Box 3: Prey perception of risk
A large literature in behavioral and sensory ecology has examined prey perception of danger based on cues that provide information on levels of enemy risk (Weissburg et al. 2014; Ehlman et al. 2019). Arthropods perceive risk using chemical (both airborne and via direct contact; Dicke & Grostal 2001; Sitvarin & Rypstra 2012; Hermann & Thaler 2014), visual (Gonçalves-Souza et al. 2008), vibratory (Castellanos & Barbosa 2006), auditory (Skals 2005), and tactile cues (Castellanos et al. 2011; Okada & Akamine 2012). Organisms often use multiple cue modalities, which can vary depending on prey perceptual ability and the types of enemies.
A primary source of risk cues is the enemy itself, whether directly as sounds, vibrations, chemical cues, or visual presence, or indirectly as chemical footsteps, feces, molts, and silk. Organisms can also respond to indicators of risk before they actually detect enemies; e.g., by responding to ‘alarm cues’ associated with other prey being attacked, injured or killed (Schoeppner & Relyea 2005; Vandermoten et al. 2012). Alarm cues can induce a range of responses and can even be shared across species (Agarwala et al. 2003; Goodale & Nieh 2012). Another cue may be habitat or microhabitat type. If certain habitat types are associated with enemy risk, then risk avoidance may drive habitat selection, regardless of direct cues from enemies or even conspecifics (Lucas et al. 2000).
Cues can vary widely in spatiotemporal extent, affecting different numbers of prey over varying timescales. For example, because chemical cues can spread widely and remain detectable for long periods, they can cause risk effects to persist long after enemies have left an area (Wilson & Leather 2012; Ninkovic et al. 2013). Theory suggests that because the cost of under-responding to risk (i.e., getting killed) is often much greater than the cost of over-responding (e.g. hiding unnecessarily and losing feeding opportunities), when cues provide imprecise information about the presence (versus absence) of predators, this uncertainty can induce strong enemy-risk effects even when predators are only occasionally present (Sih 1992). This may be true for many prey facing the risk of attack by ambush predators. In contrast, seeing or coming into physical contact with an enemy is usually a more definitive risk indicator.
The links between cue generation, detection, and anti-enemy response are complex, involving multiple steps and interactions. Environmental context can strongly affect both the strength and detection of a cue (e.g., wind may disperse a chemical cue) and the perception of risk upon detection (e.g., perceived risk may be lower if a refuge is nearby). Response to risk can be highly state-dependent; a starving organism may be more likely to accept higher risk to avoid starvation, and a larger, faster individual may assess risk differently than a smaller, more vulnerable organism. In some cases, it can take a combination of multiple cues to trigger a response (Gish et al. 2011). Recent theoretical work has suggested that cues indicating risk should be integrated with other cues indicating safety to shape responses (Trimmer et al. 2017; Ehlman et al. 2019), and supporting evidence has emerged from recent studies with desert isopods (Zaguri & Hawlena 2019).
A key insight from signal detection theory is that all cues are imperfect indicators. Cues can vary in strength; a chemical cue can be diluted or concentrated, a visual cue can be obscured by other objects, and an auditory cue can be disrupted by ambient sounds. On top of variance in cue strength, the specificity of cue modalities can vary. The visual cue of a looming shape could come from a dangerous enemy or a harmless passing organism, the chemical and tactile cue of a parasitoid could come from a species that parasitizes the pest or another, closely related parasitoid that does not (Fill et al. 2012), and cues that elicit stress and reduce population growth can come from activity of commensal organisms (Jensen & Toft 2020). The reliability of cues may change with the introduction of novel organisms (Ehlman et al. 2019) or through habituation to the cue. The consistent application of synthetic alarm pheromone may cause decreased sensitivity of aphids to the cue, but this insensitivity may in turn increase CEs by coccinellid predators (de Vos et al. 2010). Finally, synthetic predator kairomones can increase mosquito mortality synergistically with Bacillus thuringiensis applications, even when completely decoupled from real predators (Op de Beeck et al. 2016; Delnat et al. 2020). Biocontrol practices might benefit from deeper understanding of pest perception of cues associated with enemy risk.