Do animals have awareness of self?

If self-awareness is a product of language, and language is a purely human capacity, can we say that other animals do not have self-awareness? The logical answer is ‘yes’; but, as with all things cognitive, there is a significant ‘however’ to be added to that simple affirmative. We need to look at different types of awareness of self before we can say unequivocally that self-awareness does not exist in other species.

We can start with the question ‘Do other animals have awareness of their physical bodies?’ As we have seen, this is far from a given of being alive, and is certainly not an attribute of single-celled animals; but how complicated does a nervous system have to be before it is capable of physical self-awareness? Fortunately, we can approach this question from the other end by looking at only a small subset of animals, the primates.

In 1970, Gordon Gallup thought he had found a way to investigate physical self-awareness: he tested a small number of chimpanzees and monkeys to find out if they could recognise their own reflection in a mirror. His experiments involved familiarising his subjects with mirrors, then placing a mark on their face while they were anaesthetised and seeing how they reacted to the mark when they woke up. The chimpanzees began investigating their own faces when they saw the mark – they seemed to know that what they were seeing in the mirror was a reflection of themselves. Afro-Eurasian monkeys tended to ignore the mark, or attempted to investigate the face in the mirror – they did not recognise their physical self in a reflection (old-world monkeys were selected for testing because humans share a more recent common ancestor with them than with the new-world American monkeys). Chimpanzees who had not previously encountered mirrors did not pass the test, nor did human children younger than about 18 months old (Amsterdam 1972), but it did appear that the mirror test was identifying something significant in self-recognition.

The test has since been repeated with a range of animals, and it has become clear that mirror recognition is not even limited to mammals. All the great apes (bonobos, chimpanzees, gorillas, orang-utans) have been shown to pass the test (de Waal et al. 2005), as have several non-primates, such as bottlenose dolphins (Reiss and Marino 2001), orcas (Delfour and Marten 2001), an Asian elephant (Plotnik et al. 2006) and European magpies (Prior et al. 2008). To complicate matters, a more recent experiment showed that rhesus monkeys can also be trained to pass the test (Chang et al. 2015). Contrary to Gallup’s findings, these animals co-identify the reflection with their physical self if they are forced to pay attention to the face in the mirror and can link current sensations with current visual effects.

This does not mean the mirror test doesn’t work, but it does mean we should be careful about how we interpret it. As Gallup himself has said, ‘Simply because you’re acting as if you recognize yourself in a mirror doesn’t necessarily mean you’ve achieved self-recognition’ (quoted in Callaway 2015). However, if the mirror test is indeed a measure of physical self-awareness (and it seems to be), then we can at least say that this capacity is not limited to humans.

How exclusively human is cognitive social modelling? This, too, seems to be a capacity we share with other animals. Dorothy Cheney and Robert Seyfarth (2007) have shown that chacma baboons (Papio hamadryas ursinus) maintain quite complex models of their group, allowing them to identify both hierarchies of individuals within families, and the hierarchy of families within the group. It helps that there seems to be no ranking overlap between families: the lowest-ranking individual in a high-ranking family is still higher than the highest-ranking individual in the next family down. While this social modelling is social arithmetic rather than social calculus (as it is about the relationships of others to the self rather than the relationships between others), the baboons are nonetheless able to adjust their social modelling appropriately for changes of rank both within and between families. Although social modelling has been less widely explored in other species, there are indications that great apes, dolphins, whales, elephants and some birds are capable of social arithmetic; and there are clues that they may also be capable of a form of social calculus. The similarity between the list of species capable of social arithmetic and that of animals passing the mirror test may be telling us something important about social cognition – or it may just be that these are the animals whose cognition we have studied in sufficient detail. However, the extent of social modelling beyond the human species is not something that needs to concern us here; as with physical self-awareness, it is enough to be able to say that it is not exclusively human.

So is shared social calculus exclusive to the human lineage? Here, we have something that has not yet been unequivocally detected in non-human communication. This does not mean it is exclusively human; but, until we identify the exchange of social information in other species, it seems reasonable to treat it as provisionally exclusive. Sharing social calculus requires communicative strategies that leave their mark on a communication system, and some of these we have yet to identify in non-humans – but some we have identified, so we cannot yet be certain either way about the exclusivity of shared social calculus.

The first of these strategies is naming, or attributing identity labels to other group members. Unlike internal social calculus, a name-label needs to reliably identify a particular member of a group to all other members of that group, so the labels must be communally shared. This type of labelling has only been reliably identified in one species of dolphin (Tursiops truncatus) so far: it appears that every dolphin in a pod has a signature whistle (Cook et al. 2004), and they each use that whistle to indicate their presence and position to other pod members. This signature whistle remains the same when the dolphin moves to a new group (which they do often), so it is a label the dolphin uses to identify themself, not a label given to the dolphin by each group (King et al. 2018). Even more interestingly, they also use the signature whistles of other pod members to attract the attention of those others. By itself, this does not mean dolphins are sharing social calculus, but it does indicate that this group of species is worthy of further investigation.

The second requirement for sharing social calculus is signal combinatoriality: you must be able to consolidate individuals into a group by linking their names to a particular relationship; and you must have the communicative mechanisms to bring this relationship to the attention of the listeners. To put it simply, your signal must be able to indicate two individuals and the relationship between them. So, for social calculus, combinatoriality requires both segmentation (a signal must be capable of containing more than one meaning-unit) and differentiation (the meaning-units have to represent different things – individuals and relationships). This type of combinatorial social sharing has not yet been identified in non-humans, but other forms of combinatorial signalling have. For instance, Campbell’s monkeys were found to use distinctive calls for eagle or leopard predator alert calls, but they also used a suffix, the same suffix for both calls, to change their meanings to more general alerts (Ouattara et al. 2009). In contrast, the sound-units in the calls of putty-nosed monkeys seem to have no individual meaning, but the way they are combined can create warning signals of ground predators or aerial predators, or can be used to coordinate group movement to new feeding grounds (Arnold and Zuberbühler 2006).

However, these combinatorial calls may not offer sufficient proof: the interpersonal referentiality in shared social calculus is very different from that in other combinatorial signalling. In contrast to social calculus, which is explicitly about other in-group individuals as third persons, other combinatorial signals seem to reference in-group members only as receivers, or second persons, and this (except for the dolphins) only implicitly; the presence of receivers justifies the making of the signal, but the signal is for them, not about them. Other combinatorial signalling, like most signalling (including the dolphins), involves attracting attention; but any third-person reference is to out-group individuals, which can be treated as events or things rather than individuals. Shared social calculus has its origins in, and remains today, largely third-person reference about in-group individuals.

So what of the capacity to model the self – to be aware of your own selfness? Is this limited to humans? In terms of their natural social and communication systems, there seems to be no sign of awareness of selfness in non-humans; but then, because of the disadvantages that self-modelling brings for the modeller, we would expect to see it only if there were identifiable advantages to possessing it. This may be why, when we look at animals exposed to human language, we do see some indication of awareness of self: experiments in which we have tried to teach human language to non-humans have shown that, the more we expose them to a human communication system, the more humanlike their behaviour seems to become.