Uncertainty about future predation risk modulates monitoring behavior from refuges in lizards

Behavioral Ecology doi:10.1093/beheco/arq065 Advance Access publication 13 January 2011 Original Article Uncertainty about future predation risk modulates monitoring behavior from refuges in lizards Vicente Polo, a Pilar López, b and José Martín b a Departamento de Biología y Geología, Área de Biodiversidad y Conservación, Escuela Superior de Ciencias Experimentales y Tecnología, Universidad Rey Juan Carlos, Tulipán s/n, 28933 Móstoles, Madrid, Spain and b Departamento de Ecología Evolutiva, Museo Nacional de Ciencias Naturales, Consejo Superior de Investigaciones Científicas Resources invested by prey to acquire information on predator behavior from inside the refuge are crucial to minimize the risk of suffering a future fatal attack. However, most studies have only analyzed situations where information on the predator behavior is unavailable for hiding preys. Also, temporal patterns of risk may affect antipredatory behavior allocation. We simulated in outdoor terraria series of 2 types of predatory attacks (low vs. high risks) with different sequences of risks (predictable series vs. attacks where risk level changed randomly) to Iberian rock lizards. We measured time spent entirely hidden in refuges until appearing near the exit of the refuge (appearance time) and time spent leaning out of the refuge while monitoring the predator (monitoring time) after each predatory interaction. Monitoring time, irrespectively of the temporal pattern of risk, was higher after a single low-risk approach of the predator than after a direct unsuccessful attack. In addition, after multiple repeated interactions with the predator, there was a significant decrease in monitoring time with the sequence order of interactions but only when lizards fled to the refuge after low-risk approaches. Lizards spent more time monitoring the predator from inside the refuge after a low-risk approach, which may be explained because, if there has not been a clear attack, uncertainty on immediate future risk would be greater, and prey may need more time before leaving the refuge to ensure that a predator has not detected the lizard and that it is not ambushing near the refuge. Our study suggests that acquisition of information during and after an attack is important in determining refuge use as an antipredation response. Key words: Iberolacerta cyreni, monitoring behavior, predation risk assessment, refuge use, risk allocation. [Behav Ecol 22:218 223 (2011)] Resources invested by prey in order to acquire information of predator behavior are crucial for survival (Lima 1998). Prey is very often able to adjust their antipredatory responses when they find a predator and respond by hiding into safe refuges (Ydenberg and Dill 1986; Lima and Dill 1990). However, refuge use may involve some physiological or behavioral costs (Koivula et al. 1995; Dill and Fraser 1997; Martín and López 1999a; Polo et al. 2005). Thus, optimal emergence times from the refuge should balance costs and benefits of refuge use (Sih 1992, 1997; Martín and López 1999b, 2001; Polo et al. 2005; Cooper and Frederick 2007). One of the costs of refuge use may be that a hiding prey is often unable to obtain further information on predator behavior. This is important because, although some predators may search elsewhere when a prey has hidden after an unsuccessful attack, other predators may remain in ambush outside the refuge or try to flush prey from the refuge and launch new attacks (Johansson and Englund 1995; Martín and López 2001; Hugie 2003; see also model of successive attacks by Polo et al. 2005). Prey would minimize the risk of suffering a new attack in the immediate future if they were able to acquire accurate information on predator presence from inside the refuge to decide when to emerge from the refuge and resume normal activity (Sih 1992; Cooper 1998, 2008; Martín and López 1999b). Address correspondence to V. Polo. E-mail: vicente.polo@urjc.es. Received 25 February 2009; revised 8 February 2010; accepted 8 February 2010. Ó The Author 2011. Published by Oxford University Press on behalf of the International Society for Behavioral Ecology. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com Most studies of refuge use have only analyzed situations where information on the predator activities after the attack is not available for hiding prey. In these circumstances, current risk of being killed can only be estimated by prey based just on time elapsed since the attack (Sih 1992; Martín and López 1999b; Hugie 2003; Polo et al. 2005) or on predator behavior during the previous attack (Cooper 1997a, 1997b, 1998; Cooper et al. 2003). However, in some occasions, prey may be able to monitor predators from the refuge safety and use this information to decide hiding times (Martín and López 1999b; Cooper 2008). Nevertheless, monitoring predators from the refuge may also have costs, such as revealing the position of the refuge to the predator if prey has to lean out of the refuge to scan for the predator presence. Another key aspect is that antipredatory behavior decisions, and thus the time monitoring predators from the refuge, might also be dependent on previous experiences with predators (i.e., antipredatory decisions in relation to the recent history of risk; see Martín et al. 2009). This is important because prey in nature experience a broad range of temporal patterns of predation risk, and antipredatory behavior should vary with not just the immediate level of risk but also with the precedent temporal pattern of perceived risk (Sih et al. 2000; Hamilton and Heithaus 2001), which includes aspects such as predictability in the temporal pattern of risk (e.g., see Martín et al. 2009) or frequency of predatory attacks (Cooper 1998; Polo et al. 2005). That is, the time that a prey spends monitoring predator behavior from the refuge, or that is waiting for the predator to eventually reappear, may depend not only

Polo et al. Monitoring predation risk in lizards 219 on perceived risk level of the immediate past attack alone ( fixed risk assessment behavior) or on the current level of risk but also on how the patterns of risk level have been in the recent past ( flexible risk assessment behavior; see Martín and López 2004, 2005), the latter being dependent on predictability and/or frequency of attacks in the past. Previous estimations of risk may make prey change their future responses even if risk level does not actually change (see Polo et al. 2005 to show changes in antipredatory decisions of lizards that experienced successive predatory attacks with the same risk level). We used here as a model prey species the Carpetane rock lizard, Iberolacerta cyreni (formerly Lacerta monticola cyreni), a medium-sized lacertid lizard found at high altitude mountains in Central Spain. Lizards of this species select microhabitats with high cover of rocks (Martín and Salvador 1997) and escape from predators by hiding under rocks in screes or in rock crevices from where hiding lizards can later monitor predators (Martín and López 1999b). Previous studies have shown that these lizards are able to modulate their antipredatory behavior and pattern of refuge use as a function of risk level, thermal conditions, and foraging or mating expectations (Carrascal et al. 1992; Martín and López 1999b, 2000, 2001, 2003; Cooper et al. 2003; Martín et al. 2003a, 2003b; Amo et al. 2007). In addition, these lizards can also use flexible antipredatory strategies modifying their refuge use through successive attacks of the same risk level (Martín and López 2004; Polo et al. 2005) and in function of frequency and predictability of the past temporal patterns of risk (Martín et al. 2009). In this study, we examined whether the current level of predation risk and the long-time temporal patterns of predation risk level in the past affect monitoring behavior and risk assessment inside refuge by adult male lizards I. cireni. We simulated in outdoor terraria daily series of repeated predatory attacks of 2 different intensities (i.e., low-risk approach vs. high-risk direct attack), with similar overall numbers of those risk levels across a series but with different temporal patterns of presentation (i.e., predictable series vs. attacks where risk changed randomly; see MATERIALS AND METHODS for details). Then, we analyzed the effects of predation risk level and temporal patterns of risk on the time that the lizard spent leaning out of the refuge monitoring and acquiring information on the predator. We predicted that lizards should hide inside the refuge for the minimum time to ensure that predator moves on. If attacked, then lizards should hide for long enough so that the predator must move on (because its overall foraging gain from waiting would be too low even if successful). If lizards are not attacked, then they should monitor from safety for long enough to ensure that the predator does not escalate from a nonattack to an attack situation. MATERIALS AND METHODS Study animals and experimental procedures During July 2007, we captured by noosing 16 adult male I. cyreni (SVL:X 6 1 standard error [SE] ¼ 74 6 1 mm, body mass: 8.16 0.2 g) in Puerto de Navacerrada (Guadarrama Mountains, Central Spain). We limited this study to adult males, outside of the mating season, to reduce the possible sources of variation due to season, sex, and age differences (e.g., Martín and López 2003; Martín et al. 2003a). Lizards were individually housed at El Ventorrillo Field Station (Navacerrada, Madrid Province), 5 km from the capture site in outdoor opaque plastic cages (80 3 50 3 40 cm) containing rocks for cover. Terraria were placed separately from each other, such that our approaches to a terrarium did not influence lizards in other terraria. In each trial, 1 experimenter, always the same person and wearing similar clothes, approached each terrarium every 10 min, noted if the lizard was hidden entirely inside the refuge, leaning out of the refuge (i.e., the lizard stayed inside the refuge but looked outside with the snout outside or closer than 1 cm to the exit of the refuge), or entirely outside the refuge. Then, the experimenter immediately simulated a predatory attack with 1 of 2 different risk levels (see Martín and López 2004, 2005). In the low-risk situation (L), the experimenter walked slowly near (1 m) but tangentially to the lizard terrarium, looking straight ahead, and without paying attention to the lizard (see risk discrimination of direct vs. tangential approach in Burger and Gochfeld 1990). Lizards could clearly see the experimenter from their terraria, as showed by their alert behavior and, eventually, escape responses. In the high-risk situation (H), the experimenter simulated a direct predatory attack by tapping lizards close to the tail with a paintbrush to stimulate them to run and hide entirely in the refuge. With this procedure, we simulated an attack from an avian predator coming from above the lizard. If the lizard was inside or leaning out of the refuge before the attack, we touched twice the open end of the refuge with the brush simulating that the predator was trying to flush the lizard out from the refuge. We were confident that with this procedure, we simulated 2 risk levels of different intensity because antipredatory responses of lizards were clearly different between situations and similar to those observed in field experiments under similar simulated attacks (Martín and López 1999b, 2004). The same person performed all predatory attacks and then moved to a hidden position between attacks, whereas other immobile and hidden persons recorded the lizard s behavior with binoculars from a vantage point. Each lizard was tested in 2 conditions of predictability of risk levels (sequences of repeated interactions with the predator where risk level was predictable or unpredictable). To simulate this, we maintained the overall number of L approaches and H attacks in each sequence but changed the temporal patterns of distribution of risk levels. In the predictable sequences, we simulated either a H attack every 10 min during 2 h a total of 12 H attacks and then an L approach every 10 min during the following 2 h 12 L approaches ( high-low treatment ; HL) or an L approach every 10 min during 2 h 12 L approaches and then an H attack every 10 min during the following 2 h 12 H attacks ( low-high treatment ; LH). In the unpredictable situation, we simulated 12 L approaches and 12 H attacks in a random order of presentation, one every 10 min during 4 h ( random treatment ; R). Every lizard was tested in 2 replicates of each treatment made in 2 different days (i.e., 6 tests in 6 different days in total), and the order of participation in these treatments was counterbalanced between all individual lizards. The interval between tests was at least 48 h so that the 6 tests were distributed along a minimum of 11 days allowing lizards to rest between tests. From the observations of initial position of lizards taken every 10 min immediately before the new approach or attack, we calculated the average number of times that a lizard was inside, outside, or leaning out of the refuge after the H attacks and after the L approaches in each treatment. These measurements were considered as indexes of general activity level after predatory attacks or approaches. In the H attacks of each treatment, after the lizard was forced to hide, we recorded the time that the lizard spent hidden in the refuge until at least the lizard leant out of the refuge and looked outside with the snout outside or closer than 1 cm to the exit of the refuge ( appearance time ) and the time spent leaning out of the refuge and looking outside because appearance

220 Behavioral Ecology until at least half of the body of the lizard emerged from the refuge ( monitoring time ). We also recorded these times spent in the refuge when lizards hid into the refuge when the predator was passing by but not actually attacking the lizard (i.e., false alarm flight after an L approach). Thereafter, we calculated from all observations the average appearance and monitoring times of each lizard in each treatment and risk level (24 observations per lizard, treatment, and risk level). Average values of each individual were used in subsequent repeated measures analyses (see below). The capture of individuals and the experiments were licensed by the Consejería de Medio Ambiente y Desarrollo de la Comunidad de Madrid, Spain. We followed the guidelines for reptiles research as proposed by the American Society of Ichthyologists and Herpetologists (2004). We made observations of antipredatory behavior of lizards in outdoor conditions during July, from 1200- to 1600-h GMT, when lizards were fully active. All the animals were healthy during the trials, did not show any sign of stress, and at the end of the experiments, 40 days later, they were released at their initial sighting location before capture. All of them had maintained their original body mass through the experiment. Food (mealworms and crickets) dusted with a multivitamin powder and fresh water was provided to lizards ad libitum. Uneaten mealworms and crickets were removed from the terraria. Lizards were housed in their individual terraria for 2 weeks to familiarize them with the novel environment prior to testing, but we minimized contact with the experimenters to avoid habituation to human presence. We placed terraria in an open sunny location, whereas shade was provided by one of the terrarium walls and the refuge (relatively flat stones of similar size and shape under which lizards readily hid). Thus, we allowed lizards to thermoregulate and attain their preferred body temperatures for at least 2 h before the beginning of the trials (Martín and Salvador 1993). Data analyses We used two-way repeated measures analysis of variances (ANOVAs) to evaluate variations of average values per individual in the proportion of times that lizards were leaning out of the refuge before an interaction with the predator and in average time spent hidden into the refuge of individual lizards (appearance and monitoring times) in relation to risk level in each set of approaches/attacks (L vs. H) within each trial and in relation to the temporal distribution of risk treatments (HL vs. LH vs. R) (both within factors). Working this way, we aimed to accomplish powerful designs and prevent pseudoreplication in these analyses. We used all observations of time spent hidden in the refuge (appearance and monitoring times) in a three-way analysis of covariances (ANCOVAs) to examine variations of these variables between individuals (random factor) and between risk level and treatments (fixed factors) in relation to the sequence of L approaches or H attacks (covariant effect). We included the interaction in the models to examine whether differential responses to the 2 levels of risk varied depending on their temporal distribution (Sokal and Rohlf 1995). Our aim with these ANCOVAs analyses was not to evaluate variations in appearance and monitoring times with L or H sequences but just to discard differences between treatments in the variations in appearance and monitoring times in relation to the sequence of L approaches or H attacks (i.e., we tested for nonsignificance in the interaction sequence 3 treatments). Because there was no significant variation due to the sequence of L approaches and H attacks (see RESULTS ), we calculated from all observations of the same individual lizard the average appearance and monitoring times of each lizard in each sequence of L approaches or H attacks. Then, with these individual average values, we calculated the slopes of the relationships between appearance or monitoring times and sequence order (i.e., 2 slope values per each individual lizard). Finally, to prevent pseudoreplication, we used one-way repeated measures ANOVAs to evaluate differences between the slopes of each lizard in appearance and monitoring times between L and H sequences. RESULTS The average proportion of times that lizards were leaning out of the refuge immediately before an approach or attack did not significantly differ either between treatments (0.22 6 0.04 vs. 0.22 6 0.03 vs. 0.23 6 0.04, mean 6 SE of HL vs. LH vs. R treatments, respectively: F 2,30 ¼ 0.69, P ¼ 0.51) nor in relation to risk level in each set of approaches/attacks (0.22 6 0.04 vs. 0.23 6 0.04, L vs. H risk, respectively: F 1,15 ¼ 0.0091, P ¼ 0.93). These nonsignificant differences in risk level were consistent across treatments (interaction term, treatment 3 risk level: F 2,30 ¼ 0.62, P ¼ 0.54). There were no significant differences in average appearance times of lizards near the exit of the refuge between risk levels (390 6 17 s vs. 356 6 14 s, L vs. H risk levels, respectively: F 1,15 ¼ 2.08, P ¼ 0.17) nor between treatments (374 6 30 s vs. 383 6 36 s vs. 361 6 30 s, HL vs. LH vs. R treatments, respectively: F 2,30 ¼ 0.96, P ¼ 0.40). This nonsignificant difference between risk levels was consistent across treatments (interaction term, treatment 3 risk level: F 2,30 ¼ 0.73, P ¼ 0.49). Working with all data, appearance time near the exit of the refuge decreased with the sequence of repeated interactions with the simulated predator (covariate variable, sequence of risk levels: F 1,926 ¼ 49.16, R 2 ¼ 5.05%, beta ¼ 20.22, P, 0.001), being this decrease similar in the 3 treatments (interaction sequence 3 treatment: F 2,924 ¼ 2.39, P ¼ 0.091) and similar after L approaches than after H attacks (interaction sequence 3 risk level: F 1,925 ¼ 0.33, P ¼ 0.57). The slopes of the relationship between appearance time and sequence order of predator interactions were negative and nonsignificantly different between L approaches and H attacks (b ¼ 29.96 6 2.98 vs. b ¼ 28.01 6 3.33 s per increase of sequence order, slopes in L approaches vs. H attacks, respectively: F 1,15 ¼ 0.31, P ¼ 0.59; see Figure 1 for pooled data of L and H). In summary, the decrease in appearance time of lizards with the sequence of interactions was similar after L approaches or H attacks in the 3 treatments. Figure 1 Time spent by lizards inside the refuge until they appeared near the exit of the refuge (appearance time) in relation to the temporal sequence of trials. Data are the average (mean 6 SE) of 32 mean values of individual lizards (2 times per individual in HL, LH, and R treatments in L approaches and H attacks; see MATERIALS AND METHODS ) of each sequence order.

Polo et al. Monitoring predation risk in lizards 221 Figure 2 Time spent by lizards monitoring the predator presence from a secure position near the exit of the refuge (monitoring time) in relation to risk level (L white bars and H black bars) and temporal pattern of risk treatment. Data are mean 6 SE of the average in monitoring time of 16 individual lizards for each risk level and treatment. Individual lizards spent significantly longer average monitoring times, leaning out of the refuge while looking outside, presumably recording information on the predator presence and behavior, after an L approach than after a H direct attack (268 6 20 s vs. 176 6 13 s, L vs. H risk levels, respectively: F 1,15 ¼ 17.65, P, 0.001). There were no significant differences in monitoring times between treatments (242 6 18 s vs. 232 6 28 s vs. 193 6 19 s, HL vs. LH vs. R treatments, respectively: F 2,30 ¼ 1.61, P ¼ 0.21). The significant difference in monitoring time between risk levels was consistent across treatments (interaction term, treatment 3 risk level: F 2,30 ¼ 2.58, P ¼ 0.093; Figure 2). Working with all data, there was a significant decrease in monitoring time with the sequence Figure 3 Time spent monitoring and collecting information of predator presence from a secure position near the exit of the refuge (monitoring time) in relation to the temporal sequence of trials of the same risk level. Data are the average (mean 6 SE) of 16 mean values of individual lizards (times per individual in HL, LH, and R treatments) of each sequence order in L approaches and H attacks (see MATERIALS AND METHODS ). order of interactions with the predator (covariate sequence of risk levels: F 1,816 ¼ 21.38, R 2 ¼ 2.55%, beta ¼ 20.16, P, 0.001) but only after an L approach (interaction sequence 3 level: F 1,815 ¼ 10.19, P ¼ 0.0015; Figure 3), this decrease being similar in the 3 treatments (interaction sequence 3 treatment: F 2,814 ¼ 1.32, P ¼ 0.27). The slopes of the relationships between monitoring time and sequence of predator interactions were negative in L approaches but nonsignificantly different from zero in H attacks (b ¼ 214.23 6 3.12 vs. b ¼ 22.24 6 2.56 s per increase of sequence order, slopes in L approaches vs. H attacks, respectively: F 1,15 ¼ 24.63, P, 0.001). In summary, male lizards spent longer monitoring times after an L approaches than after an H attack in all treatments. However, the times monitoring the predator near the exit of the refuges decreased with the sequence order but only in L approaches and not H attacks. DISCUSSION Predation is a complex phenomenon that involves many different aspects, which require a diversity of specific antipredatory behavioral responses of prey (Endler 1986; Cooper 1999, 2000). Antipredatory responses of lizards, for example, may depend directly on the current risk level of a predator attack (e.g., single emergence times from refuges; Martín and López 1999b) or be affected by the past temporal pattern of risk (repeated emergence times under persistent attacks: Polo et al. 2005 or false alarms flights: Martín et al. 2009). But also, responses may be affected by the uncertainty of immediate future risk (e.g., see Ferrari et al. 2008 as a relevant study on the effect of predictability of risk and risk intensity on the responses of fish to predation threat). Here, we were interested on the antipredatory behavior immediately after a prey had already hidden in a safe place in response to an unsuccessful predator attack or to the mere presence of a nearby predator that had not yet attacked. The hidden prey had to assess whether it is safe to resume normal activity by assessing probability of future attacks after emergence from the refuge. Our results confirm that male I. cyreni lizards can apparently monitor adaptively current predator behavior, while still being safe inside the refuge, thus minimizing the risk of suffering future attacks. We obtained the somewhat counterintuitive finding that male lizards were able to adjust their risk assessment inside refuges by spending more time monitoring the predator after an indirect low-risk approach than after an unsuccessful high-risk direct attack. Besides, responses are also dependent of the past experience with predators. The risk allocation hypothesis (e.g., Lima and Bednekoff 1999) and the threat sensitive predator avoidance hypothesis (Helfman 1989) predict that prey should allocate the highest antipredatory effort after the highest risk situations of relatively short duration. However, our results support the idea that there will likely be greater uncertainty about levels of future risk when current risk is low. This difference would be explained because, if there has not been a clear attack, uncertainty on future risk would be greater, and prey might need more time before leaving the refuge to ensure that a predator has not detected the lizard and that it is not ambushing near the refuge. In addition, predators that make an unsuccessful direct attack should vacate the area quickly, whereas one that has not detected the lizard could take longer time to vacate the area even if not ambushing (see Hugie 2003). As a consequence, when the predator has already directed an intentional attack, risk level is already clearly high, and prey should just need to confirm whether the predator has left the area before going out from the refuge. Whereas, when

222 Behavioral Ecology the predator has been detected but has not attacked, current risk may be apparently low, but prey would require longer times to monitor predator behavior and assess actual risk level (e.g., to ensure that the predator is not ambushing rather than passing by). Therefore, different levels of uncertainty on immediate future risks, rather than just the previous risk level (Martín and López 1999b), seem to modulate monitoring time from the refuge. This result was consistent independently of the predictability of the past temporal pattern of predation risk level used (i.e., random risk stimulus vs. 2 predictable temporal variations in risk stimulus). Our experiment also confirms that lizards can use flexible antipredatory strategies (i.e., differences in antipredatory behavior in response to the same level of risk) when monitoring predators from inside the refuge. Through a series of repeated interactions with the predator, lizards modulated the duration of the successive monitoring times by decreasing progressively the intensity of this effort but only after successive interactions with predators that did not launch direct attacks (see Figure 3). This decrease in successive monitoring times may result from the continuous assessment of the temporal sequence of predatory risk levels. It should be conservative for prey to exhibit high antipredatory responses to the low-risk stimuli but only in the first approaches of predators. Researchers acting as predators disappeared from the lizard field of view immediately after the low-risk approaches. Thus, prey should also use past sequential information on predator behavior after successive low-risk stimuli to save energetic or time resources that could be used in other activities. As a consequence, monitoring times of lizards in the low-risk situation decreased until reaching, in the last approaches, the values observed after high-risk attacks. This agrees with flexible changes observed in other parameters of refuge use in this species in the field (Martín and López 2004) and in the laboratory (Polo et al. 2005). We have observed the same pattern of decrease in the proportion of false alarm fights to the refuge (i.e., the propensity to hide in refuges when the predator is close but not actually attacking) with the sequence order of approaches of predators (Martín et al. 2009). Although responding to all approaches of the predator by fleeing to the refuge and maintaining long monitoring times would minimize potential predation risk, time and energy can be saved, and costs of refuge used can be minimized if lizards responded accurately only to actual predator attacks (Ydenberg and Dill 1986; Martín and López 1999a; Cooper and Frederick 2007). The pattern of temporal distribution of predation risk levels (i.e., random vs. predictable treatments of risk level) had no effect neither on appearance time nor on the proportion of times that lizards were leaning out of the refuge immediately before an approach or attack. Similarly, Ferrari et al. (2008) failed to show an effect of risk predictability on the behavioral responses of cichlids to high-risk alarm cues but predictability did influence responses to low-risk cues. Nevertheless, the pattern of temporal distribution of risks was nearly significant in affecting monitoring time of lizards in our experiment, which tended to allocate higher antipredatory effort in response to more unpredictable risk (Polo et al. 2005; Martín et al. 2009). Our results suggested that both the difference in monitoring time between the 2 simulated risk levels and the decrease in monitoring time with the sequence of repeated low stimulus were marginally higher in the LH treatment than in the HL or R treatments (P ¼ 0.091 for the interaction sequence 3 treatment; see Figure 2). Prey should allocate the highest antipredatory effort after the highest risk situations (Helfman 1989; Lima and Bednekoff 1999; Sih et al. 2000) but also after the more unpredictable temporal distribution of risks (Polo et al. 2005; Martín et al. 2009). However, lizards, as many other animals, are able to compensate for higher predation risk or higher unpredictability in the level of future attacks by adopting conservative strategies and increasing time spent into refuges (Sih et al. 1992; Sih 1997; Cooper 1998; Martín and López 2001; Polo et al. 2005; Cooper and Frederick 2007). This conservative behavior could, however, be explained because overestimation of risk may have milder fitness consequences (Amo et al. 2007) than the cost of underestimating danger, which might result in prey being captured in a new attack (Bouskila and Blumstein 1992; but see Abrams 1994). Thus, this conservative costly strategy may be beneficial for survival when future risk is predictably high but also in the case of unpredictability of future predation risk level. Besides, low intensity of antipredatory behavior (i.e., low monitoring times) could emerge from unacceptable decrease in the level of foraging activities after intense predation pressure (Mirza et al. 2006; but see Van Buskirk et al. 2002). In summary, we have found that lizards modulate the time allocated to monitor predators from the refuge after an unsuccessful attack or approach of the predator. The monitoring time into the refuge was affected by the level of uncertainty on immediate future risk level rather than just based on risk level of the immediate past interaction with the predator. Monitoring times are longer when the level of future risk is more unpredictable regardless that the current level of risk was low. Also, lizards are able to show flexibility in this antipredatory strategy; lizards decrease monitoring times through successive repeated interactions with the predator even if current risk level remains similar, probably because uncertainty on future risk level decreases after successive interactions. FUNDING Projects MEC-CGL2005-00391/BOS, MCI-CGL2008-02119/ BOS, and MCI-CGL2008-02843/BOS. The experiments were done under license from the Madrid Environmental Agency (Consejería del Medio Ambiente de la Comunidad de Madrid). We thank two anonymous reviewers for helpful comments and El Ventorrillo MNCN Field Station for use of their facilities. REFERENCES Abrams PA. 1994. Should prey overestimate the risk of predation? Am Nat. 144:317 328. American Society of Ichthyologists and Herpetologists. 2004. Guidelines for use of live amphibians and reptiles in field and laboratory research. 2nd ed. Lawrence (KA): American Society of Ichthyologists and Herpetologists. Amo L, López P, Martín J. 2007. Refuge use: a conflict between avoiding predation and losing mass in lizards. Physiol Behav. 90: 334 343. Bouskila A, Blumstein DT. 1992. Rules of thumb for predation hazard assessment: predictions from a dynamic model. Am Nat. 139: 161 176. Burger J, Gochfeld M. 1990. Risk discrimination of direct versus tangential approach by basking black iguanas (Ctenosaura similis): variation as a function of human exposure. J Comp Psychol. 104:388 894. Carrascal LM, López P, Martín J, Salvador A. 1992. Basking and antipredator behaviour in a high altitude lizard: implications of heatexchange rate. Ethology. 92:143 154. Cooper WE Jr. 1997a. Factors affecting risk and cost of escape by the broad-headed skink (Eumeces laticeps): predator speed, directness of approach, and female presence. Herpetologica. 53: 464 474. Cooper WE Jr. 1997b. Threat factors affecting antipredator behavior in the broad-headed skink (Eumeces laticeps): repeated approach,

Polo et al. Monitoring predation risk in lizards 223 change in predator path, and predator s field of view. Copeia. 1997:613 619. Cooper WE Jr. 1998. Risk factors and emergence from refuge in the lizard Eumeces laticeps. Behaviour. 135:1065 1076. Cooper WE Jr. 1999. Tradeoffs between courtship, fighting, and antipredatory behavior by a lizard, Eumeces laticeps. Behav Ecol Sociobiol. 47:54 59. Cooper WE Jr. 2000. Tradeoffs between predation risk and feeding in a lizard, the broad-headed skink (Eumeces laticeps). Behaviour. 137:1175 1189. Cooper WE Jr. 2008. Visual monitoring of predators: occurrence, cost and benefit for escape. Anim Behav. 76:1365 1372. Cooper WE Jr, Frederick WG. 2007. Optimal time to emerge from refuge. Biol J Linn Soc. 91:375 382. Cooper WE Jr, Martín J, López P. 2003. Simultaneous risks and differences among individual predators affect refuge use by a lizard, Lacerta monticola. Behaviour. 140:27 41. Dill LM, Fraser AHG. 1997. The worm re-turns: hiding behaviour of a tube-dwelling marine polychaete, Serpula vermicularis. Behav Ecol. 8:186 193. Endler JA. 1986. Defense against predators. In: Feder ME, Lauder GV, editors. Predators-prey relationships. Chicago (IL): University of Chicago Press. p. 109 134. Ferrari MCO, Rive AC, MacNaughton CJ, Brown GE, Chivers DP. 2008. Fixed vs random temporal predictability of predation risk: an extension of the risk allocation hypothesis. Ethology. 114:238 244. Hamilton IM, Heithaus MR. 2001. The effects of temporal variation in predation risk on antipredator behavior: an empirical test using marine snails. Proc R Soc B Biol Sci. 268:2585 2588. Helfman GS. 1989. Threat-sensitive predator avoidance in damselfishtrumpetfish interactions. Behav Ecol Sociobiol. 24:47 58. Hugie DM. 2003. The waiting game: a battle of waits between predator and prey. Behav Ecol. 14:807 817. Johansson A, Englund G. 1995. A predator prey game between bullheads and case-making caddis larvae. Anim Behav. 50:785 792. Koivula K, Rytkönen S, Orell M. 1995. Hunger-dependency of hiding behaviour after a predatory attack in dominant and subordinate willow tits. Ardea. 83:397 404. Lima SL. 1998. Stress and decision making under the risk of predation: recent developments from behavioral, reproductive, and ecological perspectives. Adv Study Behav. 27:215 290. Lima SL, Bednekoff PA. 1999. Temporal variation in danger drives antipredator behavior: the predation risk allocation hypothesis. Am Nat. 153:649 659. Lima SL, Dill LM. 1990. Behavioral decisions made under the risk of predation: a review and prospectus. Can J Zool. 68:619 640. Martín J, López P. 1999a. An experimental test of the costs of antipredatory refuge use in the wall lizard, Podarcis muralis. Oikos. 84:499 505. Martín J, López P. 1999b. When to come out from a refuge: risksensitive and state-dependent decisions in an alpine lizard. Behav Ecol. 10:487 492. Martín J, López P. 2000. Costs of refuge use affect escape decisions of Iberian-rock lizards, Lacerta monticola. Ethology. 106:483 492. Martín J, López P. 2001. Repeated predatory attacks and multiple decisions to come out from a refuge in an alpine lizard. Behav Ecol. 12:386 389. Martín J, López P. 2003. Ontogenetic variation in antipredatory behavior of Iberian-rock lizards (Lacerta monticola): effects of body size dependent thermal exchange rates and costs of refuge use. Can J Zool. 81:1131 1137. Martín J, López P. 2004. Iberian rock lizards (Lacerta monticola) assess short-term changes in predation risk level when deciding refuge use. J Comp Psychol. 118:280 286. Martín J, López P. 2005. Wall lizards modulate refuge use through continuous assessment of predation risk level. Ethology. 111: 207 219. Martín J, Salvador A. 1993. Thermoregulatory behaviour of rock lizards in response to tail loss. Behaviour. 124:123 135. Martín J, Salvador A. 1997. Microhabitat selection by the Iberian rocklizard Lacerta monticola: effects on density and spatial distribution of individuals. Biol Conserv. 79:303 307. Martín J, López P, Cooper WE Jr. 2003a. Loss of mating opportunities influences refuge use in the Iberian rock lizard, Lacerta monticola. Behav Ecol Sociobiol. 54:505 510. Martín J, López P, Cooper WE Jr. 2003b. When to come out from a refuge: balancing predation risk and foraging opportunities in an alpine lizard. Ethology. 109:77 87. Martín J, López P, Polo V. 2009. Temporal patterns of predation risk affect antipredatory behaviour allocation by Iberian rock-lizards. Anim Behav. 77:1261 1266. Mirza RS, Mathis A, Chivers DP. 2006. Does temporal variation in predation risk influence the intensity of antipredator responses? A test of the risk allocation hypothesis. Ethology. 112:44 51. Polo V, López P, Martín J. 2005. Balancing the thermal costs and benefits of refuge use to cope with persistent attacks from predators: a model and an experiment with an alpine lizard. Evol Ecol Res. 7:23 25. Sih A. 1992. Prey uncertainty and the balancing of antipredator and feeding needs. Am Nat. 139:1052 1069. Sih A. 1997. To hide or not to hide? Refuge use in a fluctuating environment. Trends Ecol Evol. 12:375 376. Sih A, Kats LB, Moore RD. 1992. Effects of predatory sunfish on the density drift and refuge use of the stream salamander larvae. Ecology. 73:1418 1430. Sih A, Ziemba R, Harding KC. 2000. New insights on how temporal variation in predation risk shapes prey behavior. Trends Ecol Evol. 15:3 4. Sokal RR, Rohlf FJ. 1995. Biometry. 3rd ed. New York: WH Freeman. Van Buskirk J, Müller J, Portmann A, Surbeck M. 2002. A test of the risk allocation hypothesis: tadpole responses to temporal changes in predation risk. Behav Ecol. 13:526 530. Ydenberg RC, Dill LM. 1986. The economics of fleeing from predators. Adv Study Behav. 16:229 249.