Exploring the spatial distribution of variation in plant secondary metabolites is critical for understanding the evolutionary ecology of biochemical diversity in wild organisms. In the present study, concentrations of foliar sideroxylonal, an important and highly heritable defense chemical of Eucalyptus melliodora, displayed strong, fine-scale spatial autocorrelation. The spatial patterns observed could promote associational effects on herbivore foraging decisions, which may influence the selection pressures exerted on sideroxylonal content. Multiple chemical traits have roles in certain eucalypt-herbivore interactions, and the spatial characteristics of the herbivore foraging environment are therefore determined by these different factors. We used a model of E. melliodora intake by common brushtail possums (Trichosurus vulpecula), based on the combined effects of two chemical traits, to explore this idea and found that the spatial patterns were different to those of sideroxylonal alone. Spatial genetic autocorrelation, examined using microsatellites, was strong and occurred at a fine scale, implying that restricted gene flow might allow genetic patches to respond to selection relatively independently. Local two-dimensional genetic autocorrelation, explored using a new heuristic method, was highly congruent with the pattern of local phenotypic variation observed for sideroxylonal, suggesting that the genetic variance underlying the sideroxylonal variation is similarly structured. Our results suggest that the spatial distribution of genetic and phenotypic variation could influence both the selective pressure imposed by herbivores on eucalypt defenses and the potential of populations to respond to natural selection. Spatial context should be considered in future studies of plant-herbivore interactions.