vol. 163, no. 3 the american naturalist march 2004 Limits to Trophic Levels and Omnivory in Complex Food Webs: Theory and Data Richard J. Williams * and Neo D. Martinez † Rocky Mountain Biological Laboratory, P.O. Box 519, Crested Butte, Colorado 81224 Submitted December 9, 2002; Accepted July 1, 2003; Electronically published March 9, 2004 abstract: While trophic levels have found broad application throughout ecology, they are also in much contention on analytical and empirical grounds. Here, we use a new generation of data and theory to examine long-standing questions about trophic-level limits and degrees of omnivory. The data include food webs of the Ches- apeake Bay, U.S.A., the island of Saint Martin, a U.K. grassland, and a Florida seagrass community, which appear to be the most trophi- cally complete food webs available in the primary literature due to their inclusion of autotrophs and empirically derived estimates of the relative energetic contributions of each trophic link. We show that most (54%) of the 212 species in the four food webs can be unambiguously assigned to a discrete trophic level. Omnivory among the remaining species appears to be quite limited, as judged by the standard deviation of omnivores’ energy-weighted food-chain lengths. This allows simple algorithms based on binary food webs without energetic details to yield surprisingly accurate estimates of species’ trophic and omnivory levels. While maximum trophic levels may plausibly exceed historically asserted limits, our analyses con- tradict both recent empirical claims that these limits are exceeded and recent theoretical claims that rampant omnivory eliminates the scientific utility of the trophic-level concept. Keywords: trophic level, food chains, omnivory. The study of food chains and the trophic structure of ecosystems has long been central to ecology (Elton 1927; Lawton 1989, 1995; Wilbur 1997; Post 2002a). Food chains depict the paths through a food web that organic energy travels, beginning with basal species and ending with as- similation by a species of interest. A species’ trophic level (TL) indicates the number of times chemical energy is * E-mail: rich@sfsu.edu. † Corresponding author; e-mail: neo@sfsu.edu. Am. Nat. 2004. Vol. 163, pp. 458–468.  2004 by The University of Chicago. 0003-0147/2004/16303-20453$15.00. All rights reserved. transformed from a consumer’s diet into a consumer’s biomass along the food chains that lead to the species. Convention holds that species that eat no other organisms are basal species with , while their direct and in- TL p 1 direct consumers are at higher levels. Research on TL fo- cuses on patterns common to all ecological systems (Elton 1927; Lindeman 1942; Pimm and Lawton 1978; Pimm 1980; Cousins 1987; Lawton 1989, 1995; Yodzis 1989; Pimm et al. 1991; Martinez and Lawton 1995), patterns that distinguish types of systems (Hairston et al. 1960; Ehrlich and Birch 1967; Briand and Cohen 1987; Moore et al. 1989; Carpenter and Kitchell 1993; Hairston and Hairston 1993, 1997; Polis and Strong 1996; Post et al. 2000), and patterns that distinguish species’ roles within ecological systems (Carpenter et al. 1987; Power 1990; Ca- bana and Rasmussen 1994; Vander Zanden and Rasmussen 1996; Brett and Goldman 1997; Pace et al. 1999; Vander Zanden et al. 1999; Schmitz et al. 2000), including the role of human exploitation in marine ecosystems (Pauly et al. 1998b, 2002). Measuring TL is central to this wide range of trophic ecology (Post 2002a). Food-web research plays a promi- nent role in measuring TL on a species-by-species and whole-system basis (e.g., Pimm et al. 1991; Polis and Wine- miller 1996; Vander Zanden and Rasmussen 1996; Post 2002a). Food webs, or “who eats whom” within ecological systems, describe the food chains in these systems. When the food web includes empirical estimates of energy flows through trophic links, “flow-based TL” is measured by computing food-chain lengths and the relative energetic contributions through chains of different lengths (Levine 1980; Adams et al. 1983). Food webs usually lack such flow estimates and more simply characterize flows or “links” between species as present or absent. In this binary situation, various measures of consumers’ food-chain lengths have been interpreted as measures of consumers’ TL. Pimm (1980, 1982) preferred modal chain length but also identified the extreme measures, the longest and shortest chain to a basal species. Ecologists who argue that most energy flows through the shortest chain to a basal species (e.g., Yodzis 1984; Hairston and Hairston 1993)