Box 3: Multiple cues drive behavioral responses to multiple stressors
Responses to dangers (e.g., running from predators) are typically elicited by sensory cues relating to that danger (e.g., seeing or hearing a predator). Thus, at a detailed level, it is the cue of the danger that really acts as the stressor (in terms of eliciting a response) in trade-offs 1-4, rather than the danger per se, though these are sometimes confounded when it comes to automatic physiological responses because the danger itself can also directly act as the cue (e.g., the physiological response of a fish following an increase in water temperature or salinity). Here, we consider the case of responses to possible predators, where it is the cues that drive responses, rather than the danger directly. This may initially seem very simple; however, unlike the physiological case of heat/salinity, we consider cases where multiple cues are present and no cue is perfectly reliable.
In many cases, action should be taken before an organism is certain of their situation. There are three cases relating to the source of the cues:
A common cause, so one best action (Fig. 3A)
Multiple cues may be produced by a single danger (e.g., a glimpse of a predator, combined with a faint whiff of its scent) (Hale et al.2017) Such cases are relatively simple in that there is typically a single best response to such a danger, and the cues correlate positively with that danger, so the best response also correlates positively with the overall probability given by those cues.
Different causes, but one best action (Fig. 3B)
In some cases, there may be multiple uncertain cues of danger, each of which, alone, does not tell an organism to take evasive action, but together they do. For instance, a vole emerging from its hole may get the faint whiff of a cat (which may not be enough on its own to trigger a response of fleeing back to its hole), and a glimpse of a bird overhead (which alone, may not be sufficiently indicative of an aerial predator). Because the best response to each danger is the same (although the cues come from different sources), they can be combined to govern a stress response of fleeing (or of increasing one’s physiological readiness to flee).
Different causes, with different best actions (Fig. 3C)
The more complex case is one where the best response to one danger (e.g., running from a terrestrial predator) conflicts with the best response to another danger (e.g., an aerial predator may be more likely to spot moving prey). In this case, the uncertainties relating to the cues interact with the expected payoffs for each possible action in each situation (Brilot et al. 2012). This can often mean that an action which is not optimal in either actual situation (terrestrial or aerial danger present) is the best while gathering more information. For instance, it may be best to freeze for a short time while gathering more information, even if climbing a tree would be the best defense against one danger, whilst running back toward one’s burrow is best against another danger.
Cases (1) and (2) can usually be represented using simple (one dimensional) signal detection theory but, in general, the number of dimensions (required for a signal detection approach) increases with the number of possible dangers (even when there are only two possible actions). Thus, signal detection theory provides a robust framework that can be applied to systematically evaluate expectations for behavioral responses to the multiple- (often many-) stressor scenarios that abound in nature.