1733 Ecology, 84(7), 2003, pp. 1733–1743  2003 by the Ecological Society of America THE ROLE OF HERBIVORES IN THE MAINTENANCE OF A FLOWER COLOR POLYMORPHISM IN WILD RADISH REBECCA E. IRWIN, 1 SHARON Y. STRAUSS,SHONNA STORZ,AIMEE EMERSON, AND GENEVIEVE GUIBERT Center for Population Biology, University of California, Davis, California 95616 USA Abstract. Plant species exhibiting polymorphisms with respect to flower color are widespread. Our understanding of the selection pressures that may maintain these color polymorphisms has primarily been confined to one set of organisms—pollinators. Yet, selection on flower color may also be driven by other agents, such as herbivores, especially in cases where pollinators and herbivores are using the same or correlated traits to select plants. A wealth of studies have documented pollinator preference for anthocyanin-recessive color morphs (A-; yellow and white flowers) of wild radish, Raphanus sativus, over anthocyanin-dominant morphs (A+; pink and bronze flowers); yet, differences in pollination alone do not explain the maintenance of the flower color polymorphism. Here, we ask whether variation in flower color in R. sativus influences the preference and performance of herbivores for A- flower color morphs vs. A+ flower color morphs in four types of herbivores (generalist and specialist Lepidoptera, slugs, aphids, and thrips). We found that all herbivores except for aphids and thrips preferred flowering A- color morphs compared to A+ morphs of R. sativus. Furthermore, all herbivores except larvae of specialist and generalist Lepidoptera performed better on A- color morphs. Differences in plant secondary chemistry may play a role in differential herbivore preference and performance on the anthocyanin flower color morphs. Chemical analyses of leaf secondary compounds (indole glucosinolates) revealed that A+ color morphs produced higher concentrations of indole glucosinolates than A- morphs in the presence of herbivore damage. Therefore, herbivores may exhibit lower preference for A+ color morphs, and these morphs may support lower herbivore performance because they are heavily defended once damaged. This is the first study, to our knowledge, to document differential preference and performance of herbivores for different flower color morphs. Previous studies have shown that increased herbivore damage can have profound negative direct and indirect effects on the reproduction of R. sativus. The data presented here suggest that differential preference and performance of herbivores for R. sativus color morphs may counter selection on flower color exerted by pollinators. Key words: anthocyanins; California, USA; flower color polymorphism; herbivory; plant–insect interactions; Raphanus; wild radish. INTRODUCTION Plants in natural environments experience myriad di- rect and indirect interactions with mutualistic as well as antagonistic species, and the traits expressed by plants are likely shaped by the simultaneous integration of all of these selection pressures (Strauss and Armbruster 1997). Despite the complexity of interactions occurring in natural systems, studies in plant–animal interactions have traditionally focused on interactions between a plant and one type of visitor, for example, interactions between plants and pollinators, herbivores, seed predators, or nec- tar robbers (e.g., Inouye 1983, Proctor et al. 1996, Karban and Baldwin 1997, Chittka and Thomson 2001). Traits that are involved in one set of interactions may, however, be co-opted into other, very different interactions. For Manuscript received 15 February 2002; revised 6 September 2002; accepted 10 September 2002; final version received 4 De- cember 2002. Corresponding Editor: L. F. Delph. 1 Present address: Institute of Ecology, Ecology Building, University of Georgia, Athens, Georgia 30602 USA. E-mail: rirwin@arches.uga.edu example, work by Armbruster et al. (1997) suggests that resins, now serving as rewards for highly specialized pol- linators, may have initially functioned as antiherbivore and antipathogen defenses. Similarly, showy bracts that attract pollinators also have antiherbivore properties when they close at night and protect stamens and pistils from florivores (Armbruster and Mziray 1987, Armbruster 1997). Thus, the identities of important selective agents on particular traits are not always obvious. In this paper, we examine whether variation in flower color, tradition- ally viewed as a pollinator-selected trait, might also in- fluence the preference and performance of herbivores. Plant species exhibiting polymorphisms with respect to flower color are well documented (Kay 1978). Flow- er color has traditionally been viewed as a trait that is essential in attracting certain suites of pollinators (Grant 1950, Faegri and van der Pijl 1979, Proctor et al. 1996). Although many studies have found selective foraging by pollinators for specific color morphs (e.g., Levin 1972, Mogford 1974a, Kay 1976, Hannan 1981, Waser and Price 1981, Brown and Clegg 1984, Jones