Delayed effects of larval predation risk and food quality on anuran juvenile performance


  • Present address: David Álvarez, Departamento de Biología Funcional, Universidad de Oviedo, E-33006 Oviedo, Spain.

Alfredo G. Nicieza, Departamento de Biología de Organismos y Sistemas, Universidad de Oviedo, E-33006 Oviedo, Spain.
Tel.: +34985104788; fax: +34985104866; e-mail:


Metamorphosis can disrupt the correlation structure between juvenile and adult traits, thus allowing relatively independent evolution in contrasting environments. We used a multiple experimental approach to investigate how diet quality and larval predation risk affected the rates of growth and development in painted frogs (Discoglossus galganoi), and how these changes influence post-metamorphic performance. A high-energy diet entailed growth advantages only if predation risk did not constrain energy acquisition, whereas diet quality affected primarily the extension of the larval period. Predation risk influenced juvenile shape, most likely via the effects on growth and differentiation rates. Juvenile frogs emerging from predator environments had shorter legs and longer bodies than those from the nonpredator tanks. However, these morphological changes did not translate into differences in relative jumping performance. Neither size-adjusted lipid storage nor fluctuating asymmetry was significantly influenced by food quality or predation risk. Our data suggest that the post-metamorphic costs of predator avoidance during the larval phase are mostly a consequence of small size at metamorphosis.


A central question in evolutionary biology is how the construction of developmental systems can limit phenotypic expression. The evolution of phenotypic plasticity may be constrained if the ability to be plastic entails significant costs (DeWitt, 1998; DeWitt et al., 1998; Scheiner & Berrigan, 1998). A different kind of constraint can arise where individuals face obligate ontogenetic transitions across contrasting habitats. In this case, the metamorphosis process separates two functionally distinctive morphs in the same individuals, and therefore different suites of traits will be related to performance in different selective environments. If the characters in the pre- and post-metamorphic phases are phenotypically correlated and the fitness payoffs of the expression of these traits are negatively correlated, selection on juvenile or adult individuals can hamper the evolution of inducible defences against predators in the larval environment. This ontogenetic constraint has two notable features. One is a time-lag between the production or occurrence of the inducible defence and the implementation of the cost: it will be a ‘delayed cost’ because it derives from the suboptimal performance of a trait that simply does not exist at the time of expression of the inducible defence. The second feature concerns the degree of causal relationship between phenotypically correlated characters. Unlike direct costs, the value of the relevant trait in the post-metamorphic environment is not necessarily the result of the inducible defence itself, but the consequence of the modification of a related trait (i.e. ‘indirect’ costs).

Metamorphosis can disrupt the genetic correlation structure between juvenile and adult traits, thus allowing relatively independent evolution in contrasting environments (Ebenman, 1990). In fact, the formation of larval and post-metamorphic structures may depend on separate developmental programmes (Wald, 1981; Parks et al., 1988). Developmentally related traits can be highly correlated because of pleiotropic, epigenetic and environmental effects (Atchley, 1984; Schlichting & Pigliucci, 1998), and large positive correlations can constrain trait evolution in the different environments if selection pressures act in opposite directions (Via & Lande, 1985; Roff, 1996). However, even if metamorphosis contributes to break up these genetic constraints, physiological constraints can produce the same effect. For example, a change in some environmental factor (i.e. increased temperature) can influence developmental rate in a way that increases the chance of larval survival (e.g. through shorter larval periods) and concomitantly affect post-metamorphic traits, thus reducing juvenile survival or growth rates (Swain, 1992a,b; Blouin & Brown, 2000; Álvarez & Nicieza, 2002a). We suggest that there are two ways this can arise. First, there might be a direct effect of an environmental factor during the larval phase on the juvenile character, due either to developmental stress or adaptive plasticity (Hoffmann & Parsons, 1997; Møller & Swaddle, 1997). Secondly, the alteration of the larval trait itself can have an effect on post-metamorphic traits; this is likely to occur if shorter larval periods are achieved by truncation of development, or if faster differentiation generates developmental errors.

The goal of this study was two-fold. First, we set out to investigate the effect of predation risk on tadpole behaviour, and age and size at metamorphosis. As the effect of predation risk could be context dependent and closely associated with energy acquisition (Anholt & Werner, 1995), we manipulated food quality to see whether a change in the ability to gather energy per unit time can affect the antipredator behaviour of larvae and, if so, the consequences of the behavioural shift on metamorphosis. Differences in the length of the larval period between larvae reared with or without predator cues can arise through changes in the rate of differentiation, and truncation of development (Hanken, 1984). The second objective of this study is to investigate the deferred effects of predation risk to larvae on juvenile performance. We present a conceptual framework in which size at metamorphosis depends on differentiation and growth rates and influences juvenile lipid reserves and locomotor performance (Fig. 1). Factors affecting the larval period can influence the size, morphology, and energy reserves of the metamorphs (Pfennig, 1992; Blouin & Brown, 2000; Álvarez & Nicieza, 2002a), thus affecting locomotor performance and juvenile growth and survival (Wassersug & Sperry, 1977). To analyse the influence of predation risk on post-metamorphic performance we use an integrative approach that takes into account phenotypic plasticity and developmental stress (Fig. 1).

Figure 1.

 Proposed links between morphological, behavioural, and developmental phenotypic plasticity derived from predation risk and food quality, and juvenile performance of Discoglossus galganoi. Expected positive or negative relationships are indicated by (+/−). Developmental stress is presented as a separate source of phenotypic plasticity to distinguish between nonadaptative and potentially adaptive phenotypes. ‘Behaviour’ refers to activity or use of food-rich/risky habitat. As the energy content of food increases, the time spent in foraging to achieve a given growth rate decreases, but the reward per unit of time increases. Therefore the effect of food quality on behaviour can be either positive (take a higher risk if the reward is greater) or negative (avoid the risk if the cost of avoidance is low).

Food quality affects the amount of food needed to achieve a given growth rate, and thereby the amount of time spent in foraging required to acquire a given size at metamorphosis (Álvarez & Nicieza, 2002b). In this context, we predict that tadpoles will show spatial avoidance of predators (hypothesis 1) and this response will be more pronounced if the energy content of food is higher (hypothesis 2a). Alternatively, as growth efficiency increases with the energetic value of food, tadpoles could be more prone to predation risk under a high-quality food (hypothesis 2b). In a further step, we predict that lower activity in presence of predators will slow down growth and cause either a longer larval period (hypothesis 3a), smaller size at metamorphosis (hypothesis 3b), or time delay and larger size (hypothesis 3c). The net effect of an increased risk of predation is therefore a prolongation of the larval period (positive correlation between predation risk and larval period). A separate hypothesis involves reduction of the larval period (negative correlation between predation risk and larval period) either through faster differentiation (with lower mass at metamorphosis; hypothesis 4a) or development truncation (without change in metamorphic mass; hypothesis 4b).

We suggest that phenotypic plasticity in age at metamorphosis can affect size-independent energy reserves and size-independent locomotor ability, the latter through shape modifications derived from development truncation. In addition, developmental stress induced by predation risk could affect juvenile morphology. If so, risk perception itself could entail direct costs that have not been examined yet.


Anurans are an excellent model system for examining the ecology of phenotypic plasticity (Miner et al., 2005). Larvae exhibit strong behavioural and morphological responses to predation risk during the larval stage (McCollum & Leimberger, 1997; Nicieza, 1999, 2000; Van Buskirk & McCollum, 2000), and these have been assumed to impose costs related only to the premetamorphic or metamorphic phases (e.g. Smith, 1987; but see Relyea, 2001a,b). Reduced activity induced by predator presence causes a deceleration in growth rates (e.g. Laurila & Kujasalo, 1999; Lardner, 2000), which often results in a smaller size at metamorphosis (Lardner, 1998; Nicieza, 2000). We conducted a laboratory experiment to study the effects of predation risk and food quality on larval behaviour, metamorphosis and post-metamorphic performance in the Iberian painted frog (Discoglossus galganoi), an opportunistic, explosive breeder endemic from the Iberian Peninsula. Eggs are laid in ephemeral rain pools where tadpoles face a high risk of desiccation. Embryonic development is completed after 2–6 days depending upon temperature. Larvae are cannibalistic and can reach metamorphosis in 2 weeks (Álvarez & Nicieza, 2002b), but generally larval period extends for 20–60 days.

Experimental set-up

We collected larval D. galganoi (Gosner [1960] stage 20) from a temporary pool on 21 June 1997. The pool is part of a system of rain pools, ditches and small permanent springs located near Collada Fumarea, Asturias (43°26′N, 5°34′W; 565 m elevation). From October to March, the pools are heavily used as breeding sites by common frogs (Rana temporaria). Common newts (Triturus helveticus) are present from February to June, and adults can prey actively on recently hatched larvae of D. galganoi. In addition, these pools support a variety of insect predators including Aeshna cyanea.

The experiment was conducted in polyethylene trays (27 × 43 cm) filled with 9 L of dechlorinated tap water). We used a 2 × 2 factorial design to examine the effects of predation risk and food quality on larvae behaviour, age and size at metamorphosis, and post-metamorphic traits. Two levels of diet (high- and low-energy and protein content) were crossed with the presence and absence of a caged predator, and each combination replicated in five trays. When the tadpoles were at Gosner (1960) stage 26, we randomly assigned groups of 18 larvae to each tray. Predation risk regime consisted of a predator-absent environment (no predator, NP) and presence of caged (thus nonlethal) predacious dragonfly larvae (Aeshna cyanea) (predator, P). To simulate high-risk conditions without producing tadpole mortality we placed one Aeshna larva inside a cylindrical mesh cage (7 cm diameter) attached to one of the tray ends. We fed Aeshna larvae with one nonexperimental Discoglossus tadpole every second day. Aeshna larvae were removed every 10–15 days and replaced by newly collected larvae. Predation risk was crossed with two levels of food quality: high-energy (HE) content (Trouvit food pellets; TROW Espan̂a, S.A., Burgos, Spain; 46% protein, 22% lipids, 2% carbohydrates, 9% ash) and low-energy (LE) content (rabbit food pellets; 17% protein, 3% lipids, 15% carbohydrates, 10% ash). We fed tadpoles ad libitum (12% body mass per day). We put the pellets inside Petri dishes (6 cm diameter, 1 cm depth) near the predator cage to avoid the spread of food. Identical cage and Petri dishes were fixed in the end of the tray opposite to the predator and food compartments; those cages and dishes were empty and used only to keep the trays spatially symmetrical. By doing so, we simulated areas of high risk and food availability, and others where both the risk of predation and food concentration were low. In the low-risk treatment the two predator cages were blank (Appendix S1a: Fig. A1). Tadpoles experienced a fixed photoperiod (14 L : 10 D, lights on 06:00–20:00 h) and water temperature of 12 ± 1 °C throughout the study period.

Changes in larval predator avoidance and life history traits

To investigate how predator and food quality affect tadpole behaviour the experimental trays were observed over 6 days between 16 and 21 July (twice daily except for the last day). We recorded the number of tadpoles in the no-food half of each tray (i.e. the low risk–low food side in the predator's trays). The number of larvae in the food halves was counted to confirm the estimates of spatial avoidance. All the behavioural observations were conducted during daytime, at 12:00 and 18:00 h. Larval period was estimated as the number of days from the start of the experiment (day 1, stage 26) to Gosner (1960) stage 42 (emergence of at least one forelimb). Trays were checked every day, and stage 42 individuals collected, weighed (±0.1 mg), and placed individually in 125-mL plastic vessels until they completed tail resorption (Gosner stage 46). Absolute growth rate was calculated as [(mass at stage 42 − initial tadpole mass)/(number of days from start of the experiment to stage 42)]. Vessels were supplied with a small amount of water to avoid desiccation.

Juvenile morphology, locomotion and lipid content

We examined locomotor performance, morphology, and energy reserves of newly metamorphosed frogs to ascertain if predation risk and food quality during the larval phase can affect juvenile performance and fitness. Metamorphs were individualized in 125-mL plastic vessels and fed ad libitum with adult Drosophila melanogaster for 7–10 days prior to being subjected to the jumping trials. As a measure of locomotor performance, we used jumping ability because it can influence escape from predators and foraging success in anurans. Juvenile D. galganoi were placed on a clean, flat surface and chased them to induce the escape. For each individual, we registered the length of five jumps. In a previous study (Álvarez & Nicieza, 2002a), we found that maximum jump distance correlates better with body size than the average of several jumps. Moreover, maximum jump reduces the risk of underestimation of the jumping ability associated with low motivation in some of the jumps. Therefore, locomotion performance was estimated as the maximum of five jumps. Hops shorter than five body lengths were not considered because they were unequivocally below individual ability. Trials were carried out in an indoor room at 23.0 °C (1SD = 1.4 °C; range 21–25 °C). After the jumping trials, frogs were measured for snout-to-vent length (SVL ± 0.01 mm; two repeated measures per individual), weighed (±0.1 mg), killed by an overdose of benzocaine, photographed and frozen for further analysis of lipid content.

Digitized images of metamorphs were obtained using a binocular microscope and a videocamera connected to a computer equipped with image analysis software. From the digitized images (ventral view) we measured (±0.01 mm) six morphological distances (Appendix S1a: Fig. A2): head width (HW), head length (HL), body length (BL), femur length (FL), right tibiafibula length (RTL), and left tibiafibula length (LTL). As an indication of developmental stress we used measures of fluctuating asymmetry (FA) in tibiafibula length. Therefore, we realized three repeated measurements on each tibia. To estimate energy reserves, metamorphs were oven-dried at 40 °C for 4 days to a constant mass, and total nonpolar lipids were determined by Soxhlet extraction of dried samples using petroleum ether. This solvent is a highly efficient for the extraction of nonpolar (storage) lipids, with little removal of polar (structural) lipids (Dobush et al., 1985). Samples were weighed (±0.01 mg) before and after a 3-h extraction period, and the lipid content was calculated as the difference between fat and lean mass.

Statistical analyses

We used repeated measures analysis of variance to evaluate the effects of predation risk and food quality on tadpole predator avoidance. The observation period (n = 11) was a within-subjects factor. The response was the number of tadpoles in the half of tray containing the food (the ‘high risk–high reward’ side in the caged predator trays). Multivariate analysis of variance (manova) was used to evaluate the effects on life history traits (larval period, mass at metamorphosis, growth rate, survival to stage 42, and mortality rate between forelimb emergence and tail resorption). Separate univariate anovas for each significant effect in manova were conducted after ascertaining the homogeneity of sample variances using the Cochran's C-test (Underwood, 1981). For all the life history traits we used tank mean values to ensure data independence.

To examine the effects of predation risk and food quality on the actual juvenile jumping performance (i.e. regardless of the causal pathways involved) we used anova. In a second step, ancova with juvenile length (SVL) as a covariate was used to evaluate the effects of the predation and food independent of variation in body size. The anova results provide information about the actual consequences for tadpoles of experiencing a particular level of risk or food treatment, whereas ancova may help to understand how the effects of the tested factors are transmitted to jumping performance. ancova was conducted also to evaluate the effects of predation and food type on tadpole energy reserves. As the amount of energy reserves required to face a period of food shortage is dependent on body mass, it is not the absolute but the relative amount of lipids that matters. Therefore the ancova model included dry lean mass as a covariate. Repeatabilities of jumping performance and SVL were estimated using the intraclass correlation coefficient (Sokal & Rohlf, 1981; Lessells & Boag, 1987). Repeatability of size-corrected jumping performance was examined by analysing residuals of multiple regression of jump distances against SVL and temperature.

To compare morphology between predator and food treatments, we log-transformed all morphological measurements and conducted a principal component analysis (PCA) on the covariance matrix of six distances (HW and HL, BL, FL, and RTL and LTL) to derive a measure of general size. All traits loaded strongly and positively (0.67–0.95) on the first axis, revealing PC1 to be a size component (Bookstein, 1991). We then regressed five of these distances (left tibiafibula was excluded because of redundancy with the right side) against PC1 scores and saved the residuals. We calculated the mean trait residuals for each experimental unit to serve as size-corrected measures of juvenile morphology. manova on mean residuals was then used to examine the effects of predators and food type on general juvenile morphology. To assess the assumption that all the treatments had the same allometric relationship, we conducted separate ancovas for each of the five traits, using PC1 as a covariate. ancova revealed that the slopes of log-trait on PC1 were statistically equivalent (HW: F1,3 = 0.05, P = 0.98; HL: F1,3 = 0.14, P = 0.93; BL: F1,3 = 0.04, P = 0.99; FL: F1,3 = 0.90, P = 0.47; tibiafibula length: F1,3 = 0.01, P = 0.99).

Each tibiafibula was measured three times for asymmetry analyses. All the measurements were taken blind. The asymmetry was measured as the signed (R − L) difference and FA was estimated as the absolute value of this measure. For each predator and food treatment, a two-way anova (‘sides’ was a fixed factor and ‘individuals’ was a random factor) conducted on the repeated measures was used to test for the significance of the between-sides differences due to FA relative to the between-sides differences due to measurement error (ME) (Palmer, 1994; Palmer & Strobeck, 2003). In addition this analysis provides a test of the significance of directional asymmetry (DA). The Bonferroni sequential correction (Rice, 1989) was used to adjust significance levels where multiple related tests were conducted. An analysis of asymmetry should ideally include several descriptors of ME. We therefore calculated ME1, ME2 and ME5 (repeatability of FA; for details see Palmer & Strobeck, 2003). As measures of asymmetry we computed FA4 and FA10a (from Palmer & Strobeck, 2003). An advantage of FA10a is that it provides an estimate of FA in which ME has been factored out. ME5 and FA10a were computed from the two-way (sides × individuals), mixed model anova.

As size-dependent effects can greatly complicate the interpretation of FA variation among samples, and dogmatic correction for body size variation may yield apparent differences in developmental instability where none exist (Palmer & Strobeck, 2003), we conducted regression analyses to test for positive size dependence before applying any size-correction. Within-treatment regressions of absolute (unsigned) FA against mean tibiafibula length indicated no association between trait asymmetry and trait size; all the relationships were negative (−0.284 < r < −0.113) and nonsignificant (all F1,33 < 2.88; P > 0.099). Differences in FA among treatments were tested by two-way anova (predation × food) of |R − L| values and Levene's test for heterogeneity of variances (Palmer & Strobeck, 2003).


Larval predator avoidance

Predation risk greatly reduced the number of tadpoles in the R + R side of the containers (Appendix S1b: Table B1; Fig. 2). In presence of caged predators, the proportion of tadpoles in the area containing the food source was far below the expected at random (on average, 28.9% and 39.6% for high- and low-energy food, respectively). This contrasted with large proportions in the food side in the no-predator containers (56.8% and 68.7% for high- and low-energy food, respectively). Repeated measures anova revealed significant interactions between predation risk and time, indicating that temporal variation in the use of space and foraging activity differed among treatment groups. Groups fed different diets differed also in their propensity to occupy the two halves of the container, although the effect of food type was less strong than the predator effect (Appendix S1b: Table B1). As expected, the number of tadpoles in the feeding area was greater in the low-energy diet group than the high-energy group (Fig. 2). There was no significant interaction between predation risk and food quality.

Figure 2.

 Spatial distribution (proportion of tadpoles in the ‘food’ side) of Discoglossus galganoi larvae reared in the presence or absence of caged predators and fed two different diets. Data are mean values of five replicate tanks checked over six consecutive days.

Larval life history

The multivariate analysis of five life history traits revealed significant main effects of predator treatment and food quality, and an interaction effect (Appendix S1b: Table B2). Separate univariate tests indicated that survival was not affected by the presence of predators or food type (Appendix S1b: Table B3; Fig. 3). Survivorship during the larval phase was high, ranging between 83.3% and 94.4%. However, both the predation treatment and food quality had a strong effect on the larval period (Appendix S1b: Table B3). In the presence of predators, D. galganoi tadpoles metamorphosed at an older age than tadpoles in the no-predator tanks (Fig. 3). Moreover, HE tadpoles metamorphosed earlier than LE tadpoles. There was no significant interaction between predation risk and food quality. These effects were clearly additive: tadpoles reared with predators and fed a low-energy diet had the longest larval periods, whereas those reared without predators and fed the high-energy diet had the shortest (Fig. 3).

Figure 3.

 Spatial avoidance, survival, growth rate, larval period, and final body mass (mean ± 1SE) of Discoglossus galganoi under four larval environments. Two levels of diet (high or low energy) were crossed with two predator treatments labelled as P (predator) and NP (no predator).

Univariate anovas revealed significant main effects of predation, food, and the interaction, for size at metamorphosis and growth rate (Fig. 3). Separate anovas for each type of food indicated that predation affected mass (F1,8 = 59.24, P = 0.00006) and growth rates (F1,8 =119.83, P < 0.00001) under a high-energy diet, but not with the low-energy food (mass: P = 0.17; growth rate: P = 0.29).

Juvenile performance: jumping and energy reserves

Jumping performance and SVL were highly repeatable within individuals (Appendix S1b: Table B4). Jumping performance was not affected by predation risk (anova; F1,88 = 0.05, P = 0.822), food type (F1,88 = 0.24, P = 0.628), or the interaction (F1,88 = 0.01, P = 0.967). Maximum jump distance was positively correlated with SVL (r = 0.206, P = 0.049, n = 92) and air temperature (r = 0.246, P = 0.018, n = 92). However, there were no significant effects of predation (ancova; F1,87 = 0.44, P = 0.511), food type (F1,87 = 1.41, P = 0.238), or the interaction (F1,87 = 0.35, P = 0.557) on size-adjusted jumping ability. The inclusion of air temperature as a second covariate did not change these results (ancova; all P > 0.47; Fig. 4). Intraclass correlation coefficients were highly significant and similar to those for unadjusted data, suggesting a minor role of body size repeatability in determining the repeatability of unadjusted jump distances.

Figure 4.

 Maximum jumping distances and lipid content of Discoglossus galganoi juveniles from different larval environments with regard to predation risk (P, caged predators; NP, without predators) and food type (high- or low-energy content). Values are least-squares mean values (±1SE), adjusted for either mean snout-to-vent length (jumping) or mean lean mass (lipids).

There were significant effects of predators (anova; F1,16 = 16.57, P = 0.0009), food quality (F1,16 = 137.98, P < 0.00001), and the interaction (F1,16 = 17.19, P = 0.0008) on total dry mass of juvenile D. galganoi. The predator treatment had a strong influence on dry masses of HE frogs [ln(total dry mass), mean ± 1SE; predator: 10.32 ± 0.04; no-predator: 10.62 ± 0.04; anova, P < 0.00001], but the mean values across the LE groups were nearly identical (10.04 ± 0.03 and 10.03 ± 0.03, for predator and no-predator frogs, respectively; F1,8 = 0.01, P = 0.96). The amount (unadjusted) of lipids was affected by food type (F1,16 = 80.11, P < 0.00001), but not predator treatment (F1,16 = 1.09, P = 0.311) or the interaction (F1,16 = 2.17, P = 0.16). These results suggest a strong influence of body size, so we conducted an ancova of lipid mass with lean mass as a covariate. The data met the assumptions of homogeneity of deviations from individual regressions (Cochrans’C(4,4) = 0.39, P > 0.60) and homogeneity of regression slopes (F3,12 = 1.55, P = 0.25). ancova confirmed that relative lipid content was not affected by predators (F1,15 = 2.99, P = 0.11), food composition (F1,15 = 0.18, P = 0.67) or the interaction (F1,15 = 1.83, P = 0.19; Fig. 4).

Juvenile morphology

The PCA analysis of six morphological traits yielded a size component (PC1) and two components that represented leg and head shape. manova revealed a significant interaction of food type and predator treatment (Wilks’F5,12 = 3.67, P = 0.030). There was no overall effect of food type on size-corrected morphological traits (Wilks’F5,12 = 0.90, P = 0.511). Univariate anovas on size-corrected traits confirmed that metamorphs emerging from predator environments had short legs (tibiafibula) relative to those emerging from the no-predator treatment (Appendix S1b: Table B5; Fig. 5). They also tended to have longer bodies and heads (Fig. 5). The predator effect for these traits was close to significance, and there was a significant effect on total (SVL) length (F1,16 = 4.95, P = 0.041). There was also a significant interaction of food type and predator presence for residual HL. Separate tests for each food type revealed a significant effect of predators only when tadpoles were fed a low-energy diet (F1,8 = 7.23, P = 0.028).

Figure 5.

 Relative morphology of Discoglossus galganoi juveniles from different larval environments with regard to predation risk (P, reared in presence of caged predators; NP, reared without predators) and food type (high- or low-energy and protein content). All traits are residuals after regression of the log-transformed dimensions against an overall measure of size (PC1 scores; see Methods).

Juvenile fluctuating asymmetry

The measures of tibiafibula length were highly repeatable (Appendix S1b: Table B4). Moreover, repeatabilities of size-corrected traits were still highly significant, although somewhat lower. A two-way anova (predation risk and food type) of the signed asymmetry (R − L) revealed a significant predation × food interaction (F1,136 = 5.86, P = 0.017) suggesting differences in the degree of DA among treatments. In fact, separate two-way anovas (sides × individuals) showed significant DA (left side greater than right side) for the P-HE and the NP-LE groups (Appendix S1b: Table B6; P-values significant after Bonferroni correction). For all the four groups, the between-sides variation (FA) was significantly greater than the expected due to ME (sides × individuals: all P < 0.000001). The error variance was low for all but except the NP-LE group (Table 1). Across-treatment variation in the repeatability of FA (ME5) and ME1 paralleled this pattern. The repeatabilities of FA were high (74–83%) except for NP-LE frogs (58%). Finally, the difference between FA4a (ME included) and FA10a (ME excluded), which stands for the contribution of ME to FA, was roughly the same in the four groups (Table 1). An anova of the unsigned differences between sides (|R − L|) did not reveal significant effects of predation (F1,136 = 0.21, P = 0.65), food (F1,136 = 1.63, P = 0.20), or the interaction of these (F1,136 = 0.92, P = 0.34) on FA. A Levene's test for heterogeneity of variances confirmed that there were no significant differences in FA among treatment groups (F3,136 = 0.92, P = 0.43).

Table 1.   Descriptors of fluctuating asymmetry and measurement error derived from the results of a mixed model (sides × individuals) anova of tibiafibula length, for different predation risk and food treatments. FA10a and FA4a: descriptors of FA. ME1 reports measurement error in the original units, and ME3 represents the mean difference between replicate measurements as a proportion of mean difference between sides (mean squares of the sides × individuals interaction). ME5 expresses FA variation as a percent of the total between-sides variation (including ME), and provides a standardized measure of FA repeatability (see Palmer & Strobeck, 2003).
GroupFA10a (mm)FA4a (mm)ME1 (mm)ME1 (% of FA4a)ME3 (% of MSS×I)ME5 (repeatability)
 High energy0.12190.15760.050532.0310.260.745
 Low energy0.15140.19380.056429.118.470.783
No predator
 High energy0.17240.21850.056425.826.670.824
 Low energy0.10300.14050.061943.9919.350.581


Our study found strong evidence for direct consequences of predation risk on the life history of larval anurans. The nonlethal presence of predators and, more specifically, its interaction with the potential for energy acquisition per unit time (evaluated as the amount of energy and protein in the diet) had unambiguous effects on growth and developmental rates of D. galganoi larvae, and these effects were expressed as differences in size and age at the time of metamorphosis. Moreover, the observed changes in growth rates were mediated by an intense behavioural response (hypothesis 1: spatial avoidance of predators). Exposure to predators during the larval phase also altered the morphology and the size of energy reserves of juvenile D. galganoi, but this had no apparent effects on jumping performance. We found no evidence of a clear pattern of variation in developmental instability associated with the conditions experienced during the larval development, although there was a significant FA for all the experimental groups. Most important, the effects of predation risk were not independent of the energetic value of food. These results are important for delineating a scenario where the consequences of predation risk for anuran life history depend upon the expected rate of energy acquisition during the larval phase.

Effects on developmental rate and size at metamorphosis

A number of anuran species have been found to alter the length of the larval period in response to predation risk (Laurila et al., 1998; Altwegg, 2002a,b; Van Buskirk, 2002; Boone & Semlitsch, 2003). In most cases predation risk induces a decrease in tadpole activity which results in longer larval periods and smaller size at metamorphosis relative to control individuals (Lardner, 2000; Nicieza, 2000; Van Buskirk, 2002). This suggests a cascade of effects from behaviour, growth rate and size to age at metamorphosis.

In this study, the perception of predation risk caused longer larval periods of D. galganoi. In addition, the patterns for developmental rates and spatial avoidance of predators were very alike, which supports a cause–effect relationship between the behavioural response and developmental rate (hypothesis 3a). Tadpoles fed a high-energy food showed, on average, the strongest response of spatial avoidance (hypothesis 2a). This supports the idea of a ‘pure’ trade-off between energy acquisition and risk avoidance, as opposed to our prediction that the trade-off itself changes as a function of the energy intake per unit time (hypothesis 2b). Moreover, hypothesis 2a conforms well to previous empirical and theoretical work suggesting a prominent role of time constraints in the expression of predator-induced life history shifts and predator-avoidance behaviour (Werner & Anholt, 1993; Altwegg, 2002a).

Tadpoles exposed to predators and fed a high-energy diet were older when they reached metamorphosis and had a smaller size than nonpredator tadpoles, whereas in the low-quality environment tadpoles exposed to predators reached metamorphosis at an even older age but a similar size to the nonexposed tadpoles. These data did not support either hypothesis 3b (smaller size without change in larval period) or hypothesis 3c (larger size). The observation of the longest larval periods for tadpoles exposed to predators and low-energy food, and the absence of size differences between predator and nonpredator tadpoles from the low-energy treatments may indicate the need to exceed a minimum size for development to proceed (Wilbur & Collins, 1973). In addition, energy constraints could reduce the negative response of tadpoles to predators or, alternatively, the positive response to the release of predation pressure. In general, the effects of predation risk are weaker under conditions of high-competition or when food resources are scarce (Laurila et al., 1998; Relyea & Hoverman, 2003).

Here we used the term ‘energy constraint’ as synonymous to the amount of energy that can be used per unit time. Energy and time constraints can set an upper boundary to the total amount of energy that will be consumed prior to some endpoint in development. Therefore, different factors decreasing the expectancy of energy acquisition within a given developmental stage can have similar effects on the predator-induced life history responses. In our experimental scenario, availability of a high-energy food implies a reduction of the time spent in foraging to achieve a given size, and therefore it is equivalent to a release of time constraints. This may explain both the observed pattern of spatial avoidance of predators and the differences in final body size.

A high-protein diet can be important in oligotrophic pools or with very high tadpole densities. This can be the case for D. galganoi; this species often breeds in ephemeral rain pools with very low algal production and where initial (post-hatching) larval densities can be high (in the study area above 500–800 ind./m−2; A.G. Nicieza, personal observation). In these conditions, rapid development under a carnivorous diet can result in a reduction of the mortality associated with pool desiccation.

The prediction that tadpoles should react to predators by accelerating development (hypothesis 4a), or in any case by shortening the larval period through developmental truncation (hypothesis 4b) was not upheld in D. galganoi. This suggests a lack of a direct control on the onset of metamorphosis associated with predation risk. However, direct control on the timing of transitions between life history stages is not exceptional in vertebrates, and the presence of egg and larval predators has been shown to alter hatching time in fishes and amphibians (e.g. Sih & Moore, 1993; Warkentin, 2000; Jones et al., 2003; E. Capellán & A.G. Nicieza, unpublished). The fact that embryos are able to alter their developmental programme in response to predator presence whereas this control is absent in more developed stages may be paradoxical. However, differences in hatching time induced by predators are in a scale of hours (Laurila et al., 2002; E. Capellan & A.G. Nicieza, unpublished), whereas predator-induced differences in age at metamorphosis are often greater than 1 week (Nicieza, 2000; Relyea, 2001b; this study). In addition, hatching represents the endpoint of a differentiation phase; apart from egg-size variation, during the embryonic phase there is very little scope for hatchling-size variation. Therefore, it is unlikely that a difference of hours in the time to hatch can affect size at metamorphosis and subsequent juvenile performance. In contrast, the larval period involves both growth and differentiation; a shortening of this phase to reduce the risk of predation would require either a significant reduction of juvenile size, or an increase of foraging rates and thereby a greater ‘instantaneous’ risk of predation.

Consequences for juvenile performance

Predation risk has been shown to influence the relative size of body and hindlimb parts (Relyea, 2001b; Van Buskirk & Saxer, 2001; this study). Emerson (1986) outlined the idea that many of the differences in anuran morphology observed at different hierarchical levels of genetic diversity (among closely related species, among genotypes, or within genotypes among environments) may result from shifts in developmental timing or growth rate. Emerson pointed out that if there were a genetic correlation between hindlimb length and these life history traits, small differences in hindlimb length might reflect selection for the age at metamorphosis or larval growth rate (Emerson, 1986). In line with this, relative hindlimb length has been shown to be positively correlated with growth rate and negatively correlated with time to metamorphosis in several anuran species, regardless of whether the proximate factor causing extended larval periods is high density (Emerson, 1986; Tejedo et al., 2000) or low temperature (Blouin & Brown, 2000).

Anuran larvae exposed to predators usually grow slowly and prolong the larval period compared with nonexposed larvae. Therefore, we should expect shorter relative hindlimbs in predator environments. Larvae exposed to predators become short-legged juvenile in Rana ridibunda (Van Buskirk & Saxer, 2001) and D. galganoi (this study), thus supporting that prediction. All these studies reinforce the idea that predation risk during the larval phase can affect juvenile shape indirectly by altering the extension of the larval period.

Although the consequences for juvenile performance and survival of predator-induced changes in metamorphic morphology are not clear yet, there is growing evidence that differences in initial juvenile shape have negligible effects on juvenile growth and jumping performance (Tejedo et al., 2000; Van Buskirk & Saxer, 2001). This suggests that these changes in initial juvenile shape are not adaptive, but a byproduct of extended larval periods. Why should shape differences associated with variation in the larval period have no apparent effect on juvenile performance? First, these differences are probably too small (e.g. lower than 4% in relative or absolute hindlimb length: Relyea, 2001b; Van Buskirk & Saxer, 2001; this study) to transmit any appreciable effect. According to Emerson (1978), differences in hindlimb length must exceed roughly 10% to have significant consequences on jumping performance. Secondly, compensatory responses can minimize small initial differences in shape (Emerson, 1986) or body size (Nicieza & Metcalfe, 1997) during the first growing season.

Energy reserves can be decisive for metamorph survival before foraging abilities are fully developed during the first days in a novel environment (Pfennig, 1992). Therefore, our finding that the relative size of lipidic reserves of juvenile D. galganoi was unaffected by predators does not support the idea that predator-induced changes in larval behaviour and development could have a delayed cost. Moreover, small differences in lipid stores are unlikely to have any effect on juvenile growth and survival, as small losses of energy reserves can be recovered during short periods of unconstrained feeding (Álvarez & Nicieza, 2005).

An important question is whether the stress imposed by predators on larvae is intense enough to alter the developmental process in such a way that its consequences remain after the metamorphosis. As differentiation is largely incomplete at stage 26, stress associated with predators could have increased FA (Leary & Allendorf, 1989). Our data indicated a significant level of FA in this species, but this was not affected by predation risk or food quality. This may suggest that either the level of stress inflicted by predators or a poor quality diet was too low, or the trait used to measure FA (tibiafibula size) was inadequate to evaluate developmental stability.

We found strong DA in two of the four experimental environments (P-HE and NP-LE). Palmer & Strobeck (1992) argued that DA is not a reliable indicator of developmental instability because it is impossible to separate the asymmetry caused by this source from that which has a genetic basis. Nevertheless, DA has been induced in Drosophila melanogaster exposed to high concentrations of benzene (Graham et al., 1993). In the present study, the genetic composition of the experimental groups should be homogeneous, and therefore the observed DA seems to have a major environmental basis.


This study indicates that phenotypic modifications induced by predators on anuran larvae may have little effect on juvenile performance. This supports the Ebenman (1990) view of metamorphosis as a vehicle for disrupting the genetic correlation structure between larval and adult traits, thus permitting relatively independent evolution in contrasting environments. Alternatively, the range of plasticity can be evolutionarily limited by downstream effects on performance; if larval phenotypic plasticity has a strong negative effect on juvenile performance, then selection acting during the juvenile period could eliminate the larval plasticity responsible for causing the decreased locomotor performance. An important conclusion is that the net effect of predators on growth rates and size at metamorphosis depends upon the maximum possible rate of energy acquisition, ultimately dictated by food quality. By focusing on these interactions we will gain new insights into the evolution of predator-induced plasticity. Although changes in the rates of growth and differentiation induced by larval predators influence juvenile morphology, these have no effect on size-adjusted performance. Because it is the absolute performance that can influence survival and growth, larval predator avoidance can still impose significant costs at the terrestrial stage.


This research was supported by grants DGES-PB96-0861 from the Spanish Ministerio de Educación y Ciencia (MEC), and MCYT REN2001-2647/GLO from Ministerio de Ciencia y Tecnología (MCYT) and FEDER funds. Kate Arnold, Anssi Laurila, Miguel Tejedo and an anonymous reviewer provided valuable comments on an earlier draft of the manuscript.