Properties of the prey
Within prey guilds, species employ various means to detect (Weissburget al . 2014), evade (Moore & Biewener 2015), and resist (Creel 2011) predators. Modes of detection (acoustic, chemical, olfactory, visual, tactile) enable prey to identify risky places, for example by quantifying spatial variation in the intensity of persistent predator cues; and risky times, as when a predator’s approach is observed (Creelet al . 2008). Sensory modalities for perceiving and responding to risk are a critical source of contingency during phase one (Fig. 1 ). Prey species may lack the capacity to detect persistent evidence of a predator’s presence and thus to prepare for encounters, or instances when predator-prey spatial overlap is such that detection of one by the other is possible (Lima & Dill 1990). Alternatively, their preparation for encounters may be continuous and generalized, leading to high fitness costs and reduced efficacy (Creel et al . 2008; Creel 2018). Similarly, inability to sense the approach of a predator limits reactive responses to those triggered by an attack (e.g., physical resistance; Creel 2018). In sum, then, consideration of sensory biology should aid in predicting which members of prey guilds are least likely to be subject to non-consumptive (versus consumptive) effects of a predator, and which kinds of risk stimuli (background versus immediate) are most likely to induce defensive responses by a given prey species.
The kinds of sensory modalities used to perceive predation risk should also shape the propagation of NCEs during phases two and three (Fig. 1 ). First, different sensory modalities may mediate the type and intensity of information transferred from a risk cue to prey (Weissburg et al . 2014). Thus, sympatric prey species that use different senses to detect the same predator may respond with divergent intensity and/or specificity depending upon the pathway through which they receive and process the information (Weissburg et al . 2014). The threat level and predator identity perceived by a given prey species could influence its response and any associated risk effects (including from stress) during phase two, as well as any indirect interactions precipitating during phase three. Second, prey with multiple sensory modalities may be better able to detect predators and have an anti-predatory advantage (Munoz & Blumstein 2012). For example, access to both visual and chemical cues allowed for more accurate detection and appropriate responses to predators by mosquito fish (Gambusia holbrooki ) (Ward & Mehner 2010). Thus, members of prey guilds with multiple sensory modalities may exhibit more striking and appropriate anti-predator responses, higher vulnerability to risk effects, and greater capacity to transmit indirect NCEs to other community members than sympatric heterospecifics relying on a single means of detection.
Although some may double as routine safeguards, tactics for evading and resisting predator attacks are typically reactive countermeasures triggered by encounters with predators (Creel 2018). Thus, these ‘escape behaviors’ (Heithaus et al . 2009; Wirsing et al . 2010) usually act as drivers of contingency during the latter two phases of non-consumptive interactions, after risk is perceived. Evasive behaviors are diverse and include altered activity (Schmitz 2007), body part autotomy (Maginnis 2006), dynamic flash coloration (Murali 2018), feigning death (Humphreys & Ruxton 2018), fleeing (Moore & Biewener 2015), grouping (to confuse predators or dilute risk; Lehtonen & Jaatinen 2016), hiding/crypsis (Caro 2014), and seeking a refuge (Sih 1987). Their efficacy can be prey- and predator-specific and hinge on environmental features (Heithaus et al . 2009; Wirsing et al . 2010; Schmitz 2017; Creel 2018). The effectiveness of flash coloration as a means of visually confusing predators, for example, can depend on visual obstructions, light levels, and background colors (Murali 2018). To the extent that prey have scope to modify the effectiveness of their escape tactics, interspecific variation in evasive behaviors may lead to differences in anti-predator responses to the same risk stimuli during phase two. For example, sympatric prey species that flee predators with disparate means of locomotion may respond divergently to a shared predator by proactively seeking areas that suit their respective movement styles in preparation for an encounter or reactively shifting to these areas after an encounter has occurred. Consistent with this expectation, mule deer (Odocoileus hemionus ) and white-tailed deer (O. virginianus ) exhibited divergent proactive shifts to terrain suiting their respective running gaits (bounding and galloping) when exposed to gray wolves (Canis lupus ) (Dellinger et al . 2019; Fig. 2 ). A similar scenario characterizes NCEs of tiger sharks (Galeocerdo cuvier ) on a community of vertebrates in the seagrass ecosystem of Shark Bay, Australia (Heithaus et al . 2012; Fig. 3 ). By implication, during phase three, a predator targeting more than one sympatric prey species could impose multiple indirect effects on other community members (e.g., basal resources for the different prey species) that occur because of prey-specific forms of evasion with divergent consequences for distribution (Wirsing & Ripple 2011). This possibility has not been tested.
Forms of prey resistance may discourage predators prior to an attack or repel an attacker. Resistance may include cooperative defense (Lehtonen & Jaatinen 2016), induced chemical defense (Mukherjee & Heithaus 2013), fighting back (Mukherjee & Heithaus 2013), and honest (e.g., aposematism, pursuit deterrence; Harvey & Paxton 1981; Caro 1995) and deceptive signaling (e.g., actions that make an individual seem more difficult to capture such as increases in apparent size, mimicry; Caro 2014). As with evasion, the efficacy of resistance may be predator- and setting-specific (Mukherjee & Heithaus 2013). Chemical defenses of herbivorous insects, for example, are more effective against vertebrate than invertebrate predators, perhaps because of the latter group’s enhanced capacity to develop adaptations to tolerate or overcome prey defenses (Zvereva & Kozlov 2016). Unlike evasive behaviors, however, resistance usually manifests after the predator detects the prey, and often after an attack has been initiated. Rough-skinned newts (Tarichia granulosa ), for instance, show little behavioral response to predators (Murray et al . 2004) save to honestly signal by displaying the bright coloration of their underbelly when accosted by a would-be attacker. Hence, these countermeasures are less likely than evasion to result in either costly risk effects (e.g., diminished condition after prolonged foraging disruption) or in changes to prey activity budgets and distributions during phase two (e.g., displacement) that could indirectly affect other species during phase three. For example, adult moose (Alces alces ), which can fight back effectively against wolves (C. lupus ), show little spatial response to wolf presence (Nicholson et al . 2014). Not surprisingly, observed indirect effects of wolves on the plants that moose consume appear to be transmitted primarily by the numerical effects of direct predation rather than NCEs (Post et al . 1999). By implication, prey species relying on resistance should respond differently to predation risk, and to be less likely to be vectors of indirect NCEs, than those depending on evasive behaviors. There are studies supporting the former expectation (e.g., Lingle & Pellis 2002) but it has not been addressed broadly. The latter expectation remains unexplored.
Within populations, prey state may shape individual responses to predation risk and, consequently, propagation of NCEs (Sih et al . 2015; Schmitz 2017). States can be relatively stable (e.g., sex, behavioral type, and epigenetically or genetically derived morphs) or dynamic (e.g., age/developmental stage, current behavior, disease state, learning, nutritional condition, residual reproductive value, and stress level). An individual’s state can influence its risk taking behaviors in any of three ways. First, an individual’s capacity to recognize danger may be state-dependent, as when prey acquire the capacity to detect and respond appropriately to cues via development/growth and learning (Kavaliers & Choleris 2001). For example, large bumblebees (Bombus terrestris ) are more sensitive to spider risk while visiting inflorescences, likely (at least in part) because they possess eyes with greater visual acuity than smaller conspecifics (Gaviniet al . 2019). Working with fathead minnows (Pimephales promelas ), Ferrari et al . (2006) showed that individuals learned to recognize northern pike (Esox lucius ) as predators from a paired exposure to conspecific alarm pheromones and pike odor. Once learned, a minnow’s fear response increased with the concentration of pike odor alone. Not surprisingly, therefore, naïve individuals often differ markedly from experienced conspecifics in terms of whether (phase one) and how (phase two) they respond to predation risk (Sih et al . 2010). This form of experience-driven contingency in defensive behaviors could give rise to differences in the extent to which individuals (and populations) with divergent amounts of prior predator conditioning transmit indirect NCEs (phase three).
Second, prey state may affect vulnerability, as when individuals in different growth stages are differentially able to outpace (Diamondet al . 2019) or resist (Schmitz 2017) predators. Thus, against any predator, individuals in less susceptible states should have reduced need to invest in countermeasures and, consequently, respond differently to perceived risk than more vulnerable conspecifics during phase two. For example, Christensen (1996) observed that juvenile roach (Rutilus rutilus ) that were beyond the gape limits of their predators invested less in defense (time spent near the surface and jumping out of the water when at risk) than smaller (ingestible) conspecifics. Similarly, Dannock et al . (2019) found that blue wildebeest (Connochaetes taurinus ) eschewed chewing while being vigilant following lion (Panthera leo ) playbacks, presumably because mastication hampers predator detection. Thus, the overall pattern of anti-predator behavior characterizing a prey population during phase two, and the degree to which it transmits indirect NCEs during phase three, could hinge on the distribution of states manifested by its constituents. Indeed, where prey switch ontogenetically from being the prey to being the predator of another species (Ferrariet al . 2010), relative abundance of different developmental stages within a population could mediate the extent to which it experiences and transmits versus initiates NCEs. These hypotheses have yet to be evaluated systematically.
Third, a prey’s state may influence its willingness to respond to perceived risk, as when individuals with risk-prone behavioral types are less likely to invest in anti-predator behavior (Michalko & Řežucha 2018) or those with compromised energetic state are more willing to expose themselves to danger to avoid starvation (Clark 1994). The former mechanism is gaining support in the literature (Réale et al . 2007; Mittlebach et al . 2012; Sih et al . 2015; Moranet al . 2017). The latter, known as state-dependent risk taking, has long been recognized and is thoroughly explored in a range of taxa (e.g., Box 1 ). Both have consequences for levels of anti-predator investment and subsequent predation rates experienced by prey during phase two. For example, bold mud crabs (Panopeus sapidus ) exhibit lower refuging times relative to shyer conspecifics following exposure to predator cues, and consequently experience higher predation from blue crabs (Callinectes sapidus ) (Belgrad & Griffen 2016). Rainbow trout (Onchorhynchus mykiss ) with reduced access to food take greater risks to achieve growth and, consequently, suffered increased predation mortality (Biro et al . 2005). Thus, the extent to which any prey population is subject to consumptive versus non-consumptive predator effects may depend on its average behavioral type (Sih et al . 2004; Moran et al . 2017) or its mean energetic state (Anholt & Werner 1995; Heithaus et al . 2008; Matassa & Trussell 2014). These scenarios have only rarely been assessed under large-scale field conditions (e.g., Sinclair & Arcese 1995). The additional inference that mean temperamental or energetic states should influence the transmission of indirect NCEs in communities has, to our knowledge, not been addressed.
Finally, prey may possess constitutive (permanent) defenses that influence risk-taking behavior including armor, harmful morphology (e.g., spines), toxicity/unpalatability, and honest or deceptive advertisments of similarity to toxic/unpalatable heterospecifics (Tollrian & Harvell 1999). In theory, the effectiveness of these defenses should be inversely proportional to the need for anti-predator behavior (Dewitt et al . 1999). Freshwater snails (Physa gyrina ) with vulnerable shell shapes, for instance, exhibited greater behavioral responses (refuging, avoidance) than harder-to-kill conspecifics when confronted by cues from crayfish (Orconectes rusticus ) (Dewitt et al . 1999). By implication, taxa that are well defended constitutively should exhibit weaker anti-predator responses than other community members with less effective constitutive defenses during phase two, whether or not cues are detected in phase one, and be less likely to transmit indirect NCEs during phase three. However, the effectiveness of any constitutive defense is, itself, context dependent. For example, Pokallus & Pauli (2016) observed that, despite possessing a well-developed predator deterrent (quills), porcupines (Erethizon dorsatum ) altered their movements to reduce risk from fishers (Pekania pennanti ), a specialized porcupine predator. Hence, even prey with generally effective constitutive protections may react to and transmit indirect NCEs elicited by predators that can, under some circumstances, breach their defenses.