Insect herbivory in an experimental agroecosystem: the relative importance of habitat area, fragmentation, and the matrix Kyle J. Haynes and Thomas O. Crist K. J. Haynes (kjh1087@louisiana.edu), Dept of Biology, Univ. of Louisiana, PO Box 42451, Lafayette, LA 70504, USA.  T. O. Crist, Dept of Zoology, Miami Univ., Oxford, OH 45056, USA. Habitat area, fragmentation, and the surrounding matrix influence levels of herbivory in various ecosystems, but the relative importance of these effects has rarely been assessed. We compared levels of herbivory and densities of dominant arthropod herbivores (the hemipteran insects Agallia constricta, Empoasca fabae, Therioaphis trifolii, Lygus lineolaris and Halticus bractatus) among experimental plots that varied in the area and fragmentation of clover habitat and the composition of the matrix (bare ground or grass) surrounding clover habitat. To assess levels of herbivory, we compared clover biomass within herbivore exclosures to the biomass accessible to herbivores. The area and fragmentation of clover habitat, as well as matrix composition, significantly influenced the collective densities of herbivores, although each species exhibited unique responses to habitat structure. Herbivory was strongest in plots with large (64 m 2 ) as compared to small (16 m 2 ) amounts of clover habitat. The difference in clover biomass between the inside and outside of exclosures increased significantly with increasing density of Empoasca fabae but was unrelated to the densities of the other herbivores, suggesting that Empoasca fabae was an exceptionally important herbivore in this system. This study supports the view that herbivore densities and herbivory generally increase with increasing area of plant monocultures, but emphasizes that levels of herbivory may be driven primarily by one or a few key herbivore species. There is a large body of literature on the effects of the area and fragmentation of a focal habitat type and composition of the surrounding matrix on the distributions and abundances of species (Moilanen and Hanski 1998, Debinski and Holt 2000, Tscharntke et al. 2002, Cronin 2003). The effects of these factors on levels of herbivory have received a smaller, but still considerable, amount of attention (Thies et al. 2003, O ¨ sterga ˚rd and Ehrle ´n 2005, Valladares et al. 2006). However, few studies have evaluated the relative importance of habitat area, fragmentation, and matrix composition to levels of herbivory (but see Dieko ¨tter et al. 2007). If one or a few herbivore species are responsible for most of the herbivory in a given community, the feeding specialization of these key species may determine the manner and strength with which habitat area, fragmenta- tion and matrix composition affect levels of herbivory. Positive, negative, and neutral density responses to habitat area are frequently observed (Connor et al. 2000, Matter 2000, Zaviezo et al. 2006), but species with strong preferences for a focal habitat type are more likely to display positive densityarea relationships in the preferred habitat type than are habitat-generalist species (Hamba ¨ck et al. 2007). The fragmentation of a focal habitat type is generally thought to adversely affect populations but, in fact, positive effects of habitat fragmentation may be more likely for species that use multiple habitat types (Law and Dickman 1998). Fragmentation of a focal habitat type may have positive effects on a habitat generalist by increasing the extent to which complementary resources in the matrix occur within the organism’s dispersal range (Law and Dickman 1998, Fahrig 2003). Finally, whereas habitat specialists may respond more strongly to the area and fragmentation of a focal habitat type, habitat generalists may be more sensitive to the composition of the matrix (Brotons et al. 2003, Steffan-Dewenter 2003). For example, Haynes et al. (2007) found that (adult) densities of the polyphagous grasshopper Melanoplus femurrubrum in a focal habitat type (clover) were more strongly influenced by the availability of complementary resources in the matrix than by the area and fragmentation of the focal habitat. Assessing the relative importance of different attributes of habitat structure and composition to herbivory presents particular challenges. The habitat attributes of interest often covary in natural landscapes, hindering a researcher’s ability to isolate their independent effects. For example, Fahrig’s (2003) review revealed that most studies examining the effects of habitat area and fragmentation on population or community dynamics confounded the effects of these two factors. Furthermore, plant size or quality may covary with landscape features (e.g. matrix composition; Haynes and Cronin 2004), making it difficult to isolate the effects of herbivory from the effects of other factors on measures of plant growth (e.g. biomass, leaf area). One approach to Oikos 118: 14771486, 2009 doi: 10.1111/j.1600-0706.2009.17720.x, # 2009 The Authors. Journal compilation # 2009 Oikos Subject Editor: Daniel Gruner. Accepted 6 April 2009 1477