by Bor, Daniel
Clayton has even shown that one species, the New Caledonian crow, can use tools in a surprisingly sophisticated way. For instance, these crows can use a sequence of tools in order to win food. In one example, the birds learn to use a short tool to hook out a longer one from a narrow tube, which they use to retrieve a third, even longer hook, which the crows finally use to directly fetch a food source. This was an experiment involving wild birds that voluntarily chose to enter the testing area. The crows do have a modicum of prior practice with the tools individually, but they have never before seen this sequential tool setup, and no human has ever demonstrated the task to them. Nevertheless, one of the three crows, known as Betty, successfully used the three tools in sequence and thus retrieved the food on her very first trial. Another of the birds, Pierre, also passed the test, but not in the way the experimenters ever intended: After exploring the apparatus, Pierre momentarily left the testing site, but he soon returned with a long twig that was perfectly suited to collecting the last, longest hook. This allowed him to retrieve the food using two tools instead of three. This form of complex, adaptive tool use suggested that the crows had a clear mental picture of the concepts involved and were invoking planning.
One final, particularly impressive example of tool use in corvids relates to the old Aesop’s fable of the crow and the pitcher. The old story goes that a crow, almost dying of thirst, chances upon a pitcher full of water. But the crow can’t access the water, because its beak can’t get far down the opening at the top. After a while, it has a brain wave, and drops lots of pebbles into the pitcher so that the water level will rise and the crow can drink. This fable, it turns out, isn’t fiction at all. This time, rooks were used instead of crows, but again no example demonstration by the experimenters occurred, and the birds were left to figure out the puzzle on their own. The rooks easily learned, usually on the very first trial, to drop stones into a beaker of water in order to raise the surface and reach a floating grub. To dispel any doubts that this was mere random behavior, the experimenters also showed that the rooks always stopped placing stones in the beaker once the food was retrieved; that the rooks favored larger, more effective stones over small ones; and that they avoided pointlessly placing stones into a beaker of sand instead of water.26
Not to be outdone by their feathered counterparts, an even trickier version of this ancient example of insight has been observed in the great apes by Daniel Hanus and colleagues. In this version, a peanut floats on a small volume of water toward the bottom of a beaker and can only be retrieved if more water is added to the beaker. There is a water dispenser nearby, which is about the only “tool” available. The cleverest 16 percent of the chimps realize the water’s potential and discover an innovative solution by themselves, usually on the very first trial. They hold the water from the dispenser in their mouths and spit it out into the beaker the few times needed so that the peanut floats to the top and the tasty treat can be retrieved. One particularly inventive chimp, impatient to claim his food reward, decided to speed up this process by urinating directly into the beaker and winning the prize in this unusual way, seemingly unperturbed by its unpleasant liquid soaking, gobbling up the peanut as soon as it was available!
Intriguingly, in an earlier study the same research group found that every single one of the five orangutans tested could pass this challenge on the first trial, providing provisional evidence that this more distant cousin is in fact our smartest primate relative. Even more strikingly, four-year-old children are embarrassingly outcompeted on this task by chimpanzees and only show their superiority after reaching the age of six.
It seems almost certain that a highly active mental life replete with considerable planning and imagination is required to solve such devilish tasks. But does all this complex tool use mean that the crow family, effectively a primate in feathered form, is conscious? Are crows just as conscious as the great apes, who have similar mental skills? Based on their behavior so far, all we can definitively say is that corvids and the great apes are highly intelligent for nonhuman animals, while their seeming capacity for innovative thoughts is strongly suggestive of consciousness. But it isn’t proof. We have no access to their mental world, and so simply can’t tell for sure.
CAN A BIRD ADMIRE ITSELF IN A MIRROR?
Another way to grapple with this problem is to set the consciousness bar challengingly high: It is widely assumed that if you are self-aware to the extent that you can recognize yourself in the mirror, then by definition you are also conscious. So, if an animal passes this demanding test, then there is little doubt it is conscious.
The standard way to test this is to put a colored spot on the animal’s face in such a way that the animal can only detect it in a mirror. If the animal sees its reflection as itself, and not another animal, as many animals do, then it might try to rub the mark off its face, or at least move its body to get a better view of the mark. This is seen as positive evidence of self-awareness. The trouble is that an animal may fail the test for numerous reasons, regardless of whether or not the animal has self-awareness. It might have poor eyesight, might not be looking in the right way, or instead may well notice the mark, know that it is a spot on its own face, but not be inclined to investigate or remove it on that particular testing day. Humans don’t pass this test until they are about eighteen to twenty-four months old, on average. The other animals that have passed it include chimpanzees, orangutans, gorillas, elephants, pigs (on a modified test), and even a member of the corvid family, the magpie. To my mind, passing this test really is strong evidence that an animal is conscious. But a determined skeptic would still have a valid position if he retorted that the animal was in no way reporting to us that it was conscious, and may merely have been reacting in a sophisticated way while inside there were no experiences whatsoever, no conscious thoughts or feelings at all.
But if we are looking for absolute incontrovertible proof, then we cannot even establish that any human, except for ourselves, is conscious! I think it’s useful, therefore, to be somewhat pragmatic about the issue of animal consciousness. Although it’s equally unhelpful to be absolutely convinced from scant evidence that other animals are aware, we should see hints such as passing the mirror-recognition test as strong, if problematic, evidence of animal consciousness.
What’s missing is for an animal to report on its own level of awareness. But how can a nonhuman creature ever tell us that it is aware? In fact, some ingenious researchers have devised experiments where other animals can tell us exactly this—and given this chance, the animal even exhibits a level above pure awareness, to being aware that it is aware (in other words a kind of meta-awareness). So these animals demonstrate not just consciousness, but a sophisticated form of consciousness to boot.
For instance, recall that monkeys can be trained to tell you when their visual experience flips between one image and another on a binocular rivalry task. The stimulus is always the same, but the switch occurs within the mental life of the monkey, which it can report via a button press. The skeptic has a far greater challenge denying this form of awareness.
GAMBLING ON CONSCIOUSNESS
However, there is even more impressive evidence of monkeys being able to report on their internal mental states. In a series of trials, Nate Kornell and colleagues showed monkeys a set of dots on a computer monitor, with one of the dots a little larger than the others. The monkeys would try to choose the dot that was the odd one out. They’d then be presented with a gambling choice of two buttons, one that was a safe option, and the other that was high risk, but with high potential rewards. If they pressed the high-risk button, three tokens would be added to a tally if they were previously correct on the dots task, but if they were wrong, three tokens would be removed. When this tally reached twelve tokens, the monkeys would get a special reward of a banana pellet. But the monkeys could also, after guessing the odd one out, choose the safe option, which always added one token to the tally, regardless of whether they were right or wrong.
A sensible policy here would be to choose the high-risk button when you were confident that you had guessed the dots task correctly, and the safe button if you really weren’t sure. This sounds obvious, but grasping it requires a rather sophisticated internal mental world, where we have knowledge of the strength of our own beliefs, sometimes knowing our perceptual decisions are accurate, but at other times doubting the validity of those decisions. Monkeys, spurred on by the chance to obtain their tasty snacks, performed the gambling component of the task in appropriate ways, choosing the risky option far more when they were correct in the previous odd-one-out task, and the safe option when their perceptual guesses concerning the dots were less accurate. Not only this, but without further training on the gambling component, monkeys were easily able to transfer this ability for their confidence to track the accuracy of their perceptual decisions to other tasks, such as those involving working memory. They were also able to ask for hints when unsure, as we do, providing another strand of evidence in support of an inner mental world capable of having knowledge about your own knowledge level. For instance, when learning a complex sequence, initially the monkeys would ask for many hints, but when they became proficient at the sequence, they asked for very few.
In fact, the monkeys’ ability to track the strength of their own knowledge, and use this skill to gamble appropriately (making risky bets when knowledge is firm, and safe bets when uncertain), was highly comparable to the average level of human skill on such tasks. So on this very advanced task, which in humans we take as an unequivocal sign of consciousness, monkeys and humans are equivalent.
Further bridges between monkeys and humans can be built by looking at what is occurring in the monkey brain during such gambling tasks. If these decisions in monkeys are indeed conscious ones, a reassuring result would be to detect neural activity that directly mirrors the differing confidence levels in regions we know relate closely to consciousness in humans. So far, this has only been studied in the back portion of the prefrontal parietal network, in the posterior parietal cortex, but in this region neuronal activity exactly matched level of confidence in a perceptual decision, just as if it were strength of awareness that were being measured.
Other species that have shown similar skills, where they gamble on high-risk options when they are usually correct and stick with safe options when they are less likely to be right, include the great apes and even, in a simpler version of the task, rats.
My reading of these results is that it is almost certain that monkeys, and any other type of animal that can also pass such tests, already have an extensive form of consciousness. But because monkeys are so similar to us in this gambling ability, we need to look elsewhere if we wish to explain exactly what distinguishes the immensely rich human form of consciousness from what is most likely a relatively weaker form in other species.
ANIMAL CHUNKING
So far, these behavioral studies have attempted to show that other animals are conscious in some form or other, but what these experiments have largely avoided is the question of the quality of consciousness—What can these animals actually do with their awareness? One interesting question along these lines, based on the main arguments in this book, is how well other animals can use chunking strategies. Humans are certainly not the only animals that can learn in a structured way. Rats, for instance, have been shown to apply simple forms of chunking. If you present a rat with 12 hidden openings, comprising 4 sets of 3 different types of food, they will learn to group the openings together according to the food group and head straight for the 4 openings that are sources of their favorite food. Even pigeons have been shown to apply a rudimentary form of chunking in their learning. For example, Herbert Terrace trained pigeons to peck either a sequence of colors or a mix of colors and plain white shapes. In order to obtain food, the pigeon had to get the entire sequence right. The pigeons were painstakingly trained for up to 120 sessions to get a sequence of 5 in a row correct in order to get food—probably the equivalent human task of completing a degree. The pigeons struggled terribly if the sequence was all 5 colors, but they did far better if they could break the sequence down into 2—for instance, if the first 3 were colors and the last 2 were shapes. It wasn’t just that there were 2 different kinds of stimuli to remember: If the colors were interspersed with shapes—for instance, a shape, then a color, then a shape, then a color, and finally a shape again—then the pigeons were back to being terrible at the task. This grouping of parts of a sequence could well be analogous to human forms of chunking.
But there is a world of difference between being able to chunk at all, on the one hand, and being able to discover and use chunks in a powerful, hierarchical way, on the other. Other animals may be able to recognize themselves in a mirror, plan for future events, remember many past ones, or even be aware of their own awareness, but one key factor where humans—even toddlers—leave animals in the shade is the extent of our ability to chunk. Specifically, it seems that the number of levels of chunking on which we can operate, the height of our pyramid of meaning, easily outstrips even our closest relative, the chimpanzee.
One way of demonstrating this is to observe different species at play. In one experiment, a group of chimps from 15 months to adulthood, one adolescent bonobo, and human babies between 6 months and 2 years of age were all given a random collection of 6 objects, such as cups, rings, and sticks that could be red, blue, or yellow. The experimenters simply watched them play and recorded how they moved the objects about, combined them, and so on. Various levels of information and behavior are available here: On a basic level, these are all separate objects and can be manipulated one at a time. But far more meaning than this can potentially be extracted. For instance, the objects can be grouped by type, size, or color alone, but two categories can also be combined—for instance, by placing all the red rings together. A higher level still is even available, if all the large red rings, say, are separated from their smaller equivalent. Then there are the relations between different items. For instance, sticks and rings can be placed inside cups, and sticks can pass through rings.
When it comes to the basic skills of manipulating single objects, there is little to separate human babies from their primate cousins of the same age. As soon as you move up the information pyramid, though, developmental differences become increasingly apparent. Chimps and bonobos can learn to group items according to categories, such as color or shape, but they learn such concepts considerably later than humans do, and the complexity, or levels, of meaning they can learn are terribly limited compared to human babies. It is not uncommon, for instance, for a human baby, by the age of two, to use one hand to turn upright and then hold a cup, and use her second hand to clasp a few small spoons together out of a set of random cutlery, which includes larger spoons, forks, and so on, before finally placing the small spoons in the cup. By grouping the small spoons together, she is demonstrating at least two conceptual levels above individual items because she is picking objects based on two combined categorical features. She is then demonstrating knowledge of another level again when she puts the spoons in the cup, by linking a group with another item. These seemingly simple acts appear largely beyond any other primate, at any stage of development. But, of course, humans rapidly learn far more complex, hierarchical concepts and actions than this relatively straightforward example.
In a more formal series of experiments exploring this issue, humans between the ages of 11 and 36 months and mature chimpanzees, bonobos, and capuchin monkeys were all compared on the same simple task. All subjects were given three nesting cups, a standard children’s toy (see Figure 7). The experimenter repeatedly demonstrated what was required of the subjects: To place the smallest cup inside the middle cup, and then put both cups together inside the largest one. The cups were then dismantled and the subjects were encouraged to copy the experimenter exactly. This is a useful developmental test for human babies. When they are around a year old, all that most children can do is place on
e of the three cups inside another, leaving the third untouched—in other words, they can’t complete the task. By around 16 months, they can complete the task, putting the middle cup inside the large one and then the smaller one inside the other two. This isn’t quite what the experimenter showed them. Critically, these toddlers haven’t yet grasped the complex, hierarchical idea that you can move two cups at once (the middle one with the smaller one inside), as if they were a single compound item, and in addition that this group of cups can still be placed in any cup larger than the outside one. (I’ve watched my daughter at this age play with the same kind of toy, and even if I place the smallest cup inside the middle one to help her to solve the puzzle, she will deliberately take the smallest cup out before stacking the cups one by one [middle into largest and then smallest into middle]. For her, it seems as if the concept of groups of nesting cups is seemingly impossible. The group simply has to be dismantled for progress to be made.) Finally, many children from around the age of 20 months or so can exactly copy the experimenter. This shows that they have grasped the idea of hierarchies, so that two objects put together can in some sense be seen as one single object.27 You might suspect that part of their mastery of this progression arises simply from infants and toddlers having ample time outside of the lab to know how to manipulate these or similar items. But although that’s true of most babies in Western culture, this and a related experiment were also carried out in Zinacantecos babies and toddlers in southern Mexico, and they also exhibited mastery of hierarchy. This Mayan group has few materials in its environment on which to practice these kinds of manipulations, and the children have no toys, but the Zinacantecos children showed exactly the same pattern of development, suggesting that this stepwise acquisition of the mental machinery of hierarchical chunking is universal.