Experimental test of predator and herbivore food preference (Tables S2–S5) Feeding Assays Methods: To estimate the feeding preferences of consumers, we conducted a series of feeding assays at the University of North Carolina at Chapel Hill’s Institute of Marine Science (IMS) in Morehead City, NC, October through December of 2004. Each of the five predator species and several herbivore species used in the diversity experiment were provided with potential food sources in controlled, independent trials. Predator food sources were four macroinvertebrate herbivore species and seven species of macroalgae. Herbivores were assayed with all seven species of macroalgae (Table S2). We used two additional species of macroalgae (Ulva lactuca and Gracilaria verrucosa) in the feeding trials not included in the diversity experiment. All species used in these assays are common in local hard substrate benthic marine habitats. We included the herbivore Ampithoe longimana in these trials even though it was not abundant in the diversity experiment because the results of other studies in this system indicate that it is an important herbivore and a strong interactor (Duffy & Hay 2000). Several species that were abundant in the field at the time of the mesocosm experiment (e.g., Hyale spp. and Corophium spp.) were not abundant enough two weeks later for use in the feeding trials. All of these herbivores are macroscopic crustaceans known to consume macroalgae, microalgae or epiphytic algae, although many of the details of their feeding ecology are not well understood (Duffy & Hay 2000). We manipulated consumer and prey identity in a crossed factorial design (n = 5 for each consumer/prey combination). Containers were 2 L in volume and were maintained in indoor water tables with aeration and fresh, cycling seawater supplied from the adjacent Bogue Sound. All consumers were acclimated to the container environment for 24 hours without food prior to the experiment. We maintained the predator treatments for 24 hours and the macroinvertebrate consumer treatments for 72 hours. Time, water temperature, and predator weights and lengths were recorded for all 315 trials. At the end of each trial, we counted remaining animal prey and measured remaining macroalgae biomass. Algal wet mass in the predator assays was measured after removing excess water using a salad spinner (60 revolutions). To measure consumption by macroinvertebrate herbivores, we estimated the change in algal surface area by calculating initial and final area using digital image analysis software (ImageJ version 1.28v by NIH Image). We analyzed data with a two-way ANOVA (both factors fixed: consumer and prey identity) and LSM planned contrasts using SAS JMP 5.0.1a. The response variable was the amount of prey consumed and results are reported as the mean percent change in alga surface area over the course of the experiment. Feeding Assays Results: There is evidence for selective predation of macroinvertebrate prey species by the five predator species (Table S3). Overall, predators consumed more amphipods than isopods (LSM planned contrast: P < 0.001). The number of macroinvertebrate prey consumed varied among predator species, and L. rhomboides consumed more prey than other predator species (P < 0.001). There were no significant interaction effects between the predator and prey treatments (Table S3). Most predators did not consume algae (Table S4). There was a significant interaction between predator species and algae species. L. rhomboides preferentially consumed the green alga Enteromorpha linza and was the only predator species that consumed more than 30% of any algae (Table S3). E. linza was the most consumed algae across all predator treatments (P < 0.0001). Feeding preferences differed significantly among herbivore species (Table S3 and S5). Paracerceis spp. and A. longimana consumed brown and green algae and less red algae, while E. levi consumed red and green algae (Table S5). A. marcuzii was the most selective, only consuming the brown alga S. filipendula. Codium fragile was the least likely to be consumed by any macroinvertebrate herbivore (P < 0.0001). Across all species tested, macroinvertebrate herbivores as a group do not appear to prefer a single species or functional group of macroalgae (Table S5). Table S2. Species used in feeding experiment. Trophic level Species Common name Predator Portunus spinimanus Swimming crab Penaeus aztecus, Predator P. setiferus Shrimp Predator Lagodon rhomboides Pinfish Predator Fundulus heteroclitus Killifish Hypleurochilus geminatus, Hypsoblennius hentzi, Predator Starksia ocellata Blennies Herbivore Ampithoe longimana Amphipod Herbivore Elasmopus levi Amphipod Herbivore Ampithoe marcuzii Amphipod Herbivore Paracerceis spp. Isopod Primary producer Codium fragile Green algae Primary producer Dictyota menstrualis Brown algae Primary producer Hypnea musciformis Red algae Primary producer Sargassum filipendula Brown algae Primary producer Enteromorpha linza Green algae Primary producer Gracilaria verrucosa Red algae Primary producer Ulva lactuca Green algae Table S3. Two-way ANOVA results for single species feeding trials. Predator – macroinvertebrate prey SS df F P Consumer effect 186.34 4 6.74 0.0001 Prey effect 112.36 3 5.42 0.0019 Interaction effect 44.32 12 0.53 0.8860 Consumer effect 0.0131 4 4.58 0.0017 Algae effect 0.0654 6 15.25 < 0.0001 Interaction effect 0.0916 24 5.34 < 0.0001 Consumer effect 0.1052 3 19.17 < 0.0001 Algae effect 0.0803 6 7.31 < 0.0001 Interaction effect 0.2843 18 8.63 < 0.0001 Predator – algae Macroinvertebrates – algae Table S4. Percent of herbivores and macroalgae consumed after 24 hours. Shaded boxes indicate treatments in which at least 3 of the 5 predators consumed 30% or more of the prey. Herbivores Consumer: Shrimp Crabs Killifish Blennies Pinfish Paracerceis spp. 40 40 50 70 70 Elasmopus levi 50 50 80 100 100 Ampithoe longimana 80 60 80 90 100 Ampithoe marcuzii 70 70 80 70 100 Sargassum filipendula Dictyota menstrualis Macroalgae Ulva lactuca Enteromorpha linza Codium fragile Hypnea musciformis Gracilaria verrucosa 70 Table S5. Percent of algae consumed after 72 hours. Shaded boxes indicated treatments in which at least 3 of the 5 herbivores consumed 30% or more of the prey. Consumer: E. levi Paracerceis A. marcuzii A. longimana Sargassum filipendula 85 100 100 Dictyota menstrualis 81 76 Ulva lactuca 53 99 64 Enteromorpha linza 92 100 80 Codium fragile 73 Hypnea musciformis 96 Gracilaria verrucosa 100 41 76 Literature Cited Duffy J.E. & Hay M.E. (2000) Strong impacts of grazing amphipods on the organization of a benthic community. Ecological Monographs, 70, 237-263