SEARCH

SEARCH BY CITATION

Keywords:

  • Predation;
  • Phenotypic Plasticity;
  • Functional responses;
  • amphibian;
  • dragonfly

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  • 1
    The role of non-consumptive predator effects in structuring ecological communities has become an important area of study for ecologists. Numerous studies have shown that adaptive changes in prey in response to a predator can improve survival in subsequent encounters with that predator.
  • 2
    Prey-mediated changes in the shapes of predators’ functional response surfaces determine the qualitative predictions of theoretical models. However, few studies have quantified the effects of adaptive prey responses on the shape of predator functional responses.
  • 3
    This study explores how prey density, size and previous predator experience interact to change the functional response curves of different-sized predators.
  • 4
    We use a response surface design to determine how previous exposure to small or large odonate predators affected the short-term survival of squirrel tree frog (Hyla squirella) tadpoles across a range of sizes and densities (i.e. the shape of odonate functional response curves).
  • 5
    Predator-induced tadpoles in a given size class did not differ in shape, although induction changed tadpole behaviour significantly. Induced tadpoles survived better in lethal encounters with either predator than did similar-sized predator-naive tadpoles.
  • 6
    Induction by either predator resulted in increased survival with both predators at a given size. However, different mechanisms led to increased survival for induced tadpoles. Attack rate for the small predators, whereas handling time increased for the large predators.

Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

An organism's ability to change its phenotype in response to environmental conditions can be critical for its survival (Scheiner 1993; Via et al. 1995). For example, Cope's grey tree frog (Hyla chrysoscelis) tadpoles develop large red tails with robust musculature when reared with predator chemical cues, and these changes in morphology improve their probability of escaping predators (McCollum & VanBuskirk 1996; McCollum & Leimberger 1997). A growing body of empirical work has demonstrated the existence of such trait-mediated interactions (TMIs): interactions between species that are mediated by changes in organismal traits. In addition, theoretical studies have shown that adaptive trait change can have unpredictable impacts on multispecies interactions and may be important for determining the long-term dynamics and persistence of populations and communities (Abrams 1982, 1992; Holt 1984; Matsuda, Hori & Abrams 1994; Bolker et al. 2003).

Despite these advances, our understanding of how TMIs and predation influence the natural communities remains limited. This limitation exists, in part, because much of the empirical research on TMIs has focused upon documenting the occurrence of survival trade-offs for different phenotypes, and has largely been decoupled from TMI theory (Bolker et al. 2003). For example, understanding how adaptive trait change affects the shape of the functional response is important for defining the curvatures of cost and benefit functions associated with particular adaptive responses, and for making qualitative predictions from theoretical models (Abrams 1992, 1995, 2001; Bolker et al. 2003). Typically, most natural communities are heterogeneous in both space and time and include multiple predators feeding on multiple prey. The susceptibility of particular prey phenotypes to different predators might change as a function of prey size and density. Therefore, integrating multiple densities and sizes of prey or predators is important for understanding how prey phenotypic change affects the shapes of predator functional responses and the short-term costs and benefits of induced phenotypes (Jeschke & Tollrian 2000; Aljetlawi, Sparrevik & Leonardsson 2004).

multiple predators

Induced defences can either increase or decrease the stability of multiple-predator systems (Matsuda, Abrams & Hori 1993; Soluk 1993; Matsuda et al. 1994; Sih, Englund & Wooster 1998; Jeschke, Kopp & Tollrian 2002; Vonesh & Osenberg 2003; Vonesh & Warkentin 2006). Predator-specific defences can lead to reduced inter- and increased intra-specific effects for predators, which promotes coexistence among predators (Matsuda et al. 1993, 1994), but often has strong negative effects on prey (e.g. Soluk 1993). In contrast, generalized defences by prey have equivocal impacts on inter- and intra-specific interactions between predators (Matsuda et al. 1993, 1994) and are expected to stabilize short-term dynamics by reducing the initial slope of the functional responses of both predators. However, defensive responses that are general but differentially effective against different predators can lead to positive or negative effects, depending on the degree of asymmetry in the effectiveness of the preys’ response to each predator. Predator-specific phenotypes are most likely to occur when predators have non-overlapping foraging niches or have qualitatively different foraging styles (e.g. sit-and-wait vs. active foraging) and are less likely to occur when they increase vulnerability to other predators (Matsuda et al. 1993, 1994; Sih et al. 1998). For example, Benard (2004) found that Pacific tree frog tadpoles exposed to chemical cues from predaceous diving-beetles and bluegill sunfish (predators that do not typically co-occur) expressed different phenotypes. Tadpoles survived better in encounters with predators to which they had previously been exposed than with a novel predator. In contrast, Teplitsky et al. (2004) showed that tadpoles exposed simultaneously to chemical cues from Aeshna dragonfly nymphs and fish predators (Gasterosteus aculeatus) expressed phenotypes similar to those expressed for the predator that inflicted the highest mortality rates in foraging trials: Aeshna.

prey size

Studies of plastic responses to multiple predators have shed light on potential conflicts, but they have neglected the size-structured nature of many predator–prey systems. There is substantial evidence that prey body size affects predation risk (e.g. Osenberg & Mittelbach 1989; Semlitsch 1990; Persson & Leonardsson 1998; Alford 1999; Chase 1999a; Aljetlawi et al. 2004; Vonesh & Warkentin 2006), that predator species have different patterns of size-selectivity (Brodie & Formanowicz 1983; Osenberg & Mittelbach 1989; Vonesh et al. 2006) and that many prey change the way in which resources are allocated to somatic growth, reproduction and defence in response to predators (e.g. Law 1979; Crowl & Covich 1990; Reznick 1990; Matsuda et al. 1993; Werner & Anholt 1993; Abrams & Rowe 1996; Chase 1999a, 1999b). While the presence of predators can itself induce changes in prey growth (Vonesh & Bolker 2005), we concentrate here upon the effects of predator induction at a given size.

study objectives

We conducted experiments to examine the occurrence and consequences of predator-induced phenotypic plasticity in squirrel tree frog (H. squirella Bosc) tadpoles of experimentally manipulated sizes and densities in the presence of small (Pachydiplax longipennis Burmeister) and large (Tramea carolina Linnaeus) predators. Although this design did not allow us to examine size trajectories of tadpoles or predator-induced variation in these trajectories, it did allow us to investigate how induced defences in tadpoles within a given size class affected their risk of predation by small and large predators. First, we tested for differences in behaviour and shape of squirrel tree frog tadpoles within a given size class exposed to chemical cues from the different predators and those reared without predator cues. Secondly, we tested for differences in performance for induced and non-induced tadpoles of different sizes by fitting functional response curves for each of the predators foraging on prey from each size and induction treatment in a fully crossed design. We quantified mortality for tadpoles that were induced by P. longipennis or T. carolina, in five different size classes and at five different densities, in the potentially lethal presence of P. longipennis and T. carolina separately. This design allowed us to evaluate whether previous exposure to one predator enhanced or reduced risk to the second predator, and to evaluate the size-specificity of such trade-offs.

Methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

study system

Ephemeral ponds have been the subject of many studies in community ecology (e.g. Wilbur 1997). Much of the faunal diversity in these systems consists of amphibian and odonate species, making them model subjects for testing ecological theory. Anuran larvae are highly plastic in both their behaviour and morphology (Smith & Vanbuskirk 1995; McCollum & VanBuskirk 1996; McCollum & Leimberger 1997; Wilbur 1997; Relyea 2001a, 2001b) and there are known trade-offs associated with particular phenotypes (McCollum et al. 1996; Relyea 2002, 2003; Benard 2006).

study species

Squirrel tree frogs (H. squirella) and the odonates (P. longipennis and T. carolina) are common inhabitants of temporary wetlands throughout the south-eastern United States. Nymphs of these dragonflies can be voracious predators, and the large-bodied T. carolina are often the top predators in these systems (Wissinger 1988; Batzer & Wissinger 1996; Wellborn, Skelly & Werner 1996). In addition, both of these odonates are visual predators and therefore are more likely to consume active than inactive prey.

However, the vulnerability of tadpoles to predation probably differs for these two species of predators. P. longipennis is smaller than T. carolina and has smaller mandibles (a strong determinant of gape size); thus P. longipennis may be a serious threat only during the early tadpole larval stages, whereas T. carolina may be able to capture and consume tadpoles successfully until they reach metamorphosis.

collection and maintenance of experimental animals

Twenty-two amplecting pairs of H. squirella were collected from small ponds in Alachua County, Florida. Each pair was placed into a small plastic container and left at the collection site until eggs were laid and fertilized. Adults were released and egg masses were assigned haphazardly to nine outdoor plastic wading pools to hatch. The predatory nymphs of both P. longipennis and T. carolina were collected with dipnets from small ponds in Alachua County, Florida, including those from which squirrel tree frogs were collected.

functional response estimation

Induction treatments

Each of the nine wading pools was assigned randomly to one of three predator induction treatments – three caged P. longipennis (small) nymphs, three caged T. carolina (large) nymphs or three empty cages. Predator cages were constructed of polyvinylchloride (PVC) septic drainpipe cut into 25 cm lengths with each end secured by a fibreglass window screen. The predrilled drain holes in the pipe were also covered and secured with hot glue. These cages prevented direct interactions between predators and tadpoles but allowed free exchange of water and predator chemical cues. Each pool was fitted with a fibreglass window screen lid that was secured with a bungee cord to prevent colonization by other frogs or predatory insects. The dragonfly nymphs were each fed 10 small tadpoles every other day throughout the induction phase of the experiment (3 weeks). Tadpoles fed on naturally occurring periphyton in the pools, but were also supplemented with ad libitum additions of Spirulina algae disks throughout the experiment.

experimental design to test for survival trade-offs in tadpoles

We used a response surface design to determine the effects of predator-induced plasticity on the tadpoles’ performance with different predators (Inouye 2001). Specifically, we crossed five densities of tadpoles (5, 10, 20, 30, 40) with five size classes (tadpoles size classes: (1) inline image = 9·40, sd = 0·71; (2) inline image = 13·16, sd = 1·47; (3) inline image = 16·78, sd = 2·52; (4) inline image = 25·63, sd = 4·00; (5) inline image = 33·05, sd = 2·06). Due to logistical constraints, we chose treatment combinations that provided the most information for estimating a response surface (Fig. 1). These included those that defined the corners of the surface (i.e. size class 1 × density 5; size class 1 × density 40; size class 5 × density 5; and size class 5 × density 40) and the full series of size by density combinations that defined the shape at the centre of the surface (i.e. all size classes × density 20 and all densities × size class 3) (Fig. 1). The experiment was replicated in three temporal blocks (Fig. 1), although each time block represented an incomplete set of the treatment combinations (i.e. incomplete block design: Fig. 1). In this design treatments received different levels of replication, but provided sufficient replication to fit and compare surfaces and account for between-block variation. We crossed the density–size treatments with three non-lethal predator treatments (‘induction’) and two lethal predator treatments (‘performance’), yielding a total of 78 treatments. The induction by performance combinations included: (1) non-induced × lethal small predator; (2) non-induced × lethal large predator; (3) induced by small predator × lethal small predator; (4) induced by small predator × lethal large predator; (5) induced by large × lethal large predator; and (6) induced by large predator × lethal small predator.

image

Figure 1. Experimental design. Circles represent 13 size by density treatment combinations crossed with six induction by performance treatments. Black-filled circles were replicated within a block, grey-filled circles were replicated in all three blocks and open circles were unreplicated but are members of a full regression series.

Download figure to PowerPoint

Performance trials

Performance trials were conducted in 38-litre glass aquaria filled with 30 litres of water; the bottom of each aquarium was covered with a piece of felt outdoor carpet. The perimeter of the carpet was secured with aquarium gravel. Each aquarium was also provided a small plastic aquarium plant and one pulverized Spirulina algae wafer. All tanks, carpets, gravel and ornaments were cleaned and sterilized in 70% ethanol between time blocks.

For each performance trial the tadpoles from the three induction pools for each treatment (described above) were grouped into the appropriate size class, added to an aquarium at the appropriate density, and allowed to acclimate overnight. Predators were starved for 24 h prior to the start of each trial. At the start of each trial predators were added to the aquaria (order of predator additions was determined randomly), and were allowed to feed for 24 h. At the end of the trial predators were removed in the same order in which they were added, and all tadpoles that remained alive were counted.

morphology and behaviour

To quantify morphological phenotypic responses we sampled and photographed 15 tadpoles from each induction treatment after 3 weeks. Tadpoles were photographed from the top and side, and eight linear measurements of morphological traits were taken as described by Relyea (2001a): (1) tail fin length; (2) tail fin depth; (3) body length; (4) body depth; (5) tail fin muscle depth; (6) tail fin muscle width; (7) body width; and (8) total body length.

We quantified behavioural responses of squirrel tree frog tadpoles to these two odonate predators in a separate experiment that was performed during June and July 2005. For this experiment, six newly hatched predator-naive squirrel tree frog (H. squirella) tadpoles (collected from the same location and in the same manner as described above) were placed into 19-litre glass aquaria that contained one of three predator treatments (caged P. longipennis, caged T. carolina) or an empty cage. Caged predators were fed four tadpoles every other day and tanks were cleaned twice weekly. Tadpoles were fed pulverized alfalfa pellets every other day. We quantified activity as the proportion of time that tadpoles exhibited any of five different measures of behaviour: (1) surfacing; (2) swimming; (3) feeding; (4) darting (rapid bursts of swimming); and (5) tail undulating (i.e. movements of the tail without a significant change in position). We recorded the number of tadpoles exhibiting each behaviour every minute during a 15-min observation for each tank. Tadpole behaviour was assessed after 1 week on small tadpoles (approximately size class 2) and again after 3 weeks on larger tadpoles (approximately size class 4).

statistical analysis

All statistical analyses were performed in the r statistical programming environment version 2·3·1 (R Development Core Team 2006).

functional response estimation

Test for survival trade-offs

To evaluate differences in the performance of tadpoles from the three induction treatments with each of the two predators, we compared the attack rate (α) and handling time (h) parameters from an estimated Type II functional response (Holling 1961). However, because prey were not replaced as they were consumed in the foraging trials, we estimated α and h from the ‘random predator equation’, which accounts for prey depletion (Rogers 1972; Juliano 2001). The random predator equation is:

  • Ne = N0{1 − exp[α(hNe − T)]}(eqn 1)

where T is time, Ne is number of tadpoles eaten, N0 is the initial density of tadpoles, h is the handling time and α is the instantaneous attack rate.

We incorporated size-specificity of predation risk explicitly by allowing attack rate to vary as a function of tadpole size. Typically, size dependence in the attack rate has a hump-shaped form, because predator encounter rates are low for small prey and large prey are hard to capture (Persson & Leonardsson 1998; Aljetlawi et al. 2004; Vonesh & Bolker 2005). Size-dependent attack rates have been modelled as modified logistic functions (Vonesh & Bolker 2005), power Ricker functions (Persson et al. 1998) and standard Ricker functions. Whether the particular form of the function used to model size-dependent attack rates falls off hyperbolically (i.e. proportional to size−1 or more generally as a decreasing power function of size, size−γ) or exponentially may affect long-term predator–prey dynamics qualitatively (Persson et al. 1998; Aljetlawi et al. 2004). Unfortunately, the data obtained in this study were insufficient to distinguish among these alternatives; thus we chose initially to follow the most common practice of modelling size dependence in attack rate with a standard Ricker model (i.e. exponential decay at large sizes). However, because attack rates declined monotonically with increasing size in our study (i.e. the smallest size class tadpoles in this study was already beyond the window of vulnerability), we estimated attack rate with a declining exponential function:

  • αs = ceb·(size–minsize),(eqn 2)

where c describes the attack rate on the smallest size class of prey and b describes the decay of attack rate with increasing tadpole size (i.e. attack rate decay). When b is large, tadpole risk declines sharply with tadpole size (i.e. there is a strong size refuge effect). Previous applications of the random predator equation have used iterative methods to compute values for Ne (e.g. (Juliano 2001; Vonesh et al. 2005). Because the solution of the Rogers equation can be written in terms of Lambert's W function (a special function that provides the solution to the equation W(x)e(x) = x), we were able to use an algorithm from Corless et al. (1996) to fit our models. In particular, if W(x) is Lambert's W function, then the number of prey eaten Ne equals:

  • image(eqn 3)

where P is the number of predators (an R function implementing this algorithm is available from the authors).

We used generalized non-linear least squares to estimate the parameters for the size-specific random predator equation. Generalized non-linear least squares fits a non-linear model, assuming normally distributed variation, but allowing heteroscedasticity in the error structure. In this analysis, variance was assumed to be a power of the estimated mean predation rate; the exponent was estimated from the data as part of the estimation procedure. One limitation of this experimental design is that because only three size classes were represented in any given block we could not estimate the parameters of a true mixed-effects model to account for block effects. Therefore, we tested for block effects by fitting a fixed parameter model to pooled data and then using analysis of variance tested for a significant block effect in the residuals. However, we found no significant between-block differences in the residuals (P = 0·52), indicating that time blocks did not influence the model fit significantly. Thus we estimated parameters without this term included in the model.

Because we were interested in the interaction between predator and prey size, we focused our hypothesis tests on the parameters for handling time (h) and attack rate decay (b). The attack rate on the smallest size class of prey, c, did not vary significantly among treatments in preliminary analyses, so we held it constant when estimating parameters for all other comparisons. Specifically, we first tested the following three predictions.

  • • 
    Handling time was expected to be longer for the small predator.
  • • 
    Handling time was expected to be longer for induced prey.
  • • 
    Attack rate decay b was expected to be larger for the small predator.

We also tested three predictions that varied depending on the nature of the trade-offs and level of adaptive responses.

  • • 
    If tadpole responses to predators are adaptive then attack rate decay b would be larger for tadpoles that had prior experience with the lethal predator than for predator-naive tadpoles or those induced by the opposite predator.
  • • 
    If tadpole responses are predator-specific, then attack rate decay b would be smaller for large predators feeding on tadpoles induced by small predators than for large predators feeding on predator-naive tadpoles.
  • • 
    Similarly, if the tadpole responses are adaptive but generalized, then attack rate decay b should be indistinguishable for tadpoles induced by either predator and should be larger than for predator-naive tadpoles.

morphology and behaviour

Because differences in the sizes of morphological traits can often be explained partly by differences in overall body size, quantitative comparisons of morphological traits typically require size correction. We used common principal components analysis (CPCA) and Burnaby's back projection method (CPCA/BBPM) to obtain size-corrected trait values. This approach accounts for error in the estimated size axis (the first common principal component, CPC1) that is propagated into the back-projected (i.e. size-corrected) data and incorporates it into an analysis of variance (McCoy et al. 2006).

We used a generalized linear model with quasibinomial error structure (Crawley 2007) to test whether predator cue affected tadpole behaviour.

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

performance trials

There was a significant effect of lethal predator (predator experienced during the performance phase: P = 0·002, d.f. = 1, F = 10·09) and a significant interaction between inducer (i.e. predator experienced during the induction phase) and predator (P = 0·042, d.f. = 2, F = 3·24) on the estimate of attack rate decay (b) (Figs 2 and 3). Specifically, attack rate decay (b) was 3·4 times larger on average for small predators feeding on tadpoles that had been induced by either predator than for small predators feeding on predator-naive tadpoles (Figs 3 and 4). Similarly, the change in attack rate decay (b) was 1·6 times larger for the large predators feeding on tadpoles that had been induced than for large predators feeding on predator-naive tadpoles (Figs 3 and 4). However, attack rate decay (b) did not differ significantly for tadpoles induced by small or large predators.

image

Figure 2. Attack rates as a function of size estimated for each predator treatment.

Download figure to PowerPoint

image

Figure 3. Fitted surfaces for each predator treatment. In these ‘lollipop’ plots, observed predation rate (i.e. the number of tadpoles eaten) is represented by spheres and the residuals from the fitted surfaces are represented by solid black lines. The large variance of the residuals from the fitted surface at small size and high density is accounted for in the generalized least squares model used to estimate parameters. (a, c, e) Tadpoles exposed to small predators (Pachydiplax longipennis). (b, d, f) Tadpoles exposed to large predators (Tramea carolina).

Download figure to PowerPoint

image

Figure 4. Estimates of the attack rate decay parameter (b) from generalized non-linear least squares fits of the random predator equation. Tadpoles from small and large inducers are pooled. Error bars represent one standard error.

Download figure to PowerPoint

The handling times of the small predators were longer on average than large predators (P = 0·042, d.f. = 2, F = 3·24, Figs 3 and 5). However, the small predators took 27% longer to handle induced tadpoles, whereas it took the large predators 84% longer to handle induced tadpoles (P = 0·002, d.f. = 1, F = 10·09, Figs 3 and 5). The effects of induction on handling time (h) were opposite to the effects of induction on the attack rate decay (b).

image

Figure 5. Estimates of handling time (h) from generalized non-linear least squares fits of the random predator equation. Tadpoles from the small and large inducers are pooled. Error bars are one standard error.

Download figure to PowerPoint

morphology and behaviour

We first tested whether the morphological traits of tadpoles from different treatment groups had shared allometry, and then removed the potential effects of differences in organism size to evaluate whether our predator induction treatments resulted in differences in sizes of morphological traits. Tadpoles from all three predator-induction treatments had a common first principal component, suggesting that all shared the same underlying allometry of morphological traits (McCoy et al. 2006). We used Burnaby's back projection method to remove differences in the sizes of traits due to differences in body size (McCoy et al. 2006). Once body size was removed, there were no detectable differences in tadpole morphology (P > 0·2 for all traits).

In contrast, there were large differences in the activity levels of tadpoles reared with chemical cues from predators and those that were in predator-free tanks and a significant interaction between time of observation (week 1 vs. week 3) and predator treatment (P = 0·042, F2,42 = 3·4419). During week 1 (size class 2) tadpoles in control treatments were more than 85% more active than tadpoles in the large predator treatment and 98% more active than tadpoles in the small predator treatment. During week 3 (size class 4) tadpoles in control treatments were 64% more active than they were in week 1, whereas tadpoles in large predator tanks were 77% more active and tadpoles in small predator tanks were 98% more active during week 3 than in week 1. However, control tadpoles were still 79% and 64% more active than tadpoles in the large and small predator treatments, respectively, during week 3 (Fig. 6).

image

Figure 6. Behavioural plasticity by squirrel treefrog tadpoles in response to chemical cues from two different caged odonate naiads (small = Pachydiplax longipennis, large = Tramea carolina).

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

In this study, we demonstrated that squirrel tree frog tadpoles of a given size with prior exposure to waterborne chemical cues from two different species of larval odonate predators (P. longipennis and T. carolina: i.e. predator-induced tadpoles) survived subsequent encounters with these predators better than predator-naive tadpoles. In addition, the tadpoles’ size and previous experience with the predators changed the shapes of both predators’ functional response surfaces. As predicted, small predators had longer handling times than large predators across all prey sizes and predator induction increased handling time for both predators. Tadpole size had a greater effect on the attack rate decay b for the small predator than for the large predator. The larger attack rate decay b for induced than for predator-naive tadpoles suggests an adaptive response of tadpoles to chemical cues from these odonate predators. However, b was indistinguishable for tadpoles induced by either predator, suggesting a generalized adaptive response. Although previous exposure to chemical cues from the two predators did not result in significantly different functional response surfaces, induction affected tadpole survival via two different mechanisms. Induced tadpoles survived better with small predators through an increased size refuge (larger b) and reduced attack rates. In contrast, increased handling times drove the survival advantage of induced tadpoles in the presence of large predators (Figs 2–5).

In this study, handling times were estimated from model fits rather than directly from behavioural observations. Thus, our estimates of handling time reflect how rapidly per capita tadpole mortality decreased with increasing tadpole density. Therefore, we cannot say for certain that decreases in consumption at high density were due solely to the time required for predators to manipulate and consume prey. Our models allowed attack rate to vary as a function of body size, but not with tadpole density. Thus, density-dependent differences in attack rates might also affect our estimates of handling time. For example, if tadpoles decrease activity levels in response to the number of predation events that occur around them, lowering apparency and hence attack rates, our estimation procedure would interpret this change as an increase in handling time. Thus, the greater change in handling time observed for large predators might be explained partly by actual time required to manipulate prey and partly by more consumption events by large predators. Indeed, graded responses to predation risk are predicted by theory (Lima & Dill 1990; Peacor 2003) and have been observed for behavioural and morphological traits in other anuran species (Van Buskirk & Arioli 2002; McCoy 2007).

The differences in attack rates and handling times observed at a given size for induced and non-induced tadpoles in this study could indicate greater sensitivity to chemical cues by induced tadpoles. Lower attack rates for the small predator and longer handling times for the large predator might also reflect differences in the predators’ foraging behaviours. Although both species of predators used in this study can hunt actively, P. longipennis are less active foragers than the larger T. carolina (personal observation). Thus, predator-induced reductions in tadpole activity could have reduced encounters with the sit-and-wait predator, but not the active forager. In addition, the small predator may have been less successful in attacking larger tadpole prey. Thus, the combined effects of fewer encounters and lower probability of capture success could explain the lower attack rates observed for P. longipennis when feeding on predator-induced prey. In contrast, handling times for large predators on induced prey may have increased because reduced prey activity led to lower perceived effective prey density and hence a lower giving-up density for exploiting a captured prey item. If a captured prey item is a resource patch then its relative value, and the time spent extracting nutrients from it, should increase as patches (prey) become more rare and the expected time taken to travel to another patch (capture another prey) increases (Charnov 1976). If less active tadpoles are encountered less frequently or are more difficult to capture, then predators may shift their foraging behaviour from partial prey consumption to complete prey consumption, which would increase estimates of handling time (e.g. Aljetlawi et al. 2004). Partially consumed carcasses were observed commonly in this study; however, these data were not recorded in sufficient detail to test this hypothesis.

Changes in attack rates and handling times might be expected when the prey undergo morphological transformations in response to predator cues (e.g. McCollum et al. 1996; Wiackowski & Staronska 1999; Laforsch & Tollrian 2004; Kishida & Nishimura 2006). We expected a priori that the large predator species would be less gape-limited and would have lower handling times. Thus, adaptive predator responses by the tadpoles were expected to exploit the gape limitations of the small predator (e.g. Wiackowski & Staronska 1999; Jeschke et al. 2002; Laforsch & Tollrian 2004; Kishida & Nishimura 2006). However, we were unable to detect any significant differences in the shape of tadpoles between the predator and no predator induction treatments (i.e. there was no predator-induced morphological plasticity). We may not have measured the appropriate morphological traits. However, it is unlikely that this species’ morphological responses would be so different from other anuran amphibians, including closely related species (e.g. Lardner 2000; Relyea 2001a, 2003; Van Buskirk 2002), that they would be undetectable in any of the traits measured here.

More generally, these predators may also induce changes in individual growth rates by which prey attempt to exploit the functional constraints (e.g. gape limitation) of small predators. While these kinds of adaptive responses are clearly important as part of the overall suite of morphological and life-history responses to predation (Vonesh & Bolker 2005), our study design focused instead on the effects of changing behaviour and shape on vulnerability at a given size: clearly, a full understanding of prey responses will have to integrate changes in growth rate as well as the behavioural and morphological characters studied here.

Larval amphibians also change behaviour adaptively to reduce predator risk. Tadpole activity and competitive ability are correlated strongly. Active tadpoles are more likely to encounter and be killed by predators (Morin 1983; Fauth & Resetarits 1991; Anholt & Werner 1995; Wilbur 1997; Schiesari, Peacor & Werner 2006). Indeed, in this study and in numerous other studies tadpoles reduce levels of activity in response to chemical cues of predators. In fact, tadpoles reduced activity more when they were small than when they were large, and large tadpoles tended to be more active in the presence of the small predators than in the presence of large predators. Thus, improved survival of induced tadpoles observed in this study may have resulted from predator-induced changes in prey activity.

Behaviour is typically considered a highly labile phenotypic trait that can be changed on very short time scales (West Eberhard 1989). Thus, unlike predator-induced changes in prey morphology, induced behaviours (activity in this study) are typically not expected to persist and affect survival in future encounters. However, we found that previous experience with predators resulted in improved survival in the absence of any significant change in shape. Thus, expression of adaptive behaviour (reduced activity) in H. squirella tadpoles might include a component of learning or the physiological mechanisms that lead to (or result from) reduced activity may not be quickly reversed. For example, if tadpoles become sensitized to recognize predator cue via exposure, they may be able to recognize and respond more quickly to predator threats experienced later. Several studies have shown that fish that experience chemical cues from predators and alarm signals from conspecifics learn to recognize predators as threats (reviewed in Kelley & Magurran 2003). Alternatively, it may be adaptive for antipredatory behaviour to decline gradually to prethreat behaviour, or antipredator behaviour may less plastic than assumed typically (e.g. it may require a significant change in neurophysiology).

Although predator-specific morphological responses are observed commonly in anuran tadpoles (e.g. Relyea 2001a, 2004; Teplitsky, Plenet & Joly 2004; Benard 2006), there was no strong evidence for predator-specific morphological responses to the two different predators in this study. This may be due to the functional similarity of the two predators studied. Both species use qualitatively similar predation strategies, and thus may select for similar defences. In addition, these species of odonates are both common inhabitants of temporary ponds in the south-eastern United States and squirrel tree frog tadpoles are likely to encounter both species in a given pond.

Our findings are consistent with theoretical predictions that generalized adaptive responses are more likely when predators co-occur. However, the long-term implications of these adaptive responses are harder to intuit from theoretical models. For example, the observed reductions of attack rates associated with both increasing size and antipredator phenotypes of tadpoles might be expected to stabilize interactions between the two odonate predator species (Matsuda et al. 1993, 1994), but reduce or limit population growth rates for the anurans. In contrast, the increases in handling time that also stem from increases in body size and induction by predators might be expected to reduce long-term coexistence of the two predators, and increase population growth rates for the anurans. These findings highlight the need for more studies focused upon understanding how predator induced changes in traits affect the shapes of predator functional responses and the importance of prey size in these interactions. Future studies should focus upon both quantifying changes in phenotypic traits in response to predators and on understanding how those trait changes affect survival across density and size gradients.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

We thank Craig Osenberg, Bob Holt, Colette St Mary, C. Kenneth Dodd, Krista McCoy, Peter Abrams and an anonymous reviewer for helpful comments. Members of the St Mary–Osenberg–Bolker laboratory group provided helpful discussions and comments on previous drafts. Krista McCoy, Alex Jahn, Casey Cassidy, Alison Amick and several high school students assisted with experiments. We thank the staff of the Ordway–Swisher biological station for assistance and use of facilities. Nici Schraudolph wrote a matlab function implementing the algorithm of Corless et al. 1996, which we translated into r. This work was funded partially by Sigma Xi GIAR and SICB GIAR grants to M.W.M., and NSF OCE 0242312 to C.W.O., B.M.B. and C. St Mary).

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References