Introduction

Predation shapes the behaviour, life history, morphology and distribution of prey animals over both evolutionary and ecological timescales [1], [2]. In order to survive, prey must carefully balance the costs of predator avoidance with the benefits of other fitness-promoting activities such as foraging and reproducing [3], [4]. To complicate matters, predators are highly variable in the threat they pose. Predator activity and hence risk to prey may vary depending on a predator's body size, their foraging preferences, and the place and time during which they focus their foraging effort [5].

On a temporal scale, the risk of being eaten may fluctuate on a seasonal, lunar, daily, or even a minute-by-minute basis [1], [6]. Yet there is some evidence that predation may be predictable enough to allow prey to adaptively respond to temporal patterns of predation risk [7]. For example, rodents [8], [9] and storm petrels [10] are known to reduce their activity during periods of bright moonlight when nocturnal predators are increasingly active. Ants, Pheidole titanus, avoid aboveground foraging activity during times of the day when their predators, the dipteran parasitoids are more active [11]. Also, predation pressure during dawn and dusk ‘crepuscular’ periods is thought to drive species specific sheltering times for diurnal reef fishes [12], [13] as well as the timing of group migration between resting and feeding areas in nocturnally foraging grunts (Haemulidae) and other fish species [14], [15]. Yet it is unknown whether these adjustments are a result of innate recognition or temporal threat-sensitive learning in ecological time [16], [17].

Many prey do not show innate recognition of their predators [18]. Learning provides a means by which prey can identify novel predators [19] and respond to changes in the predator community structure as it fluctuates through space and time [20]. Aquatic prey are known to rely on a number of sensory modalities to detect predation-related cues, and amongst them, chemosensory detection appears to be the most widespread [18]. More specifically, a wide diversity of aquatic taxa from corals to larval amphibians, rely on damage released chemical alarm cues (hereafter alarm cues) to assess the level of predation risk in their local environment [18]. Alarm cues are released from the damaged epidermis of prey animals attacked/captured by predators, hence providing a reliable indicator of predation threat to conspecifics and some heterospecifics [21]. Alarm cues can also mediate associative learning of predators via the simultaneous pairing with a predator cue (sight, smell or sound) [22]. Learned predator recognition mediated via alarm cues occurs in a wide variety of prey taxa [18].

The threat-sensitive predator avoidance hypothesis [5] predicts that prey should match the intensity of their antipredator response to the level of predation risk they experience. Recent studies demonstrate that learned predator recognition allows prey to gauge the level of threat posed by novel predators. For instance, predator-experienced fathead minnows have been shown to distinguish the diet [23], size [24], density and proximity [25] of predators using predator odour only. The first step for such threat-sensitive assessment is for prey to use the concentration of alarm cues as a proxy for risk assessment [18]. So far, threat sensitive learning in relation to temporal patterns of risk has been demonstrated in the embryos [26], [27] and larvae [16] of one amphibian, Rana sylvatica. Surprisingly little is known about the temporal foraging periodicity of predators and their prey in environments with highly complex predator assemblages, such as tropical rainforests and coral reefs.

Piscivorous fishes on coral reefs are abundant, diverse [28], and often distributed unevenly among habitat patches. [29]. Most reef fishes have a bipartite life cycle that includes a planktonic larval stage [30] and as a result, larvae rarely settle to a reef with the same composition and density of predators as their natal reef. As juveniles grow, they will enter the size selection ranges of new predators and so will need to maintain their ability to learn and adapt to novel predation risks [31], [32]. Being able to rapidly recognise the daily temporal foraging patterns of different predators should allow prey to maximise trade-offs between predator avoidance and foraging, leading to higher fitness and survival [16].

There is evidence that predation pressure as a whole on coral reefs may be non-uniform over 24 h periods [33], [34], [35]. Some predators are thought to have relatively predictable activity patterns, such as highly diurnal wrasses (Labridae) and trevally (Carangidae) or nocturnally active snappers (Lutjanidae) and grunts (Haemulidae) [36]. Predators may also forage during specific period of the day, for example, the lionfish, Pterois volitans which displays greater activity during dawn and dusk [37]. Hence, many coral reef fishes would benefit from learning the temporal foraging patterns of the predators commonly found in their habitat patches.

The current study investigates whether juvenile lemon damselfish, Pomacentrus moluccensis, can learn temporal patterns of risk, and subsequently respond in a threat-sensitive manner to reduce their risk of predation. Coral reef fishes have been shown to use alarm cues to learn about multiple unknown predators as well as the level of predation threat posed through associative learning [38], [39]. In the present study, groups of juvenile P. moluccensis (“morning-risk” groups hereafter) were conditioned with a predictable pattern of risk consisting of alarm cues paired with predator odour in the morning (predator feeding and dangerous – high-risk) and predator odour paired with seawater in the evening (predator present but not feeding – low-risk). Other groups (“evening-risk” groups) received the opposite treatment, i.e. a low-risk in the morning followed by a high-risk in the evening. After the conditioning period, we tested individual P. moluccensis of both groups for their response to the predator cue alone (in both the morning and the evening) to determine whether individuals displayed different antipredator responses based on the schedule of risk they experienced during the previous 6 days.

Materials and Methods

Results

The non-parametric 3-factor MANOVA revealed a statistically significant interaction among the conditioning treatment, testing time and testing stimulus (Pillai's trace:_{6, 394} = 0.08, p<0.05). The 3-factor ANOVAs revealed a significant 3-factor interaction between conditioning treatment, testing time and testing stimulus for each of the variables: foraging rate (F_{2, 198} = 5.03, p<0.01), line crosses (F_{2, 198} = 3.06, p<0.05) and relative distance from shelter (F_{2, 198} = 3.52, p<0.05).

To investigate the nature of the interaction, analyses were performed on each cue separately. No statistically significant interaction between conditioning risk and testing time was found for P. moluccensis tested with alarm cues or seawater for any of the response variables (p>0.05). In other words, fish constantly responded to the risk cues and never responded to the control cues, regardless of conditioning risk or testing time. In contrast, a statistically significant interaction was apparent between conditioning and testing time for prey fish tested with predator odour: foraging (F_{1, 67} = 20.42, p<0.001), line crosses (F_{1, 67} = 6.50, p<0.05), and relative distance (F_{1, 67} = 8.88, p<0.005).

Tukey's HSD post-hoc comparisons revealed that fish previously exposed to both morning and evening conditioning treatments exhibited a statistically significant (p<0.05) decrease in foraging at the testing time in which they had previously been conditioned with risk, as opposed to the testing time in which they had previously been conditioned with predator odour alone (Fig. 1). The following discussion thus focuses on foraging rate, as this was the only variable for which a statistically supported description of the response pattern was possible.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Change in foraging rate for Pomacentrus moluccensis in response to cues of varying threat.

Mean (± SE) percentage change in strike rate (per 3-min observation) from the pre-stimulus baseline for Pomacentrus moluccensis tested with: A. predator odour (PO), B. alarm cues (AC) or C. seawater (SW) at one of two times: morning (dark grey bars) or evening (light grey bars). P.moluccensis were previously conditioned for six days with either morning risk (high-risk in the morning, low-risk in the evening) or evening risk (high-risk in the evening, low-risk in the morning). * Indicate significant differences at p<0.05, between P. moluccensis tested in the morning and evening.

https://doi.org/10.1371/journal.pone.0034535.g001

Discussion

The present study shows that juvenile coral reef fish have the ability to learn to respond to temporal patterns of predation risk. This experiment clearly establishes that P. moluccensis have the capacity to develop threat-sensitive responses to predator contingent on the temporal pattern of predation risk that they receive. Individual P. moluccensis showed a significant reduction in foraging to predator odour when tested at the time of day to which they had previously been conditioned to receive higher risk. In contrast, those P. moluccensis tested with alarm cues alone did not vary their foraging response in respect to conditioning and testing treatments, highlighting the importance of alarm cues as perpetual indicators of imminent risk [18].

P. moluccensis did not display a significant reduction in line crosses or distance from shelter when presented with predator odour at the time of risk conditioning, although the pattern of reduction in these measures was similar to that found for foraging rate. Therefore, it appears that in contrast to foraging, reduced activity level and distance from shelter were not key antipredator responses for juvenile P. moluccensis in this instance.

Learning temporal patterns of risk has been reported in amphibians [16], [17], however this is the first study to identify the associative learning of temporal patterns in a marine fish. This is also the first study to investigate temporal associative learning in a highly predator diverse environment.

Like most prey species, coral reef fish must effectively balance predator avoidance with other fitness-promoting behaviours (i.e. foraging) [1]. The threat-sensitive predator avoidance hypothesis suggests that prey should trade-off predator avoidance against other fitness promoting activities, and should do so in an appropriate graded manner [5].

The results of the current experiment uphold predictions from the threat sensitive avoidance hypothesis as prey demonstrated a stronger antipredator response to the introduction of predator odour when the time of testing corresponded with the time of high-risk conditioning. The response was categorised by a substantial decrease in foraging rate, a common antipredator behaviour exhibited by a number of species including P. moluccensis [39]. Given the cost involved in reducing foraging effort, the behaviour displayed by P. moluccensis suggests that being vigilant during a predator's preferred foraging time may be a small price to pay in comparison to the probable benefit of reducing mortality [26].

Like most demersal marine fishes, P. moluccensis has a dispersive larval stage in the pelagic environment, and settle on the reef around the time of metamorphosis [30]. Learning of resident predator species is likely to be critical at this life history transition where mortality is extremely high [45], [46]. However, juvenile P. moluccensis in our study were already established on the reef for ∼1 month and are therefore likely to have had prior knowledge of temporal foraging and activity patterns exhibited by C. cyansostigma. To account for this, prey fish were randomly assigned to one of two conditioning risk treatments, which provided high-risk either in the morning or evening. This design allowed us to conclude with confidence that the behaviour of juvenile P. moluccensis was a result of learning at the time of experimentation and not due to innate or previously learned temporal behaviour of predators.

Temporal threat sensitive learning has been explored in one other species of fish: Galaxias muculatus [47]. Reebs (1999) found that G. muculatus were unable to learn to be in a specific time and place to escape predation (time-place learning), raising questions about whether predators are predictable enough both in space and time to allow learning to occur [17], [48]. In contrast, the current study shows that learning temporal patterns of risk can develop in juvenile P. moluccensis within a period of six days. While little was known about the nature of predation risk in the study by Reebs (1999), common resident predators of P. moluccensis, such as P. fuscus and Cephalopholis spp. do have foraging peaks during particular times of the day (Feeney et al. submitted; Fishelson et al. 1989). Both P. fuscus and C. cyanostigma are sedentary, patchily distributed, and are known to occupy relatively small territories in the vicinity of planktivorous damselfish [29], [41]. Similarly, juvenile P. moluccensis are extremely site attached with tagging studies showing that fish rarely move further than 2 m [49]. While more transient fishes may be inclined to shift their foraging habitat to avoid predators, juvenile P. moluccensis will be forced to live with any predator species that takes up residence in their habitat patch. This situation is likely to promote shifts in foraging time, which oppose the diel foraging pattern of predators.

Learning is thought to be advantageous for prey where the environment may fluctuate and predator composition vary [48], [50]. In a complex environment such as a coral reef, it is highly likely that predation risk may change through time [1], [6]. One way that coral reef fishes will experience variation in predation threat is through immigration and emigration of resident predators from their habitat patch. For instance, even C. cyanostigma, a relatively site attached predator, may move distances of between 21 and 48 m over longer time scales (three months to two years) [41]. As diel foraging patterns vary among predator species, a flexible mechanism allowing prey to respond to temporal variability in predation risk would be highly beneficial.

In complex coral reef environments, it is also possible that predators may temporarily alter their predominant foraging patterns even whilst remaining within their habitat patches. If we consider that both predators and prey are active participants in behavioural interactions there may come a point where too much predictability will be detrimental for predators that specialise in few prey species [51]. Notwithstanding any physiological constraints, predators may increase variation in their temporal foraging patterns to increase success. For example, recent studies by Roth and Lima [51] have found that hawks need to attack at random if they are ever to successfully catch pigeons. Predator-prey interactions can thus be likened to a dynamic behavioural game, whereby if risk is too predictable, prey will be able to learn too efficiently, eventually causing predators to change their attack schedule to be more efficient [51], [52].

There may also be particular times of the day that are simply more dangerous for diurnal reef fishes as a whole. On coral reefs, predation pressure is considered to be strongest during dawn and dusk [33], [34], [53] with predators thought to possess physiological adaptations for operating at lower light levels [54]. Diurnal reef fishes, such as juvenile P. moluccensis, may therefore find it innately easier to respond to predation risk provided at dawn and dusk than other times within the diel cycle. We strongly encourage further studies that investigate species-specific diel foraging patterns within coral reefs and other complex systems to allow us to determine any overarching patterns.

While various studies have investigated the effect of temporal patterns in predation risk on prey, a largely unappreciated aspect of the relationship concerns how prey allocate behaviours when temporal risk varies [6]. The risk allocation hypothesis (RAH) is one of the first models to actually consider how temporal variation in predation risk through time (days, weeks or months) may influence prey antipredator behaviour [6]. In contrast to the threat sensitive model, the RAH looks specifically at how background levels of predation risk over days to weeks may influence how prey allocate their foraging efforts and hence exposure to predation risk [6]. While we did not specifically test the RAH in this study, the ability for prey animals to assess temporal predation risk through learning is likely crucial for them to display effective and cost-efficient antipredator responses, including strategies following the RAH. Further studies on how coral reef fishes allocate their risk-taking behaviours through time will be essential if we are to achieve a holistic understanding of temporal periodicity in risk and its effect on predator-prey dynamics in complex ecosystems.

Despite the importance of diel patterns of predation risk in influencing prey behaviour, the link between predator and prey activity schedules has often only been inferred, and in very few cases has the mechanism maintaining this link been determined [7], [16]. Our study provides some of the first evidence that even in a highly complex system, prey have the ability to learn and aptly respond to predictable temporal patterns of predation risk. Although we have only begun to scratch the surface of how pervasive and important these effects may be, such abilities may play an important role in structuring prey behaviour in a wide range of ecosystems.

Acknowledgments

We thank A. Vail and A. Fennell for assistance in collecting fish for this study. We thank the staff of Lizard Island Research Station (Australian Museum) for logistical support. Research was carried out under approval from the Great Barrier Reef Marine Park Authority, and under James Cook University ethics guidelines.