Murph, who is 10, sits in front of his computer screen playing a game that consists of classifying images. He finds out how he is doing after several rounds, with happy whoops for right and buzzers for wrong. He doesn’t seem to like being told he is wrong but has learned that he can avoid this by choosing to pass when he is not sure of the correct answer. Natua plays a similar game, but he has to discriminate between sounds, choosing whether they are high or low-pitched. He doesn’t like being wrong either, and when the task gets really tough he also chooses to pass rather than guess.
We all know that feeling of not knowing. When an answer pops into our heads we get a buzz of confident recognition, an “aha” moment. If we don’t know, we might feel a frustrating tip-of-the-tongue sensation, a mild form of panic or embarrassment, perhaps even gut-wrenching anxiety when we realise that we don’t have a clue.
Maybe that explains why Murph and Natua would rather pass and admit they don’t know than get it wrong. People do this all the time, so what’s the big deal? Well, Murph is a monkey, and Natua a dolphin. Their game-playing has got some researchers very excited because it suggests that these animals might know something about what’s on their own minds.
Advertisement
Knowing what you know, and what you do not, may not sound that clever, but it is a very important mental skill. Philosophers have long debated the significance of being able to think about thinking or know about knowing. This type of abstract thought, which is called metacognition, is at the heart of Descartes’s pronouncement “I think, therefore I am”. Now some researchers even believe metacognition could be an important first step on the ladder of consciousness: knowing what’s on your mind might be a prerequisite for knowing about whose mind it is, which is important for having a concept of self, self-awareness and, eventually, full reflective consciousness.
In practical terms, the ability to think about what you know is a big advantage. Getting things wrong can be costly, even if it is just wasteful of time, and metacognition allows you to pause, reflect and seek more information if you need it. With something so useful, you might assume other animals would have the ability too, but metacognition has never been demonstrated in rats or pigeons – the workhorses of the psychology lab – and it is hotly debated even in primates. It is still often assumed that abstract thought needs some form of language. No wonder the apparent discovery of metacognition in monkeys and dolphins is rattling a few cages.
It all began in the early 1990s, when David Smith, a psychologist from the University at Buffalo, part of the State University of New York, was given the opportunity to study a bottlenose dolphin, Natua, in a harbour in Florida. He and his colleagues began testing Natua’s ability to discriminate between sounds as a prelude to exploring ideas about metacognition. Natua listened to a high-pitched tone, then another tone that he had to classify as either high or low in pitch by pressing one of two buttons with his nose to receive a food reward for a correct answer.
At first the pairs were very different – a major seventh apart. Gradually Smith introduced sounds that were more and more similar until they seemed to be around the threshold of Natua’s ability to discriminate. He noticed that when the answers were easy the dolphin sometimes seemed super-confident, enthusiastically lunging for the button and soaking the precious electronic equipment.
His ingenious solution was to waterproof the switches with condoms, which gained him quite a reputation at local stores – if you ever need to purchase vast numbers of condoms, Smith advises, note that it doesn’t help to say you need them for your dolphin research. It may have ruined his social life around the marina, but it also gave him an idea.
If Natua could appear confident of knowing and more hesitant when the task became difficult, could he also report when he didn’t know? Could he master a third button to say “I don’t know” and show a sign of metacognition? Smith added the new button, which switched off the sound and moved to the next trial without a reward. Then he waited for Natua to discover it. “We just had to wait until one day he got a hard trial, and said ‘this sucks’, and chose to use the move-to-next-trial button instead of answering,” said Smith. It took a while, but from then on, Natua would use the third button to pass on a particular trial whenever he was uncertain. Unfortunately Smith never got a chance to try other tasks, because just as Natua had mastered this one, his film career took off, and he went away to make a movie.
Back at the lab, Smith decided to continue his experiments with rhesus monkeys, working with colleagues Michael Beran, Josh Redford, Wendy Shields and David Washburn. Murph and his chums learned to use a joystick to choose between various patterns and shapes displayed simultaneously on the screen. Sometimes they had to choose the densest pattern, sometimes the pattern with the smallest or largest number of objects. They were given food pellets for correct answers and a “time out”, when the screen would go blank for a few seconds, for incorrect ones. The monkeys seemed to love playing the game and became agitated when the blank screen stopped them. So they soon learned to choose an on-screen icon for “pass” when the test got too tough. This didn’t give a reward or time out but simply moved to the next trial.
In 2003 Smith and his team wrote a detailed account of their findings, and it was published along with comment from a dozen or so other researchers (Behavioural and Brain Sciences, vol 26, p 317). The commentators all agreed that the discovery of metacognition in animals would be very exciting and significant, but some remained unconvinced that this was what Smith had demonstrated. Several pointed out that by choosing the “pass” icon when things get hard, monkeys were not penalised by time out and so would get another chance at a reward more quickly. Herb Terrace, a psychologist from Columbia University in New York, points out that by passing on a test a monkey might not be saying “I’m not sure”, but just “this way I don’t have to wait”. It could be a simple conditioned response, whereby the monkeys have learned the behaviour that leads to the greatest reward.
If this interpretation is correct you might expect animals with less complex brains to respond in the same way. Try as he might, however, Smith could not get rats ever to use a “pass” option for opting out of difficult trials. And when Sara Shettleworth of the University of Toronto in Canada tried the same strategy with pigeons she found they just used the “pass” option at random. A better test came earlier this year, however, when Smith changed the regime for rewarding and punishing his monkeys, so that they only received the feedback after a group of trials. It would be hard to imagine how simple reinforcement could explain the uncertain behaviours in this case, he says. Not all the monkeys mastered the task this time, though – leaving a glimmer of doubt in the minds of some researchers.
Smith’s latest experiments will surely go some way to dispel this. He has now shown that monkeys who learn to use the “pass” button in his original test can immediately transfer that response to a brand new task that also creates difficulty and uncertainty for them. This is an important step because it implies that the response cannot simply be a conditioned one.
Remembering and forgetting
Following a slightly different line of enquiry, another group has published a paper entitled “Rhesus monkeys know when they remember” (Proceedings of the National Academy of Sciences, vol 98, p 5359). Robert Hampton, now of Emory University in Atlanta, Georgia, and his colleagues showed monkeys an image, then a short while later let them choose whether or not to take a memory test. If a monkey chose to go ahead, it then had to pick out the image from a selection of four pictures. As a control, to make sure they were not just passing at random, sometimes the researchers presented a blank screen instead of an image, making the task impossible. Monkeys almost always passed on these tests, implying that they had checked on what was in their memory, found it empty and so treated the test as one in which they had forgotten the initial image. They really did seem to know when they didn’t know.
Psychologist Janet Metcalfe of Columbia University believes these findings are even more convincing than the perceptual tasks that Smith’s monkeys were doing. “This is not just a judgement about what is in the environment,” she says. In Hampton’s experiment, the stimulus isn’t present any more – it must be remembered, so “this is definitely going on inside the mind rather than outside in the world”.
If that still leaves any doubt, new work by Terrace and colleagues Nate Kornell and Lisa Son, due to be published in January in the journal Psychological Science, combines all these methods. It shows that monkeys that learn a “not sure” response in a perception task can then apply it to other perceptual and memory tasks. First the researchers trained rhesus monkeys to make a judgement about which of two lines on a touch-sensitive screen was longest (or, for some monkeys, shortest). Once the monkeys had got the hang of that, they were taught to show their confidence in their answers by betting on them. The monkeys had a pile of virtual tokens pictured in a “bank” in the corner of the screen and were allowed to place either a high-risk bet of three tokens or a low-risk bet of one before giving each answer. Once they had earned a total of 12 tokens, they were given a food reward. “They were not rewarded for being right or wrong, they were rewarded for making an accurate judgement of how well they knew the answer,” says Terrace.
After this the researchers switched to a different perceptual task that the monkeys had already learned, picking out the greatest number of objects from pairs of groupings. The animals had no trouble combining this with the betting, which implies that this is not a conditioned response but instead a confidence judgement based on the animal knowing something about its own state of mind. As a final test the researchers gave the monkeys a memory test too. They were shown a sequence of pictures and the task was to indicate whether a subsequently presented image had been in the sequence, betting on the answer as before. Again, they had no difficulty, suggesting that they could make judgements based on some kind of general sense of confidence, whatever the task.
That still leaves the question of what exactly is going on inside the animals’ heads. Terrace points out that we can never know whether they “feel” confident or unsure in the same way that we do. He wouldn’t push the argument, he says, but adds: “It’s interesting how upset the monkeys get when they lose three tokens.” There were many tantrums.
“It’s interesting how upset the monkeys get when they lose three tokens. There were many tantrums”
Smith agrees that the interpretation of all these experiments can only be taken so far. “I can’t claim these monkeys show fully fledged consciousness, but I have shown the exact cognitive analogy to what we have in humans, and for us it is conscious.” At the very least, he believes tests for metacognition could be another tool for comparative studies of animal cognition, equivalent to the mirror self-recognition test, tool use or language skills (see Diagram), and he plans to look for it in other species, including chimps and the less cognitively savvy new-world monkeys. There are already some studies showing that chimps seek more information when they need it to solve a problem and some anecdotal reports that orang-utans can monitor their own actions – for example, Rob Shumaker from the Great Ape Trust of Iowa in Des Moines says that orangs will sometimes refuse to continue doing a selection task if they think they have made a wrong choice, holding out until they are allowed to try again.
Metcalfe also sees metacognition as a key step on the ladder to full consciousness. “This capability to reflect upon something that is mental, not in the environment, is the beginnings of something called high-level consciousness,” she says. She points out that most animals seem instead to be trapped in the moment by a particular stimulus, unable to get out of their situation, while humans reflect upon memories, and imagine how things could be different. “Metacognition is the first glimmering of that,” she says. What’s more, when we reflect on our memories we feel certain that they are ours and not someone else’s. This leads Metcalfe to suggest that knowing what you know could be the platform on top of which develops a sense of self or agency that can connect you to that knowledge. That would make metacognition a bridge to full reflective consciousness.
“Knowing what you know could be the platform on which develops a sense of self”
Thought without language
Hampton is more circumspect about whether the ability to introspect shows a level of conscious awareness. “In humans it’s the main capacity associated with the experience we call consciousness, but there is nothing an animal can do to tell us about the quality of its private experience.” For Terrace too, speculation about consciousness is a sideline. He believes the real news from this work is that animals can think without language. “It should now be up there in neon lights: contrary to Descartes, you don’t need language to think. Cognition is one thing and language is another.”
Smith also believes his findings have additional implications. “These animals have active cognitive lives, and that raises the bar towards making sure we treat them respectfully,” he says. Clearly they have close analogues to conscious uncertainty. Maybe it will not be too big a step to get them to report when they are uncomfortable or suffering too. “My research makes you think that it’s not out of the question to get animals to report ‘I hurt’,” Smith says. We know that animals feel pain and try to avoid it, but do they reflect on pain and suffer emotionally as a consequence? That, says Hampton, may finally become clear as a result of this work. If so, prepare for a revolution in animal welfare.
