Predator identity and trait-mediated indirect effects in a littoral food web

Oecologia (2001) 129:139 146 DOI 10.1007/s004420100705 Randall J. Bernot Andrew M. Turner Predator identity and trait-mediated indirect effects in a littoral food web Received: 23 October 2000 / Accepted: 27 February 2001 / Published online: 23 May 2001 Springer-Verlag 2001 Abstract Perturbations to the density of a species can be propagated to distant members of a food web via shifts in the density or the traits (i.e. behavior) of intermediary species. Predators with differing foraging modes may have different effects on prey behavior, and these effects may be transmitted differently through food webs. Here we test the hypothesis that shifts in the type of predator present in a food web indirectly affect the prey s resource independent of changes in the density of prey. We assessed the importance of predator identity in mediating the grazing effects of the freshwater snail Physa integra on its periphyton resources using field and mesocosm studies. Field observations showed that Physa used covered habitats more in ponds containing fish than in ponds containing crayfish or no predators at all. A field experiment confirmed that snail behavior depended on predator identity. Physa placed near caged pumpkinseed sunfish (Lepomis gibbosus) selected covered habitats, but Physa placed near caged crayfish (Orconectes rusticus) moved to the surface of the water. The effects of predator identity on periphtyon were then examined in a mesocosm experiment, using caged predators. Habitat use of Physa was similar to their habitat use in the field experiment. In the presence of caged sunfish, periphyton standing crop in covered habitats was reduced to 34% of the standing crop in the presence of crayfish. In contrast, periphyton in near-surface habitats was 110% higher in the presence of fish than in the presence of crayfish. Thus, the effects of predator identity on Physa behavior cascaded through the food web to affect the abundance and spatial distribution of periphyton. R.J. Bernot ( ) A.M. Turner Department of Biology, Clarion University of Pennsylvania, Clarion, PA 16412, USA e-mail: rjb5578@ksu.edu Tel.: +1-785-5326814, Fax: +1-785-5326653 Present address: R.J. Bernot, Division of Biology, Kansas State University, 232 Ackert Hall, Manhattan, KS 66506, USA Keywords Predator identity Trait-mediated indirect effects Physa Littoral food webs Predator avoidance Introduction Changes in the abundance of predators affect not only the population size of foragers, but also their behavioral, morphological, life-historical, and physiological traits (reviews in Kerfoot 1987; Lima 1998). Because the per-capita effect of foragers on other species may depend on their phenotypic traits, perturbations to the abundance of top predators can be transmitted through a food web via shifts in the traits of intermediary species, independent of any density effects (Turner and Mittelbach 1990). Interactions in which the per-capita effects of one species on another depend on the abundance of a third species have been called higher-order interactions (Abrams 1983; Werner 1992), interaction modifications (Wooton 1994), and traitmediated indirect interactions (Abrams 1995; Abrams et al. 1996). Theory suggests that trait-mediated indirect effects may be common and important in structuring food webs (Abrams 1992, 1995; Werner 1992), and a few empirical studies demonstrate that the strength of trait-mediated indirect effects can rival that of density-mediated indirect effects (Peacor and Werner 1997; Schmitz et al. 1997; Peacor and Werner 2000; Relyea 2000). To predict the occurrence of trait-mediated indirect effects, community ecologists must first consider the specific features of predators that induce trait shifts in foragers. Most studies on trait-mediated indirect interactions have focused on predator-induced shifts in habitat use and foraging activity (Turner and Mittelbach 1990; Peacor and Werner 1997; Turner 1997; Kuhara et al. 1999). Because predators vary in their hunting mode and distribution, predators with differing foraging modes have different effects on prey behavior (Sih et al. 1998; McIntosh and Peckarsky 1999; Turner et al. 1999). Thus, different predators will likely vary in the food web effects they create due to differing effects on prey traits (McIntosh and Townsend 1996).

140 Several lines of evidence are required to demonstrate that trait-mediated indirect effects are important in food webs, and that they vary among predators. First, an investigator must demonstrate that the trait is variable (i.e. plastic) under field conditions. Second, it is necessary to show that the plasticity is induced by the predator. Third, the response of the prey must lead to effects on other species in the food web. Finally, the prey response and the food web response must differ in response to different predators. Here we present an integrated set of observations and experiments designed to evaluate whether predator identity, by itself, alters the impact of grazing snails on their periphyton resources. First, we conducted a field observational study to examine the extent of variation in habitat use among populations of Physa under differing predator regimes. Then, we contrasted the non-lethal effects of fish and crayfish on Physa habitat use in a field experiment. Finally, we conducted a mesocosm experiment to test whether contrasting effects of fish and crayfish on Physa behavior have indirect effects on the abundance and spatial distribution of periphyton. Because we used caged predators fed equal number of snails to simulate predation risk, we were able to partition the effects of predator identity from the effects of variation in prey density or level of imposed risk. Materials and methods Study system The traits of freshwater snails, including the widespread genus Physa (=Physella), are quite plastic. Exposure to predators affects several aspects of Physa s phenotype including shell morphology (DeWitt 1998), growth rate (Turner 1997; DeWitt 1998), size and age at reproduction (Crowl and Covich 1990), and habitat use (Alexander and Covich 1991; Turner et al. 1999, 2000; McCarthy and Fisher 2000). These effects are induced by waterborne chemical cues associated with predation risk (Dodson et al. 1994; Turner 1996). Freshwater snails can affect the species composition, standing crop, and productivity of attached algae (Lowe and Hunter 1988; Hill et al. 1992). If the effect of snails on periphyton depends on their phenotype (i.e. habitat use), then the non-lethal response of snails to predators should also affect periphyton, independent of any density effects (a trait-mediated indirect effect). Recent mesocosm experiments with snails have shown that changes in the level of predation risk by a single predator species does influence periphyton standing crop, independent of any density effects (Turner 1997; McCollum et al. 1998). Two important predators of Physa and other freshwater snails in North America are fishes and several species of crayfish (Decapoda, Astacoidea) (Osenberg and Mittelbach 1989; Hanson et al. 1990). These two predators often have fundamentally different foraging modes. Fishes, such as the pumpkinseed sunfish (Lepomis gibbosus), are visually oriented predators, and are unable to feed on snails that move under cover (Turner 1996). Crayfish are primarily benthic, rely on tactile and olfactory cues to locate prey, and are often found under cobbles, woody debris, and other covered substrates (Hamrin 1987). Thus, the behavioral response of Physa to predation risk may depend on the type of predator present. Field habitat use observations We monitored the habitat use of Physa integra in nine small ponds located near the Pymatuning Laboratory of Ecology (Linesville, Penn., USA). Physa were abundant (>20 snails/m 2 ) in each pond, and were the only species of snail present in all nine ponds. Two ponds lacked any apparent snail predators, three ponds contained crayfish (Orconectes rusticus and Cambarus robustus) but no fish, and four ponds contained both crayfish and fish. We established whether crayfish were present in a pond through visual sightings and minnow trap catches. Fish presence was established by electroshocking each pond. Creek chub (Semotilus atromaculatus) were the most abundant fish species in each fish pond. Field studies have shown that creek chub include snails in their diet (Gilliam et al. 1989), and our own feeding trials in aquaria show that creek chub >100 mm SL prey on Physa (R.J. Bernot, unpublished data). Each of the ponds underwent seasonal fluctuations in water level and was situated near other water-bodies, so it is likely that predator species composition varied from season to season. We assessed snail habitat use using artificial substrates placed into the ponds. Substrates were constructed from two square ceramic tiles (20 20 cm) and separated from each other by four 2- cm-tall legs. Snails using the top surface were classified as using the open habitat, whereas snails occupying the bottom surface of the top tile or the bottom tile were classified as using the covered habitat. Four substrates were placed in each pond, and snail habitat use was then censused once a day for 10 days. Substrates were positioned near the edge of the ponds, in water 8 11 cm deep. The overall pattern of snail habitat use in a pond was calculated by pooling the data from each of the four substrates and averaging across the 10-day period. We analyzed the effect of pond type on Physa habitat use with one-way ANOVA (ponds served as the unit of observation), and we determined which pond types differed from each other with Tukey s multiple comparison. Field experiment We manipulated predator identity in a field experiment in July 1998 to test whether among pond variation in snail habitat use could be induced by the different types of predators in the ponds. We placed caged fish and crayfish into the field and monitored the habitat use of neighboring Physa. Cubical cages (40 cm each side) were constructed from 1-mm mesh netting stretched over the sides and bottom of a PVC frame. We then bundled pairs of cages together so that chemical cues could freely pass from one cage to another. One cage in each pair was denoted as a predator cage, and the other cage a snail cage. We placed a 30 30 cm ceramic tile on the PVC pipes framing the bottom of the snail cage to provide a standardized substrate with which to evaluate snail habitat use (the tile was about 2 cm above the cage bottom). Predator cages were fitted with a mesh lid to prevent animals from escaping. We placed cages along the shoreline of Pymatuning Reservoir (Crawford County, Penn.) in water 30 46 cm deep, 5 m apart. Ten Physa collected from Pymatuning Reservoir were placed in each snail cage. Predator cages were assigned to one of three treatments: (1) a sunfish treatment: one pumpkinseed sunfish was fed four Physa daily; (2) a crayfish treatment: two crayfish (O. rusticus) were fed four Physa daily; and (3) a no predator treatment. By feeding equal numbers of snails to pumpkinseed and crayfish, we attempted to examine the effects of predator identity on snail behavior without any confounding influence of different levels of risk posed by these predators. Both pumpkinseeds and crayfish ate all snails offered to them. We censused two aspects of snail habitat use to assess the behavioral responses of Physa to treatments: (1) the proportion of the snail population beneath the ceramic tile covering the bottom of the cage (covered habitat use); and (2) the proportion of the population within 2.5 cm of the surface of the water (near-surface habitat use). Predators were fed in the early morning (0700 hours). Censuses were taken in the late morning (1030 1200 hours) and in the afternoon (1400 1600 hours) over an 8-day period, resulting in 16 habitat use observations for each cage.

141 We analyzed predator identity effects, time-of-day effects, and day-of-observation effects on snail habitat use with repeated-measures ANOVA (Winer 1971). Time-of-day and day-of-observation were both repeated factors. We adjusted the P-values for the repeated factors to account for a heterogeneous correlation structure among dates (Geisser and Greenhouse 1958). When a significant interaction between day-of-observation and the predator treatment was identified, we used polynomial contrasts to identify whether the predator treatment effect changed in a linear fashion over days. Differences between pairs of treatments were identified with Tukey s multiple comparison. Mesocosm experiment We contrasted the non-lethal effects of pumpkinseed sunfish and crayfish on periphyton standing crops in a mesocosm experiment designed in a manner similar to our field experiment. We placed 30 polyethylene pools (1 m wide 2 m long, 525 l) outdoors at the Pymatuning Laboratory of Ecology and filled them with water from nearby Sanctuary Lake. Pools were arranged in 10 rows of three, and rows served as blocks in the analysis. Each pool received a covered habitat in the form of a corrugated vinyl sheet (60 60 cm) supported by six PVC legs (4 cm tall). Covered habitats were placed in one-half of the rectangular pools, and cylindrical predator cages (75 25 cm diameter, 1 mm mesh) were placed at the opposite end. We then stocked twenty Physa (mean shell length =9.9 mm ±1.12 SE), collected from a nearby pond containing both fish and crayfish, into each pool. We imposed one of three treatments on each pool: (1) a sunfish treatment in which one pumpkinseed sunfish (L. gibbosus) was housed in the predator cage and fed four Physa daily, (2) a crayfish treatment in which two crayfish (O. rusticus) were housed in the predator cage and fed four Physa daily, and (3) a no predator treatment in which the predator cage remained empty. The experiment began on 11 July 1998 and continued for 8 days. We fed predators every morning (0730 hours) and censused snail habitat use three times per day (at approximately 1000 hours, 1400 hours, and 2000 hours, minus one night census on the 6th day of the experiment) for a total of 23 observations per pool. We recorded the proportion of the snail population in three different microhabitats: (1) under cover, (2) within 2.5 cm of the water s surface, and (3) in the open. Each day, after censusing, any missing or dead snails were replaced to maintain a constant snail density. We also placed three pairs of ceramic tiles (15.2 5 cm) in each pool: one pair was placed on the bottom under the covered habitat, one pair was placed on the bottom but in the open, and one pair was suspended just under the water s surface on the pool walls. Tiles were first incubated in a recirculating artificial stream system for 14 days, where algae edible to snails accumulated on the tiles. Because algal growth was not completely uniform among tiles, they were grouped according to the amount of periphyton visually estimated to be present, then stratified into the three habitats. Tiles with the greatest amount of periphyton (mean of two random tiles =9.61 g/cm 2 ) were placed in covered habitats, tiles with the least amount of periphyton (mean =2.45 g/cm 2 ) were placed in open habitats, and tiles with an intermediate amount of periphyton (mean =6.80 g/cm 2 ) were placed in near-surface habitats. At the conclusion of the experiment, we scraped algae from a 0.45 cm wide strip running the length of the tile. The sample was vacuum filtered onto a glass fiber filter (Whatman 25 mm GC/F), dried at 55 C for 24 h, weighed, ashed at 550 C for 30 min, and weighed again to determine the ash-free dry mass (AFDM). Because algae accumulated on the sides of the pools over the course of the experiment, we also sampled periphyton from the sides of the pools at the conclusion of the experiment. We partially drained the pools, delineated four 4 4 cm quadrats randomly positioned around the sides of the pools (each located 10 cm below the water line), and scraped the periphyton from the quadrats. These samples were then pooled into one overall sample per pool. The ash-free dry mass of each sample was measured in the manner described above. Treatment effects, time of day effects, and day of census effects on snail habitat use were analyzed with repeated-measures ANOVA in the same manner as in the field experiment. Treatment effects on periphyton AFDM were analyzed with one-way ANOVA. In both cases, we used Tukey s multiple comparison to identify treatment differences. Results Field habitat use observations Physa habitat use differed among pond types (ANOVA: F 2,6 =28.10, P<0.01; Fig. 1). Physa occupying ponds with fish were generally found under cover, whereas Physa in ponds without fish were usually in the open, regardless of whether crayfish were present or not (Fig. 1). Thus, the presence of fish was associated with a 4-fold increase in the use of covered habitat. Next, we explored whether fish were the cause of the pattern. Field experiment The presence of caged predators had a strong effect on the covered habitat use of Physa (predator effect: 2,10 =197.16, P<0.001; Fig. 2a). Physa used cover more often in the presence of pumpkinseeds than in no-predator treatments (Tukey s: P<0.001), but they used cover less often in the presence of crayfish than in no-predator treatments (Tukey s: P=0.02). Treatment effects changed over time (day-of-experiment by treatment interaction: F 14,70 =3.97, P<0.01; interaction linear over time, P<0.01), but inspection of the time series shows that the difference between crayfish and sunfish treatments grew larger over the course of the experiment (Bernot 1999). Treatment effects on use of cover did not depend on the time of day (time-of-day by treatment interaction: F 2,10 =0.82, P=0.46). Fig. 1 Habitat use of Physa integra in ponds under three different predator regimes. Covered habitat use is the proportion of snails (mean ± SE) under two-level artificial substrates placed in ponds. Different letters indicate significant differences (P<0.05) between treatments (Tukey s multiple comparison). n=2 ponds without predators, 3 ponds with crayfish, and 4 ponds with fish and crayfish

142 Fig. 2 Covered (a) and near-surface (b) habitat use of Physa integra populations in a field experiment with three predator treatments (overall mean ± SE, n=6 enclosures per treatment). Different letters indicate significant differences (P<0.05) between treatments (Tukey s multiple comparison) Use of near-surface habitat by snails also depended on predator identity (predator effect: F 2,10 =39.26, P<0.001; Fig. 2b). Significantly more Physa occupied the near-surface habitat in crayfish treatments than in either no predator (Tukey s: P<0.001) or sunfish treatments (Tukey s: P<0.001). Predator effects on near-surface habitat use depended on the day of observation (day-of-experiment by treatment interaction: F 14,70 =3.46, P<0.001) but polynomial contrasts showed that the effect of time on the strength of the treatment effects was not linear (P>0.10). Predator effects on near-surface use by prey did not depend on the time of day (time-of-day by treatment interaction: F 2,10 =2.49, P=0.13). Mesocosm experiment Grazer habitat use Caged predators had a strong effect on Physa s use of covered habitats, but this effect depended on predator Fig. 3a,b Habitat use of Physa integra populations in a mesocosm experiment with three predator treatments (overall mean ± SE, n=10 pools per treatment). a Proportion of snails using covered habitats. b Proportion of snails using near-surface habitats. Different letters indicate significant differences (P<0.05) between treatments (Tukey s multiple comparison) identity (Fig. 3a). Covered habitat use was highest in the presence of sunfish and lowest in the presence of crayfish (predator effect: F 2,18 =361.58, P<0.001). Predator effects also depended on the day of observation (day-ofexperiment by treatment interaction: F 14,126 =3.21, P=0.001), but this relationship was non-linear (P>0.10). Caged fish and crayfish also had contrasting effects on the use of near-surface habitats by Physa (predator effect: F 2,18 =19.71, P<0.001; Fig. 3b). Relative to no predator pools, significantly more Physa were near the water s surface in crayfish pools (Tukey s: P<0.001), but fewer Physa were near the surface in sunfish pools (Tukey s: P<0.001). Predator effects depended on the day that observations were taken (day-of-experiment by treatment interaction: F 14,126 =25.33, P<0.001), and the interaction was linear with time (P<0.001), which probably resulted from weak treatment effects on the fifth through seventh days of the experiment (Bernot 1999). However, treatment effects on the eighth and final day of the experiment mirrored the overall time-averaged ef-

143 Fish and crayfish also had contrasting effects on periphyton near the surface of the water. Periphyton standing crop on tiles near the surface of the water differed significantly at the conclusion of the experiment (overall treatment effect: F 2,18 =4.46, P=0.027; Fig. 4b) with significantly more periphyton remaining on tiles in sunfish pools than in crayfish pools (Tukey s: sunfish vs crayfish: P=0.039). Periphyton on the sides of each pool also differed significantly among treatments (overall treatment effect: F 2,18 =10.5,P<0.001). Periphyton standing crops were more than twice as high in sunfish pools as in either crayfish pools or pools without a predator. Taken together, these results show that the presence of caged sunfish can enhance periphyton standing crops in nearsurface habitats, but the presence of crayfish can lead to lower periphyton standing crops in near-surface habitats. Discussion Fig. 4a c Periphyton standing crop (g/cm 2 ) at the conclusion of the mesocosm experiment (mean ± SE, n=10). a Periphyton standing crop on tiles placed under cover. b Periphyton standing crop on tiles placed near the surface of the water. c Periphyton standing crop on the sides of the pools. Different letters indicate significant differences (P<0.05) between treatments (Tukey s multiple comparison) fects (Bernot 1999). Predator effects also depended on time-of-day, with overall predator effects being strongest at night (time-of-day by treatment interaction: F 4,36 = 24.43, P<0.001). Periphyton standing crops Periphyton standing crop in covered habitats differed among treatments (treatment effect: F 2,18 =13.07, P<0.001) with the highest periphyton standing crops found in the presence of crayfish and the lowest periphyton standing crops found in the presence of fish (Fig. 4a). Predators can affect the abundance of species with which they have no direct trophic links (Kerfoot 1987; Wooton 1994; Pace et al. 1999), but few studies establish the mechanism of these indirect effects. In theory, the magnitude of indirect effects can depend on variation in prey density, variation in level of risk imposed on prey, and variation in the type of risk imposed on prey. Previous studies have explored the roles of prey density and risk. In contrast, our results show that indirect effects can depend on the type of risk predators impose on prey. The spatial patterns of the treatment effects on periphyton standing crops support the hypothesis that the periphyton effects were mediated by shifts in snail habitat use, and not some other factor. There was a negative correspondence between snail habitat use and periphyton standing crops in both covered and near-surface habitats. Any study of predator indirect effects on algae must consider the potential role of increased nutrient availability associated with prey mortality (e.g. Mulholland et al. 1991; Rosemond et al. 1993; McCollum et al. 1998). With respect to the fish versus crayfish contrasts we minimized any confounding effects of nutrient remineralization by feeding fish and crayfish equal numbers of snails. In addition, we observed substantial resource depletion in prey refuges (Fig. 4a, fish versus no-predator controls, Fig. 4b, crayfish vs no-predator controls), which is consistent with behaviorally-mediated mechanisms but not with nutrient-mediated mechanisms. Although these indirect effects of predator identity on periphyton were demonstrated in a mesocosm study, several lines of evidence support the hypothesis that predator identity plays an important role in the dynamics of littoral food webs. Our field observations show snail habitat use is highly variable among ponds, and is related to the type of predators present in a pond. We have also documented small-scale spatial variation in snail habitat use along a depth gradient within a pond (Turner et al. 2000), which is likely related to shifts in the level and type of predation risk along the depth gradient. The

144 field experiment in this study showed that changing predator identity induced habitat shifts quite similar in magnitude to the differences in habitat use between ponds with crayfish alone and ponds with fish (e.g., compare Fig. 1 and Fig. 2). Because the field experiment was performed with the natural backdrop of olfactory cues produced by background levels of predation, these results provide strong evidence that much of the spatial variability in snail habitat use in the field is driven by variation in the type of predators confronting the snails. The responses of Physa to predation risk in the mesocosm study do not appear to be an artifact of the experimental setup, as we observed similar magnitudes of snail habitat use shifts in both the field and mesocosm experiments. The mean use of covered habitats in the presence of sunfish was 73% and 65% for the field and mesocosm experiments respectively, and the mean use of near-surface habitats in the presence of crayfish was 40% and 44% for the field and mesocosm experiments. This congruence of behavior in field observations, field experiment, and mesocosm experiment suggests the cascading effects of predator identity on periphyton standing crop shown in the mesocosm experiment are likely similar in magnitude to those effects operating in natural food webs. Since a large number of studies performed in natural and semi-natural conditions show that snails have large effects on periphyton standing crops and species composition (Lowe and Hunter 1988; McCollum et al. 1998), the treatment effects on periphyton standing crops observed in the mesocosm experiment are likely representative of the interactions operating in littoral communities. Thus, any variation in snail grazing activity will likely result in significant effects on the periphyton resources of the snails, or on other consumers of those periphyton resources. Although we found significant treatment effects on periphyton species composition in our mesocosm study (Bernot 1999), these effects were small in magnitude. The extent to which predator identity has strong food web effects in other communities will depend on the specificity of the cues with which prey assess predation risk. Although few studies have contrasted the effects of two or more predator species on the behavior of a single prey species, these studies generally agree that prey are able to discriminate between predators with differing foraging modes (McIntosh and Townsend 1994; Watt and Young 1994; Turner et al. 1999). All of these studies involve systems in which prey assess predation risk via chemosensory cues (olfaction). In general, the ability of prey to employ highly specific responses to different types of predators will be limited by perceptual constraints, but chemical cues provide animals with a reliable means to accurately assess the level and type of risk posed by a predator (Kats and Dill 1998). Physa from different populations apparently respond to somewhat different cues (see also McCarthy and Fisher 2000). Physa from a Michigan lake crawled under cover in response to crushed conspecifics, and were oblivious to whether a fish predator was present or not (Turner 1996). Physa from a Pennsylvania marsh, however, did not react strongly to crushed conspecifics, but did show strong habitat shifts in the presence of actively feeding fish or crayfish (Turner et al. 1999). These differences may be related to the predator regime faced by each population, and are consistent with the hypothesis that prey should use more generalized (and perhaps sensitive) cues when confronted with just a single type of predator. The Michigan population faced predation primarily from pumpkinseed sunfish, whereas the Pennsylvania population faced predation from both pumpkinseeds and crayfish. We expect that factors such as heterogeneity in selective agents and gene flow will shape the extent to which animals use highly specific cues, which will in turn affect species interactions. Here, we have focused on contrasting the individual effects of fish and crayfish on snail behavior and periphyton standing crops, but we have also examined the combined effects of fish and crayfish on snail behavior and trait-mediated interactions (Turner et al. 2000). In our previous study, we found that the combined effects of fish and crayfish were generally intermediate to their individual effects. For example, Physa s use of covered habitats in the presence of both fish and crayfish was intermediate to their covered habitat use in the presence of each individual predator (a non-additive effect of multiple predators), because crayfish by themselves had little effect on covered habitat use, but fish had a weaker effect when crayfish were present (Turner et al. 2000). Those results show that predicting the non-lethal effects of multiple predators may not be straightforward and will require detailed knowledge of the costs and benefits of prey behavioral options (see also Sih et al. 1998). The results presented here have important implications for how we model species interactions in communities, and suggest that the effects induced by changes in the identity of predators are a potentially important class of trait-mediated indirect interactions. Most studies which show the importance of trait-mediated indirect interactions in food webs have induced shifts in prey phenotypes by manipulating the level of risk (e.g. Turner and Mittelbach 1990; Peacor and Werner 1997; Schmitz et al. 1997; Relyea 2000). Often, the primary effect of increasing risk is to reduce overall foraging activity, allowing the prey s resources to increase in abundance. This trait-mediated indirect effect mimics any numerically mediated effects, though it may occur over shorter time scales. Predator-mediated habitat shifts should lead to lower resource levels in prey refuges (e.g. Fig. 4a, contrast no-predator control to fish treatment), an effect that would run counter to and weaken the density-mediated indirect effect, though few studies have actually demonstrated resource depletion in prey refuges (but see Mittelbach 1981, 1988). Thus, trait-mediated indirect effects induced by variation in risk and density-mediated indirect effects are always correlated (assuming that foragers accurately perceive risk) making it possible to construct an accurate model of species interactions that did not explicitly distinguish between behaviorally transmit-

ted and numerically transmitted effects (though it would have to incorporate spatial structure). Unlike variation in risk, predator identity can vary independently of prey mortality rates, and is not necessarily linked with the numerical effects of predators on prey. Consider, for example, a food web in which the top predator is replaced with another predator, which imposes exactly the same death rate on exactly the same array of prey species, but employs a different foraging mode. The shift in predators can change prey foraging patterns, which in turn can affect the spatial distribution of the resources, independent of any variation in prey mortality rates. Such dynamics can only be captured by explicitly incorporating prey foraging behavior as a function of predator identity into models of species interactions. A consideration of these results also suggests that the criteria by which we assign species to functional groups deserve to be reconsidered. In order to describe community structure and predict community dynamics, ecologists often simplify complex food webs by aggregating species into trophic groups of functionally similar species (e.g. Pimm and Rice 1988; Paine 1992). The role of species in food webs has traditionally been assessed based on feeding habits: species that share similar predators and similar resources are assumed to be functionally similar (Pimm and Rice 1988; Yodzis and Winemiller 1999). More sophisticated approaches take into account variable interaction strengths (Osenberg and Mittelbach 1989; Paine 1992; McPeek and Peckarsky 1998) but implicitly assume that the function of a species in a community can be completely described by its per-capita effects on other species. Our study shows, however, that two predator species (pumpkinseed sunfish and crayfish) who share the same resource can have very different food web effects, even if they have identical per-capita effects on the survivorship of their prey. In short, we assert that species who share the same prey are not necessarily functional equivalents, because they may differ in the sorts of trait shifts they induce. Acknowledgements We thank Christina Boes for helping with all aspects of the project; Kathleen Gibson, Debra Miller, and Jon Russell for logistical support, Phil Brautigan, Brooke Fails, Tiffany Knight, and Emily Greiner for help with cage construction, and the Linesville Fish Hatchery, Pennsylvania Fish and Boat Commission, for use of land to set up pools. Craig Osenberg, Michelle Evans- White, Melody Kemp, Mike Quist, the Ecology Seminar Group at Clarion University, and two anonymous reviewers commented on previous drafts of this manuscript. This project was funded by Sigma Xi, the McKinley Research Fund of the University of Pittsburgh, and NSF grant IBN-9982196 to A.M. Turner. This is contribution number 123 of the Pymatuning Laboratory of Ecology. References Abrams PA (1983) Arguments in favor of higher order interactions. Am Nat 121:887 91 Abrams PA (1992) Predators that benefit prey and prey that harm predators: unusual effects of interacting foraging adaptations. Am Nat 140:573 600 Abrams PA (1995) Implications of dynamically variable traits for identifying, classifying, and measuring direct and indirect effects in ecological communities. Am Nat 146:112 134 145 Abrams PA, Menge BA, Mittelbach GG, Spiller D, Yodzis P (1996) The role of indirect effects in food webs. In: Polis GA, Winemiller KO (eds) Food webs: integration of patterns and dynamics. Chapman and Hall, New York, pp 371 395 Alexander JE, Covich AP (1991) Predation risk and avoidance behavior in two freshwater snails. Biol Bull 180:387 393 Bernot RJ (1999) The consequences of predator-specific defenses on trophic interactions in the littoral food web. M.S. Thesis, Clarion University, Clarion, Pa. Crowl TA, Covich AP (1990) Predator-induced life-history shifts in a freshwater snail: a chemically mediated and phenotypically plastic response. Science 247:949 951 DeWitt TJ (1998) Costs and limits of phenotypic plasticity: tests with predator-induced morphology and life history in a freshwater snail. J Evol Biol 11:465 480 Dodson SI, Crowl TA, Peckarsky BL, Kats LB, Covich AP, Culp JM (1994) Nonvisual communication in freshwater benthos: an overview. J North Am Benth Soc 13:266 267 Geisser S, Greenhouse SW (1958) An extension of Box s results on the use of the F distribution in multivariate analysis. A Math Stat 29:885 891 Gilliam JF, Fraser DF, Sabat AM (1989) Strong effects of foraging minnows on a stream benthic invertebrate community. Ecology 70:445 452 Hamrin SF (1987) Seasonal crayfish activity as influenced by fluctuating water levels and presence of a fish predator. Holarct Ecol 10:45 51 Hanson JM, Chambers PA, Prepas EE (1990) Selective foraging by the crayfish Orconectes virilis and its impact on macroinvertebrates. Freshw Biol 24:69 80 Hill WR, Boston HL, Steinman AD (1992) Grazers and nutrients simultaneously limit lotic primary productivity. Can J Fish Aquat Sci 49:1257 1267 Kats LB, Dill LM (1998) The scent of death: chemosensory assessment of predation risk by prey animals. Ecoscience 5: 361 394 Kerfoot WC (1987) Cascading effects and indirect pathways. In: Kerfoot WC, Sih A (eds) Predation: direct and indirect impacts on aquatic communities. University Press of New England, Hanover, pp 57 70 Kuhara N, Nakano S, Miyasaka H (1999) Interspecific competition between two stream insect grazers mediated by non-feeding predatory fish. Oikos 87:27 35 Lima SL (1998) Nonlethal effects in the ecology of predator-prey interactions. Bioscience 48:25 34 Lowe RL, Hunter RD (1988) Effect of grazing by Physa integra on periphyton community structure. J North Am Benth Soc 7:29 36 McCarthy TM, Fisher WA (2000) Multiple predator-avoidance behaviours of the freshwater snail Physa heterostropha pomila: responses vary with risk. Freshwater Biol 44:387 397 McCollum EW, Crowder LB, McCollum SA (1998) Complex interactions of fish, snails, and littoral zone periphyton. Ecology 79:1980 1994 McIntosh AR, Peckarsky BL (1999) Criteria determining behavioural responses to multiple predators by a stream mayfly. Oikos 85:554 564 McIntosh AR, Townsend CR (1994) Interpopulation variation in mayfly antipredator tactics: differential effects of contrasting predatory fish. Ecology 75:2078 2090 McIntosh AR, Townsend CR (1996) Interactions between fish, grazing invertebrates and algae in a New Zealand stream: a trophic cascade mediated by fish induced changes to grazer behavior? Oecologia 108:174 181 McPeek MA, Peckarsky BL (1998) Life histories and the strengths of species interactions: combining mortality, growth and fecundity effects. Ecology 79:867 879 Mittelbach GG (1981) Foraging efficiency and body size: a study of optimal diet and habitat use by bluegills. Ecology 62:1370 1386 Mittelbach GG (1988) Competition among refuging sunfishes and effects of fish density on littoral zone invertebrates. Ecology 69:614 623

146 Mulholland PJ, Steinman AD, Palumbo AV, Elwood JW, Kirschtel DB (1991) Role of nutrient cycling and herbivory in regulating periphyton communities in laboratory streams. Ecology 72:966 982 Osenberg CW, Mittelbach GG (1989) Effects of body size on the predator-prey interaction between pumpkinseed sunfish and gastropods. Ecol Monogr 59:405 432 Osenberg CW, Mittelbach GG, Wainwright PC (1992) Two-stage life histories in fish: the interaction between juvenile competition and adult performance. Ecology 73:255 267 Pace ML, Cole JJ, Carpenter SR, Kitchell JF (1999) Trophic cascades revealed in diverse ecosystems. Trends Ecol Evol 14: 483 488 Paine RT (1992) Food web analysis through field measurements of per capita interaction strength. Nature 355:73 75 Peacor SD, Werner EE (1997) Trait-mediated indirect interactions in a simple aquatic food web. Ecology 78:1146 1156 Peacor SD, Werner EE (2000) Predator effects on an assemblage of consumers through induced changes in consumer foraging behavior. Ecology 81:1998 2010 Pimm SL, Rice JC (1988) The dynamics of multispecies, multilife-stage models of aquatic food webs. Theor Popul Biol 32: 303 325 Relyea RA (2000) Trait-mediated indirect effects in larval anurans: reversing competition with the threat of predation. Ecology 81:2278 2289 Rosemond AD, Mulholland PJ, Elwood JW (1993) Top-down and bottom-up control of stream periphyton: effects of nutrients and herbivores. Ecology 74:1264 1280 Schmitz OJ, Beckerman AP, O Brien KM (1997) Behaviorally mediated trophic cascades: effects of predation risk on food web interactions. Ecology 78:1388 1399 Sih AP, Englund G, Wooster D (1998) Emergent impacts of multiple predators on prey. Trends Ecol Evol 13:350 355 Turner AM (1996) Freshwater snails alter habitat use in response to predators. Anim Behav 51:747 756 Turner AM (1997) Contrasting short-term and long-term effects of predation risk on consumer habitat use and resources. Behav Ecol 8:120 125 Turner AM, Mittelbach GG (1990) Predator avoidance and community structure: interactions among piscivores, planktivores, and plankton. Ecology:2241 2254 Turner AM, Fetterolf SA, Bernot RJ (1999) Predator identity and consumer behavior: differential effects of fish and crayfish on the habitat use of a freshwater snail. Oecologia 118:242 247 Turner AM, Bernot RJ, Boes CM (2000) Chemical cues modify species interactions: the ecological consequences of predator avoidance by freshwater snails. Oikos 88:148 158 Watt PJ, Young S (1994) Effect of predator chemical cues on Daphnia behavior in both horizontal and vertical planes. Anim Behav 48:861 869 Werner EE (1992) Individual behavior and higher order species interactions. Am Nat 140:S5-S32 Winer BJ (1971) Statistical principles in experimental design. Mc- Graw-Hill, San Francisco Wooton JT (1994) The nature and consequences of indirect effects in ecological communities. Annu Rev Ecol Syst 25:443 466 Yodzis P, Winemiller KO (1999) In search of operational trophospecies in a tropical aquatic food web. Oikos 87:327 340