Introduction

Behaviour plays a decisive role in shaping the outcome of predator–prey interactions (Lima and Dill 1990; Lima 1998; Ferrari et al. 2010). Predation is an important selective force acting on the behaviour of prey species and, in order to minimise predation risk, many organisms have evolved a variety of predator-avoidance behavioural defences (reviews in Lima and Dill 1990; Kats and Dill 1998). A decrease in activity level is one of the most common and effective behavioural antipredator responses that reduces vulnerability to predation (Lima and Dill 1990; Kats and Dill 1998). Spatial avoidance of predators also acts as an antipredator defence by reducing the rate of predator–prey encounters and, consequently, predation risk (Laurila et al. 1997; Relyea 2001; Nicieza et al. 2006). However, these behavioural shifts often incur costs to animals due to reduced resource acquisition, which can alter growth, development and life-history patterns (Werner and Anholt 1996; Lima 1998; but see Steiner 2007). Therefore, there should be strong selection on prey to recognise dangerous predators, to accurately determine predation risk and to adjust antipredator responses accordingly.

Invasive predators are a worldwide threat to biodiversity (Blackburn et al. 2010). Freshwater ecosystems are amongst the most invaded and are particularly vulnerable to introduced predators (Lodge et al. 1998; Cox and Lima 2006). For instance, many amphibian population declines have been associated with the introduction of exotic aquatic predators (review in Kats and Ferrer 2003). When predators invade areas outside their historical geographic range, native prey species are likely to suffer heavy predation due to the lack of a common evolutionary history with the exotic predator (Cox and Lima 2006; Gall and Mathis 2010). This is because the evolutionary naïveté of prey may either cause a failure to recognise and respond to novel predation threats, or result in inappropriate or ineffective antipredator defences (Cox and Lima 2006; Strauss et al. 2006; Sih et al. 2010). The degree of naïveté can depend on the similarity of the invader to native predators, since phylogenetic relatedness is often associated with ecological similarity (Cox and Lima 2006; Sih et al. 2010). For instance, distantly related aquatic predator species are likely to have dissimilar chemical signatures. Hence, a novel predator with no related species in the native community may pose a higher threat, as it is less likely to be recognised as dangerous by native prey (Strauss et al. 2006; Gall and Mathis 2010; Sih et al. 2010). Several studies have reported a lack of native prey responses to aquatic invasive predators, which may cause profound changes in the invaded ecosystems (e.g. Gamradt and Kats 1996; Knapp 2005). A classic example is the introduction of the Nile perch (Lates niloticus) in Lake Victoria, which caused severe declines in native cichlids, probably due to lack of predator recognition (Witte et al. 1992).

In aquatic ecosystems, predator chemical cues are particularly important for prey in assessing predation risk (Kats and Dill 1998). The chemicals to which prey respond may be predator-specific odours, cues that are actively or passively released by injured or consumed conspecifics or, more frequently, a combination of both (Chivers and Smith 1998; Schoeppner and Relyea 2009; Fraker et al. 2009; Hettyey et al. 2010). Several studies have shown that fed predators commonly elicit strong antipredator defences, while starved predators often do not (Stirling 1995; Slusarczk 1999; Schoeppner and Relyea 2005, 2009; but see Petranka and Hayes 1998; Van Buskirk and Arioli 2002). As a predator may become chemically ‘labelled’ by its diet, recognition of a novel predator can be facilitated if diet chemical cues are associated with it (reviewed in Ferrari et al. 2010). However, studies on prey responses to invasive predators generally do not consider the role of chemical cues originating from consumed conspecifics and their potential importance for enabling predator recognition (but see Marquis et al. 2004).

Freshwater crayfishes, which have been widely introduced outside their native range, can have major impacts on native prey species and cause severe population declines (Hobbs et al. 1989; Lodge et al. 1998; Cox and Lima 2006; Larson and Olden 2010). In the southwest Iberian Peninsula, most freshwater habitats have been invaded by Procambarus clarkii (red swamp crayfish), an exotic crayfish endemic to northeastern Mexico and south-central USA (Hobbs et al. 1989; Gherardi 2006). This crayfish was introduced in Spain in 1973, and by the 1990s it was already abundant in southwestern Portugal, an area previously devoid of freshwater crayfishes or functionally similar species (Habsburgo-Lorena 1983; Almaça 1991). This area holds a remarkable endemic anuran community; out of nine species, three are endemic to the Iberian Peninsula and three others have a restricted distribution outside Portugal and Spain (Gasc et al. 1997). P. clarkii readily preys upon eggs and larvae of all these anurans (Cruz and Rebelo 2005). Due to its high fecundity and rapid growth, it can quickly build extremely large populations, especially in fish-free habitats, and become a serious threat to amphibian populations (Gherardi 2006; Larson and Olden 2010).

The ability of anuran tadpoles to detect and respond to P. clarkii is largely unknown, despite the importance of this information for understanding the extent to which native amphibians are able to cope with this invasive predator. In this study, our main goal was to understand if anuran tadpoles from a community in the southwest Iberian Peninsula that was invaded by P. clarkii approximately 25 years ago are able to exhibit behavioural defences in the presence of chemical cues from this crayfish. For this, we performed a laboratory experiment in which we assessed changes in tadpole activity level and spatial avoidance of the predator in the presence of chemical cues from the invasive crayfish, and compared them with those elicited in the presence of a common native predator, the larval dragonfly Aeshna sp. We also investigated the role of chemical cues from consumed conspecifics by comparing behavioural responses in the presence of starved and conspecific-fed predators. A previous study has shown that many of our study species may change their behaviour under direct predation risk by P. clarkii (Cruz and Rebelo 2005), but no study has compared these responses to those elicited in the presence of a native predator, nor investigated the role of predator chemical cues or cues from predated conspecifics. Finally, we evaluated how these behavioural responses change over larval development.

The anuran community in southwestern Portugal consists of nine species: Iberian water frog (Pelophylax perezi), Mediterranean tree frog (Hyla meridionalis), European tree frog (Hyla arborea), common toad (Bufo bufo), natterjack toad (Bufo calamita), Iberian parsley frog (Pelodytes ibericus), Western spadefoot toad (Pelobates cultripes), Iberian painted frog (Discoglossus galganoi) and Iberian midwife toad (Alytes cisternasii). Except the two Bufo species, these species (or very closely related species) show behavioural plasticity in the presence of caged aeshnid dragonflies (Van Buskirk 2002; Nicieza et al. 2006; Richter-Boix et al. 2007; Gomez-Mestre and Díaz-Paniagua 2011), and we expect them to show behavioural responses in the presence of this native predator. Behavioural responses to aeshnids have not been studied for P. cultripes; we predict that behavioural defences may be unnecessary, at least late in development, because the large body size of these tadpoles acts as an important antipredator mechanism (Tejedo 1993). Concerning responses to the exotic crayfish, we predict most of the species to not detect and respond to this novel predator; however, chemical cues from consumed conspecifics may elicit responses. Cruz and Rebelo (2005) have shown that these anuran species differ in vulnerability to predation by P. clarkii, the two Bufo species, P. cultripes and D. galganoi being the most susceptible species. This may indicate that these species lack appropriate defences towards this crayfish. Finally, we predict antipredator defences to be strongest in the beginning of tadpole development because prey vulnerability and responses to predators decrease as prey size increases (e.g. Eklöv and Werner 2000; Hettyey et al. 2010).

Materials and methods

Study species and its maintenance

This study was conducted in the surroundings of the Sado River basin (southwest Portugal), an area that was colonised by P. clarkii in the 1980s (R. Rebelo pers. obs.). In this area, P. clarkii can be found in all types of water bodies (e.g. streams, temporary ponds), including small and shallow ones (Cruz and Rebelo 2007). Aeshna larvae are voracious native predators of tadpoles, widely abundant in water bodies throughout this area. All anuran species included in this study co-occur with both these predators.

The experiment took place in the laboratory of the Centro de Biologia Ambiental field station (Grândola, southwest Portugal, 38°06.482′N, 8°34.140′W), from 15th December 2007 to 15th November 2008. This extended experimental period was due to differences in breeding phenology of the study species. Several egg masses of each species were collected from ponds located in the Alentejo region, southwest Portugal (Table 1). In A. cisternasii, males show parental care until tadpoles become free-swimming, and the tadpoles used in the experiment had been very recently released in the ponds. All the ponds had established crayfish populations, with evidence of crayfish reproduction (juveniles or brooding females). Clutches were kept in several species-specific 5 L aquaria filled with spring water from the field station, until tadpoles reached Gosner stage 25 (operculum closure over gills; Gosner 1960). Larvae were fed commercial fish food and boiled lettuce ad libitum every 2 days. Throughout the experiment, water was changed every 5 days. Water temperature was 18.0 ± 0.17 °C (mean ± SE) and the photoperiod was 12L:12D.

Table 1 Number of clutches collected, collection site (type of water body and coordinates), date of their collection and dates of start and end of the experiment for each of the nine studied species
Full size table

Predators were collected from local streams or ponds. Adult crayfish were captured using baited funnel traps and late instar dragonfly larvae using dip-nets. Predators were transferred to the laboratory and kept either in 40 L aquaria (crayfish) or individually in 1.2 L plastic boxes (dragonflies). We fed crayfish with commercial fish food and small invertebrates, and dragonflies with tadpoles and Ephemeroptera larvae.

Experimental setup

We performed a factorial experiment using the nine anuran species and five predator treatments, each combination being replicated once in each of five spatial blocks. This resulted in a total of 225 experimental units, each consisting of a plastic tank (39 × 28 × 28 cm) filled with 10 L of water. At the start of the experiment (day 0), ten tadpoles at developmental stage 25 and a predator cage were added to each tank. The five predator treatments were: fed crayfish, fed dragonfly, starved crayfish, starved dragonfly and control (no predator). We used both fed and starved predator treatments in order to investigate whether behavioural responses towards the exotic predator are facilitated by the presence of chemical cues emitted by consumed conspecifics. The predators were placed in the cages 1 day after the tadpoles. A single predator was placed in a cylindrical opaque cage, tethered at the surface in the middle of the container. The sides of the cage were covered with fine mesh netting, allowing chemical cues to flow out of the cage. Only chemical predator cues (no visual or tactile cues) were available for the tadpoles. Crayfish and dragonfly cages had a diameter of 85 and 62 mm, respectively. In the control treatments, we randomly used empty predator cages from one of the two sizes. While inside the predator cage in each experimental container, predators from the ‘fed’ treatments were fed three focal species tadpoles every other day. Starved predators were not fed during the experiment. Prior to entering the experiment, all predators were starved for at least 48 hours. All the predators were replaced every 2 weeks so that the starved predators could be fed.

Response variables and statistical analyses

We recorded tadpole behaviour (activity level and spatial avoidance) at the beginning (time period 1), middle (time period 2) and close to the end (time period 3) of larval development. As the length of larval period differs among different species, the dates of behavioural observations—as well as head/body length and Gosner developmental stage (Gosner 1960) at each date—also varied (Table 2). On each observation day, tadpole activity and spatial avoidance were estimated by counting the number of active tadpoles and the number of tadpoles close to the predator cage in each tank at five repeated occasions between 8 a.m. and 12 p.m., separated by at least 30 minutes. A tadpole was considered active when it was actively swimming (either slowly or with fast bursts of speed), feeding (even if not substantially altering position) or simply undulating its tail (without actively swimming). Tadpoles were considered to be close to the predator cage if they were either in physical contact with the cage or not more than 1 cm away from it.

Table 2 Days of the experiment, Gosner developmental stage (according to Gosner 1960) and head/body length when behaviour was sampled in the three time periods (early, middle and late in development) for each species
Full size table

When making comparisons across taxa, there is a need to examine whether the focal phenotypic traits are correlated with the phylogenetic history of the focal species. Since nine different anuran species were used in this study, we followed Abouheif’s (1999) recommendation of testing for phylogenetic trait independence prior to analysing our data. If no significant phylogenetic autocorrelation is detected among species (failure to reject the null hypothesis of phylogenetic independence), phylogenetically comparative methods do not have to be used and conventional statistical analyses can be applied (Abouheif 1999). To test for the assumption of phylogenetic independence within our data set, we first calculated the extent of predator-induced behavioural plasticity for each species and predator treatment separately. This was estimated as the proportional change in either activity level or spatial avoidance in each of the four predator environments relative to the no predator environment [e.g. (behaviour in fed dragonfly presence—behaviour in predator absence)/behaviour in predator absence; Richardson (2001); Van Buskirk (2002). Using these values, a Test For Serial Independence (TFSI) was performed for each treatment, using the program “Phylogenetic Independence, Version 2.0” (Abouheif 1999; Reeve and Abouheif 2003). The phylogenetic tree used was constructed based on Duarte et al. (2012) (Online Resource 1). Phylogenetic autocorrelation was calculated in the form of a C statistic and topology of the original data was randomised 10,000 times to estimate the null hypothesis.

Since we did not find a significant phylogenetic autocorrelation for behavioural plasticity either in activity level (C ≤ 0.177, P ≥ 0.22) or spatial avoidance (C ≤ 0.353, P ≥ 0.09) in any of the predator treatments, subsequent statistical analyses were not phylogenetically corrected. To examine differences in species behavioural responses to the exotic and native predators over their development, we used a Generalised Linear Mixed Model (GLMM) with a Binomial error distribution and a logit link function. As behaviour was recorded in three different time periods, the tank (i.e., the experimental container) was considered a random factor in the model to restrict the independence of tanks in the different time observations. For the GLMM, the proportion of active tadpoles or tadpoles close to the predator cage per container, estimated as the number of tadpoles active/close to the predator cage divided by their total number in a tank, was used as a response variable. This value was the average of the five measurements recorded per container in each time period. The analyses were performed at the community level, using all nine species. They were followed by multiple comparisons, in which the responses of each species in each time period were compared. The specific comparisons used were chosen in order to answer the following questions: (1) which species are responding to the native and/or to the invasive predators (control treatment vs. all others); (2) do cues from consumed conspecifics play a role in this (fed vs. starved treatments); and (3) how similar is the response to the two predators (fed dragonfly vs. fed crayfish and starved dragonfly vs. starved crayfish).

In order to understand if species responding more strongly to the native predator also respond more to the exotic predator and if these responses are mediated by cues from consumed conspecifics, Pearson correlations between predator-induced behavioural plasticity in the two starved and in the two fed predator treatments were calculated (see above for calculations of behavioural plasticity). We used data from time period 1, as this was the period when the strongest responses were observed. If species antipredator responses are mainly mediated by consumed conspecifics cues, we expect to find a positive correlation between the two fed predator treatments and no correlation between the two starved treatments. Alternatively, if species require the presence of predator-specific cues to respond, we expect a strong response to the dragonfly and a weak or no response to the crayfish (due to the lack of recognition of crayfish kairomones), resulting in either a negative or no correlation between the two fed and the two starved treatments. To investigate which amount of predator-induced behavioural plasticity in the fed treatments was due to the presence of cues from consumed conspecifics (hereafter plasticity due to CC cues) we estimated, for each predator and species, the proportional change in activity level or spatial avoidance in the fed predator treatment relative to the starved predator treatment [(behaviour in presence of fed predator—behaviour in presence of starved predator)/behaviour in presence of starved predator]. We then performed a correlation between the plasticity due to the CC cues in the presence of dragonfly and crayfish. If all plasticity is attributed to the CC cues, plasticity should be the same in the presence of the two predators and a near-perfect correlation should be observed, since the number of tadpoles fed to the two different predators was equal. All the analyses were performed using IBM SPSS statistics 20.

Results

Time period was an important factor in determining activity level among the nine anuran species (Table 3). In general, species increased their overall activity with time, and D. galganoi was the only species that decreased activity over time (Fig. 1). A significant interaction between time period, species and predator treatment was found, indicating species-specific complex alterations in tadpoles’ behavioural responses to different predator regimes throughout their development (Table 3). The multiple comparisons between different predator treatments performed for each species showed that dragonfly presence (together with predated conspecifics) strongly affected tadpole activity level, with activity decreasing in all species except P. perezi, B. calamita and H. arborea. In most species, this was observed in all time periods (Table 4; Fig. 1). For P. cultripes, an initial response of reduced activity to fed dragonfly later disappeared, while for P. perezi a higher proportion of active tadpoles was observed in the starved dragonfly treatment early in ontogeny (Table 4; Fig. 1). B. bufo was the only species showing a continuous and strong response to both dragonfly (fed and unfed) and crayfish (fed) predators, although this response was stronger in the presence of the native predator (Table 4; Fig. 1). An initial significant response to the fed crayfish and a later response to the starved dragonfly arose in A. cisternasii, while in D. galganoi these responses appeared later in development and were stronger in the fed treatments. Both hylid species were less active in the presence of the crayfish predator late in their development, either when the predator was fed (H. meridionalis) or unfed (H. arborea) (Table 4; Fig. 1). For B. calamita, no differences in activity level were detected between any predator treatments.

Table 3 Generalised Linear Mixed Model for the activity level and spatial avoidance of the nine anuran species when subjected to different predator treatments in the three different time periods
Full size table
Fig. 1
figure 1

Proportion of active tadpoles (mean ± SE) of the nine anuran species a Alytes cisternasii, b Bufo bufo, c Discoglossus galganoi, d Hyla meridionalis, e Hyla arborea, f Pelobates cultripes, g Pelophylax perezi, h Pelodytes ibericus and i Bufo calamita) in presence of five predator treatments during three different time periods. Plots of different species, top to bottom, are organised from the most to the least responsive species. Note different y axis scales in the graphs due to species variable baseline activity levels

Full size image
Table 4 Results of multiple comparisons between predator treatments for tadpole activity level and spatial avoidance after analysis shown in Table 3
Full size table

The overall response across species in predator spatial avoidance did not vary significantly across time periods (Table 3; Fig. 2). Still, we again found a significant interaction between time period, species and predator treatment, indicating among-species differences in antipredator responses through ontogeny (Table 3; Fig. 2). For all species, significant responses of spatial avoidance consisted of moving away from the predator cage. D. galganoi, B. bufo and H. meridionalis strongly avoided the cage with either fed or unfed dragonfly predators, while H. arborea and P. ibericus avoided the predator cage only in the presence of the fed dragonfly (Table 4; Fig. 2). Tadpoles of A. cisternasii showed continuous and strong avoidance behaviour to both the native and the exotic predators (both fed and starved) throughout development. This was the only species significantly avoiding the predator cage in the crayfish treatment (Table 4; Fig. 2). Tadpoles of P. cultripes, P. perezi and B. calamita did not show spatial avoidance of the predator cages in the presence of any predator (Table 4; Fig. 2).

Fig. 2
figure 2

Proportion of tadpoles near the predator cage (mean ± SE) of the nine anuran species in presence of five predator treatments during three different time periods. Note different y axis scales in the graphs

Full size image

No correlation was detected between species-level plasticity in activity level in the starved dragonfly and starved crayfish treatments (r = 0.269, N = 9, P = 0.484; Fig. 3a). However, there was a significant positive correlation between species plasticity in the fed dragonfly and fed crayfish treatments (r = 0.717, N = 9, P = 0.03; Fig. 3b). A positive correlation was also found between the two predator treatments for plasticity in activity attributed to the CC cues (r = 0.699, N = 9, P = 0.036; Fig. 3c), suggesting that the correlation between plasticity in the two fed treatments was due to among-species variation in plasticity to CC cues. Concerning behavioural plasticity in spatial avoidance of the predator, no significant correlations were found between fed (r = 0.317, N = 9, P = 0.406) or starved treatments (r = 0.09, N = 9, P = 0.889), or for plasticity to the CC cues (r = 0.039, N = 9, P = 0.920).

Fig. 3
figure 3

Relationship between behavioural plasticity in activity level of the nine anuran species in the presence of dragonfly and crayfish, in time period 1 (solid black line): a plasticity in presence of the starved predators: for each predator axis data points represent [(activity with starved predator present—activity with predator absent)/activity with predator absent]; b plasticity in presence of the fed predators: for each predator axis data points represent [(activity with fed predator present—activity with predator absent)/activity with predator absent]; c plasticity attributed to the consumed conspecifics chemical cues: for each predator axis data points represent [(activity with fed predator present—activity with starved predator present)/activity with starved predator present]. The dashed line represents the expected relationship if species respond to cues exactly in the same way in the dragonfly and crayfish treatments. Plasticity values below zero indicate a decrease in activity in the presence of predators and positive plasticity values reflect an increase in activity with predators present. Ac stands for Alytes cisternasii, Bb for Bufo bufo, Dg for Discoglossus galganoi, Hm for Hyla meridionalis, Ha for Hyla arborea, Pc for Pelobates cultripes, Pp for Pelophylax perezi, Pi for Pelodytes ibericus and Bc for Bufo calamita

Full size image

Discussion

This study provides the first community-wide assessment of antipredator behaviour of anuran larvae in response to chemical cues of the invasive crayfish P. clarkii, the most cosmopolitan crayfish species in the world (Gherardi 2006). Although the introduction of P. clarkii into our study area is relatively recent, we found that five out of nine prey species responded behaviourally to this predator. As expected, these responses were greatly mediated by chemical cues resulting from predation events.

Behavioural responses to the native and introduced predators

In the presence of dragonfly, five species had a strong behavioural response (A. cisternasii, D. galganoi, B. bufo, H. meridionalis and P. ibericus), three had a moderate response (P. cultripes, H. arborea and P. perezi), while one species had no significant response (B. calamita). In general, activity levels were greatly reduced and strong spatial avoidance of the predator was observed in the presence of the native predator. These are common and usually effective antipredator responses (reviews in Lima and Dill 1990; Kats and Dill 1998), often reported for larval amphibians exposed to odonate predators (e.g. Werner and Anholt 1996; Relyea 2001; Richardson 2001; Nicieza et al. 2006). The strong responses observed here indicate that, as expected, most of these anuran species do have the potential of exhibiting behavioural strategies to avoid predation. The significant behavioural response of B. bufo to the dragonfly was somewhat unexpected, since most studies report a lack of behavioural antipredator responses in this species (e.g. Laurila et al. 1997; Richter-Boix et al. 2007). However, although bufonids are known to rely on chemical toxic deterrents as defences against predators, several invertebrate predators (including P. clarkii) are resistant to them, which may cause tadpoles to use alternative defences (Semlitsch and Gavasso 1992; Cruz and Rebelo 2005). Laurila et al. (1998) also reported a weak decrease in B. bufo activity in the presence of a dragonfly.

We found that five tadpole species (A. cisternasii, B. bufo, D. galganoi, H. meridionalis and H. arborea) significantly altered activity level and/or showed spatial avoidance in the presence of P. clarkii. Although these responses were generally weaker than the ones elicited towards the native predator, this indicates that the ability to respond to the invasive crayfish is quite pervasive in this amphibian community. These results are consistent with previous studies demonstrating that amphibians alter their behaviour in response to chemical cues of non-native predators, such as P. clarkii (e.g. Pearl et al. 2003; Marquis et al. 2004; Gall and Mathis 2010). Pearl et al. (2003) found different responses in two native anurans to cues from an introduced predator which, together with our study, show that species from the same community may commonly respond differently to exotic predators.

Differences in the ability of prey species to respond behaviourally to the crayfish could have arisen due to several aspects of species ecology, such as their habitat use or body size (and associated swimming abilities). Anuran species face different selective environments along the pond permanency gradient, ranging from permanent lakes and ponds to ephemeral pools (Wellborn et al. 1996). Since permanent water bodies have a higher predator abundance and diversity than ephemeral ones, species using the former habitats (P. perezi, H. arborea, B. bufo and A. cisternasii in our case) are likely to have more generalised (as opposed to predator-specific) antipredator defences that allow responses to a wide array of predators, likely facilitating responses to novel invasive predators (Richter-Boix et al. 2007; Smith and Awan 2009; Gómez and Kehr 2011). Indeed, in this study, three of the four species typically inhabiting permanent water bodies responded to the exotic predator (A. cisternasii, B. bufo and H. arborea). Nonetheless, P. perezi, also a permanent pond species, did not respond to the crayfish, indicating that the hydroperiod gradient does not explain all the differences in species antipredator responses. Alternatively, prey that have small body size and relatively weak swimming ability, such as the bufonids and D. galganoi, should elicit antipredator defences in the presence of many different predators. Two of these species (the exception being B. calamita) indeed responded to the crayfish predator. However, species with larger body size and good swimming ability, such as the hylids, also responded to the exotic crayfish. The evolutionary history of a species can also be important in defining differences in species antipredator responses (Richardson 2001); however, we did not find any phylogenetic autocorrelations in the behavioural responses of these species, indicating that evolutionary history does not seem to play an important role here (see also Richter-Boix et al. 2007). On the whole, since none of the previously mentioned factors entirely explains differences in behavioural responses to the exotic predator, it is possible that all these factors interact to define the final species-specific responses.

The strong behavioural responses by B. bufo and D. galganoi towards P. clarkii are surprising, considering that an earlier study found that these species were the most vulnerable to P. clarkii’s predation (Cruz and Rebelo 2005). This might indicate that in Cruz and Rebelo’s study, tadpoles did not have enough time to develop behavioural defences against the crayfish (it only lasted 48 hours) or, alternatively, that the behavioural responses shown here may be a non-effective response towards this exotic predator. For example, because crayfish rely much more on chemical rather than on visual cues to detect prey, decreased activity in response to crayfish may render inactive tadpoles to a risk equal to that experienced by active tadpoles (Aquiloni et al. 2005).

The importance of chemical cues from consumed conspecifics

In this study, most species significantly changed behaviour in the presence of fed predators, suggesting that cues from consumed conspecifics play a very important role in inducing anuran behavioural responses. However, six out of nine species also altered behaviour when exposed to chemical cues from the starved dragonfly, indicating that predator-specific odours from a native predator are, in many cases, sufficient to launch behavioural defences.

In the three species that responded strongly to the exotic predator (A. cisternasii, B. bufo and D. galganoi), behavioural plasticity (considering activity level) to the fed crayfish was mostly explained by plasticity to the CC cues (cues from consumed conspecifics), indicating that the response to the exotic predator was mediated by cues resulting from predation events. These results are in accordance with those by Marquis et al. (2004), who suggested the presence of conspecific alarm cues to be crucial in chemical detection of predators by B. bufo. Injured tadpoles of this species are known to release alarm cues—chemicals released by specialised cells in the skin—that induce antipredator responses, such as a reduction in activity, in conspecifics (Chivers and Smith 1998; Hagman 2008; Fraker et al. 2009). Responding to cues from injured conspecifics is frequent among anuran species (Chivers and Smith 1998; Hagman 2008; Ferrari et al. 2010) and may also be the case for A. cisternasii and D. galganoi (S. Amaral, M.J. Cruz and R. Rebelo, unpublished data). Responding to broad and general alarm cues seems to be a critical factor in allowing tadpoles to elicit behavioural defences to novel invasive predators, and makes common evolutionary history between predator and prey unnecessary for responses to be elicited (Sih et al. 2010).

A. cisternasii and H. arborea showed a significant behavioural response in the presence of the starved crayfish. This indicates that these species can launch defences towards the exotic crayfish even when only information about the predator identity is present, which likely reflects the detection of a threat. This may indicate a simple reaction to an unknown chemical stimulus, but it can also indicate predator recognition since, given enough time, native species may develop the ability to recognise and respond to cues from novel invasive predators (Strauss et al. 2006). This can be highly beneficial because it would allow species to respond to novel predators even when predation events have not yet occurred (Kats and Dill 1998; Schoeppner and Relyea 2005).

Behavioural responses across ontogeny

As predicted, anuran species showed different antipredator responses to predator chemical cues over ontogeny. All the species responding to the native predator elicited behavioural defences in the first developmental period, indicating an early development of antipredator behaviour, which is in accordance with other studies (e.g. Petranka and Hayes 1998; Laurila et al. 2004). However, not all these species elicited defences in time period 3; at times, the antipredator response disappeared as the tadpoles grew larger. This indicates that, as tadpole size increased, vulnerability to predation probably decreased, reducing the importance of behavioural defences (e.g. Eklöv and Werner 2000; Hettyey et al. 2010). Alternatively, it may also indicate that tadpoles altered their defensive strategies over ontogeny, shifting from behavioural to morphological defences later in development (Relyea 2003). In the case of P. cultripes, overall activity level greatly increased at the same time as the antipredator response disappeared, suggesting that in this case large body size and high activity levels are linked with decreased behavioural responses towards predators. On the other hand, since predation risk did not increase with time in this experiment, we cannot exclude the possibility that, in some of the species, tadpoles might have stopped responding due to habituation (Magurran 1990).

Behavioural plasticity correlations

We found a positive correlation between predator-induced behavioural plasticity in activity level in the presence of the fed native and exotic predators. Since there was no relationship between plasticity in the two starved treatments, the former correlation was probably a result of the correlation between plasticity to the CC cues (cues from consumed conspecifics) in the two predator treatments. This reinforces the idea that the antipredator responses were greatly mediated by chemical signals resulting from predation events upon conspecifics. Further, it suggests that the magnitude of antipredator responses to native predators, when greatly mediated by CC cues, can be a good proxy for predicting the potential antipredator responses to novel predators in amphibians and possibly even in other species. Still, the correlation between plasticity to the CC cues in the two predator treatments was not close to perfect, as expected if all the observed plasticity was due to the CC cues. Instead, species plasticity to the CC cues was stronger in the presence of dragonfly than in the presence of crayfish, indicating that additional information from the predator contributed to an increased plastic response when the dragonfly was present. Once again, there seems to be a difference in species responses depending on the length of coexistence with the predator.

The fact that we found no significant correlations between predator treatments for plasticity in predator spatial avoidance was probably due to many species not altering this trait in predator presence and to the wide variation in responses often observed. This suggests that plasticity in this trait is less widespread among species from this community than plasticity in activity level. Alternatively, it may also indicate that, for some species, chemical cues are not enough to induce alterations in this behavioural trait, maybe because it implies more risk or is more costly.

Conclusions

Since P. clarkii arrived in southwest Portugal in the 1980s, larval anurans have been exposed to high predation pressure imposed by this exotic predator (Cruz and Rebelo 2005). Our experiment showed that five out of nine anuran species present in this area exhibited behavioural responses to this novel predator. We suggest that these species, provided the behavioural responses are effective, will be better able to persist with the continued expansion and establishment of this introduced predator. However, since failure to recognise an exotic species as a predator is probably one of the most damaging forms of prey naiveté (Cox and Lima 2006), the remaining four species—unless eliciting alternative defences—will be highly vulnerable, which may lead to population declines or even local extinctions. Since one of these species, P. ibericus, is an Iberian endemic with a very limited distribution area (Loureiro et al. 2008), serious conservation problems may arise in the near future. Evaluating which species rely on chemical cues from predation events in their responses to predators may be a useful tool for understanding the potential of native prey species to respond to invasive alien predators and help direct conservation efforts to the most susceptible species.