Individuals are at risk when communicating because conspicuous signals attract both conspecifics and eavesdropping predators. This predation cost of communicating has typically been attributed to signalling individuals because of their conspicuous role, and is a core concept within sexual selection and communication ecology. But, if predators are attracted to signals, then receivers, both intended or otherwise, may also find themselves at risk of predation. Here, we review the theoretical basis and empirical evidence that receiving also carries a risk of predation. We distinguish between the risks of receiving and responding to signals, and we argue that receivers of signals that are long lived, are highly predictable in time or place and/or cannot be received quickly are likely to be at greater risk of predation compared to receivers of signals without these properties. We review recent empirical evidence from a variety of taxa that supports the hypothesis that receivers (including heterospecific prey) are aware of these risks and that they modify their behaviour to balance the risks against the benefits of receiving under predation threat. We also discuss the wider implications of risky receiving for receiving and signalling behaviour in prey, as well as for the prey's predators.
A general introduction to signal reception and interception
Since Darwin (1871), it has been recognised that conspicuous communication signals such as colourful mating displays and complex courtship songs are also highly attractive to predators and parasitoids (reviewed in Magnhagen 1991; Zuk & Kolluru 1998). This attraction arises because of the inherent properties of signals. Generally, signallers are unable to promote reception by intended receivers while simultaneously preventing interception by other, unintended receivers (Box 1; see Table 1 for a glossary of communication terms). Predation, therefore, presents a significant risk to reproduction and other activities that rely upon signalling, and signallers (typically males) respond to this risk by altering conspicuous signal traits or signalling at times and/or locations when predators are least active (e.g. Endler 1980; Ryan et al. 1982). These trade-offs potentially affect a signaller's fitness and can lead to population-level changes in the expression of signalling traits (Zuk & Kolluru 1998). How signallers trade-off the benefits of signalling, such as attracting mates and deterring rivals, with the costs of attracting unintended receivers such as eavesdropping conspecifics, predators and parasitoids, has been the subject of considerable research effort (Zuk & Kolluru 1998).
Table 1. Glossary of terms relating to the roles of sending, receiving and responding to animal communication signals. We use the term receiver to indicate any individual that detects and/or intercepts a signal. Intended receivers are individual(s) for whom the signal was intended, in contrast to eavesdroppers (definitions following Peake 2005) that exploit the information contained in signals (interceptive eavesdroppers) or signalling interactions (social eavesdroppers). Previously, the terms bystander (Earley & Dugatkin 2002) and audience (Matos et al. 2005) have referred to those individual(s) that are present during signalling interactions, but do not take part. Here, we use the more general term ‘passive receiving’ to include individuals that are in detection range of the signal, but are not actively trying to obtain the information contained within it; this definition may include intended receivers as well as individuals that subsequently become eavesdroppers
‘The provision of information by a sender to a receiver, and the subsequent use of that information by the receiver in deciding how to respond’ (Bradbury & Vehrencamp 1998).
A change in the environment caused by one individual (the signaller) which can convey information to another (the receiver) (Endler 1993).
An individual producing a signal. Individuals can alternate between being signallers and receivers. We use the term signaller to refer to all individuals that emit a signal, rather than a categorical distinction between individuals who have a high rate of signalling compared to individuals with a low rate of signalling.
Animal engaged in behaviours that facilitate signal production and transmission.
Signal content (strategy)
Information contained in the signal about the signaller (not necessarily honest) (Endler 1993).
Signal design (efficacy)
Properties of the signal that increase the efficacy of transmission between signaller and receiver (Endler 1992).
The process of detecting and/or intercepting a signal.
Any individual that detects and/or intercepts a signal; includes intended and unintended receivers.
Individual(s) for whom the information contained in the signal is intended.
Behaviours that promote signal reception and information acquisition.
The process of signal interception simply as a consequence of being within detection range of the signal. Passive receivers do not perform behaviours that promote the signal's reception, or intend to acquire, or act on information contained within the signal.
‘The use of information in signals by individuals other than the primary target’ (Peake 2005).
The interception (and use of information) in signals intended for another individual, usually to the cost of the signaller (usually heterospecifics) (Peake 2005).
‘Acquiring information on others by attending to their signalling interactions with conspecifics’ (Peake 2005).
A change in the behaviour of the receiver/eavesdropper resulting from information contained in the signal.
Importantly, empirical research reveals that conspicuous signals can also place non-signalling individuals (i.e. receivers; Table 1) at an increased risk of predation. Classic work by Sakaluk & Belwood (1984) reported that geckos (Hemidactylus tursicus) orient towards the calls of male crickets (Gryllodes supplicans), but more commonly prey on female crickets phonotactically orienting to the male calls. A number of other studies also provide tantalising anecdotal or indirect evidence that receiving may be associated with an increased risk of predation (see Box 2 for details). Much of this research has focused on the costs that predation risk imposes on (generally female) mate choice, such that an increased risk of predation while choosing between potential mates is now considered a fundamental cost of reproduction (reviewed by Jennions & Petrie 1997).
Each step of the mate-choice process, including searching for, assessing, and then responding to a signal, may expose an individual to the risk of predation. We suggest, however, that if predators can target non-signalling individuals that are in close association with a conspicuous signal, then an individual's decision about if, when, and where to receive a signal (hereafter collectively referred to as ‘receiving’; Table 1) should be considered separately from its decision to respond to a signal (e.g. engage in mating). Moreover, we suggest that predation risk imposes a significant, general cost on the reception of all signals and not only those related to mating, and that all individuals will be at increased risk while receiving. Accordingly, we predict that all receivers should trade-off the benefits of participating in communication with the potential risks of predation while doing so. We also predict that nuances in these trade-offs may enable individuals to optimise reception under an increased risk of predation. The way(s) in which receivers make these trade-offs have implications for their own fitness, the fitness of those whose signal(s) they are receiving and the foraging strategies of their predators.
A simple conceptual framework illustrating the potential risks of predation while sending, receiving and eavesdropping on signals
In a dyadic interaction, (a) a signaller (S) transmits a signal (black arrow) to a single intended receiver (R) that is actively engaged in receiving. An eavesdropping predator (P) that is located within range of the signal (indicated by the dashed lines) may detect the signal (grey arrow) and target the signaller (white arrow). The predator may also target the active receiver (white arrow) because it is in close proximity while receiving the signal. In a communication network, (b) a conspecific (social eavesdropper, SE) is obtaining information by observing the interaction between the signaller and intended receiver. In this scenario, the predator may target the signaller, intended receiver or social eavesdropper, whose attention may be diverted by the ongoing social interaction between the signaller and receiver. Individuals within range of at least one of the signals, but not actively receiving (i.e. passive receivers, RP) may also be targeted by predators because of their proximity. More complex communication networks (c) arise when there are two or more signallers (e.g. S1 and S2) and/or two or more intended receivers, social eavesdroppers or passive receivers (only one of each shown for clarity), as the number of signals and signalling interactions increases. Eavesdropping predators may target any (and/or all) of these individuals. Non-predatory heterospecifics (not shown) engaged in interceptive eavesdropping on the signals or signalling interactions may also be targeted by eavesdropping predators. In species that deposit scent signals directly onto the substrate (d), the signal (grey S) remains attractive after the signaller has left the area and the predator can therefore target the intended receiver, social eavesdropper or passive receiver, but not the signaller.
Active and passive signal reception
The theoretical extension of risky communication to include all receivers stems from the concept of communication as a network of multiple animals that are within signalling and receiving range of each other (Box 1; McGregor 2005). Receivers can be the intended recipients of signal(s), but the network may also contain a suite of unintended conspecific and heterospecific receivers (Box 1; McGregor 2005). The extension of receiver costs from intended receivers to all these signal recipients represents a significant generalisation of the risks of receiving, and it is therefore worth considering each of the categories of signal recipients in more detail.
We use the term receivers to collectively refer to all individuals that detect (or intercept) a signal (i.e. the process of receiving), while intended receivers are those individuals for whom the information contained in a signal was intended (Table 1). In addition, we suggest that the act of receiving can be active or passive (Table 1). Active receivers engage in behaviours that promote signal reception and information acquisition, such as attentiveness (e.g. watching, listening, smelling). Passive receivers, on the other hand, are within detection range of the signal but do not intend to acquire or use the information contained within. Passive receivers might include non-reproductive females that incidentally hear male courtship calls, or other species (e.g. heterospecific females) for which the signal and its content are irrelevant.
Unintended receivers of signals are often referred to as ‘eavesdroppers’, although this definition does not confine eavesdropping to auditory signals, and examples of eavesdropping are found in all signalling modes (e.g. auditory, olfactory, tactile and visual signals) and from a wide variety of taxa (Peake 2005). ‘Social eavesdropping’ (Peake 2005) on signalling interactions by conspecifics enables the eavesdropper to gather absolute or relative information on signalling conspecifics without engaging directly in potentially costly interactions themselves (McGregor & Dabelsteen 1996). In contrast, ‘interceptive eavesdropping’ (Peake 2005) describes the tendency for individuals to eavesdrop on the signals of heterospecifics. This form of eavesdropping is typically viewed as an intertrophic level interaction, such as predator and parasitoid attendance to prey signals to increase their foraging efficiency, and conversely, prey use of predator and parasitoid signals to evade these enemies. Nevertheless, interceptive eavesdropping can also occur within a trophic level, to gather information on shared resources, space, or predators (see Goodale et al. 2010 for a recent review).
Receiving vs. responding
In contrast to the act of receiving, responding to a signal involves a change in the behaviour of the receiver as a result of the information contained in the signal. This response can be movement towards the signaller or the signal source, or responding by becoming a signaller (e.g. returning the call). Movement away from a signaller or signal source is also a form of responding, provided that the decision is based on signal content. Yet, while receiving and responding are conceptually distinct, it is not always simple to distinguish between them in practice because they can occur simultaneously or as a sequence of closely alternating behaviours. During sequential mate assessment, for example, information on the quality of the potential mates (or their signals) attained so far is integrated into the decision of whether or not to evaluate additional potential suitors. Thus, the process of locating an additional (potential) mate's signal could be categorised as responding to signal(s) already received, and also the reception of a new signal. In cases such as this, we define receiving as the ongoing process of signal search and interception.
In this review, we explore the notion that individuals face an increased risk of predation while receiving signals. We first evaluate how differences in signal form might affect the level of risk incurred by a receiver. On the basis of this evaluation, we suggest that olfaction is an ideal modality in which to examine the concept of risky receiving. Unlike other sensory modalities, the behaviours involved in receiving (passive and active) and responding to olfactory signals are readily distinguishable and associated with differing levels of potential predation risk. Finally, we discuss the implications of risky receiving for prey behaviour, including potential mechanisms for prey to optimise receiving behaviour under a risk of predation, and the feedback these behaviours will generate on the evolution of signals and predator hunting strategies. We focus our review on the reception of social signals, including those used for mating, and do not include the risks of receiving signals from mimics or other illicit signallers, or on the more general predation risks of associating with other prey or their cues (e.g. sounds made inadvertently while moving).
Signal form and the potential for predation on receivers
Eavesdropping and the risks to receivers are phenomena that can occur for all sensory modalities, but will vary with a signal's properties (Table 2). The modality of the signal influences its basic properties such as range, longevity and specificity to the intended receiver, i.e. the likelihood that a signal will only be detected by the intended recipient. The collective properties of signals determine how easily eavesdropping predators (hereafter referred to simply as predators) can locate a signaller and/or receiver. Particular attributes of the signal, such as the duration and frequency of an acoustic signal (e.g. Ryan et al. 1982) or the number and size of colour patches comprising a visual signal (e.g. Endler 1980), influence the signal's conspicuousness to predators. The hunting mode and sensory capability of the predator (e.g. its auditory or visual sensitivity) and the environmental conditions in which the signal is transmitted (e.g. background noise or visual complexity) also play an important part in determining signal detectability (Endler 1992, 1993). Importantly, however, the environmental and signal properties that increase the risks to signallers will not always coincide with those most risky to receivers (Table 2). In the following sections, we consider how particular signal properties determine their vulnerability to interception by predators.
Table 2. Properties of signals produced in five sensory modalities that determine the locatability of both the signaller and receiver. For example, visual signals are medium range and strongly associated with the signaller (e.g. its colouration) rendering their senders highly locatable. In contrast, olfactory signals are long ranging and less readily associated with the signaller. However, the cost of receiving olfactory signals is potentially high, particularly when attaining information from these signals requires close spatial proximity. Modified from Bradbury & Vehrencamp (1998)
Specificity (likelihood of detection by the intended receiver only)
Signal duration (longevity)
Variable; can be long-lasting
Variable; can be long-lasting
Transmittable around barriers?
Energetic cost to signaller
Low, but high to maintain
Use in dark
Locatability of signaller
Locatability of receiver
Signal range & specificity
The probability and frequency of signal interception by receivers, intended or otherwise, increases with signal range, and signals that are detectable over a large geographical area should therefore be more vulnerable to interception. The area over which many signals can be perceived increases non-linearly with the straight-line distance they travel (e.g. Vergassola et al. 2007), however, and conspecific receivers of long-range signals are therefore less likely to be close to an eavesdropping predator than are receivers of short-range signals. The effects of any small errors in taxis are also magnified with distance from a signal's source, impeding an eavesdropping predator's ability to locate the source and nearby receivers. Long-range signals that have imbedded directional information may partially counter the difficulties receivers face in locating a distant signaller, although these are still difficult signal-detection tasks and highly influenced by local conditions. For instance, while male almond moths (Cadra cautella) use the concentration gradient of a conspecific female's pheromones during chemotaxis, the male's flight path and the probability that he locates the emitting female is strongly affected by local wind turbulence, speed and pulse rate (Mafra-Neto & Cardé 1994).
Signal specificity also matters, and signals that are transmitted from a signaller to an intended receiver without ‘leakage’ of information to nearby predators will pose fewer risks to both participants. As a result, short-range signals should be less risky for both signallers and conspecific receivers, as the likelihood of a predator being within signal range is lower. Nevertheless, detection of a signal at short-range by a predator may have dire consequences for both the signaller and intended conspecific receiver(s), and also for any conspecific eavesdroppers and passive receivers. Tactile signals and those that involve substrate vibrations are two examples of short-range signals with limited scope for eavesdropping (Table 2), yet even short-range vibration signals may expose prey to greater risks of attack than those relying on direct touch.
In a series of experiments, Laumann et al. (2007, 2011) recently demonstrated that scelionid egg parasitoid wasps (Telenomus podisi) eavesdrop on the plant-restricted sexual vibratory signals of their preferred stink bug (Euschistus heros) hosts. They showed that the parasitoids selectively responded to the vibratory calls of female hosts over those from males, male–female duets or control vibrations (Laumann et al. 2007). Furthermore, the parasitoids distinguished between (and preferred) the vibratory signals of female E. heros hosts over those from females of alternative-host (Piezodorus guildinii) and non-host (Chinavia impicticornis) species (Laumann et al. 2011). Of course, multimodal signals that combine a conspicuous, longer range advertisement signal with a more informative but shorter range signal may be particularly effective at attracting receivers (e.g. Roberts et al. 2007). In the case of the above-mentioned stink bugs, the risks associated with the short-range vibrational signals may be increased by the stink bugs' concomitant release of long-range pheromone attractants (Aldrich et al. 1995).
Signal longevity & complexity
The temporal scale over which a signal persists profoundly affects its likelihood of being received, and signallers can increase the probability that their signal will be intercepted if they extend its duration. Male mating success of many acoustically signalling species increases with time spent calling (e.g. Snedden & Sakaluk 1992), for example, but longer lasting signals or signalling displays are also more easily located by predators and parasitoids (Magnhagen 1991; Zuk & Kolluru 1998). Signallers can compensate for this increased risk by behaving more cautiously or adjusting the time and place at which they signal. For example, male field crickets (Gryllus integer) in the presence of a predator cue take longer to emerge from a safe shelter and also have longer latencies to resume calling if they have longer, more conspicuous calls than if their calls are shorter and less conspicuous (Hedrick 2000). Receivers are also sensitive to the predation risks posed by long-lasting signals. Although female field crickets tend to prefer males with longer calls, females will phonotactically orient at greater rates towards males with shorter calls if the short calls emanate from safe locations while the longer calls come from dangerous locations (Hedrick & Dill 1993; see also Box 2 and Table 3).
Table 3. A selection of studies referred to in the text that have examined the predation and parasitism risks associated with receiving communication signals, organised by sensory modality, signal content, risks to the signaller and receiver and the behaviours that are potentially traded against the risk of predation. Note that many of these studies examined the combined risks of searching, receiving and responding to signals. This list is not intended to be exhaustive, and does not, for example, include many studies of altered female mate choice under an increased risk of predation
Sensory modality and study system
Risks to signaller
Risks to receiver
Potential trade-offs with predation risk
Crickets (Gryllodes supplicans) and gecko (Hemidactylus tursicus) predators
Signals or signalling interactions that combine longevity with receiver proximity should particularly increase receiver risks. Females often spend a considerable period of time receiving extended or multiple male signals at close range before deciding whether to respond, and the evolution of cryptic colouration has been interpreted as a mechanism to reduce females' risk of predation while receiving. Golden-collared manakins (Manacus vitellinus) clear lekking courts to increase their conspicuousness to females, for example, but the females are themselves highly cryptic against this background (Uy & Endler 2004). The wealth of studies documenting reduced female choosiness for male mates under an increased risk of predation also suggest that lengthy signal assessments are perceived as risky (e.g. Godin & Briggs 1996; Booksmythe et al. 2008; see also Box 2 and Table 3).
Performing difficult tasks increases an individual's risk of predation because of the limited amount of attention they have to devote to given stimuli (reviewed in Dukas 2002). Animals are unable to simultaneously search for cryptic prey and attend to approaching predators (e.g. Milinski 1984; Dukas & Kamil 2000), for example, and Dawkins & Guilford (1991) hypothesised that reduced vigilance would similarly increase the risk of predation to receivers of sexual signals. Female wolf spiders (Schizocosa uetzi), for instance, are less sensitive to simulated predatory visual stimuli when attending to complex male courtship vibrations (Hebets 2005). More generally, however, we suggest that all receivers may be at an increased risk of predation if receiving a complex signal(s) reduces their vigilance for predators. Thus, complex auditory calls or visual displays should pose a greater risk of predation to active receivers such as the intended signal recipients than will simpler signals. Social eavesdroppers must be attentive towards the behaviour of signallers and receivers if they are to gain information from these interactions, and so should also compromise their vigilance. In contrast to these active receivers, passive receivers should be at a lesser risk of predation, as their capacity for vigilance should not be reduced to the same degree.
The risks to receivers arising from the spatial and temporal persistence of signals may be exacerbated by signal predictability, especially if predators can form a search image for signals that are typically associated with successful predation (Carthey et al. 2011). Nevertheless, prey frequently use signals that are predictable in time or space as they facilitate orientation within a home range and allow for the relative assessment of competitors (e.g. Brown & MacDonald 1985). Predictable signals can also reduce receiver search costs and increase signal receivership, especially if the signals have only a small range or the spatial behaviour of receivers is relatively stereotyped. Individuals actively receiving predictable signals, both intended recipients and eavesdroppers, should have a greater likelihood of being predated upon than passive receivers, due to their more frequent and/or closer proximity with a signal. Leks are just one example where receivers should incur increased risks of predation because they are spatially predictable (i.e. the same sites are used over multiple days and breeding seasons), good indicators of animal activity (e.g. repeated visits by males and females) and are also open to eavesdropping by predators. Brown water snakes (Lycodonomorphus rufulus) are attracted to leks of calling male painted reed frogs (Hyperolius marmoratus broadleyi), for example, where they predate on female reed frogs searching for calling males (Grafe 1997). Gibson & Bachman (1992) similarly showed that female sage grouse (Centrocercus urophasianus) were more likely to be predated upon by golden eagles (Aquila chrysaetos) when visiting leks. Parasitoids also exploit signals that are good predictors of host space use to find new hosts: klipspringer antelope (Oreotragus oreotragus) are parasitised by ticks (Ixodes neitzi) while inspecting conspecific antorbital secretions, because ticks are attracted to fresh antorbital scents and congregate on them in anticipation of a visiting klipspringer host (Rechav et al. 1978).
Signal content, the information or message conveyed by a signal (e.g. an individual's attractiveness, social status or distress), can affect the risk of predation to receivers in several ways. First, signal content can influence receiver risks via its effect on the behaviour of nearby individuals. Signals that attract prey towards one another (e.g. mating signals) may be riskier for receivers than those that repel prey (e.g. aggressive territorial signals), for instance, because the former provide a more reliable foraging cue for eavesdropping predators. Thus, the risks of receiving attractive signals should generally be greater than the risks associated with receiving repellent signals, assuming predators are not able to take advantage of fleeing or otherwise repelled prey.
Second, signal content can influence receiver risks not because of the information the signal contains, per se, but because content is often tightly coupled with a signal's physical properties. Aggressive territorial signals or mating signals that attract the attention of dispersed conspecifics have properties that increase the range over which they are detected, for instance, whereas sexual signals that are received just prior to mating (e.g. at leks) have properties that confer conspicuousness at short-range only, or under limited conditions (e.g. Endler & Théry 1996). In addition, some signals have evolved such that a significant investment of time (e.g. attentiveness) is required to receive them, whereas others are more rapidly received. Most mating signals, for instance, cannot be received and assessed quickly, while it is critical that alarm signals elicit an immediate response. For the reasons previously discussed, these properties of signal range and longevity (among others) will affect the risks to receivers.
Predator identity, hunting mode and perception
The relative risks to signallers and receivers will also be affected by the particular complex of predators within a community, their habitat use, hunting mode (sit-and-wait, sit-and-pursue, or actively hunting; Schmitz 2005) and perceptual capabilities (Schmitz et al. 2004). Receivers will be at an increased risk of predation from actively hunting predators if a signal requires lengthy periods of assessment with little movement by prey, whereas signals that are site-specific and repeatedly visited may expose receivers to increased risks from sit-and-wait predators. In addition, both signallers and receivers will be at increased risk if the signal is exploited by predators with different hunting modes and/or using different signal modalities. The ultraviolet reflective properties and odorous nature of vole urinary signals, for example, exposes them to both olfactorily and visually hunting predators (Viitala et al. 1995).
The spatial scale at which a predator can detect a signal and its ability to distinguish signals from a background of ‘noise’ will determine its hunting behaviour. Olfactorily searching predators require small-scale patchiness in prey cues to detect them from background odours (Carthey et al. 2011), for example, while the visual complexity of the background impedes the detectability of prey by visual predators (Dimitrova & Merilaita 2012). Environmental conditions also affect the persistence and directionality of some signals and thus determine a predator's foraging success. Blue crabs (Callinectes sapidus) often follow prey odour plumes when foraging, for instance, however, the crabs' chemoreceptive abilities and foraging success decline as flow velocity and turbulence increase and the plume becomes increasingly mixed and diluted (Weissburg & Zimmerfaust 1993). Finally, receivers (as well as signallers) may be at particular risk from predators that use multiple signals or cues to detect prey. For example, while some acoustic prey signals attract predators over larger spatial scales, these predators typically use prey movement, often that of the receiver, to locate and catch prey (e.g. Grafe 1997).
Thus, although all receivers may be at an increased risk of predation while receiving in a communication network, receivers of signals that function to attract conspecifics, require a considerable period of assessment, are relatively long-lived and stimulate predictable spatial and/or temporal patterns of behaviour are likely to be at an increased risk of predation than receivers of signals without these properties. Many species use signals that fall into one or more of these categories, and we have provided numerous examples where prey behave as though receiving is risky (Table 3). Yet, for many of these examples, simultaneous or rapidly alternating receiving and responding makes it difficult to disentangle the separate costs of the two behaviours, and there are few clear examples where the risks of receiving and are not conflated with the risks of responding (see also Box 2). The unique properties of some olfactory signals (Table 2), however, provide an opportunity to independently examine the risks that prey perceive while receiving vs. responding.
Olfactory signals: an ideal modality to examine risky receiving
The olfactory signals used by terrestrial organisms can be broadly classified into those whose source is either spatially mobile or fixed. The pheromone plumes of insects are perhaps the most familiar of the spatially mobile signals. Many vertebrates also use pheromones in mate attraction, however, they generally mark their territories using spatially fixed olfactory signals in the form of urine, faeces or other glandular secretions deposited onto a substrate (hereafter scent marks; Brown & MacDonald 1985). These scent marks are unique among signals as they are designed to persist in the absence of the signaller, up to days or weeks for some species (Alberts 1992). The low volatility of scent compounds in these marks facilitates signal longevity, however, it also limits the range over which the scent signals can be detected. To maximise the likelihood that a scent signal is received by a conspecific, marks are often placed at strategic or conspicuous locations throughout an individual's territory, or the non-volatile fraction of scents is supplemented with highly volatile and attractive compounds (see Alberts 1992 for a review). Nevertheless, close or direct contact with a scent is often still required for individual-specific information to be communicated (Alberts 1992; Nevison et al. 2003; see also Box 3). As fresh marks are often placed close to older scents, the result is a patchy accumulation of conspicuous signals and prey activity.
Olfactory signals such as substrate scent marks are therefore characterised by attributes that increase the vulnerability of signal receivers to eavesdropping predators. Furthermore, while signallers leave the immediate environment after depositing their scent, it is individuals receiving the signal at some later time that may encounter a simultaneously attracted eavesdropping predator. The risk of this occurring will be lower for species whose rate of scent investigation is low and unpredictable, as their scents may be too unreliable to guide predator search effort. But, many prey species have small territories or behaviour that is highly stereotyped around scented areas, and their scents should be a valuable foraging cue for predators (e.g. Hughes et al. 2010b). Moreover, the need for receivers to make close or direct contact with a scent means that this act of receiving can be distinguished from any subsequent response, such as a change in the spatial distribution of activity or the deposition of a new scent.
Recent experimental evidence that receiving is risky
In spite of an apparent susceptibility to predation while communicating, until recently only two studies had examined the effect of predation risk on olfactory communication, and both had focussed on the response of signallers. In a laboratory experiment, Roberts et al. (2001) demonstrated that male house mice (Mus musculus) with initially high rates of scent marking (signalling) reduced marking rates when in the presence of a predator odour (ferret urine, Mustela putorius furo), whereas males with initially low marking rates did not. In contrast, a field experiment by Wolff (2004) clearly demonstrated that scent marking rates by prairie voles (Microtus ochrogaster) and woodland voles (M. pinetorum) did not vary in response to predator (mink urine, Mustela vison) or control (rabbit faecal pellets, Oryctolagus cuniculus, and no odour) scents. These results were later confirmed in the laboratory using male prairie voles (Wolff 2004).
Recently, however, several studies have provided direct evidence that receivers of olfactory signals are sensitive to predation risks. In the presence and absence of a predator cue (cat, Felis catus, urine), Hughes et al. (2009) presented individual, adult, male house mice (Mus domesticus) with conspecific social scents. These scents were predicted to be of low or high social value for the male mice to receive; low- and high-value scents were their own scent marks and those of an unknown male conspecific, respectively, where a signal's value related to the probable fitness benefits (e.g. territory defence) to an individual of receiving a signal. Hughes et al. (2009) found that receivers traded-off the perceived social benefits of receiving conspecific signals with the perceived predation risks of doing so: mice significantly reduced their receiving of low-value conspecific signals in the presence of the predator cue, however, they maintained higher receiving rates of the high-value signal in spite of the perceived risks (Fig. 1).
A further study (Hughes & Banks 2010) showed that the perception of risk by prey extends beyond receiving individual signals (i.e. visiting individual scent marks) to encompass how signals are distributed throughout the broader landscape. When presented with conspecific scent marks in a clumped, random or regular spatial configuration and in the presence of a predator cue (cat urine), the overriding response of (individually tested) adult male house mice was to maintain their overall rates of visitation to individual conspecific scent signals, but to alter the spatial distribution of their activity at a larger scale. Specifically, while mouse activity was relatively evenly distributed across the experimental enclosure in the absence of the predator cue, mouse activity was more spatially restricted when predator odour was present, presumably reducing the males' chances of encountering an actively hunting predator.
Reducing mobility under perceived predation risk is a common anti-predator strategy by prey (Kats & Dill 1998) as there can be a positive relationship between mobility and the likelihood of encountering a predator (Lima 1998). But, while restricting movement is likely to decrease an individual's risk of detection by an already present predator, concentrations of scents are probably more conspicuous to predators, than random or regularly distributed scents (Banks et al. 2000). Hughes & Banks (2010) found some evidence for this, as mice dispersed their activity when scents were clumped (and the perceived risk of predation was high) but showed no change in behaviour when scents were distributed regularly or randomly. Thus, it appears that by incorporating information on the spatial distribution of scent signals over a variety of scales, mice can moderate their activity to reduce their risks of encountering an actively hunting predator, while maintaining the receiving of olfactory signals integral to their social lives.
There is evidence that the perceived predation risks of receiving also extend to heterospecific signals. In a large enclosure study, Hughes et al. (2010a) showed that field voles (Microtus agrestis) readily eavesdropped upon the olfactory signals of both conspecifics and sibling vole (M. rossiaemeridionalis) competitors when the risks of predation were low. The benefits of eavesdropping on heterospecific signals were outweighed by the potential predation costs when at risk from an olfactorily hunting predator, the least weasel (Mustela nivalis nivalis), however. Field voles reduced their activity at heterospecific scents in the presence of least weasels, but they did not make the same behavioural trade-offs for the conspecific scent signals. These responses indicate that individuals maintain an awareness of not only the risks posed by the conspicuous signals of nearby conspecifics, but also those of the complex of heterospecifics with which they coexist. This sensitivity to the predation risks of receiving heterospecific signals likely has implications for the structuring of communities, such as the degree of spatial overlap between species, or information on competition for shared resources.
Intriguingly, a recent study suggests that it is not only small mammals that perceive risks to receiving scent signals. Adult great barred frogs (Mixophyes fasciolatus) are relatively sedentary, leading to accumulations of scents in shelter and calling sites. Using a series of odour preference trials, Hamer et al. (2011) showed that M. fasciolatus were highly attracted to conspecific scent accumulations, and also those from sympatric competitors, striped marsh frogs (Limnodynastes peronii; Fig. 2). But, M. fasciolatus also perceived predation risks from associating with frog odours relative to control odours, significantly reducing their relative attraction to both frog species' scents upon the addition of odours from a shared scent-hunting predator, the red-bellied black snake (Pseudechis porphyriacus). Surprisingly, however, male M. fasciolatus maintained some attraction to unfamiliar conspecific scents in presence of the predator odours, suggesting that they perceive benefits of olfactory communication despite the risks (Fig. 2). These responses strongly parallel those seen in small rodents (Hughes et al. 2009, 2010a; Hughes & Banks 2010) and suggest that the perception of receiving as potentially risky may be widespread across taxa that use olfactory signals for social communication.
Implications of optimal receiving behaviour by prey
This review clearly demonstrates that individuals should perceive an increased risk of predation while receiving signals. Although similar risk-sensitive behaviour has previously been reported for individuals responding to signals (e.g. during mate choice; see Table 3 for examples), we suggest that these trade-offs are a general property of signal reception and can occur well before receivers become responders. We have also highlighted that the relative risks to individuals while receiving can vary with their role within the communication network and that some signal properties will place active receivers at greater risk of predation than individuals that receive these signals passively. We note that the number of studies examining the behaviour of receivers (and not also responders) under predation risk is limited, and that studies evaluating the risks to eavesdroppers and passive receivers are especially sparse. Nevertheless, the examples we have provided indicate that the presence of a conspicuous signal prompts receivers to modify their behaviour in a way that is predicted to reduce their risk of predation while receiving. Importantly, however, the mechanisms adopted by receivers will not necessarily optimise their behaviour from a signaller's perspective. The choice of a particular strategy to optimise receiving may therefore also have important implications for selection on signalling traits.
From the receivers' perspective, signals are patchily distributed resources (of information) of variable quality within their environment. Extending this resource-based analogy, we suggest that valuable parallels exist between optimal foraging strategies and the behavioural trade-offs underlying ‘optimal receiving’. Prey wishing to forage optimally on patchily distributed resources must balance the benefits of acquiring food (e.g. for growth, reproduction etc.) with the risks of predation while foraging. As such, optimally foraging prey adjust what to eat, where and how long to forage, whether to forage alone or in a group, the size of a group and vigilance (among other things) with their hunger state and the perceived level of predation risk (see Stephens et al. 2007 for a comprehensive discussion). In the following section, we consider how these intrinsic and extrinsic factors might influence optimal receiving strategies. As the concept of optimal receiving implies that receiving is purposeful, we focus our discussion on active receivers, but make reference to potential strategies by passive receivers where relevant.
The marginal value of information
Individuals are motivated to receive signals to gain information about some aspect of their social or physical environment, such as a potential mate's quality or an opponent's social status. The value of each additional piece of information obtained declines as its ability to reduce uncertainty decreases, however. As individuals incur costs while receiving (e.g. missed opportunity costs, increased risks of predation etc.), this marginal value of information should promote the development of optimal receiving strategies. In turn, the marginal value of information to a receiver will depend on a number of factors, including the familiarity of the signaller and their signal, signal content, the receiver's state such as their social and reproductive status and the perceived risk of predation.
Familiarity with the signal or signaller
In species where recognition of individuals is common, the marginal value of a signal should partially depend on a receiver's familiarity with the signaller, the nature of their recent social interactions, as well as the relative importance of the signal's content. Many territorial species respond less aggressively to known neighbouring conspecifics than unknown or intruding conspecifics (the ‘dear enemy’ hypothesis; Fisher 1954), thereby avoiding costly territorial interactions with regularly encountered individuals. As the signals of close neighbours should also be familiar, we predict that the benefits obtained by receiving these signals will decline with familiarity (assuming no change in their relationship or interactions). Evidence that receiving is risky (e.g. Hughes et al. 2009) supports this idea, as it demonstrates that where there was the opportunity for signals to become familiar (by repeated encounters), receivers were less willing to maintain high receiving rates under an increased risk of predation. In contrast, receiving signals from unknown individuals or from neighbours along a contested territorial boundary will provide valuable information on the probability of attempted sneaky copulations with nearby females or the threat of territorial intrusion. Receivers should therefore confer greater value to these signals and be willing to risk more to receive them (see also Box 4). As the majority of signals within an individual's immediate environment are likely to be from familiar individuals, we predict that this familiarity provides receivers with a mechanism to reduce risky (receiving) behaviours under an increased risk of predation.
Individual perceptions of state
An extensive body of research has highlighted the importance of an organism's state (e.g. energy reserves) and life history on the balance between foraging and risk-taking behaviour (see Alonzo 2002 and references within). Hungrier prey take greater risks when foraging, for example, spending more time searching for food and foraging in riskier habitats (e.g. Berger-Tal et al. 2010). Similar effects of state, including an individual's age, dominance status or probability of future reproduction are expected to affect trade-offs between and reproduction and risk-taking behaviour (Magnhagen 1991), including the reception of risky signals. In support of this idea, Lafaille et al. (2010) recently demonstrated that both acoustically signalling male moths (Achroia grisella) and phonotactically orienting conspecific females took progressively greater risks to locate mates as their probability of future reproduction declined. If individuals with high current reproductive success are willing to take fewer risks in the presence of a predator cue, then this might also provide an opportunity for less competitive rivals to move in and contest a territory or pursue illicit mating opportunities.
Most studies of risks to receivers (examined both directly and indirectly) have focused on mating signals and the reproductively active individuals who should be highly motivated to receive them. This focus naturally reflects the role of sexual selection research on the development of communication and signalling theory, and points to why there have been fewer investigations of non-reproductive individuals. In some species, however, non-reproductive individuals also actively receive signals emitted ultimately for mating purposes; the active inspection of conspecific scent signals by juvenile and subadult house mice (Hurst 1989) is just one example. Non-reproductive individuals generally do not attend as closely to sexual signals as other, more mature conspecifics, however, and we predict that they should also be less willing to bear significant risks to receive them. Similarly, passive receivers should be less willing to accept risks than those actively engaged in receiving. Behavioural responses to risky signals might therefore be stronger in some non-reproductive individuals than in reproductively active individuals and in some passive receivers as opposed to actively receiving individuals, although this remains to be tested (but see earlier discussion on interactions between risk taking behaviour and the probability of future reproductive opportunities). Individuals within juvenile or subadult categories often numerically dominate prey populations, thus avoidance of risky signals by these animals may have implications for overall prey distributions, with knock-on effects for predator foraging success.
Risk allocation and predator foraging strategies
Studies of signaller and receiver behaviour under predation risk have traditionally imposed a constant, often dichotomous level of risk, such as the presence or absence of a predator or their cue. Predation risk is rarely constant, however, and instead varies enormously across spatial (e.g. with habitat complexity) and temporal (e.g. across seasons, lunar cycles, or daily temperature fluctuations) scales. Lima & Bednekoff (1999) developed the ‘risk allocation hypothesis’ to examine how animals feeding under temporal variation in predation risk should optimise their anti-predator behaviour across different states of risk. This concept has since received substantial attention, but mixed support within the optimal foraging literature (see Ferrari et al. 2009 for a review). Yet to our knowledge, the effect of risk variability has not been explicitly examined for animal communication. Nonetheless, if foragers and signal receivers respond similarly to variation in predation risk, then the risk allocation hypothesis predicts that individuals should stop receiving during brief periods of high risk, instead concentrating signal reception during extended periods of low risk. However, if periods of high risk are frequent or extended, then prey are predicted to invest heavily in receiving during brief periods of low risk, although they may also be forced to receive signals under high-risk conditions.
Temporal and spatial variation in risk will also change with the identity, habitat preference and hunting mode of predators affecting a prey species. Schmitz et al. (2004) suggested that a predator's hunting mode and level of activity will provide prey with different cues to their presence and that this should affect the prey's habitat use. Cues from less mobile sit-and-wait predators may provide a better indication of the immediate risk of predation, for instance, than cues from actively hunting predators that roam widely. If this is the case, then we predict similar effects of a predator's hunting mode on prey receiving behaviour, including both the probability that the prey will alter its receiving behaviour and the optimal receiving and predator avoidance strategies it adopts. The study of predator facilitation and small rodents provides support for this. Field voles, for example, adjust their microhabitat use and activity level according to the foraging strategy of the predators present (Korpimäki et al. 1996). Specifically, the voles shift their activity to open microhabitats in the presence of scent-hunting least weasels, but utilise densely vegetated microhabitats and reduce their activity in the presence of kestrels (Falco tinnunculus), or when both kestrels and weasels are present (Korpimäki et al. 1996). We predict that similar switches in receiving behaviour (e.g. changes in receiving rate according to location) should be observed in the context of predators that utilise different hunting modes.
Implications of optimal receiving for signallers
Our review reveals that rather than an ‘all or nothing’ approach to receiving under an increased risk of predation, receivers have a multitude of options available to them to optimise their receiving behaviour. Adjustments in the timing, length and location at which prey choose to receive signals lie at the core of these options and, where signals are related to sexual advertisement, it is obvious that the particular behaviour(s) adopted to optimise receiving will constrain the evolution of signalling behaviour (Jennions & Petrie 1997). Likewise, the receiver strategies outlined here cannot be considered in isolation from signaller behaviour. Sexual conflict theory predicts that signallers will have evolved a range of ways to counteract the effects of ‘optimal’ receiving behaviour where these behaviours are also not optimal for a signaller (Arnqvist & Rowe 2005). Indeed, the examples of altered receiving behaviour under predation risk that we have provided throughout this review (Table 3) suggest that this will represent the majority of cases. It is unfortunately beyond the scope of this review to discuss in detail the consequences for signallers or the strategies they might adopt to counter receiver trade-offs, however, and we instead refer readers to other relevant works (e.g. Jennions & Petrie 1997; Ryan 1998; Kokko et al. 2003; Arnqvist & Rowe 2005; see also Dall et al. 2005; Schmidt et al. 2010).
Implications of optimal receiving for predator behaviour
While we have considered how receivers might optimise receiving behaviour under increased risk from predators, the evolution of these strategies is inextricably linked to the behaviour of conspecific signal emitter(s), eavesdropping predator(s) (and potentially other heterospecifics), environmental attributes (e.g. habitat complexity) and the dynamic interplay between these variables across a variety of scales. However, we found that few studies have investigated the consequences of behaviourally responsive prey and predators on prey signalling and receiving, or on predator's use of these signals when searching for prey (see also Lima 2002). Consequently, although Mitchell & Lima (2002) proposed a decade ago that the optimal strategy for prey depends on the predator's behaviour, we still know little about how predators use prey signals when searching for prey or the scale at which they perceive patchily distributed signals. Research on the scale at which prey should adjust their anti-predator behaviour to responsive predators is also lacking. For the vole, weasel and kestrel system referred to earlier, for instance, the scale at which weasels and kestrels perceive the conspicuous scent marks of their vole prey should determine the predators' foraging strategies (e.g. purposeful unpredictability, see Roth & Lima 2007). These strategies will in turn influence if and how voles alter their receiving (and other) behaviour at the scale of a single scent mark, a patch of marks (e.g. around preferred food resources) and/or at the level of a home range or larger; the relative risks to actively and passively receiving voles will also be affected. Subsequent changes in the spatial distribution and behaviour of voles and their scent signals will then feed back into the predator's foraging strategy. Research on dynamic interactions such as these will provide insights into the mechanisms driving short-term interactions between predators and prey, and also provide further detail on the long-term processes governing the evolution of behavioural defence and optimal receiving strategies in both intended and eavesdropping individuals.
The bias of empirical research on costly communication towards sexual signals has largely restricted our understanding of risky receiving to attractive mating signals. Nevertheless, we have shown here that a number of signal properties are likely to place prey at an increased risk of predation while receiving a variety of different signal types (i.e. not only those used for mate attraction). This general cost of communication should generate strategies that act to optimise receiving decisions under risk of predation, the particulars of which will vary with a receiver's position within a communication network. These insights have implications for the evolution of signalling systems, the maintenance of dominance hierarchies and state- and risk-sensitive mate choice. In particular, the processes we have outlined in this review suggest that there is still much to be learnt from explicitly incorporating the balancing act in which receivers, and not just signallers, find themselves while engaging in social communication.
We thank three anonymous referees for their thoughtful comments and excellent insights on earlier versions of this manuscript. NKH is presently a postdoctoral fellow with the Fund for Scientific Research – Flanders (FWO). JLK would like to acknowledge the Universities of New South Wales and Western Australia for postdoctoral fellowship funding. PBB was supported by funding from the Australian Research Council (DP0881455).
Statement of authorship
All persons entitled to authorship have been included. All authors have read and approved the submitted version of the manuscript and are aware they each hold responsibility for the accuracy, integrity and ethics of the manuscript and work described therein. All authors participated in extensive discussion and literature review before we wrote this paper. NKH wrote the first draft of the manuscript and all authors contributed substantially to revisions.
Indirect evidence that receiving carries a risk of predation
In addition to the direct evidence presented here, there is also substantial indirect or anecdotal evidence that receiving individuals are exposed to a risk of predation. An early example is that of Belwood & Morris’ (1987) report that foliage-gleaning bats were attracted to calling male katydids, however, at least half of the katydid prey that were consumed were non-calling females. Predators may also focus their attention on receivers rather than signallers if they are a more profitable prey item. The pike cichlid (Crenicichla alta) is attracted by the colourful mating displays of male guppies (Poecilia reticulata), for example, but in mixed-sex groups it is the drab, but larger females upon which they selectively prey (Pocklington & Dill 1995). Studies where females reduce their levels of choosiness for males under risk of predation (e.g. Forsgren 1992; Grafe 1997; Booksmythe et al. 2008; Rundus et al. 2011), reduce or alter mate sampling (e.g. Koga et al. 1998; Bonachea & Ryan 2011) or their attraction to otherwise preferred, conspicuous males after exposure to a predator (e.g. Hedrick & Dill 1993; Gong & Gibson 1996; Johnson & Basolo 2003), or increase thresholds of mate attractiveness for mating after exposure to predator cues (Demary et al. 2006) also suggest that receiving male signals incurs a risk of predation.
Signals other than those used for mate attraction may also elevate a receiver's risk of predation. Checkered beetles (Thanasimus dubius), for example, are attracted to the aggregation pheromones of their pine beetle (Ips pini) prey. However, these predators consume more of the pine beetles that are attracted to the pheromones than they do on the pine beetles that emit them (Aukema & Raffa 2004). Although counter-intuitive, some alarm signals may actually increase a receiver's risk of predation. Alarm signals function to alert conspecifics (and also heterospecifics, e.g. Templeton & Greene 2007) to a predator, prompting a variety of anti-predator behaviours such as mobbing, grouping and fleeing. Some predators, however, use the alarm calls elicited in response to another predator to locate prey patches. These attracted predators then wait until the prey alarm response (and therefore heightened vigilance) has subsided before striking (Krama & Krams 2005).
Olfactory communication in house mice
The complex of odours emanating from a territorial male house mouse's scent marks has previously been described as the chemical equivalent of a peacock's tail as these scents are both attractive and costly to produce (Penn & Potts 1998). Conspicuous, volatile compounds in scent marks alert conspecifics to the mark's presence, communicating the sexual and social status, health and diet of the signalling individual (Hurst & Beynon 2004). Reception of these volatile signals may promote further investigation of the scent, or prompt a withdrawal from the scent-marked area, depending on the status of both the signaller and receiver, and the history of interaction between them (Hurst & Beynon 2004). Yet, while volatiles are responsible for alerting a receiver to a scent mark's presence, direct contact with the scent is required to receive information detailing the identity of the signalling individual (Nevison et al. 2003).
Males advertise their dominance by distributing their marks throughout their territory and especially along its boundaries. However, scent marks are relatively ineffective in excluding competitors, who readily transgress a competitor's territory boundaries and deposit their own scent marks. The presence of intruder scent marks weakens a resident male's perceived dominance, and intruder scent marks are therefore countermarked immediately and repeatedly, especially along territory boundaries (see Hurst & Beynon 2004 for a review).
Although the majority of marks are deposited by dominant, territorial males, all individuals investigate and deposit scent marks to some extent (Hurst 1990a); to avoid doing so is to risk injury or death from conspecifics. Olfactory communication in house mice therefore promotes the conspicuous concentration of signals and activity that is likely to attract the attention of passing predators (Hughes et al. 2010b). This suggests that all individuals visiting a scent mark, whether they intend to deposit an additional mark themselves or otherwise, are placing themselves at an increased risk of predation from simultaneously attracted predators. Thus, while individual dominant males might incur costs for the production and maintenance of their conspicuous scent signals, all conspecific receivers suffer an increased risk of predation while investigating the dominant male's (and others') signals.
Parallels between predation risk and infection risk
In addition to pathological effects, many infections induce a range of physiological and behavioural changes relevant to social signalling. Infection can reduce testosterone levels (Wingfield et al. 2001) and scent-marking rates (Zala et al. 2004) in males, and the chemical profile of odours emanating from an infected individual's body, saliva or excretory products may also be altered (Penn & Potts 1998). These infection-induced changes in odour can reduce the sexual attractiveness of a male's scent to females (e.g. Kavaliers & Colwell 1995; Penn et al. 1998), and males also alter their response to the urine odours of other males infected with endoparasites or ectoparasites (Kavaliers et al. 2004).
Many infectious agents are shed and potentially transmitted via the same excretory products used for social signalling (e.g. Kallio et al. 2006). Several authors have pointed out that infection-induced changes in odours (and especially volatiles) could therefore warn individuals of a potential source of infection, facilitating the avoidance of infected individuals or potential contaminants such as urine (e.g. Able 1996). However, if olfactory signals provide cheat-proof displays of an individual's health (Able 1996; Penn & Potts 1998; Zala et al. 2004), then they should also be used to assess a rival's ability to defend its territory. As with the trade-off between predation risk and communication, selection would predict a trade-off between a competitor's willingness to investigate an infected odour and risk infection themself, with the potential competitive advantages of investigating and overmarking infected scents. If this occurs, we would predict that receiving rates of olfactory signals from infected individuals would be maintained in those individuals that have the most to lose from not doing so or the most to gain by taking the risk. Although we are not aware of any studies which have examined this hypothesis directly, higher infection rates have been seen in old, breeding males (e.g. Tersago et al. 2011), a pattern that would be predicted in species with male territorial dominance and associated high rates of scent marking and investigation. Continued, deliberate exposure to infected social signals may therefore be another factor behind the male-biased parasitism seen in some species (e.g. Schalk & Forbes 1997). The prevalence of such a strategy and its consequences for social processes such as territoriality, however, are yet to be examined.