Moving forward
In a recent review, Sheriff et al. (2020) emphasized the need to better understand how ecological and environmental context interact with prey responses to predation risk. Focusing on anti-predator behavior, we address this knowledge gap in two ways. First, our review sheds new light on NCEs by showing when and how contingency can arise from properties of the prey, the predator, and the setting as these effects unfold across three phases (prey risk perception; prey responses to perceived risk; impacts of these responses on other species). Second, our synthesis of the ‘hunting mode-habitat domain’ and ‘evasion landscape’ concepts offers a unified framework for predicting the form and magnitude of anti-predator behavior during phase two. Looking ahead, we highlight two knowledge deficiencies that require attention if we are to develop a coherent framework for predicting how NCEs propagate through ecosystems. First, there is insufficient exploration of context-dependent indirect NCEs during phase three. Second, there is need for research focused on the ways in which direct and indirect NCEs are shaped simultaneously, or even interactively, by multiple drivers of context dependence.
Drawing from a broad literature spanning diverse taxa and ecosystems, our review reveals how contingencies in NCEs can arise as a result of many factors. It is hardly surprising, then, that studies have revealed so much variation with respect to whether, and in what way, NCEs manifest in communities (Moll et al . 2016; Gaynor et al . 2019; Prugh et al . 2019). We clarify these factors by grouping them into three broad categories: (1) prey properties influencing detection of and responses to risk; (2) predator properties shaping their detectability and lethality; and (3) properties of the setting influencing the prey’s scope for predator detection and countermeasures. We also emphasize that there is great potential for interplay among them. For example, divergent responses to predators with disparate hunting modes could disappear if declining food supply limits prey capacity for defensive investment. Similarly, because prey often have multiple defenses whose efficacies are context-specific (Brittonet al . 2007; Wirsing et al . 2010; Creel 2018), sympatric prey may respond divergently to a shared predator in one setting but similarly in another, depending on the availability of landscape features facilitating particular responses (i.e., the evasion landscape). Moreover, the latter two give rise to an emergent fourth driver, (4) the timing of predation risk, and prey properties then determine how individuals respond to this temporal dimension of danger (Box 3 ). By implication, predictions based on one driver of contingency, or a single NCE pathway (Preisser & Bolnick 2008), may provide an incomplete picture of the impacts of predation risk on prey populations and communities. Rather, examination of NCEs requires thorough consideration of the functional properties of interacting predator and prey species, as well as the circumstances under which these interactions occur (Heithaus et al . 2009; Creel 2011; Schmitz 2017). Fortunately, many of these natural history or environmental details are attainable (Wirsing et al . 2010), especially given new approaches (e.g., animal-borne video, camera traps, drones) that facilitate placing behavioral data in context (Mollet al . 2007; Wirsing & Heithaus 2014).
Our review also highlights the staged manner in which NCE contingencies can manifest. Namely, prey anti-predator investment may vary intra- and inter-specifically as a function of differences in sensory perception (phase one) and the form of any deployed countermeasures (phase two); contingent outcomes during either of the first two phases then determine if, and how, indirect NCEs emerge during phase three. Across taxa, then, prey with greater sensory ability should experience and transmit larger NCEs. Furthermore, the phase in which context dependence arises shapes how the outcome of non-consumptive predator-prey interactions will respond to perturbation. For example, landscape changes that reduce prey sensory ability are likely to diminish NCEs, whereas those raising the frequency of encounters with predators by restricting prey habitat domains may elicit increased anti-predator defense during phase two (Schmitz et al . 2004) and elevate the potential for indirect NCEs in phase three. Thus, studies exploring phase-specific mechanisms by which prey, predator, and landscape properties shape anti-predator investment are crucial to forecasting NCEs in a changing world.
By synthesizing the work and concepts of Heithaus et al . (2009) and Schmitz et al . (2017a), we present a new framework that integrates prey, predator, and landscape traits to anticipate the form and magnitude of anti-predator behavior. This framework is broadly applicable, as evidenced by its ability to retrospectively explain differences in behavioral countermeasures that have been observed in the field across a range of taxa. Consistent with scenario one (Fig. 4c ), for example, prey species whose habitat domains are nested within those of tiger sharks manifest chronic vigilance and space use that facilitates their escape strategies (Heithaus et al . 2012), save when in depressed energetic states (Heithaus et al . 2007). Similarly, white-tailed deer whose domains fall within the larger movements of gray wolves exhibit space use changes within their home ranges facilitating their means of predator evasion (Dellinger et al . 2019). By contrast, sympatric mule deer practice chronic predator avoidance by shifting to refugia within their domains that are little used by wolves (scenario three; Fig. 4b ). For both ungulates, the consumptive effects of wolves appear to be limited (Dellingeret al . 2018). In the Greater Yellowstone Ecosystem, USA, elk (Cervus canadensis ) and wolves have large, overlapping domains, leading to low encounter rates (Cusack et al . 2020). Thus, consistent with scenario four (Fig. 4d ), elk in this system appear to predominantly experience the consumptive effects of wolves (Peterson et al . 2014) and typically exhibit evasive behavior only during risky times (e.g., Cusack et al . 2020). Larger elk survive many encounters with wolves via resistance (Mech et al . 2015), further contributing to their tendency to experience consumptive rather than non-consumptive wolf impacts. In an African system with multiple sympatric predators, prey consistently select for habitats offering a lower probability of lethal predator encounters, suggesting that chronic evasive behavior (under scenarios one and three) may be common where overlapping predator domains preclude outright avoidance (Thaker et al . 2011). Accordingly, it underscores characterization of habitat domains and evasion landscapes as a critical first step in forecasting the extent to which, and how, prey should respond behaviorally to perceived risk during phase two and transmit indirect NCEs in phase three. Our framework also highlights the need to discriminate among prey individuals relying principally on evasion versus resistance, given that prey expressing the latter group of behaviors are less likely to respond to the threat of predation unless the cue is acute and, consequently, to experience and transmit NCEs. Finally, it gives rise to new hypotheses. For example, in any scenario where predators cannot be avoided spatially and encounters are high enough to warrant anti-predator investment, we might nevertheless expect vigilance and space use that facilitates evasion to relax in prey species that are instead able to avoid the predator(s) temporally (Kohlet al . 2019).
Our survey revealed two knowledge gaps that represent fruitful directions for future research. First, whereas there is ample evidence for context dependence during phases one and two, few studies have rigorously examined contingency in the propagation of indirect NCEs. There are notable examples, including the role of predator hunting mode in shaping indirect NCEs of spiders on plant and soil properties (Schmitz et al . 2017b), and the impact of prey refugia on indirect non-consumptive relationships between crabs and barnacles (Trussell et al . 2006). These studies offer a template for expanded scrutiny of contingencies in NCEs during phase three, which will improve our understanding of when and how predators initiate indirect effects by altering prey traits.
Second, a growing literature underscores the importance of simultaneously considering multiple drivers of contingency in NCEs. For example, anti-predator investment by mud crabs varied with their personality (bold versus shy) and predator hunting mode (actively hunting blue crabs versus sit-and-wait toadfish, Opsanus tau ) (Belgrad & Griffen 2016). Thaker et al . (2011) showed that small members of an African ungulate guild avoided all predators whereas their larger counterparts avoided sit-and-pursue but not active hunters. More work is needed, however, particularly on the importance of three-way interactions among factors drawn from the aforementioned groups.
There are also studies suggesting that interactive impacts of multiple contingent drivers may act collectively to shape indirect NCEs during phase three. For example, Murie & Bourdeau (2019) speculated that, compared to the strong effects initiated by slow-moving sea stars, the absence of direct and indirect non-consumptive effects of crabs and octopuses on snail grazing and kelp, respectively, might owe to the inability of snails to escape these vagile predators. Thus, more mobile prey species with greater scope for avoidance may have responded equivalently to all three predators, yielding similar rather than predator-specific cascades of NCEs. The possibility that interactions between context dependent factors might modify cascading NCEs has not been tested empirically, however, and thus remains as an exciting research frontier.