Successfully recognizing and avoiding predators can have immense fitness consequences (1), but individual variation in anti-predator behaviour remains poorly understood. One well-studied factor is learning to identify predators, which is important to effectively focus anti-predator behaviour on potentially novel threats and to decrease costs of wasted defensive behaviours (2, 3). Learning can occur at an individual level, providing accurate information at a high risk to oneself, or in a social context from others, providing potentially less accurate information at a lower risk (4). Differences in the recognition of and response to predators are further amplified by individual variations in learning accuracy and personality (5, 6). Considering the importance of predation-avoidance, such individual differences may have considerable fitness impacts (2).
A less-studied contributor to individual variation in anti-predator behaviour concerns social dynamics. Social factors such as sex or dominance might heavily influence individual motivation to participate in anti-predator behaviour (7). A better understanding of the importance of social dynamics on motivational variation is interesting in its own right, and would also allow better control for motivation when studying variation in learning accuracy. Studies on predator learning to date were either conducted at an individual or a group level (8–12). In the absence of social partners, individual testing may provide similar levels of experienced threat, and therefore similar motivation to engage in anti-predator behaviour, for all subjects. On the other hand, group testing can examine social dynamics and their impact on motivational levels, but they cannot distinguish whether an individual has not learned to recognise a predator, or is simply unmotivated to respond to it. Studies conducted in the wild also face additional difficulties in recognizing individual study subjects (e.g. (13); but see (14, 15)). Only by combining both group and individual setups for the same clearly identified individuals can we tease out the specific role of social dynamics on the motivation to engage in anti-predator behaviour.
The current study examined an important anti-predator behaviour – alarm calling – in common ravens (Corvus corax), a member of the corvid family. When confronted with potential predators, corvids produce harsh alarm vocalisations directed at the predator (“scolding”), presumably to harass the predator into leaving, but also to recruit conspecifics for social support (2). Such group mobbing can provide learning opportunities for inexperienced individuals (16), and has been shown to showcase alarm callers’ status in several corvid species (white-throated magpie-jays (17); hooded crows (18); black-billed magpies (19)). While for common ravens this has yet do be shown, we know they encode individual information in other call types (20–22) and respond stronger to alarm calls of adults than those of juveniles (23).
In a series of elegant studies on wild American crows (Corvus brachyrhynchos), Marzluff and colleagues demonstrated social learning about the potential threat of particular humans, the spread of this information within the local population, and its vertical transmission across generations (13, 24). In those studies, human experimenters could be distinguished via facial masks and their threat level was manipulated via their initial participation, or absence, at catching and banding of crows. Using a similar design, we could previously show that two captive groups of ravens can remember a ‘dangerous’ human for multiple years (25). Interestingly, individuals showed a large variation in scolding response, and dominance status was a strong predictor for their behaviour. Indeed, dominant individuals took the lead in most scolding bouts together with their closest affiliates, indicating strong social dynamics effects (25).
But why should dominant ravens differ from subordinates in scolding? A recent study on jackdaws found that the more individuals give an anti-predator response, the more attractive the display becomes to others to join (26) and, presumably, the more likely the predator is to leave. Given that ravens would profit from conspecifics participating in anti-predator defence in similar ways, the described variation in scolding seems puzzling. One possibility is that in our previous study not all of the ravens were knowledgeable about the predator stimulus, and that subordinates in particular had not yet learned that the masked human represents a risk. Another possibility is that all ravens knew about the predator, but some individuals were “free-loading” on the anti-predation efforts of others, typically dominants (27). Furthermore, it is possible that dominant individuals could afford to show more scolding than subordinates, simply because they were in a better physical condition (see (7)). The ravens’ anti-predator behaviour would thus serve as an honest signal, indicating the callers’ quality (see (28)). Finally, it is possible that dominants actively suppress calling in subordinates to show-off or exaggerate their own quality. In fact, hindering others from calling is both energetically costly and takes time away from engaging in the actual anti-predator response, thus counteracting the effects of group mobbing. Hence, such a costly behaviour should occur only in low to moderately risky situations and/or when potential mates are in the audience. Similar status-signalling effects have also been hypothesized for raven recruitment calls at rich but defended food sources (29), where high-status individuals within the non-breeders tend to produce more calls.
In the current study, we experimentally investigated the effect of social dynamics on individual variation in ravens’ scolding behaviour. We followed up on our previous study, in which we trained two groups of eight ravens each to recognize a human wearing a particular mask (Fig. 1) as a potential novel “predator” (25). During training, the masked person carried a dead raven in the hand, simulating the result of a predation event (30), but all tests were carried out with the masked person only, without any dead raven. We tested each bird in both group and individual settings, and compared scolding responses during group trials, where motivational levels might be heavily impacted by social dynamics, to the responses in the separation trials, where any direct social interactions are absent. We based our hypotheses on the considerations mentioned above, specifying effects due to individual learning (or not) and social influences (or their absence). Specifically, our two main hypotheses are:
- Hypothesis 1 Low scolding durations by some individuals while in the group are based on a failure to learn, resulting in some individuals simply not perceiving the artificial predator as a threat.
- Hypothesis 2 Individuals with low scolding durations in the group learned to recognize the artificial predator as a threat, but their scolding expression is decreased due to social dynamics, specifically:
-
2a) non-scolding individuals free-load on already scolding individuals (e.g. to conserve energy or minimize risk).
-
2b) scolding individuals showcase their quality by participating in risky behaviour, which low-quality individuals cannot afford (to the same extent).
-
2c) scolding individuals showcase their quality by actively preventing others from scolding.
If hypothesis 1 is correct, we predict no differences in individual scolding intensity between group and separation trials. If hypothesis 2 is correct, we predict individuals with low scolding intensity in the group should increase their scolding intensity when tested in separation. Within hypothesis 2, all three sub-hypotheses would predict an effect of dominance. Notably, dominant individuals should invest more in scolding (status show-off hypotheses 2b, 2c), whereas subordinate individuals might free-load on the dominant’s anti-predator behaviour (hypothesis 2a). If dominant individuals even suppress the participation of subordinates (hypothesis 2c), we would expect to see aggression towards subordinates when they engage in scolding and/or a fine-tuning of subordinates in when to call.