Are non-human animals conscious?

Metaphysical/Non-adaptive Argument:

Some people think that consciousness is an epiphenomenon. That is, it 'sits above' observable events, without interfering. As such it is not part of the mechanism for controlling behaviour: it has no direct, detectable effect, for example a testable increase in an animal's inclusive fitness. By this token, any animal could be conscious and we would not know it because we would be unable to test for it. And this same reasoning applies to humans, as we are animals. I know that I am conscious, but I have no reason to believe that any other human is, since consciousness does not manifest itself. Under this theory every single suggested 'function' of consciousness could just as well be carried out by an unconscious organism or by a machine programmed to behave in a sufficiently complex way. For example, if consciousness is thought to have something to do with learning to avoid painful situations and the actual sensation of feeling pain (hurting) is supposed to help this process, why couldn't unconscious animals learn just as well by simply avoiding situations that had damaged them in the past? No conscious experience of pain is strictly necessary - just a learning rule and a means of detecting damage to the body.

Why, therefore, have a consciousness if it has no use? How could a structure with no adaptive advantage be maintained by natural selection? It could arise as a separate phenomenon in certain circumstances (for example whenever a brain reaches a certain level of complexity, or just in humans, or whatever your particular view) (1). Anyway, if is plausible that most of the things we do, we do better if we are not conscious of what we are doing. If I touch a hot object a reflex action causes me to withdraw my hand. If I had to think about it - use my conscious awareness of self - then my hand would be burnt. Similarly, a musician that played directed by though processes alone would be a very poor musician. Indeed, consciousness might often interfere, for example by making the performer feel afraid. People are able to play musical instruments well because humans, like other animals, condition very easily. We become conditioned - we are creatures of habit. And if this is true of humans, who we 'think' are conscious, how much will it be true of other animals?

Churchland criticises this explanation of consciousness, which he calls 'metaphysical' because it imagines that consciousness transcends the physical. To use Nagel's example of 'What is it like to be a bat?' just because we cannot know how it is to be a bat does not mean that the bodily phenomena involved are beyond the explanation of physical science (3). In other words, the consciousness of an object may not be fully discernible to an observer, but it is the product of physical processes, i.e. it is a physical phenomenon. And therefore it can, and should, be investigated by science. This philosophical argument is complimentary to the biological, which argues that consciousness is an adaptation.

Physical/Adaptive Argument:

If consciousness is an adaptation, then, rather than suddenly 'appearing', it must have evolved by a Darwinian process. Darwinian theory states that all adaptations evolved due to the action of natural selection. That is, adaptations increase an organism's inclusive fitness. This raises three important conclusions. Firstly, consciousness must be detectable, i.e. it must be possible to test for it, if only we do the right experiments. Secondly, since evolution tends to occur by gradual steps our ancestors must also have been conscious, to some degree. Therefore, organisms that we share recent common ancestors with might also show some level of consciousness. And thirdly, since selective pressure for the evolution of consciousness exists, it is possible that other animals, unrelated to us, have evolved a form of consciousness, an example of convergent evolution.

Do we have any reason to prefer this explanation over the metaphysical one? Well, we are here today. Therefore we come from an unbroken succession of successful ancestors. And we think that we are conscious, so perhaps consciousness is a successful adaptation. If this is so, what is its function? Identifying its function, or survival value, may help us define the selection pressures that have shaped its evolution, so that we can look in similar animals for signs of consciousness.

Functions of Consciousness:

The psychologist Nicholas Humphrey argued in 1976 that for most species it is not the physical world that demands particularly great intelligence, but the social world. Bryne and Whiten suggested that this social intellect theory was a possible selection pressure that would have led to the evolution of intelligence in social animals - the so-called Machiavellian intelligence hypothesis (1,2). This points to a very important possible function for consciousness. Conscious processors have are better at dealing with novel situations, unpredictability or any situation where something has to be worked out afresh. Such effects are characteristic of the social interactions of animals. Being able to put yourself in someone else's position to such an extent that you can predict and even manipulate what they do needs a very special sort of intelligence and one which, the social intellect theory proposes, a conscious awareness of the sort 'this is what I would do in the same situation' would be a huge advantage (1). The function of consciousness would thus be to enable the organism possessing it to work out what to do in constantly changing social situations where complexity was being compounded by counterintelligence in its allies and opponents, all using their consciousness to enable them, in turn, to work out what 'self' was likely to do.

Social cleverness may not be the only function of consciousness. Indeed, it might have a more general use in any situation that demands knowledge of what the 'self' has done or might do and the effects this would have on subsequent events. Consciousness might be involved whenever the brain effectively follows a self-referring instruction. An awareness of self might be especially important when we change our minds or adjust to changing circumstances in our lives.

These functions suggest that consciousness would be an adaptive advantage in social animals and in those that are generalists (because they most often have to adapt to novel situations).

A conscious animal is aware of its surroundings and can adapt to it (including by making predictions), an unconscious one can only respond to events, albeit in complex learnt and genetically channelled ways. This implies that only conscious animals can behave in novel ways. Their responses will be quicker than the responses of unconscious animals, which can only respond by trial and error within their lifespan (a type of learning, or improvement of conditioning) and across generations (evolution). Consciousness would therefore be of particular benefit to animals that often meet new situations: for example, generalists who live in various environments and must rely on a variety of food sources, possibly of seasonal or unpredictable availability. But the most complex and novel situations arise due to interactions with other beings in close social groups. In these situations also a conscious ability would be of functional value. Therefore it is in these sorts of animals that consciousness might evolve.

Evidence for consciousness in non-human animals.

Of course detecting consciousness is very difficult. However, from the discussion above it is possible to identify some tests that might indicate consciousness.

Insight:

Understanding how things work:

Experiments with animals have shown that gorillas, and other great apes, have an idea of object permanence - i.e. that objects continue to exist in time and space even when they are invisible (2). This concept is important if mechanical skills are to be developed. For example, on one occasion a wild chimpanzee is recorded as coming across a very hard nut. A few days earlier it had been seen breaking a similar nut with a rock. The chimpanzee was seen to return to this place, retrieve the stone it found there and used it to crack open the nut. This required an understanding of the basic attributes of objects and a cognitive map of its surroundings to be able to go directly to the site where the rock had been previously left. These observations suggest that the chimpanzee was aware of its surroundings.

Another example this time with a captive chimpanzee, Sultan, shows an animal using insight to solve a problem with objects. Sultan was given the problem if raking in a reward with sticks that were not long enough for the task. After many attempts, the chimpanzee gave up and began to play around with the sticks. As it was playing, it joined two together. It then suddenly jumped up and used the joined pair to solve the puzzle. It appears that stick-joining is a normal part of chimpanzee play (2), but Sultan was able to recognise that a relationship he had created (putting two sticks together) during play was relevant to the problem. In this case, intelligence seems to be a result of noticing a solution as it came by, not calculating it by a logical purpose. In order to notice a useful solution, Sultan had to be fully aware of the past problem (of stick length) and an adequate solution (a rigid rod of sufficient length). This novel problem was solved using the experience of joining the sticks.

Sultan shows an understanding of the working of objects, how to use them as tools and how to make tools. Many animals from insects to apes use detached objects as aids in their actions. However, tool use cannot be taken as a sign of consciousness. Natural selection might also act to cause tool use in a species, or an individual may learn by random trial. But a repertoire of tool using skills (so that an animal must select between them according to the situation) and, in particular, novel use of tools to solve new problems is indicative of consciousness. To solve a new problem an animal must first understand a problem, and then fashion a tool to solve it. Few animals appear capable of this, but chimpanzees (like Sultan) maybe can. For example, in the wild chimpanzees use bendy lianas or grass stems to 'fish' for ants (in armies or colonies) and termites, to dig open honeycomb, rocks are used to open hard nuts, banana leaves are used as umbrellas, to wipe blood from a wound or to wipe faeces from an individual's bottom and captive chimpanzees have used poles as ladders, and short sticks as pitons, to aid their efforts to cease being captive (to name a few examples). Tool use is an essential adaptation to the chimpanzees' generalist lifestyle. The tools can be made. For example, an individual was observed carefully positioning a wedge beneath a loosely seated anvil stone to make it level, before starting to crack nuts with a hammer stone (2). Without any instruction or demonstration, a pygmy chimpanzee in captivity, Kanzi, has learned to produce useable stone tools. The technique he uses, throwing the rock down on a hard surface, may have been in regular use by the hominids who made Oldowan tools 1.5 My ago.

Understanding of minds:

Conscious use of objects is one way to manipulate the features of your environment. Another way is to manipulate other beings. This does not require consciousness. But if we act as though other individuals are animate causal agents with minds, and we treat them as having mental states, we are using theory of mind and assuming that other individuals are also conscious. As explained above we have to recognise that it is possible to understand that an object is in full view of a person and yet no appreciate what that person will know about it as a consequence. We must distinguish between an ability to understand another's visual perspective, and ability to understand their knowledge and mental perspective. Only the latter is evidence of consciousness. This is very hard to test.

Povinelli (1992) designed a task that required two chimpanzees to cooperate in order to receive a reward. To succeed, one individual had to indicate the correct choice of handle to a second individual, who could see the handle but not tell if it was correct. Chimpanzees succeeded in cooperating to complete the task and were able to swap roles without further training. This suggests that chimpanzees can understand what the other needs to know in order to succeed; i.e. they can know about another's knowledge. Other chimpanzees, when shown a picture of an individual with a problem (e.g. a shivering man), have selected the correct object, from a series of pictures, to solve the problem. A further experiment of Povinelli, again with chimpanzees, demonstrated that an individual was more likely to choose a person who 'accidentally' spilt its food to feed it over a person who deliberately spilt the food. These experiments suggest that chimpanzees can judge intentions and needs of humans in experiments (and hence presumably chimpanzees in nature) (2).

Another facet of the process of coming to understand the distinction between other and self is the ability to 'see ourselves as others see us' - literally. This has been tested with animals using mirrors. Most animals, when they see themselves in a mirror, give a reaction appropriate to meeting a strange member of their species and sex. Gallup has shown that chimpanzees and orang-utans (and recently a gorilla) are able to interpret their reflection in a mirror correctly. The animals are painted on their forehead whilst unconscious: when they see their reflection they attempt to rub the paint off. By contrast, monkeys will not, even though they will investigate spots of paint on parts of their body visible to them. Gallup has used mirror self-recognition as an indication that an animal has a self-concept. Without an 'I-concept' label, we would not be able to understand exactly what was on the mirror surface (2).

Second-order intentional deception or teaching is hard to test for in any animal. The distinction between (for example) "I want to do X" and "I want him to think Y, which is wrong, so I can do X" is difficult to observe. This example describes an act of intentional deception involving other individuals: as objects to be manipulated, as social tools to manipulate others, or as resources to be gained. Where individuals are long-lived intentional deception may be rare, because of retaliation by other individuals. Nevertheless, cases have been described. Plooij, while studying chimpanzees, was groomed by an infant. He tricked the youngster into going away by feigning interest in a distant bush. The young chimpanzee left to investigate, but having found nothing she 'walked over to me, hit me over the head with her hand and ignored me for the rest of the day'. The use of the tactics of counterdeception was seen during Menzel's experiments, again with chimpanzees. The action involved two individuals, a dominant male called Rock and a female called Belle. Belle was shown where food was hidden in an enclosure, replaced in the group and all the chimpanzees let out into the enclosure. Initially, Belle would lead them to the food, but Rock used his position to monopolise the reward, so Belle resorted to deception. Rock, it appears, understood these deceptive actions and countered them. A behavioural 'arms-race' of deception and counter-deception ensued, which suggested that each chimpanzee understood the other's intentions (1,2). Another example is of the chimpanzee Lucy, trained in sign language, who had deposited faeces on the carpet of her trainer. When asked 'Whose dirty, dirty?" she replied "Sue's". Challenged by "It's not Sue's. Whose is it?" she tried another human, Roger. Challenged again, she conceded "Sorry, Lucy".

Intentional teaching is harder to prove, however there are examples, again from chimpanzees. Washoe, who had been taught to sign, taught her adopted son to sign using demonstration (with careful attention to his gaze direction) and moulding of his hands (as she herself had been taught). Washoe and her son did not see a human sign, nor receive any encouragement from humans for 5 years, by which time the son, Loulis, reliably used 51 signs. Similar teaching skills have been observed in wild chimpanzees (for example being taught how to crack open nuts). Interestingly, intentional teaching has also been described in dolphins (2).

Anticipation:

Animals with a theory of mind and a concept of the self might possibly be called conscious. However, all the above examples relate to events that immediately impact on the animal's life. An ability to plan ahead beyond current needs and desires, and produce novel or inexplicable behaviour in anticipation of possible future events is indicative of an even greater awareness of the animal's environment: consciousness. There are two possible cases in chimpanzees.

Chimpanzees occasionally 'wage war' against neighbouring communities. 'Commando' raids can result in brutal hand-to-hand fighting between males, for a clear reward: the females of depleted neighbouring communities transfer to that of the aggressors. However, these rewards are long-term. No mating is observed during the raid, and females often transfer after the raid is complete (they are rarely coerced). That is, the rewards are not immediate, so the behaviour of the aggressive males cannot reflect their bodily needs; does it reflect their long-term plans? An alternative view is that this behaviour is genetically channelled (just like infanticide in Lions). However, the chimpanzees show very special behaviour when on these dangerous patrols that implies they are aware of the dangers of noisiness. Given that patrols are rare events, and so risky that 'punishment' for error may be death such behaviour must either be genetically channelled or a result of mental connections. Obviously the former is possible, but it is pertinent to say that the raids are very similar to the intercommunity raids of the Yanomamö people of the Amazon (2).

Another event that was observed by Bryne (2) is unique, but less subject to genetic explanation. The event involved the killing of a young leopard cub by a band of male chimpanzees. The cub was defended by its mother, but even so a male chimpanzee risked entering the cave where it was hidden to seize it. The female leopard did not emerge to challenge the chimpanzees, so their estimate of the risk was presumably a good one, even though the action was highly dangerous. The chimpanzees' efforts were directed at the cub not the mother, and towards its death, rather than its possession and consumption (which might give immediate benefit to the individuals involved). Bryne asks the question to what end was this behaviour directed? Such action, reducing the population of dangerous carnivores and deterring the mother, is very similar to some human behaviour. It is not easy to account for the chimpanzee's actions without attributing similar goals to them, which implies long-term anticipatory planning, awareness of their environment. Yes, the behaviour could be genetically channelled, but it is of sufficient complexity that maybe it is indicative of consciousness.

Complexity:

This does not mean to suggest that all complex behaviour implies consciousness but rather, that the complexity of behaviour and the ability to adapt to changed circumstances are characteristics of a conscious mind. Computers are complex, and yet they are not conscious. This begs the question: could we build a computer complicated enough to be conscious?

There is no doubt that much animal behaviour is very complex, often more so that it fist appears (1). Ostriches are able to recognise their own eggs out of a cluster of many other ostriches'. Rats 'appear' to be able to recognise poisoned food. But complexity is no evidence for consciousness. If food is poisoned then cautious rats will have a strong advantage. There is the case of Clever Hans, who 'appeared' to be able to do mental arithmetic, but was actually recognising inadvertent signals from his trainer. The 'Clever Hans error' argument has been used to explain many complex behaviours, including the learning of language (5). Griffin (4) has argued for the inverse 'Clever Hans error'. Horses may not be able to do mental arithmetic (it may not be very useful for them to add four and four; we should not judge another individual's consciousness by our own culture). However, what has been overlooked is the possibility that Clever Hans was consciously thinking something simple but directly relevant to his situation - perhaps something like: "I must tap my foot when that man nods his head."

Marian Dawkins identifies two other sorts of error that can be made when studying animal behaviour (in addition to the Clever Hans effect). These are the failure to specify what is expected before an experiment and a lack of consideration for simple explanations.

So complexity may not lead to consciousness at all. One example frequently used is the brain. The brain can be thought of as fulfilling two roles. It is a sort of 'on-board computer' (Dawkins 1976) that governs intelligent function (2) or a 'telephone-exchange' of sensory and motor neurons involved in automatic responses. Under the second role larger animals would require larger brains; a blue whale's brain is far larger than a human's, but this is not indicative of consciousness. However, under the first role a larger brain would suggest a more intelligent, conscious, animal. The brain fulfils both roles, but it is hard to separate the volume dedicated to each (if indeed the brain can be dissected in this way at all). The ratio of neocortex size relative to the rest of the brain has been taken as a measure for the 'thinking part of the brain' (2). By this measure, humans are more 'intelligent' than other animals.

What this suggests is that only a particular sort of complexity leads to intelligent, conscious thought. So would a complicated computer be conscious? Probably not, but Churchland (3) argues that with the correct neurocomputational hardware maybe it could. He identifies seven salient dimensions of human consciousness and describes how machines might be built to do each of these. But consciousness is a collective phenomenon - a product of the whole. And this reductionism may not be helpful. Nevertheless, in examining the likely mechanisms of consciousness by building our own machines, rather than observing its effects or functions, we are investigating the physiological requirements a brain would have to be conscious. Therefore, it is possible that neural complexity per se might not lead to consciousness (6). It is unlikely, for example, that the loss of conscious awareness seen in many neuropsychological syndromes such as blindsight is just due to loss of complexity. A conscious brain might well have to be complex, but it must be complicated in the right, qualitative, way. This suggestion is borne out by the observations of neocortex size in animals. Large animals have large complicated brains but this does not infer that they are conscious. Bryne (2) has identified two changes in brain structure during the evolution of primates. The first was a quantitative shift in cognitive ability resulting in the ability to learn and adapt quickly. This, he says, accounts for the learning abilities of many monkeys. The quantitative shift to an enlarge neocortex, allowing enhanced learning in social contexts, may have been an essential prerequisite for a more fundamental, qualitative development: the evolution of insight (see above). This possibly evolved at least once in the great ape ancestor.

Language:

There is little surprise, therefore, that the great apes are almost alone in the animal kingdom in learning linguistic skills. Humans are, of course, the prime example of this, but other apes have learnt to use sign language and languages involving symbols like Yerkish (1,2,4). These skills are surprising, given especially that other apes come from different cultural backgrounds to ourselves and lack the correct morphological structures (a complex larynx) to produce complex sounds. But language is not necessarily an indicator of consciousness. Bees have are able to communicate abstractly to each other the location of food, but this behaviour is almost certainly genetically channelled. Of course, this is not true for apes, but the initial experiments have been criticised for not ruling out the 'Clever Hans' effect. Terrace raised a young chimpanzee and taught him to sign, only to conclude that cueing (i.e. the interpretation by the chimpanzee of inadvertent cues made by the trainer) was almost certainly happening on a massive scale (1). During the last year of the chimpanzee's training only 10 per cent of his gestures could be genuinely said to be spontaneous or entirely his own and not imitations of his teacher.

But more recent studies have tested the chimpanzees' capabilities more stringently. Kanzi is a young pygmy chimpanzee that learnt to sign through observing humans signing to each other (simulating the way a human child would learn). He has picked up a large vocabulary without being taught, and even understands the meaning of spoken English words. When he was tested 'blind' (i.e. the experimenter did not know which words Kanzi was listening to on a pre-recorded tape), his comprehension was listed at 150 spoken words by the age of six. He was able, for example, to discriminate word order, e.g. "Make the doggie bite the snake" from "Make the snake bite the doggie". Further, he could understand the embedding of a clause, as in answering "Go get the tomato that's in the microwave", a task which required going past several other tomatoes. Kanzi is also able to use syntax. These experiments suggest that failures to teach other chimpanzees may be because they were older. Human children deprived of a normal environment also never reach normal language competence. Even so, at the age of eight, Kanzi's skills matched those of a two year old, so his capabilities are limited (2).

The linguistic skills of apes are not surprising given that they are our closest relatives, and also are generalists that live in complex social groups. Other animals that also live in complex social groups also appear to learn human languages. For example, Dolphins have successfully learnt to respond to auditory presented commands with a linear order syntax.

Are non-human animals conscious?

So are non-human animals conscious? Maybe. Consciousness probably has a functional value, so it is possible to test for it, although care must be taken. These tests show that probably some animals are conscious, but perhaps not to the 'extent' that we are, or maybe they are conscious in a different way that reflects their lifestyle. The language abilities of dolphins suggest that we should expand our search for consciousness beyond primates. If consciousness confers a selective advantage then it might be expected to evolve in any generalist, social animal. Consciousness is not necessarily a result of increasing complexity per se, but maybe caused by qualitative changes in the neural network. Are bees conscious? Maybe, if the extended effects of the same genes in all the individuals in a hive produce a collective awareness that extends beyond the body of an individual bee.

There is a need for accurate testing of consciousness. Weiskrantz (6) has suggested a regime based on a combination of anticipation plus rational adaptation. Consciousness awareness is having a thought about a thought, which may not be rational. Weiskrantz's proposal would only detect rational thoughts about thoughts, but it is an improvement on previous techniques.

References:

(1) Dawkins, M.S. (1993) Through Our Eyes Only?

(2) Bryne, R. (1995) The Thinking Ape: evolutionary origins of intelligence.

(3) Churchland, P.M. (1995) The Engine of Reason, the Seat of the Soul.

(4) Griffin, D.R. (1992) Animal Minds.

(5) Kennedy, J.S. (1992) The New Anthropomorphism.

(6) Weiskrantz, L. (1995) Behavioural Brain Research 71, 171-175.

(7)

(8)

(9)

(10)

2/11/98 Tom Clements

9/9