Abstract.-Recent theoretical studies have argued that plant-herbivore coevolution proceeds in a diffuse rather than a pairwise manner in multispecies interactions when at least one of two conditions are met: (1) genetic correlations exist between plant resistances to different herbivore species; and (2) ecological interactions between herbivores sharing a host plant cause nonadditive impacts of herbivory on plant fitness. We present results from manipulative field experiments investigating the single and interactive fitness effects of three types of herbivory on scarlet gilia (Ipomopsis aggregata) over two years of study. We utilize these data to test whether selection imposed by herbivore attack on date of first flowering is pairwise (independent) or diffuse (dependent) in nature. Our results reveal complex patterns of the fitness effect of herbivores. Simulated early season browsing had a strong negative fitness effect on plants and also reduced subsequent insect attack. Surprisingly, this ecological interaction did not translate into significant interactions between clipping and insect manipulations on plant fitness. However, we detected a significant interaction between seed fly and caterpillar herbivory on plant fitness, with the negative effect of either insect being greatest when occurring alone. These results suggest that herbivore-imposed selection may have pairwise and diffuse components. In our selection analysis of flowering phenology, we discovered significant pairwise linear selection imposed by clipping, diffuse linear selection imposed by insects, and diffuse nonlinear selection imposed by clipping and insect attack acting simultaneously. Our results reveal that the evolution of flowering phenology in scarlet gilia may be in response to diffuse and pairwise natural selection imposed by multiple herbivores. We discuss the evolution of resistance characters in light of diffuse versus pairwise forms of linear and nonlinear selection and stress the complexity of selection imposed by suites of interacting species.Key words.-Diffuse coevolution, Heliothis phloxiphagus, Hylemya, lpomopsis aggregata, multispecies coevolution, pairwise coevolution, plant-herbivore interaction, scarlet gilia.Received November 10, 1997. Accepted June 25, 1998.A central question in plant evolutionary ecology is how plants and animals coevolve. Despite a wide theoretical interest in coevolution (reviewed in Thompson 1994), we have little empirical information on how interacting species impose selection or the dynamics of reciprocal evolutionary responses. This is particularly true of multispecies systems. Plants rarely interact with single herbivore species in isolated two-species interactions, but rather must cope with the collective damage from a suite of diverse herbivores. (Strong et al. 1984;Karban 1989;Maddox andRoot 1987, 1990;Marquis 1990 Marquis , 1992. These herbivores are often taxonomically disparate, attack plants at different life stages, damage different plant tissues, vary in their impacts on plant fitness, a...