Figure Legends
Fig. 1 . Flow chart, adapted from Figure 1 in Lima & Dill (1990), conceptualizing the process by which direct and indirect non-consumptive predator effects (NCEs) may manifest. (a ) Phase one. Each point in space and time is characterized by some value of intrinsic predation risk, or danger, defined after Lank & Ydenberg (2003) as the inherent probability that an individual will become a prey item given no, or a standard amount of, anti-predator investment. Danger may or may not be perceived; in the latter case, no NCE will precipitate from the danger cue in question. (b ) Phase two. Given that the forager perceives risk cues, does it respond? Danger that is perceived may nevertheless fail to elicit a response of sufficient magnitude to trigger a NCE. Though not the focus of this review, prey individuals that do respond to perceived danger may experience stress, which may in turn affect fitness and consequently lead to risk effects. Furthermore, prey individuals that perceive danger may seek to manage their risk of predation through behavioral modifications, whose costs in terms of time and energy determine the magnitude of any associated risk effects. (c ) Phase three. Given that the forager responds to the cues, does the response induce an indirect interaction? Risk effects flowing from predator-induced stress and risk management can reduce prey population size and, in turn, trigger indirect interactions if changes to prey abundance affect other members of the community. The nature and strength of predator-induced risk management by prey can also determine whether and how other species in the community are affected indirectly; namely, if additional species are impacted by prey risk management, then NCEs can propagate through communities in the form of indirect interactions that are transmitted by prey behavior.
Fig. 2 . Observed (solid arrows) and hypothesized (dashed arrows) relationships between gray wolves (Canis lupus ) and two sympatric ungulates – mule deer (Odocoileus hemionus ) and white-tailed deer (O. virginianus ) – in areas of eastern Washington, USA, located outside (a , c ) and inside (b , d ) wolf pack territories. Non-consumptive effects of wolves on prey behavior (relative to wolf-free sites; a ) are depicted in b , whereas c and d display baseline and wolf-influenced tropic relationships between the herbivores and the plants they target, respectively. Increasing effect size corresponds with arrow thickness. Mule and white-tailed deer are morphologically similar but have different running gaits (Lingle 1993). When threatened, mule deer flee by stotting, a bounding gait that limits speed on flat ground but facilitates navigation of uneven terrain and obstacles. White-tailed deer flee danger by galloping, a swift means of moving over gentle terrain that is less effective where the ground is more sloped or broken. This disparity explains differences in the space use of these two deer species that emerge when they are exposed to the risk from wolf predation during phase two (b versus a ). Working in a system in eastern Washington, USA, Dellinger et al . (2019) found that wolf presence elicited elevated use of sloped terrain by mule deer (b ; heavy arrow), presumably because the uneven ground characterizing these uplands confers an advantage to bounding prey seeking to escape coursing wolves. White-tailed deer space use differed comparatively little as a function of wolf presence, with individuals exposed to wolf risk manifesting small-scale shifts within their home ranges toward flat ground and roads that actually led to increased overlap with wolves (b ; thin arrow). By inference, white-tailed deer were able to manage risk ‘in place’ because of spatial synchrony between the effectiveness of their galloping means of escape and the space use pattern of their coursing predator. Notably, this form of risk management is expected whenever the safety benefits of matching predator distribution that accrue from escape facilitation outweigh the costs associated with elevated encounter probability (Lima 1992). These divergent anti-predator responses raise the possibility of recolonizing wolves triggering prey-specific indirect NCEs on plants during phase three (c versus d ). In this ecosystem, mule and white-tailed deer exhibit considerable dietary overlap, though mule deer rely more heavily on upland shrubs (e.g., serviceberry;Amelanchier spp), and white-tailed deer exploit lowland riparian vegetation (e.g., willow; Salix spp.) to a greater degree (A. Craig, unpublished data ). Given that they elicit broad-scale spatial shifts by mule deer, wolves may dampen the impact of mule deer on lowland plant species (d ; thin dashed arrow) while strengthening this species’ effects on upland plants growing in areas with steeper slopes (d ; thick dashed arrow). By contrast, the absence of a strong spatial response by white-tailed deer in areas occupied by wolf packs suggests that wolves may have modest and localized (i.e., within existing home ranges) indirect effects on the plants exploited by this deer species (similarity in the thickness of the solid and dashed arrows in c and d ).
Fig. 3 . Observed (solid arrows) and hypothesized (dashed arrows) relationships between tiger sharks (Galeocerdo cuvier ), their air-breathing prey – dugongs (Dugong dugon ), dolphins (Tursiops cf. aduncus ), green turtles (Chelonia mydas ), sea snakes (Disteria major ; not pictured), pied cormorants (Phalacrocorax varius ) – omnivorous fish (Pelates octolineatus ), and seagrasses within shallow (<4.5m water depth) habitats in Shark Bay, Western Australia. Species interactions are depicted during times when tiger sharks are present and absent from the bay, and interaction effect sizes correspond with arrow thickness. When tiger sharks are present, they preferentially spend time over shallow banks (Heithaus et al . 2002). Within these shallow habitats, they spend more time over bank edges compared to interior areas of banks (Heithaus et al . 2006). Non-consumptive direct effects of sharks on prey behavior (phase two) are black lines, whereas indirect relationships between tiger shark prey and lower trophic levels are gray lines (phase three). Dugongs (Wirsing et al . 2007), cormorants (Heithaus et al . 2009), dolphins (Heithaus & Dill 2006), and sea snakes (Wirsing & Heithaus 2009) distribute themselves between edge and interior portions (microhabitats) of shallow banks roughly proportional to food abundance when tiger sharks are absent. When sharks are present, by contrast, these species, along with green turtles (Heithaus et al . 2007), shift among the two microhabitats to enhance safety. Their spatial shifts during phase two, however, are based on species-specific escape tactics. Green turtles, dugongs, and dolphins escape through sub-surface flight and rely on maneuverability that is constrained over interior portions of banks. Accordingly, these species move into bank edges when tiger sharks are present to facilitate escape even at the cost of higher encounter rates with sharks (Heithaus et al . 2009). Conversely, sea snakes, which are unlikely to escape a tiger shark, and cormorants, which escape by flying away, shift toward interior areas of banks where shark encounters are minimized. For green turtles, habitat use is state-dependent with turtles in better condition selecting safer areas of banks with less food (Heithaus et al . 2007). Experimental studies of herbivory (Burkholder et al . 2013; Bessey et al . 2016) show that these spatial shifts cascade to seagrass communities during phase three.
Fig. 4 . Results of meta-analyses using two effect size metrics, Hedges’ d (black boxes) and log response ratios (gray boxes), to examine relationships between compromised energetic state and four dimensions of risk taking by prey under threat of predation: activity (n = 30 studies making 34 species-specific contrasts); exposure (versus use of a refuge; n = 68 studies making 71 contrasts); foraging investment (versus apprehension; n = 37 studies making 41 contrasts); isolation (versus grouping/shoaling; n = 10 studies presenting 10 contrasts). For each dimension, positive effect size values quantify the extent to which risk taking by energetically compromised prey individuals subjected to a predation threat exceeded that of conspecifics in better condition. Mean effect sizes and accompanying 95% confidence intervals were calculated using random effects models in OpenMEE ; confidence intervals not overlapping zero were considered to indicate significant effects.