increasing distance from a pre-existing source population. Plant architecture and phenological stage were measured for each plant and the numbers of the herbivore Tephritis conura and parasitoid Pteromalus elevatus recorded. 4. Herbivore numbers increased with host-plant density per patch, but parasitoid numbers and parasitism rates were unaffected. Patch distance from the source population did not influence insect abundance or parasitism rates. Parasitoid abundance was positively correlated with host insect number, and parasitism rates were inversely density-dependent. Host-plant phenological stage was positively correlated with herbivore and parasitoid abundance, and parasitism rates at both patch and host-plant scales. 5. The differential response between herbivore and parasitoid to host-plant density did not lead to a spatial refuge but may contribute to the observed inverse density-dependent parasitism. Heterogeneity in patch quality, mediated by variation in host-plant phenology, was more important than spatial habitat structure for both the herbivore and parasitoid populations, and for parasitism rates.