The effects of sika deer (Cervus nippon) browsing and understory productivity on web-building spiders' density were experimentally examined. We particularly focused on the nonlinear effects of these 2 factors and their underlying mechanisms. Our field experiments controlling deer density and understory productivity revealed that the deer affected the densities of 2 abundant spider species nonlinearly: the deer effect was unimodal or constant in lower productivity plots and negative in higher productivity plots. The nonlinear effects seemed to occur because the response trend of the spiders' limiting factors to deer impact differed between lower and higher productivity plots. In lower productivity plots, the density of the most abundant spider species, Prolinyphia longipedella, was limited by dicot cover, and both spider density and dicot cover had unimodal-shaped responses to deer density. This spider species was limited by the availability of twigs in higher productivity plots and responded negatively to deer density. Deer seemed to affect web-building spider density in understory vegetation by changing the number of available sites for webs through browsing. However, the precise limiting factors differed with understory productivity. Since forest understory productivity is expected to show large variation at various spatial scales due to differences in canopy tree density, the availability of nutrients, and other environmental factors, considering nonlinear effects is important when predicting the impact of large herbivores on invertebrates in forest understories.