1

The long-term impacts of predators on prey: inducible defenses, population dynamics, and indirect effects

Jason T. Hoverman and Rick A. Relyea

J. T. Hoverman (jason.hoverman@colorado.edu) and R. A. Relyea, Dept of Biological Sciences, Univ. of Pittsburgh, Pittsburgh, PA 15260, USA. Current address for JTH: Dept of Ecology and Evolutionary Biology, Univ. of Colorado, Boulder, CO 80309, USA.

Despite the amount of research on the inducible defenses of prey against predators, our understanding of the long-term significance of non-lethal predators on prey phenotypes, prey population dynamics, and community structure has rarely been explored. Our objectives were to assess the effects of predators on prey defenses, prey population dynamics, and the relative magnitude of density- versus trait-mediated indirect interactions (DMIIs and TMIIs) over multiple prey genera-tions. Using a freshwater snail and three common snail predators, we constructed a series of community treatments with pond mesocosms that manipulated trophic structure, the identity of the top predator, and whether predators were caged or uncaged. We quantified snail phenotypes, snail population size, and resource abundance over multiple snail generations. We found that snails were expressing inducible defenses in our system although the magnitude of the responses varied over time and across predator species. Despite the expression of inducible defenses, caged predators did not reduce snail population size. There also was no evidence of TMIIs throughout the experiment suggesting that TMIIs have a minimal role in the long-term structure of our communities. The absence of TMIIs was largely driven by the lack of predator-induced reductions in resource consumption and the lack of consistent reductions in population size with predator cues. In contrast, we detected strong DMIIs associated with lethal predators suggesting that DMIIs are the dominant long-term mechanism influencing community structure. Our results demonstrate that although predators can have significant effects on prey phenotypes and sometimes cause short-term TMIIs, there may be few long-term consequences of these responses on population dynamics and indirect interactions, at least within simple food webs. Research directed towards addressing the long-term consequences of predator–prey interactions within communities will help to reveal whether the conclusions and predictions generated from short-term experiments are applicable over ecological and evolutionary timescales.

Prey are not passive participants in predator–prey inter-actions but instead possess a variety of adaptive defenses. Across their geographic ranges, prey experience different predator species that vary in foraging or hunting mode (Polis and Strong 1996, Chalcraft and Resetarits 2003) and, con-sequently, many prey form predator-specific defenses that reduce predation risk with different predators (McIntosh and Peckarsky 1999, Schmitz and Sokol-Hessner 2002). In recent years, significant advances have been made in our under-standing of how prey respond to complex predation scenar-ios such as differences in predator density, temporal variation in predation risk, and combinations of different predator environments (Tollrian 1993, McIntosh and Peckarsky 1999, Relyea 2004). However, these advances have largely come from relatively short-term experiments (i.e. 1 gen-eration). While effective at documenting predator-induced plasticity, this approach has limited our understanding of how the dynamics of trait induction play out over longer time scales. Short-term experiments may not be repre-sentative of longer-term effects for several reasons: 1) prey may initially over-react to an environmental manipulation but exhibit weaker responses over time (e.g. habituation),

2) prey may change defensive strategies over time, and 3) prey plasticity decisions can be affected by fixed time horizons (e.g. for metamorphosis or reproduction; Abrams 1991, Lima and Bednekoff 1999, Luttbeg et al. 2003, Werner and Peacor 2003, Hammill et al. 2010). To understand the importance of predator-induced plasticity within ecological communities, there is an urgent need to examine the long-term effects of predators on prey responses.

Predator environments can induce prey to delay or accel-erate reproductive maturity and such changes could translate into changes in the rate of population increase, final popula-tion size, or population cycling (Boonstra et al. 1998, Mitchell and Carvalho 2002). For example, when predators induce prey to delay maturity in favor of growing into a size refuge, the prey population may grow more slowly and achieve a smaller final size (Mitchell and Carvalho 2002). In contrast, when predators induce prey to accelerate maturity because a size refuge cannot be achieved, the prey population may grow at a faster rate; although the population may be limited by the carrying capacity of the system. However, in most cases we have no idea if induced phenotypes actually affect populations (Black 1993, Kusch 1998, Riessen 1999, Hammers-Wirtz and

Oikos 000: 001–012, 2011 doi: 10.1111/j.1600-0706.2011.19792.x

© 2011 The Authors. Oikos © 2011 Nordic Society Oikos Subject Editor: Matthijs Vos. Accepted 13 September 2011

2

Ratte 2000, Nakaoka 2000). To determine the actual impact of inducible defenses at the population level, we need to examine predator-induced prey populations for multiple generations.

The interactions between predators and their prey have important implications for larger communities where spe-cies interact directly and indirectly with each other. The direct effect of predators on prey density can lead to density- mediated indirect interactions (DMIIs). While ecologists have traditionally focused on changes in species density, there is a growing appreciation that changes in the traits of spe-cies can indirectly affect communities (i.e. trait-mediated indirect interactions, TMIIs; Abrams 1995). The challenge is to determine the conditions under which TMIIs will play an important role in structuring communities. The few studies that have been explicitly designed to examine the relative contributions of DMIIs and TMIIs in structur-ing communities have shown that although TMIIs vary in magnitude they can be just as strong as DMIIs (Werner and Peacor 2003). However, these studies were conducted over relatively short time scales (i.e. 1 generation) preventing an assessment of the long-term importance of DMIIs versus TMIIs (Werner and Peacor 2003). The importance of TMIIs over the long term will largely depend on the types of traits prey change in response to predators, the influence of trait changes on resource consumption and population growth, and the structure of the food web. Empirical and theoreti-cal research on TMIIs has shown that behavioral responses are an important driver of TMIIs due to the impacts on prey resource acquisition or demand (Bolker et al. 2003, Werner and Peacor 2003, Schmitz et al. 2004, Belovsky et al. 2011). However, the initiation of TMIIs via morphological plastic-ity is expected to depend on the topology of the food web (Werner and Peacor 2003). In simple linear food chains, changes in morphology alone are not expected to transmit TMIIs unless resource consumption rates are simultaneously altered (Abrams 2008). However, predator-induced changes in morphology can lead to TMIIs within more complex food webs (Relyea 2000, Trussell et al. 2002). Ultimately, for TMIIs to be important over the long term, population growth rate must change, which may be mediated by changes in resource acquisition, resource demand, or a suite of physio-logical stress mechanisms (Abrams 2008, Sheriff et al. 2009, Hawlena and Schmitz 2010a, b). Unfortunately, short-term studies prevent an assessment of life history responses and their impacts on population dynamics. Thus, TMIIs could be quite different over the long term, even to the point that they become unimportant relative to DMIIs (Persson and De Roos 2003). The few long-term investigations that exist suggest that TMIIs may exist over longer time periods (Chase 1999, Raimondi et al. 2000, Schmitz 2000), but we have too few data to draw any generalities. The objectives of our study were to assess the effects of predators on prey defenses, prey population dynamics, and the relative magni-tude of DMIIs and TMIIs over multiple prey generations.

Study system

Our model system consisted of the freshwater snail Helisoma trivolvis and three common snail predators (water bugs Belostoma flumineum; crayfish Orconectes rusticus; sunfish Lepomis gibbosus). In response to the non-lethal presence of

predators (i.e. caged predators), H. trivolvis form invasion-resistant shells with water bugs, crush-resistant shells with crayfish, and increase their use of refuges to avoid detection by fish (Hoverman and Relyea 2007, 2008, 2009). While water bugs do not appear to affect the egg production of H. trivolvis, crayfish induce a 45–68% reduction in total egg production over four months compared to predator-free environments (Hoverman et al. 2005). Snails are typically 22–33% smaller when reared with caged fish compared to no-predator treat-ments due to strong behavioral responses that reduce resource consumption (Hoverman and Relyea 2007, 2008). Although we have not examined egg production of snails reared with caged fish, slower growth is expected to reduce investment in egg production, which should lead to slower population growth. To date, we have limited information on how induc-ible defenses in snails alter long-term population dynamics and interactions within the community.

Based on our research in this system and existing theory (Abrams 2008), we expected that each predator would have different effects on prey trait induction, population dynam-ics, and indirect effects within our communities (i.e. top predator, snails, and an algal resource base). With caged water bugs, we expected an induction of morphological defenses, no induction of behavioral defenses that would affect resource consumption, and no effect on total egg pro-duction; therefore, we predicted no effect of caged water bugs on snail population size (i.e. no TMII). Given the low predation rate of water bugs ( 5 snails d21; Hoverman and Relyea 2007), we expected lethal water bugs to cause a weak DMII. With caged crayfish, we expected an induction of morphological defenses, no induction of behavioral defenses that would affect resource consumption, but a delay in matu-rity and a substantial decline in total egg production; there-fore, we predicted caged crayfish to cause a reduction in snail population size leading to a long-term TMII. Given the high predation rate of crayfish ( 150 snails d21; Hoverman and Relyea 2007), we expected lethal crayfish to cause a strong DMII. With sunfish, we expected a minimal induction of morphological defenses, but a strong induction of behavioral defenses that reduces resource consumption, and a reduction in total egg production; therefore, we expected caged fish to cause a reduction in snail population size leading to a TMII over both the short- and long-term. The high predation rate of lethal sunfish on snails should cause a strong DMII.

Methods

We employed a modified version of the experimental design of Peacor and Werner (2001), which uses a combination of hand-thinning, caged-predator, and lethal-predator treat-ments to estimate the relative magnitude of density- and trait-mediated effects within pond communities. The experiment was a completely randomized design with 12 treatments replicated four times for 48 experimental units (Table 1). The first treatment represented the base community and con-sisted solely of periphyton, phytoplankton, and zooplankton (i.e. no snails). This treatment provided an estimate of the periphyton biomass (i.e. resource abundance) without her-bivores and predators. The next two treatments contained the herbivore trophic level (i.e. snails). We examined the role

3

of density-mediated effects without predators by randomly hand-thinning snails at two rates (no hand thinning and hand thinning). For the remaining treatments, we added the predator trophic level using three different predator species (fish, water bugs, and crayfish) crossed with three predator manipulations: 1) a lethal predator that can produce preda-tion cues and thin the snail population, 2) a caged predator that can produce predation cues but not thin the snail popu-lation, and 3) a caged predator that can produce predation cues combined with experimenter applied hand-thinning. Together, these treatments provided multiple ways to exam-ine the separate and combined effects of trait-mediated and density-mediated effects on snail traits, snail population size, and resource abundance in the system (Table 1).

The experiment was conducted at the Univ. of Pittsburgh’s Pymatuning Lab of Ecology in Linesville, PA. On 14 May 2005, we collected 400 adult Helisoma trivolvis from a population known to express inducible defenses against our three predators (Hoverman and Relyea 2007). We placed 40 adults into each of 10 pools filled with 200-l of well water to oviposit. Egg deposition began immediately and continued until the adults were removed from the pools on 22 May. The snails began to hatch on 7 June and were fed rabbit chow ad libitum until the start of the experiment.

On 7–10 July, we filled 48 cattle tanks with 700-l of well water to serve as our experimental units. To provide natural substrate, we added 45 kg of soil (29% sand, 21% clay, and 50% silt) to each tank. The tanks were left un-manipulated for two weeks to allow the soil particles to settle to the tank bottom. We then added a 300 ml aliquot of pond water containing phytoplankton, periphyton, and zooplankton obtained from surrounding ponds. Nutrients (e.g. nitrogen, phosphorus) were not added to the tanks and no measurements of nutrients were made during the experi-ment. We provided structure for the snails by placing a clay tile platform (32 32 cm tile supported by a 16 16 cm

tile base) in the center of each tank. The platform rested approximately 3 cm above the substrate to provide space for snails to move under the tile. A periphyton sampler was added to the south-facing corner of each tank. The periphyton sampler was a weighted structure composed of eight plastic specimen cup lids (7.6 cm diameter) pinned in a linear fashion to 80 cm of well pipe. The sampler rested 5 cm from the substrate bottom when submerged in the tank. While these mesocosms were relatively simple (i.e. lacking macrophytes or depth gradients), similar setups have frequently been used to examine snail–predator interactions in the context of inducible defenses (Turner et al. 1999, 2000, Turner 2004).

We added two predator cages to each tank. Caged preda-tors that are fed snails release chemical cues that diffuse throughout the water without allowing the predators to kill the focal snails (Hoverman and Relyea 2007). One cage, designed to house fish, was constructed from 30 30 cm corrugated pipe capped with fiberglass window screen on each end. The other cage, designed to house water bugs or crayfish, was made from 10 10 cm corrugated pipe and was capped with shade cloth. We placed a shade cloth lid over each tank to prevent colonization by insects and amphibians during the experiment.

On 3 August, 100 snails were added to each tank (mean mass 1 SE 90.5 7.4 mg). The snail density (55 snails m22) is near the upper limit of observed snail den-sities in natural ponds (Hoverman et al. 2011). However, this density was used to ensure sufficient snails for the lethal predators and to conduct our hand-thinning treatments during the first month of the experiment before egg produc-tion. On the following day, we added the caged and lethal predators. The lethal predators were caged for 1 d and fed 1 snail to provide the experimental snails the opportunity to acclimate to the predator’s presence. The caged predators were fed 1 g (∼ 6–12 snails) of snail biomass each week, which is sufficient to elicit phenotypic responses in snails (Hoverman

Table 1. Description of the 12 experimental treatments and the appropriate comparisons used to estimate the separate and combined effects of predator cues (i.e. trait-mediated effects) and thinning (i.e. density-mediated effects) on snail traits, snail population dynamics and resource levels.

Treatment Description

No snails Baseline resource dynamics in the absence of snails or predatorsNo predator, no thinning Baseline responses for snail traits, snail population dynamics, and resource levelsNo predator, hand thinning Effects of artificial density reduction on snail traits, snail population dynamics, and resource levelsCaged predator1, no thinning Effects of predator cues on snail traits, snail population dynamics, and resource levelsCaged predator1, hand thinning Combined effects of predator cues and artificial density reduction on snail traits, snail population

dynamics, and resource levelsLethal predator1 Combined effects on predator cues and natural thinning rates on snail traits, snail population dynamics,

and resource levels

1 The predator treatments consisted of a single water bug, crayfish or fish.

Effect Appropriate comparisons

Predator cues (i.e. trait-mediated) Caged predator, no thinning vs No predator, no thinningCaged predator, hand thinning vs No predator, hand thinningLethal predator vs No predator, hand thinning

Thinning (i.e. density-mediated) No predator, hand thinning vs No predator, no thinningCaged predator, hand thinning vs Caged predator, no thinningLethal predator vs Caged predator, no thinning

Predator cues and thinning Lethal predator vs No predator, no thinningCaged predator, hand thinning vs No predator, no thinning

4

end of each tank. All hatchling and adult snails were recorded in the survey but egg masses were excluded. Although egg masses were not counted, we observed three distinct repro-ductive bouts in the tanks based on the presence of egg masses. In August 2005, April 2006 and June 2006, egg masses were present in all the tanks except the lethal crayfish and fish treatments. Few egg masses were observed in the latter two treatments throughout the experiment. Given that H. trivolvis has a perennial life history with overlapping generations, the snail population within each tank likely represented three snail generations by the end of the experi-ment. In addition to recording the number of snails in the transect, we recorded the number that were within 2.5 cm of the water’s surface and the number under the tile platform. We calculated the proportion of snails using structure and the proportion using the surface by dividing our counts by the total number of snails observed in the transect for each tank. Given that no more than three snails were ever found at the water’s surface, we did not analyze this response variable.

On five of the eight months (September, May, June, July and August) that we sampled periphyton, snail population size, and snail behavior, we also removed a sample of snails from each tank for morphological analysis (in addition to snails removed from the hand-thinning treatments). For the lethal fish and lethal crayfish treatments, there were gener-ally fewer than 10 individuals available for sampling, which would lead to unreliable estimates of mass and morphology. Thus, these tanks were not sampled. On the dates that we did not sample morphology (November, October and April), there were predominately hatchling snails in the tanks that would not have had time to morphologically respond to pre-dators and are extremely difficult to measure due to their small size. However, behavioral responses were measured during these months. To avoid small snails, we selected 10 individuals from the entire tank that were 0.8 cm in shell width. We used digital calipers to ensure that the snails were within the size range and preserved them in 10% formalin. After the sampled snails were dried at 80°C for 24 h, they were weighed to the nearest milligram and measured for shell width (Fig. 1 in Hoverman et al. 2005) using digital imaging software. We also used digital calipers to measure shell thick-ness at the leading edge of the aperture.

Statistical analyses

Our response variables included snail traits (mass, behav-ior, and morphology), snail population size, and periphyton biomass. For snail traits, we excluded the lethal predator treatments from the analysis because there were too few snails to reliably estimate responses. To analyze snail behav-ior, we conducted a repeated-measures analysis of variance (rm-ANOVA) to test for the effects of time, caged predator treatment, hand-thinning treatment, and their interactions. The second analysis examined the effects of caged predators and hand-thinning on snail mass and relative morphology. When analyzing morphological plasticity, it is important to first address the allometric relationship between morpho -logical traits and size. Shell width and shell thickness showed positive relationships with mass; mass was log10-transformed to improve the linearity of the relationships. To account for

and Relyea 2007, 2009). All empty cages were lifted briefly out of the water to equalize disturbance. Predator survival was checked every three days during the experiment and any dead predators were immediately removed from the tanks and replaced. An important consideration in the experiment was the effects of winter conditions on our predators. During the winter months (December to March) in Pennsylvania, the upper portion of the water column in cattle tanks typically freezes, which would kill our caged predators. Moreover, we expected minimal snail feeding, growth, and reproduction due to the cold temperatures. Thus, we decided to remove all predators from the experiment from December to March.

Our goal with the hand-thinning treatment was to mimic the predation rates of our lethal predators. However, it was only feasible to implement a single hand-thinning treatment rather than three hand-thinning treatments to match each lethal predator. Given that fish and crayfish consume more snails than water bugs (Hoverman and Relyea 2007), we expected these two predators to dramatically reduce snail population size. Matching our hand-thinning rate with the rate of predation by crayfish and fish would bias the results towards detecting large DMIIs in the system rather than TMIIs. Thus, we decided to mimic the predation rate observed in the lethal water bug treatment, which was expected to be lower than the other predators. To accom-plish this goal, we visually monitored snail population size in the lethal water bug treatment over the first three weeks and observed a predation rate of 20 snails week21. Hand thin-ning was conducted until the predators were removed from the tanks in December. At the start of the spring 2006, we resumed hand thinning at 20 snails week21 until the end of the experiment. We randomly selected snails that were 1 cm in shell width during the thinning events. Snails 1 cm were assumed to be in a size refuge from predation and, consequently, not an appropriate target for the thin-ning treatment (Hoverman and Relyea 2009). After remov-ing the snails, we obtained the total wet biomass of snails removed from each tank. The total biomass removed from the hand-thinning treatment did not differ between caged predator treatments (F3,12 1.5, p 0.262).

On a monthly schedule from September to November 2005 and April to August 2006, we sampled periphyton, estimated snail population size, and quantified snail behav-ior. To sample periphyton biomass in each tank, we removed one lid from the periphyton sampler, and scraped it with a brush into a tub of water to remove all attached periphyton. We then filtered the water through a pre-weighed Whatman GF/C filter that had previously been dried at 80°C for 24 h. After filtration, the filters were dried for 24 h at 80°C and weighed to determine the biomass of attached periphyton on the lid.

Snail population size and snail behavior were estimated on the same dates that periphyton was sampled. We surveyed the number of snails that were within a 30-cm wide transect extending from the water’s surface, down the side of the tank, across the bottom of the tank, and to the tile platform in the center (i.e. 20% of the tank). In addition to visually survey-ing for snails within the transect, we used our hands to gently probe for snails buried in the substrate. Snails detected in the substrate were lifted briefly from the water to confirm survival and replaced. All observations were done on the west

5

then within the hand-thinning treatment. We tested for DMIIs in the system by comparing the no-hand-thinning and hand-thinning treatment for each predator and by comparing the lethal predator treatments to the caged-predator plus no-hand-thinning treatments. We compared across the predator species to deter mine how each preda-tor impacted DMIIs versus TMIIs. We conducted mean comparisons using Fisher’s LSD test. To further explore DMIIs, we conducted a Pearson’s correlation (r) between log10-transformed periphyton biomass and snail population size. The assumptions of linearity, normality, and homo-scedascity were satisfied. A negative correlation would sug-gest that larger snail populations were reducing periphyton biomass. For both variables, we used treatment means for each sample month (n 8 treatment21) for a total n 88 in the analysis. We excluded the no-snails treatment in the analysis because there were no snails present.

Results

Effects on snail behavior

For snail behavior, there were significant effects of time, caged predator, and the time caged predator interaction (Table 2, Fig. 1). The interaction was driven by caged preda-tor effects early in the experiment (September, October 2005) and again in June 2006, but not in the other months (p 0.17). In September, mean comparisons found no differences between the no-predator and each of the three caged-predator treatments (p 0.12). However, caged fish induced a greater use of structure than caged crayfish and caged water bugs (p 0.006). In October, caged fish induced greater use of structure compared to the other three treatments (p 0.001). There were no differences among the remaining treatments (p 0.21). In June, there were no differences between no-predator and the three caged-predator treatments (p 0.10). However, there was greater use of structure with caged fish and caged water bugs com-pared to caged crayfish (p 0.018). To assess whether snail population size could be influencing snail behavior in the caged-fish treatment, we a conducted a Pearson’s correlation (r) between log10-transformed population size and snail behavior size using the caged-fish treatments across all sam-ple dates. There was a negative correlation between refuge use and population size within the fish treatments over time (r 20.395, p 0.001, n 64), suggesting that when the snail population was large, a smaller the proportion of the snails were found using the refuge.

size variation, we used analysis of covariance (ANCOVA) with log10-transformed mass as the covariate (Hoverman et al. 2005). A critical assumption in the ANCOVA procedure is that the treatments share a common slope of their regres-sion lines (i.e. a similar allometric relationship; McCoy et al. 2006); this assumption was satisfied (F12,80 1.6, p 0.118). From the ANCOVA, we used the mass-adjusted treatment mean and residuals from the within-treatment regression to calculate each individual’s size-adjusted trait value. We then calculated the mean response for each experimental unit within each sample and used these as our morphologi-cal response variables. While there are several different size- correction methods, this is a powerful approach for obtain-ing size-adjusted morphology based on our previous work (Hoverman et al. 2005, Hoverman and Relyea 2007). We then conducted three rm-ANOVAs for mass, shell width, and shell thickness to test for the effects of time, caged- predator treatments, hand-thinning treatments, and their interactions. Given that three simultaneous tests were con-ducted, we employed a Bonferroni correction prior to inter-preting test significance. We conducted mean comparisons using Fisher’s LSD test if a response variable was significant.

We used rm-ANOVAs to assess treatment effects on snail population size over time. Due to the incomplete factorial design of the experiment, we conducted two analyses with the data to address our hypotheses. In the first analysis, we con-ducted a two-way rm-ANOVA with predator treatment (no predator and the three caged predator treatments) and thin-ning treatment (no thinning and hand thinning) as the main effects. Our main interests were whether the caged predators reduced snail population size, which would indicate a trait-mediated effect on population dynamics, and whether our thinning treatment was effective at reducing snail population size (i.e. density-mediated effect). In the second analysis, we used a one-way rm-ANOVA to assess the effects of the lethal predators on snail population size. We included the three lethal predator treatments and the no-predator, no thinning treatment in the analysis. Given that two tests were conducted, we employed a Bonferroni correction prior to interpreting test significance. Population size was log10-transformed to meet the assumption of homogeneity of variance.

We used a one-way rm-ANOVA to test for treatment effects on periphyton biomass (log10 transformed). Given that there was not a significant treatment time interac-tion, we averaged across time for our subsequent tests to determine if TMIIs and/or DMIIs occurred in the system (Table 1). We tested for TMIIs in the system by compar-ing periphyton biomass among the four caged-predator treatments within the no-hand-thinning treatment and

Table 2. Results of repeated-measures ANOVA (Wilks’ l) on the effects of caged predator and hand-thinning treatments on snail behavior (% using structure) over time.

DF F p

Within subjects Time 7,18 5.6 0.001Time Predator 21,52 2.0 0.025Time Hand-thinning 7,18 2.1 0.086Time Predator Hand-thinning 21,52 1.7 0.056

Between subjects Predator 3,24 4.9 0.008Hand-thinning 1,24 0.3 0.613Predator Hand-thinning 3,24 0.2 0.892

6

Effects on snail population size

In the first analysis of snail population size, we examined the effects of predator treatment (no predator and the three caged predator treatments) and thinning treatment (no thinning and hand thinning) over time. There were only effects of time (F7,18 10.2, p 0.001) and hand thinning (F1,24 5.8, p 0.024) on snail population size (Fig. 3a, b). Averaged across time and predator treatments, snail popula-tion size was 27% smaller in the hand-thinning treatment compared to the no-hand-thinning treatment. The lack of a predator effect suggests that the caged predators were not affecting snail population size despite the expression of inducible defenses by the snails.

In the second analysis, we compared the three lethal-predator treatments to the no-predator treatment without hand-thinning to determine the effects lethal predators on snail population size (Fig. 3c). There was an effect of treat-ment (F3,12 65.5, p 0.001) on snail population size, but not time or the time-by-treatment interaction. Snail popula-tion size was lowest in the lethal fish treatment, intermediate in the lethal crayfish treatment, and highest in the no-predator

Effects on snail mass and morphology

Mass was significantly affected by time, caged predator, and the time hand-thinning interaction (Table 3). Snails reared with caged crayfish and caged fish were 30% and 20% larger, respectively, compared to the no-predator treat-ment (Fig. 2a). In addition, snails reared with caged fish were 9% larger than snails reared with caged water bugs. The time hand-thinning interaction was driven by a shift in the direction of the response to hand thinning over time. Averaged across all caged-predator treatments, snails in the hand-thinning treatment were 24% larger in September and May (p 0.023), similar size in June (p 0.203), and 25% smaller in July and August (p 0.015) compared to the no-thinning treatment (data not shown).

Shell width and thickness were significantly affected by time and caged predators, but exhibited no time-by-treatment interactions (Table 3). Caged water bugs induced 7–9% wider shells compared to all other treatments (Fig. 2b), while caged crayfish induced 27–33% thicker shells compared to all other treatments (Fig. 2c).

Table 3. Results of repeated-measures ANOVAs (Wilks’ l) on the effects of caged predators and hand-thinning treatments on snail traits (mass, shell thickness, and shell width) over time. p-values 0.017 (indicated in bold) are significant following Bonferroni correction for conducting three simultaneous tests.

Mass Shell width Shell thickness

DF F p F p F p

Within subjects Time 4,21 43.2 < 0.001 4.9 0.002 4.5 0.009Time Predator 12,56 1.4 0.195 1.9 0.054 1.6 0.128Time Hand-thinning 4,21 12.3 < 0.001 2.4 0.088 0.4 0.831Time Predator Hand-thinning 12,56 0.7 0.790 1.0 0.446 0.8 0.630

Between subjects Predator 3,24 4.6 0.011 31.1 < 0.001 8.6 < 0.001Hand-thinning 1,24 0.7 0.425 6.2 0.020 0.5 0.473Predator Hand-thinning 3,24 0.7 0.565 0.5 0.674 0.3 0.812

% u

sing

stru

ctur

e

Figure 1. The effects of caged predators on snail behavior (% using structure) over time. Closed circles, open circles, open triangles, open diamonds represent the no-predator, caged water bug, caged crayfish and caged fish treatments, respectively. Data (least-squares means 1 SE) are averaged over hand-thinning treatments. Predators were removed from the tanks during the winter (December to March).

7

size resulting from hand thinning did not cause a DMII. Comparing the caged- and lethal-predator treatments, there was no evidence of a DMII with lethal water bugs, but there was evidence of a DMII with lethal crayfish and lethal fish (e.g. 80% more periphyton). Periphyton biomass in the latter two treatments did not differ from the no-snails treat-ment. Last, we found a strong negative correlation between periphyton biomass and snail population size (r 20.829, p 0.001, n 88) providing additional support for DMIIs in the system.

Discussion

We found clear evidence that snails were forming inducible defenses against caged predators in our study. The morpho-logical responses were maintained over multiple generations whereas the behavioral responses were strong early in the experiment but not evident later in the experiment. Despite the ubiquity of predator-induced defenses, caged predators did not reduce snail population size during the experiment. As a result, we found no evidence of TMIIs in our experi-ment, but did find evidence of DMIIs; lethal crayfish and lethal fish dramatically reduced snail population sizes and this lead to increased periphyton biomass. Below, we dis-cuss the effects of our community treatments on inducible defenses, population dynamics, and indirect effects over multiple snail generations.

The snails in the experiment expressed predator-induced defenses; water bugs induced wide snail shells, crayfish induced thick snail shells, and fish induced greater refuge use. These morphological and behavioral responses to pre-dators are largely consistent with previous work (Alexander and Covich 1991, McCarthy and Fisher 2000, Hoverman and Relyea 2007). Importantly, this is one of the few experi-ments to track predator-induced defenses of prey over multi-ple generations (Altwegg et al. 2004, Verschoor et al. 2004). While the morphological responses to water bugs and cray-fish were maintained over multiple generations, the behav-ioral response to fish was not maintained; snails did not use refuges after the winter months. However, snails were consistently larger in the fish treatment over time compared to the control. Size-mediated predator avoidance behavior does not appear to be a mechanism driving this pattern (Alexander and Covich 1991). According to this mecha-nism, large adult snails may increasingly forage outside of refuges whereas younger cohorts predominately use refuges (i.e. express behavioral plasticity). Once hatchlings started to appear in the population, they would compensate for the movement of adults away from the refuge with movement into the refuge. However, we observed hatchling snails over-lapping with the adults early in the experiment when behav-ioral responses to fish were still observed. Previous studies have observed the relaxation of behavioral defenses follow-ing the formation of morphological defenses (DeWitt et al. 1999, Relyea 2003, Hammill et al. 2010). However, snails were consistently larger in the caged fish treatment regard-less of the observed behavioral responses, suggesting that trait compensation was not occurring. Another possibility is that increases in population size may have caused greater competition forcing snails to alter their behavioral decisions.

and lethal water bug treatments. There was no difference between the no-predator and lethal water bug treatment (p 0.217).

Indirect effects on periphyton

The analysis of periphyton biomass found significant effects of time (F7,30 9.4, p 0.001) and treatment (F11,36 17.0, p 0.001) but not their interaction (F77,187 0.8, p 0.897). Given that the lack of an interaction confirms that responses among treatments were similar over time, we focus on the treatment effect and refer readers to Appendix 1 for detailed responses over time.

Mean comparisons found differences among the 12 treat-ments. For example, periphyton biomass was more than six times more abundant in the treatment lacking snails com-pared to the caged-predator treatments (averaged across thinning treatments), confirming that the snails were having an impact on the periphyton (Fig. 4). Comparing the no-predator treatment to the caged-predators treatments (with and without hand thinning), there was no evidence of TMIIs on the peri phyton. Comparing the two thinning treatments, periphyton biomass did not differ for each predator treatment. This suggests that the 27% reduction in snail population

Figure 2. The effects of caged predators on (A) mass, (B) size- adjusted shell width, and (C) size-adjusted shell thickness. Data (least-squares means 1 SE) are averaged over hand-thinning treat-ments and time. Treatments sharing lower case letters are not significantly different from each other based on pairwise compari-sons using Fisher’s LSD test (p 0.05).

8

were observed ( 1 month). Snails may be allocating resources to reach a size refuge from fish predation but due to behavioral responses that reduce foraging rates, attaining the size refuge requires at least a month to reach. Together, these results demonstrate that short-term observations of predator-induced plasticity may not reflect long-term dynamics for some predator–prey interactions.

Although caged predators altered the traits of the snails, there was no evidence that caged predators reduced snail pop-ulation size. The lack of an impact on population size may be the consequence of the relatively high reproductive output of snails coupled with density-dependent population growth (Dillon 2000). Our laboratory experiments have shown that snails exposed to caged water bugs and crayfish are capable of producing 50–200 eggs per snail in just four months (Hoverman et al. 2005). Despite the high reproductive output of the snails, snail population size rarely exceeded 200 snails (i.e. log10-population size of 2.3) and peaks in population size were not sustained suggesting that the snail populations

In support of this mechanism, there was a negative correlation between refuge use and population size within the fish treat-ments over time. However, alternative mechanisms include that: 1) snails, regardless of size, may have habituated to the presence of fish over time and increasingly foraged outside of the refuge, 2) later snail generations may have lost sensitivity to fish cues due to constant exposure starting in the egg, and 3) the removal of fish during the winter altered behavioral responses the following spring. Although our study cannot assess the importance of these mechanisms, it is clear that this shift in behavior has important implications for address-ing the potential for indirect effects. We also found that caged fish and crayfish consistently induced larger snail size compared to the no-predator treatment. While we have pre-viously observed this response to caged crayfish, the response to caged fish is in the opposite direction compared to our previous experiments (Hoverman and Relyea 2007, 2008). The main difference between our past research and the present study is the short time scale over which responses

Figure 3. The effects of (A) no-hand-thinning, (B) hand-thinning, and (C) lethal predator treatments on log10-transformed snail abun-dance (number of individuals per sample) over time. The no-predator treatment without hand thinning is shown in panel (C) for com-parison. Data are least-squares means 1 SE. Symbols are as described in Fig. 1. Predators were removed from the tanks during the winter (December to March).

9

threat of predation to initiate a TMII over the long term. Ultimately, the magnitude of TMIIs may jointly depend on the costs associated with altering foraging rates and resource availability (Van Buskirk 2000, Luttbeg et al. 2003, Wojdak and Luttbeg 2005). At low resource levels, the magnitude of TMIIs may be diminished due to increased risk taking behavior by prey faced with starvation (Luttbeg et al. 2003). Alternatively, prey may respond more to predation risk at low resource levels if opportunity costs associated with risk taking behavior are lower leading to strong TMIIs (Van Buskirk 2000). In support of this latter theoretical pre-diction, Wojdak and Luttbeg (2005) documented a negative relationship between the strength of a behaviorally-mediated TMII and initial resource levels in a freshwater food chain. While our experiment did not manipulate initial resource levels to assess the above theoretical predictions, resource levels may have been reduced to levels that minimized the likelihood of observing a TMII. Despite observations from short-term experiments that our predators impact the behav-ior, time to reproduction, or egg production of snails, it is clear that these responses have minimal effects on popula-tion dynamics and trophic cascades within our simple linear food chains.

We observed strong DMIIs associated with lethal cray-fish and fish but not water bugs, which supports our predic-tions based on the differences in predation rate between the predators (Hoverman and Relyea 2007). Lethal crayfish and lethal fish reduced snail population size by 95% relative to the comparable caged-predator treatments that lacked hand thinning. These reductions in snail population size caused a 2- to 15-fold increase in periphyton biomass. Both preda-tors may also have contributed to periphyton growth due to consumer-mediated nutrient recycling (Vanni and Layne 1997). By increasing nutrient inputs or altering nutrient stoi chiometry, predators can alter the abundance and struc-ture of producer communities (McCollum et al. 1998). The difference in the magnitude of the DMII between the lethal fish and lethal crayfish treatments is most likely a consequence of crayfish, which are omnivorous, consum-ing greater amounts of periphyton as snail population size decreased.

were displaying density-dependence. The negative correla-tion between periphyton biomass and snail population size further supports our assertion that high densities of snails were competing for limited resources. Also, while our previ-ous experiments have observed delayed time to reproduction with caged predators, this delay was generally less than 13 days (Hoverman et al. 2005). Considering our experiment was conducted for 1 year, a relatively brief delay in time to repro-duction appears to be insufficient to reduce long-term snail population size under mesocosm conditions.

Although many environmentally induced traits can affect the growth and reproduction of organisms, it has never been clear whether these changes affect population dynamics. The magnitude of life history costs associated with induc-ible defenses, the reproductive output of individual prey, and the carrying capacity of the environment will largely deter-mine whether long-term population effects are observed. Investigations of the long-term effects of predator induction on population dynamics provide equivocal support for the induction of smaller population sizes (Gilbert 1980, Black 1993, Kusch and Kuhlmann 1994, Riessen 1999, Nelson et al. 2004). While short-term plasticity experiments often conclude that induced traits should affect future survival and reproduction, long-term studies are needed to determine whether such extrapolation is reasonable.

We found no evidence of TMIIs in our study system, which is counter to the growing body of literature demon-strating predator-induced TMIIs in a variety of different systems including streams (Huang and Sih 1991, Peckarsky and McIntosh 1998), ponds (Turner 1997, Relyea 2000), old fields (Beckerman et al. 1997, Gelwick 2000, Belovsky et al. 2011), tropical forests (Gastreich 1999), and inter-tidal communities (Trussell et al. 2002). While the absence of TMIIs within the water bug communities was expected, these results were not expected in the crayfish and fish com-munities. However, our results support theoretical models that have demonstrated the absence of TMIIs when induc-ible defenses do not reduce resource consumption or popula-tion size in three-species food chains (Abrams 2008). Based on the results from the lethal predator treatments, snail population size would have to be reduced by 75% for the

Figure 4. The effect of treatments on log10-transformed periphyton biomass. The predator treatments are: no predator (NP), water bug (W), crayfish (C), and fish (F). Data (least-squares means 1 SE) are averaged over time. Treatments sharing lower case letters are not signifi-cantly different from each other based on pairwise comparisons using Fisher’s LSD test (p 0.05).

10

communities with different food web topologies also will be vital for addressing how indirect effects can be transmitted through different communities. Given that many species do not solely express behavioral plasticity, there also must be a greater emphasis on indirect effects that are transmitted by means other than behavior. This is especially important in more complex communities in which species may be interacting in a multitude of ways. In sum, research directed towards addressing the long-term consequences of predator– prey interactions within communities will help to reveal whether the conclusions and predictions generated from short-term experiments are applicable on ecological and evolutionary timescales.

Acknowledgements – We thank N. Diecks, J. Lam and D. Lutomski for their assistance with the experiment. We thank M. Vos for helpful comments on the manuscript. The Conchologists of Amer-ica, the Pennsylvania Academy of Science, Sigma Xi, the Univ. of Pittsburgh McKinley Research Award, the Malacological Society of London Centenary Research Award, and the National Science Foundation (DEB 05-08277) provided financial support.

References

Abrams, P. A. 1991. Strengths of indirect effects generated by optimal foraging. – Oikos 62: 167–176.

Abrams, P. A. 1995. Implications of dynamically variable traits for identifying, classifying and measuring direct and indirect effects in ecological communities. – Am. Nat. 146: 112–134.

Abrams, P. A. 2008. Measuring the impact of dynamic antipredator traits on predator–prey-resource interactions. – Ecology 89: 1640–1649.

Alexander, J. E. J. and Covich, A. P. 1991. Predation risk and avoidance behavior in two freshwater snails. – Biol. Bull. 180: 387–393.

Altwegg, R. et al. 2004. Dynamics of an inducible defence in the protist Euplotes. – Arch. Hydrobiol. 160: 431–446.

Beckerman, A. P. et al. 1997. Experimental evidence for a behavior-mediated trophic cascade in a terrestrial food chain. – Proc. Natl Acad. Sci. USA 94: 10735–10738.

Belovsky, G. E. et al. 2011. Prey change behaviour with predation threat, but demographic effects vary with prey density: experi-ments with grasshoppers and birds. – Ecol. Lett. 14: 335–340.

Black, A. R. 1993. Predator-induced phenotypic plasticity in Daphnia pulex: life history and morphological responses to Notonecta and Chaoborus. – Limnol. Oceanogr. 38: 986–996.

Bolker, B. et al. 2003. Connecting theoretical and empirical stud-ies of trait-mediated interactions. – Ecology 84: 1101–1114.

Boonstra, R. et al. 1998. The impact of predator-induced stress on the snowshoe hare cycle. – Ecol. Monogr. 68: 371–394.

Chalcraft, D. R. and Resetarits, W. J. 2003. Predator identity and ecological impacts: functional redundancy or functional diver-sity? – Ecology 84: 2407–2418.

Chase, J. M. 1999. To grow or to reproduce? The role of life-history plasticity in food web dynamics. – Am. Nat. 154: 571–586.

DeWitt, T. J. et al. 1999. Trait compensation and cospecialization in a freshwater snail: size, shape and antipredator behaviour. – Anim. Behav. 58: 397–407.

Dillon, R. T. 2000. The ecology of freshwater molluscs. – Cam-bridge Univ. Press.

Gastreich, K. R. 1999. Trait-mediated indirect effects of a theridiid spider on an ant–plant mutualism. – Ecology 80: 1066–1070.

Gelwick, F. P. 2000. Grazer identity changes the spatial distribution of cascading trophic effects in stream pools. – Oecologia 125: 573–583.

We also assessed DMIIs in our system using hand-thinning treatments that were based on predation rates typical of water bugs. We observed a 27% reduction in snail population size with hand thinning across the predator treatments, yet no indirect effect on periphyton biomass was observed. The lack of a DMII was likely driven by compensatory growth of the remaining snails; early in the experiment snails in the thin-ning treatment were 25% larger than snails in the no thin-ning treatment. Due to logistical constraints, we could not apply additional hand-thinning rates to mimic the effects of lethal fish and crayfish, which would be expected to lead to significant DMIIs. Additional hand-thinning treatments also would provide insights into whether productivity (i.e. rate of periphyton growth) played a role in the lack of a DMII. Although the presence of snails in the tanks reduced periphy-ton biomass, high rates of periphyton productivity may have sustained snail populations due to feedbacks between produc-ers, prey and predators (Vanni and Layne 1997). While the hand-thinning treatments provided little support for DMIIs in the system, we found support for DMIIs in the negative correlation between periphyton biomass and snail population size. This demonstrates that periphyton biomass was directly associated with fluctuations in snail population size.

Overall, DMIIs appear to be the dominant factor driv-ing indirect interactions within our system. However, our ability to assess interactions between TMIIs and DMIIs was limited by our implementation of a single thinning treat-ment. Using a broader range of hand-thinning rates and caged predator densities, Peacor and Werner (2001) were able to demonstrate that the total indirect effect of lethal dragonflies on tadpole communities could be attributed to independent effects of density and trait changes as well as their interaction. While the absence of TMIIs in our sys-tem confirms that trait changes had no independent effect on the observed cascade in the lethal predator treatments, there is the possibility that TMIIs could manifest as snail densities are reduced. Thus, the implementation of a gradi-ent in hand-thinning rates including rates that mimic lethal crayfish and fish will help to determine if TMIIs and DMIIs interact within our system.

Conclusions

It is clear that environmental variation can have dramatic effects on the phenotypes expressed by nearly all organ-isms. While a wealth of information regarding the ecologi-cal and evolutionary importance of phenotypic plasticity has been obtained from short-term experiments, there is a clear need for more long-term studies that explore its role in structuring communities. In particular, multi-generation, long-term studies that track population parameters (e.g. r) of the species within communities will make significant strides towards our understanding of whether phenotypic plasticity has long-term effects on community structure and its rela-tive importance to changes in species density. Moreover, the emerging physiological stress framework may provide greater insight into the general mechanisms underlying the effect of predator-induced stress on prey populations and food web interactions (Sheriff et al. 2009, Hawlena and Schmitz 2010a, b). Studies that compare TMIIs and DMIIs across

11

Peckarsky, B. L. and McIntosh, A. R. 1998. Fitness and community consequences of avoiding multiple predators. – Oecologia 113: 565–576.

Persson, L. and De Roos, A. M. 2003. Adaptive habitat use in size-structured populations: linking individual behavior to population processes. – Ecology 84: 1129–1139.

Polis, G. A. and Strong, D. R. 1996. Food web complexity and community dynamics. – Am. Nat. 147: 813–846.

Raimondi, P. T. et al. 2000. Processes structuring communities: evidence for trait-mediated indirect effects through induced polymorphisms. – Oikos 91: 353–361.

Relyea, R. A. 2000. Trait-mediated indirect effects in larval anurans: reversing competition with the threat of predation. – Ecology 81: 2278–2289.

Relyea, R. A. 2003. Predators come and predators go: the reversibil-ity of predator-induced traits. – Ecology 84: 1840–1848.

Relyea, R. A. 2004. Integrating phenotypic plasticity when death is on the line: insights from predator–prey systems. – In: Pigliucci, M. and Preston, K. (eds), The evolutionary biology of complex phenotypes. Oxford Univ. Press, pp. 176–194.

Riessen, H. P. 1999. Predator-induced life history shifts in Daphnia: a synthesis of studies using meta-analysis. – Can. J. Fish. Aquat. Sci. 56: 2487–2494.

Schmitz, O. J. 2000. Combining field experiments and individual-based modeling to identify the dynamically relevant organiza-tional scale in a field system. – Oikos 89: 471–484.

Schmitz, O. J. and Sokol-Hessner, L. 2002. Linearity in the aggre-gate effects of multiple predators in a food web. – Ecol. Lett. 5: 168–172.

Schmitz, O. J. et al. 2004. Trophic cascades: the primacy of trait-mediated indirect interactions. – Ecol. Lett. 7: 153–163.

Sheriff, M. J. et al. 2009. The sensitive hare: sublethal effects of predator stress on reproduction in snowshoe hares. – J. Anim. Ecol. 78: 1249–1258.

Tollrian, R. 1993. Neckteeth formation in Daphnia pulex as an example of continuous phenotypic plasticity: morphological effects of Chaoborus kairomone concentration and their quantification. – J. Plankton. Res. 15: 1309–1318.

Trussell, G. C. et al. 2002. Field evidence of trait-mediated indirect interactions in a rocky intertidal food web. – Ecol. Lett. 5: 241–245.

Turner, A. M. 1997. Contrasting short-term and long-term effects of predation risk on consumer habitat use and resources. – Behav. Ecol. 8: 120–125.

Turner, A. M. 2004. Non-lethal effects of predators on prey growth rates depend on prey density and nutrient additions. – Oikos 104: 561–569.

Turner, A. M. et al. 1999. Predator identity and consumer behav-ior: differential effects of fish and crayfish on the habitat use of a freshwater snail. – Oecologia 118: 242–247.

Turner, A. M. et al. 2000. Chemical cues modify species interac-tions: the ecological consequences of predator avoidance by freshwater snails. – Oikos 88: 148–158.

Van Buskirk, J. 2000. The costs of an inducible defense in anuran larvae. – Ecology 81: 2813–2821.

Vanni, M. J. and Layne, C. D. 1997. Nutrient recycling and her-bivory as mechanisms in the “top–down” effect of fish on algae in lakes. – Ecology 78: 21–40.

Verschoor, A. M. et al. 2004. Inducible defences prevent strong population fluctuations in bi- and tritrophic food chains. – Ecol. Lett. 7: 1143–1148.

Werner, E. E. and Peacor, S. D. 2003. A review of trait-mediated indirect interactions in ecological communities. – Ecology 84: 1083–1100.

Wojdak, J. M. and Luttbeg, B. 2005. Relative strengths of trait-mediated and density-mediated indirect effects of a predator vary with resource levels in a freshwater food chain. – Oikos 111: 592–598.

Gilbert, J. J. 1980. Further obervations on developmental polymor-phism and its evolution in the rotifer Brachionus calyciflorus. – Freshwater Biol. 10: 281–294.

Hammers-Wirtz, M. and Ratte, H. T. 2000. Offspring fitness in Daphnia: is the Daphnia reproduction test appropriate for extrapolating effects on the population level? – Environ. Toxi-col. Chem. 19: 1856–1866.

Hammill, E. et al. 2010. Precise time interactions between behav-ioural and morphological defences. – Oikos 119: 494–499.

Hawlena, D. and Schmitz, O. J. 2010a. Herbivore physiological response to predation risk and implications for ecosystem nutrient dynamics. – Proc. Natl Acad. Sci. USA 107: 15503–15507.

Hawlena, D. and Schmitz, O. J. 2010b. Physiological stress as a gundamental mechanism linking predation to ecosystem functioning. – Am. Nat. 176: 537–556.

Hoverman, J. T. and Relyea, R. A. 2007. The rules of engagement: how to defend against combinations of predators. – Oecologia 154: 551–560.

Hoverman, J. T. and Relyea, R. A. 2008. Temporal environmental variation and phenotypic plasticity: a mechanism underlying priority effects. – Oikos 117: 23–32.

Hoverman, J. T. and Relyea, R. A. 2009. Survival tradeoffs associ-ated with inducible defences in snails: the roles of multiple predators and developmental plasticity. – Funct. Ecol. 23: 1179–1188.

Hoverman, J. T. et al. 2005. Putting prey back together again: integrating predator-induced behavior, morphology and life history. – Oecologia 144: 481–491.

Hoverman, J. T. et al. Environmental gradients and the structure of freshwater snail communities. – Ecography, in press.

Huang, C. and Sih, A. 1991. Experimental studies on direct and indirect interactions in a three trophic-level stream system. – Oecologia 85: 530–536.

Kusch, J. 1998. Long-term effects of inducible defense. – Ecoscience 5: 1–7.

Kusch, J. and Kuhlmann, H. W. 1994. Cost of stenostomum-induced morphological defense in the ciliate Euplotes octocarinatus. – Arch. Hydrobiol. 130: 257–267.

Lima, S. L. and Bednekoff, P. A. 1999. Temporal variation in dan-ger drives antipredator behavior: the predation risk allocation hypothesis. – Am. Nat. 153: 649–659.

Luttbeg, B. et al. 2003. Prey state and experimental design affect relative size of trait- and density-mediated indirect effects. – Ecology 84: 1140–1150.

McCarthy, T. M. and Fisher, W. A. 2000. Multiple predator-avoidance behaviours of the freshwater snail Physella heterostropha pomila: responses vary with risk. – Freshwater Biol. 44: 387–397.

McCollum, E. W. et al. 1998. Complex interactions of fish, snails, and littoral zone periphyton. – Ecology 79: 1980–1994.

McCoy, M. W. et al. 2006. Size correction: comparing morpho-logical traits among populations and environments. – Oecologia 148: 547–554.

McIntosh, A. R. and Peckarsky, B. L. 1999. Criteria determining behavioural responses to multiple predators by a stream mayfly. – Oikos 85: 554–564.

Mitchell, S. E. and Carvalho, G. R. 2002. Comparative demo-graphic impacts of ‘info-chemicals’ and exploitative competi-tion: an empirical test using Daphnia magna. – Freshwater Biol. 47: 459–471.

Nakaoka, M. 2000. Nonlethal effects of predators on prey popula-tions: predator-mediated change in bivalve growth. – Ecology 81: 1031–1045.

Nelson, E. H. et al. 2004. Predators reduce prey population growth by inducing changes in prey behavior. – Ecology 85: 1853–1858.

Peacor, S. D. and Werner, E. E. 2001. The contribution of trait-mediated indirect effects to the net effects of a predator. – Proc. Natl Acad. Sci. USA 98: 3904–3908.

12

The effects of predators and thinning treatments on log10-transformed periphyton biomass over time (means 1 SE). Closed circles, open circles, open triangles, open diamonds represent the no-predator, water bug, crayfish and fish treatments, respec-tively. Predators were caged within the no-hand-thinning (A) and hand-thinning treatments (B). The no-snail treatment (open squares) is shown in each panel for comparison.

Appendix 1