Ongoing anthropogenic changes are affecting the abundance, richness, and spatial distribution of consumers, often leading to regime shifts and the degradation of ecosystem processes (Silliman et al. 2013). This is especially true for insect herbivores, which influence numerous ecological and evolutionary processes, including functional attributes relating to plant production and pollination (Crawley 1989; Branson et al. 2006; Schmitz 2010; Agrawal et al. 2012; Silliman et al. 2013; Borer et al. 2014). There is widespread evidence for anthropogenic-based changes to insect herbivores, with signals including reduced species richness, invasion, outbreaks, and selection towards generalist feeding strategies (Tscharntke & Brandl 2004; Branson et al. 2006; Clavel et al. 2010; Martinson & Fagan 2014). What remains unclear are the main drivers of these changes, given that human disturbances may directly affect herbivore diversity and composition (e.g., habitat loss, stand perturbation) but also do so indirectly via trophically mediated effects on plants and predators (Halpern et al. 2005; Shurin et al. 2012; Rzanny et al. 2013). As with all consumer groups, theoretical models for the regulation of insect herbivores generally emphasize interactions between resource-based bottom-up and predatory top-down processes (Strong et al. 1984; Crawley 1989; Schmitz 2008; Price et al. 2011). Both processes can be affected by human activity, but the consequences for herbivore diversity and composition can be unclear (Silliman et al. 2013). Eutrophication, for example, can elevate production of insect herbivores by increasing plant productivity (Polis et al. 1997), but simultaneously select against herbivore richness, especially specialist feeders, because nutrient-rich plant communities tend to be species-poor (Stevens et al. 2004) and support fewer feeding guilds (Haddad et al. 2000, 2009; Borer et al. 2012). Loss of apex arthropod predators (Hendrickx et al. 2007; Shochat et al. 2008), however, can have similar effects, with reduced predation increasing herbivore abundance but also lowering richness if one or a few herbivores dominate (Holt 1977; Oliver et al. 2009). There may also be strong interactions between human-induced changes to resources and predators: nutrient enrichment may drive consumer birth rates that far exceed predator-driven herbivore mortality, or cause system instability by creating top-heavy feeding webs (Polis et al. 1997; McCann 2011; Shurin et al. 2012; Tunney et al. 2012). Food web dynamics can also be influenced by anthropogenic landscape transformations through factors such as habitat loss and patch isolation (Kruess & Tscharntke 1994; Polis et al. 1997; Valladares et al. 2006). This can affect herbivores spatial turnover directly by dispersal constraints, but also indirectly by constraining their resources and predators. Habitat loss, for example, can maximize either stochastic influences on plant community assembly (i.e., different plant communities on different islands due to random dispersal), or deterministic ones by favoring the same subsets of species with traits for long-distance colonization (Arroyo-Rodríguez et al. 2013; Harvey & MacDougall 2014). Habitat loss can also modify predation pressure by concentrating predators in smaller remnant areas, creating predation-free patches via dispersal limitation (Kruess & Tscharntke 1994; Hein & Gillooly 2011), or reducing predator diversity on smaller patches because of food limitation (Holt 1997; Gravel et al. 2011). Here, I test how multiple stressors on consumers and resources affect insect herbivore diversity, composition and spatial turnover in grasslands. I use a large-scale factorial mainland-islands food web assembly experiment, simulating a range of co-occurring changes associated with anthropogenic impacts (eutrophication, stand perturbation, and habitat fragmentation) and examining their impacts on the regulation of insect herbivore communities. My main objective is to test whether these typical anthropogenic impacts associated with global change affect insect herbivores mainly via direct effects or indirectly by their effects on plant or predator composition and diversity. I meet this objective in three complementary steps: (i) I test for the main effects of habitat size, isolation, eutrophication and defoliation on herbivore diversity, composition and among-islands spatial turnover, (ii) I use a path model comparison approach to test whether these effects on herbivores occur through direct or food web mediated effects, and (iii) I contrast how these changes compare to herbivore dynamics in unperturbed and continuous mainland habitat. This approach allows me to separate the direct effects of perturbation on herbivores, versus those mediated indirectly by how perturbations affect plants, predators, and spatial constraints on herbivores.