Chapter 5

True Belief

'There's no use trying,' she said: 'one can't believe impossible things.' 'I dare say you haven't had much practice,' said the Queen. 'When I was your age, I always did it for half an hour a day. Why, sometimes I've believed six impossible things before breakfast.'

Lewis Carroll Alice Through the Looking-Glass

Within a month, /Ere yet the salt of most righteous tears/Had left the flushing in her gallèd eyes,/She married. O, most wicked speed, to post/With such dexterity to incestuous sheets! 

moans Hamlet about the speedy marriage of his mother to his uncle Claudius whom Hamlet correctly suspects has killed his father (Act I, scene ii). Hamlet's former love, Ophelia, has been sent to discover the source of Hamlet's apparent madness. Neither knows that the King and his counsellor, Polonius are listening, but halfway through their conversation, Hamlet becomes suspicious. He mistakenly believes his mother is eavesdropping, and so, for her benefit, he maligns Ophelia and denounces her sex as a way of hurting the Queen.

I have heard of your paintings too, well enough. God hath given you one face, and you make yourselves another. You jig and amble, and you lisp. You nick-name God's creatures and make your wantoness your ignorance. Go to, I'll have no more on't. It hath made me mad. I say we will have no more marriage. Those that are married already - all but one - shall live. (Act III, scene i).

However, Hamlet is wrong; it is the King who is listening and who benefits from Hamlet's false belief (that it is Hamlet's mother who is listening) as he realises that Hamlet plans to murder him in revenge for his father. How and when we begin to understand another person's false beliefs are to be the topic of this chapter.

In chapter three, we saw how young children and animals first begin to understand the language of the eyes, initially by detecting the direction in which a person is looking and by the dawning realisation that looking at another's eyes can be informative. In chapter four we saw how eyes acted as a salient clue when guessing someone's emotions and how an understanding of what someone is paying attention to can lead to an understanding of their desire. In this chapter, I want to draw these two strands together, first by returning to the eyes and showing how, in older children, 'seeing' becomes a much more complex activity. By understanding what someone else can see, it is possible to understand what they know: this first glimmer of another person's thoughts filters through a child's consciousness at the age of three. Second, we will discover how and when beliefs are understood, and finally, decipher the litmus test of Theory of Mind: comprehending another being's false belief about the world.

We all take the process of seeing so much for granted that we are no longer consciously aware that looking at an object leads to acquiring information about it. Yet children younger than three or four, and most autistic people do not have this insight. A very simple test is enough to prove this. Imagine that two people are standing next to a cupboard, one is leaning against the cupboard, and the other one has opened the door and is looking inside. Surprisingly, neither young children nor autistics can tell you which person knows what is in the cupboard. In order to know about our environment, you have to have information about it, and to do this, you need to see it. It is this link between knowing and seeing that is not understood by small children or those with autism.

Young children, people with autism and animals can link gaze to a person's or an animal's actions, but they don't necessarily mentally connect gaze with actions. The fact that an organism can appreciate the properties of seeing as I outlined in chapter three, does not mean that it understands the mental consequences of its perception. This comprehension usually appears in young children between the ages of three and four. Prior to this, children conflate the idea that seeing will lead to knowing with their burgeoning comprehension of other people's desires; for example, they think you must know all about an event or a situation if you are interested in it. In the cupboard example, if the person leaning against the cupboard and unable to see inside acts enthusiastically, a two year old will think that this person knows more about the cupboard’s contents than the person who is actually looking inside but is not particularly excited about the left-over can of beans she can see.

Although young children cannot understand that seeing leads to knowing, they can grasp the fact that eating leads to satiation or that starving leads to hunger. At first, this distinction, made by Josef Perner, from Salzburg University in Austria, and Jane Ogden, from the University of Sussex, seems fatuous, but hunger and knowledge have much in common. Both are internal states which cannot be seen, both rely on internal organs (the stomach and the brain), both are 'fed' with external products (food and information) and their state can be inferred from the situation a person is in and how he is behaving. For instance, if we see a boy peer into a cupboard which contains his football, and we find out he wants to play football, we assume that he knows where his ball is and will fetch it. Similarly, if someone is prevented from eating dinner, we assume that he or she is going to be hungry. There are degrees of hunger and knowledge: a person who has just eaten two tubs of ice-cream is less hungry than someone who has only eaten a Mars bar; a child who looked through a box thoroughly will have a greater knowledge of what's inside than her friend who has only given the box a cursory glance.

The major difference between these two internal states is that eating leads to satiation and can be physically felt. We never seem to get full of knowledge - there's always room for another fact - and unlike the phrenologists, we cannot point to our right temple and say, 'There, that's the place where I hold the knowledge that Aunt Martha is hungry'. Three year olds understand perfectly that if Dad won't let you have any tea, you'll be hungry, but fail to comprehend that because Jasmine saw the ball in the garden, that's where she'll look for it. This is not because they do not understand the interaction between observable events and internal states, but for quite another reason. Josef Perner thinks children's can't understand knowledge because they can't represent mental states. We saw in the previous chapter that young children have difficulty comprehending desires and it is only later that they develop a more explicit mental understanding. Until this time their responses consist of attitudes towards situations. This still allows young children to have a rich and complex mental repertoire: they understand their own behaviour and other people's in terms of their goals, fears, dislikes, wishes, and so on, and can imagine various scenarios, provided they are linked to concrete situations.

Although children younger than three can tell you what's going on in their head, they are incapable of justifying why they do or do not know a fact. Children of this age think adults know more than they do, indeed, more than adults could possibly know, and they think other children know less than they could know. Neither do they realise that if they are told a secret, they possess unique knowledge. They act, instead, as if everyone knows, as many a parent has found out to his embarrassment. It is not that they are completely egocentric, it's just that they don't understand the conditions under which knowledge is acquired or that misleading information leads to false beliefs.

Autistic children have even more of a problem with understanding knowledge; about fifty per cent of autistic children will fail a simple, limited knowledge task. The test, devised by Alan Leslie and Uta Frith went as follows: one person and the child watched whilst an experimenter hid an object - a yellow counter - in full view of both of them. The person who was watching then left the room. The experimenter produced an identical counter and asked the child to hide it. The child was asked if the person who had left the room knew where the counter was that the child had hidden, and, if that person came back, where she would look for the counter? Obviously the answer is that she doesn't know where the hidden counter is because she didn't see it being hidden, and she'll look for the counter she saw the experimenter hiding, yet autistic children are unable to answer these questions.

Between the ages of four and eight, children's understanding of what other people know when they see objects changes drastically. Husband and wife team, John and Eleanor Flavell, and their colleague, Frances Green, are all from Stanford University, California. They've shown that at four years old, children show little comprehension that a person who is preoccupied with some visual task, such as staring intently at a photograph to work out if they know anyone pictured in it, will not be paying attention to other inconspicuous or obviously irrelevant objects that happen to be in that person's visual field. So, for instance, the children would say the woman staring at the photograph is also paying close attention to the gilt- edged picture frame and the piano in the corner of the room. By the age of eight, children have a sophisticated grasp of how seeing can determine what someone is paying attention to and the process of thinking itself. Eight year olds understand that the picture frame is not of no interest, but perhaps fragmentary attention will be paid to it whilst the woman focuses most of her attention on the photograph. While their friend is thinking about what movie she went to last week, she might appear to be looking at the piano, but she is not attending to it. Six year olds were bang in the middle: they are better than four year olds at this test but not as confident as eight years olds. The Flavells put it this way: 'Four year olds may implicitly conceive of the mind as more like a lamp than a flashlight, that is, a device that can radiate attention and thought in all directions at once rather than in only one direction at a time.' Four year olds don't assume that people are paying attention to everything in the room, it is just that they haven't figured out how attention and perception work. They overestimate people's attentional capacity in the same way that they overestimate brain power; they often exaggerate how many things they can remember at one go, and it is only later that they realise how depressingly few things one can memorise. This lack of understanding of perception parallels unfamiliarity with the process of acquiring knowledge and their incomprehension when faced with information that is false or misleading.

It is only after the age of five that children can refer to the brain as an organ for thinking and talk about its mental functions, such as dreaming, remembering and imagining. Autistic children have no idea that the brain is used for thinking. To them it is an organ like any other. When asked what the brain does, they say things such as, 'It makes you move.' Uta Frith once conducted an experiment on reading with some autistic children. When one child did particularly well, she asked, quite by accident, 'Oh, how did you know that?' He replied, 'By telepathy.'

It's almost impossible to determine whether animals think and actively know they are thinking. The only animals you can ask are ones that will answer. Sue Savage-Rumbaugh was driving through the woods with Kanzi's sister, Panbanisha, when she noticed that the bonobo appeared to be very quiet and pensive. 'I was moved to ask her what she was thinking - a question I generally avoid since I have no means of validating the answer, nor even if an ape understands the question. However, at this moment Panbanisha looked literally lost in thought, so I dared. She seemed to reflect upon the question for a few seconds and then answered "Kanzi". I was very surprised as she almost never uses Kanzi's name.'

We have looked at how understanding vision leads to understanding knowledge, and explored children's mental understanding of attention. Can animals understand that seeing leads to knowing? Daniel Povinelli and Timothy Eddy wanted to find out whether young chimpanzees aged between four and five could understand that vision shows what a person or another chimp is paying attention to: could the chimps interpret visual perception as a device that connects them to others and to events and objects in the world? Would they gesture only to those people who could see their gesture? If chimps cannot understand that when a person looks at them, they are paying attention to them, they should gesture to anyone, regardless of whether they were looking at them or not. Povinelli and Eddy trained the young chimps to beg for food from a trainer. They then gave them a choice of two trainers, one of whom could see the chimps, and one whose vision was restricted. The simplest way the trainer was prevented from looking at the chimps was by standing with her back to them. Other ways that Povinelli restricted the trainer's vision was to get her to cover her eyes with her hands, or a blindfold, or put a bucket or a paper bag over her head. The two trainers swapped roles so that the one who could not see the chimps was not always the same person.

The young chimps showed no immediate disposition to gesture towards the person who was looking at them, although they rapidly learnt from whom to beg for food. Towards the end of the study they were correctly able to ask for food from the person who was able to see them from the very first trial. These results could be interpreted in several ways. Because the chimps learnt rapidly, they may have understood that a person is only mentally connected to you if they can see you, but, because of their youth and the fact that it was a rather alien experimental set-up, it took them a few attempts to work out what was being asked of them. Another hypothesis is that chimpanzees develop mentally much more slowly than human children; it is possible that chimps of this age still had not reached the same level as three- year-old humans. Indeed, in other experiments on older chimpanzees, Povinelli seems to have shown that chimps are capable of a far more sophisticated approach to seeing and knowing.

An alternative explanation is that the chimpanzees rapidly learnt a rule, such as, 'gesture to the person holding a bucket, not the person with a bucket over their head'. This rule may then have become generalised: 'Gesture to person whose eyes are not covered.' Since they were rewarded with food each time they gestured to the correct person, these rules are likely to be picked up quickly. But when Povinelli gave this test to three-year-old children, from their first attempt, they had no problems indicating which person was looking at them in order to get their sweet reward. That the chimps had learnt a set of rules very quickly now seems the most likely explanation, for, two years later, Povinelli repeated the same experiment on the same chimpanzees. Although they could immediately choose the person facing them as opposed to the person who had her back to them, on any other combination of seeing and not-seeing they were right back at chance levels - yet in the intervening two years these chimpanzees had all participated in experiments that were connected to gaze-following and attention. 'I was stunned, to be honest,' says Povinelli, 'If they had a mentalistic understanding of attention, why on earth do they fall apart when they're tested again? They had to learn all the rules again - it's exactly as if the apes became primed to rapidly learn what the relevant cues were.' He adds that he doesn't think their joint attention behaviour is mentalistic. 'It's not that they don't explore body posture and gaze, they clearly do - they're very sensitive to gaze direction. But they learn rules, their gaze-following is not a window into the mind.'

Andy Whiten, from the University of St Andrews in Scotland, disagrees. Whiten worked with Sue Savage-Rumbaugh's chimps to find out whether they understood that attention was like a 'psychological spotlight'. The experiment was simple. Normally when researchers want the symbol using chimps to name an object, a researcher will hold up an object and ask the chimp to name it. Whiten sat next to a chimp surrounded by every day objects. He obtained eye contact with the chimp and then looked at one of the objects and said, 'What's that?' The chimps spontaneously named the object in a rather eerie fashion; some of the chimps have now been trained to use a computer with a joystick that moves a cursor around the screen. When the chimp moves the cursor to the desired symbol for the object, the computer speaks. Whiten appeared to be glancing at objects, asking what they were, and the chimps would respond by saying out loud, 'String,' or 'Television'!

Whiten tried other scenarios such as looking at objects placed behind the chimp, or sitting with his back to them - the chimp had to guess what he was looking at using only the back of his head. The only thing the chimps weren't good at was determining what Whiten was looking at when he moved his eyes but not his head. This ties in with Povinelli's work which suggested that for chimpanzees the eyes alone were not indicators of attention. This may be because chimps will naturally turn their head to look at objects rather than only move their eyes and, in addition, the white sclerota that surrounds the iris cannot normally be seen. When we glance to one side, it is very easy to see what we are looking at, but harder when a chimpanzee does since the whole surface area of their eye is dark. Whiten was able to train the chimps to respond to eye gaze alone, but they seemed to interpret this as something quite separate from the 'mentalistic understanding of attention' they were using to name the objects Whiten was attending to.

A few years before Whiten’s experiments, Povinelli conducted some rather more sophisticated tests on chimps (albeit older ones) which were designed to test whether they could understand that seeing leads to knowing and not-seeing leads to ignorance - in other words, if you haven't seen an event, but pretend to know, will chimps understand that you are merely guessing? One person hid food in one of four cups, whilst a second waited outside the room. Both people then pointed to a cup. The person who had hid the food was the Knower - he had seen where the food was and consequently pointed to the correct cup. The Guesser, who had been outside the room whilst the food was hidden, pointed to any cup at random. If a chimp gestured towards the Knower, the Knower would share the hidden food reward with the chimp. The two people playing the Knower and the Guesser swapped roles so that a chimp watching would have to understand the connection between seeing and not-seeing, rather than learning that one person would always point to the cup with food in it.

The chimpanzees Povinelli tested were aged between seven and twenty-eight and they quickly learnt to point to the Knower. They seemed to have grasped the complex mental difference between the state of mind of a person who knows, and one who is ignorant: when a third trainer hid the food whilst the Knower watched, and the Guesser remained in the room, but with a bag over his head so he couldn't see, the chimps still chose the Knower. However, the chimps could have rapidly learnt a rule that enabled them to choose the right person, just as they did in the seeing-knowing experiment.

Povinelli and his colleagues also tried this same test on rhesus macaques but none of them was able to gesture to the Knower. Their inability certainly did not stem from lack of trying on Povinelli's part: most of the monkeys had over 600 goes at this experiment, and one, a male called Tuck, had more than 800. Povinelli then made the trainer who played the role of Knower wear a pink glove, and the monkeys were able to indicate correctly who 'knew' where the food was. All that they were doing, in fact, was using the pink glove as a clue to discriminate between the right person to pick and the incorrect one. The monkeys had learnt a simple rule: to get food, point to person wearing pink glove. Any rules the chimps might have learnt would have been far more sophisticated and complex since they were not provided with any salient and consistent cue like the pink glove.

Although the term 'Theory of Mind' was coined by psychologist David Premack following observation of a chimp, Sarah (who, incidentally, was the twenty-eight-year-old subject of Povinelli's experiment), thereafter, the bulk of experiments have been conducted on children and most of the tests carried out on animals have been adapted from ones previously done on children. Povinelli works in the opposite way. He usually conducts his experiments on chimps first and then on children. He found that all the children he tested who were older than three were able to point to the Knower. However, none of the three year olds - but most of the four year olds, was able to tell him how they knew who was the right person to point to. Indeed, one child said she knew where the sweet was because her mother fed her lots of carrots so she could see very well! Povinelli concludes that even though three- year-old children have been pointing and responding to pointing for nearly two years, they do not appear to be aware of how this gesture functions to create knowledge in themselves and others. Children of this age have yet to understand the link between perception and knowledge: they still don't fully understand that seeing leads to knowing.

Even more unusually, other psychologists picked up on Povinelli's experiment with primates and tried it out on adults. Joseph Gagliardi, Mark Blumberg and colleagues from the University of Iowa, in Iowa City, conducted the experiment on 'thirty-seven undergraduate students (Homo sapiens) enrolled in introductory psychology courses.' Unsurprisingly, the adults quickly, within a couple of trials, correctly chose the Knower.

The same skill - understanding someone else's state of knowledge - has been demonstrated by Sue Savage-Rumbaugh and two of her symbol-using chimpanzees, Austin and Sherman. On the first trial, Austin followed Savage-Rumbaugh to the fridge where she put some banana slices in an opaque container and put the lid on. She led the chimp over to a colleague, Sarah Boysen, who had not seen the transaction. Boysen asked Austin to identify what was in the container. He pressed the symbol for 'banana'. Sherman, who had watched Austin disappear into the kitchen and return with the container, quickly also lit up 'banana' on his keyboard. Savage-Rumbaugh says, 'While we had no reason to believe Austin recognised on the first trial that he was telling Sherman what was in the container, Sherman seemed to believe that Austin was describing the container's contents [. . .] The singular fact that the chimp who did not know what was in the container nearly always waited and watched until the other one revealed its contents was sufficient to show that the chimp recognised differing states of knowledge, based on observing or not observing the baiting process.'

Although the chimp that hadn't seen the food being hidden seemed to understand that the other one knew what was in the container, the 'knower' did not act as if he understood that he needed to tell his companion what was in the container. In any case, it didn't matter to him whether the other chimp knew or not - they both got the goodies in the end. Savage-Rumbaugh imposed the constraint that both chimps had to signal correctly what was in the container before it was opened and the food was shared. This meant that the 'informer' needed to ensure that the other chimp was paying attention. It took the chimps a few attempts before they realised what was being requested of them. On one occasion, Sherman, who had seen the container being baited, pressed the key for 'apple'. Austin, who prefers bananas, perhaps out of wishful thinking, pressed 'banana'. The container was opened, and Austin immediately tried to correct his mistake to 'apple'. After that, Sherman monitored Austin very closely and if ever he looked hesitant, Sherman would quickly repeat what the food was. But again, Sherman and Austin could have learnt a complex set of rules, motivated by the fruit.

On another occasion, the researchers put a banana in a cardboard tube with covered ends and left it in a cage. A chimp called Mercury discovered the fruit. The next day he was again let into the cage on his own. He raced across to the tube expecting another treat. What he found inside make him practically jump out of his skin with fright. A live iguana was curled up inside the tube. Mercury gave 'waa-waa barks', his hair stood on end and he rushed back to Panzee, Lana and Sherman. They reacted by barking anxiously. When they entered the cage that Mercury had been in, they were extremely cautious and approached the cardboard tube with some trepidation. The researchers assumed that Mercury had 'told' the others about the iguana, understanding that they did not know about the scare awaiting them in the next room. However, Mercury's fear may simply have frightened the other chimps. They, in their heightened state of fear, were expecting to see something awful in the next cage.

We've seen that children only from the age of four onwards can understand that seeing leads to knowing and that only children over this age can grasp the concept of knowledge as a mental state. Chimpanzees over the age of six may also be able to understand that if you are visually connected to a situation, you are paying attention, and that seeing leads to knowing, whereas not seeing leaves a person in ignorance. However, it is not clear whether chimps are using a complex set of rules, or whether they do have a full mental awareness. In the previous chapter, we saw how children younger than four cannot understand beliefs, and hence, even though they can normally understand another person's desire, they cannot predict a character's desires when that character's belief is involved. At around the time when they are beginning to grasp what it means to know something, young children also start to understand a fundamental mental concept: simple beliefs. For example, given a story such as 'Sam wants to find his puppy, but the puppy is lost. It might be hiding in the garage or under the porch. Sam thinks his puppy is under the porch,' children of three to four are able to predict where Sam will look for this missing puppy. Sam's belief is correct, the puppy is under the porch. However, at this age, children still have problems with false beliefs.

A false belief is where a character's belief differs from the situation in the real world and consequently affects their actions. At around age four and a half, children are able to answer a simple false belief task where the child is explicitly told that the central character holds a belief that is wrong. In the case of Sam and his puppy, the child would be told that although Sam thinks the puppy is under the porch, it's really in the garage.

Heinz Wimmer and Josef Perner were the first psychologists to show that children under the age of between four and five cannot understand false beliefs. They devised a task now called the Sally-Ann test. Children were told a story about two dolls, Sally and Ann. Sally has a basket and Ann has a box. Sally places her marble in her basket and goes out. While she's out, naughty Ann moves Sally's marble from the basket to her own box, then she goes out. Sally comes back in.

The children were asked where Sally would look for her marble. Of course, Sally will look for it where she thinks the marble is - in her basket. Sally has a false belief about the marble's location. Children are not able to give the correct answer to this question until they are older than four or five. This understanding of false beliefs is what truly opens the gates to a full comprehension of other people. False beliefs are a hallmark of Theory of Mind because in order to understand another person's false belief, it is necessary to comprehend their internal, invisible mental world rather than referring to the situations in the real world. It is a skill understood the world over, from North America to Japan, and even amongst preliterate people living an ancient hunter-gatherer life-style. Jeremy Avis and Paul Harris, from Oxford University, have shown that the Baka pygmies who live in the rain forests of south-eastern Cameroon can understand beliefs and desires. The children normally play in a miniature village, Ndabala, behind the main camp, where they practise hut-building, spear-shaping and fire-making. One child was summoned from there to a star-shaped fire. The two experimenters, Mopfana and Mobissa, cooked wild mango kernels, a favourite delicacy. Mopfana put his kernels in one of the cooking pots and covered them with a lid. He then went to mbanjo, the male meeting place, to smoke. Mobissa and the child hid the kernels in a different pot. Mobissa asked the child where Mopfana would look for his mango kernels when he came back, and also asked, 'After he lifts the lid, will his heart feel good or bad?' The children were able to answer both questions correctly, the first about Mopfana's false belief, and the second on his desire - at the same age as Western children. 'The fact that belief-desire reasoning emerges at approximately the same age in such diverse settings strengthens the claim that this mode of reasoning is a universal feature of normal human development,' says Avis.

But although children with Down's syndrome can pass this task, the majority of autistic children cannot. Simon Baron-Cohen, Uta Frith and Alan Leslie gave a non-verbal version of this test to autistic children by telling children a story in pictures and asking them to arrange it in the right order. 'I was looking for ways to test what Daniel Dennett calls the intentional stance,' says Baron-Cohen, 'Imagine you're trying to make sense of a social interaction frame by frame: without the intentional stance, without the notion that someone wants something,- you wouldn't see the links, you'd just see slices of behaviour.'

The stories showed a causal sequence where a girl trips and hurts her knee; a desire scenario, where one child wants the other child's ice cream; and a false belief picture sequence, in which one child steals another’s teddy when she’s not looking. 'The autistic kids were okay at the physical causal stories, but they had a lot of difficulty trying to find a meaningful sequence for the mentalistic stories,' says Baron-Cohen. 'They never knew where to put the final picture in the false belief story. The child is surprised that her teddy has disappeared but the autistic children said she was tired, or she was yawning.'

The Sally-Ann test is essentially the same as the Smarties task, where the child is asked what is inside the tube, only to discover that it does not contain sweets, but is full of pencils, or birthday cake candles. As Janet Astington and Alison Gopnik discovered, children younger than five can neither understand someone else's false belief, nor report their own false belief. The very instant the lid is snapped off that Smarties tube and its true contents are revealed, small children are unable to report their own false belief. This seems an extraordinary finding. In Western philosophy, we've always treated the mind as if it were private and as if we, alone, had access to our thoughts. As we said in the previous chapter, it runs contrary to our own belief system to discover that children do not know their own minds.

Perner says that it's because children of this age cannot metarepresent: in other words, they are unable to hold a representation about a representation in their heads. He thinks that because young children react to the world as if it were composed of situations, they cannot hold two conflicting views of the world - one must be right! Thus in a child's mind there is no hierarchy between different versions of the truth. When they are asked to search their memory for a belief, they look for a description of a situation and run into two conflicting situations - one in which they did not know what was in the Smarties tube and one in which they did. Unable to deal with two different modes of reality, they fall back on what is currently true - they know what is in the Smarties tube now.

Children have a similar problem distinguishing between an object's appearance and what it really is when they are younger than four. For instance, if you show a child a sponge which looks like a rock, only a five year old will be able to say, 'I thought it was a rock until I touched it; now I know it is a sponge, but it still looks like a rock'. A particularly striking example of young children confusing appearance and reality is seen if you give a child a white flower and place a blue filter above it. Children now think the flower really is blue. Whilst the flower is under the filter, one of its petals is removed and pulled out from underneath the filter. Another petal, a blue one, is placed next to the white petal. If children are now asked which petal came from the flower, they point to the blue one. Again, they are battling with two different modes of reality, and they can't hold them both in their mind at the same time.

Autistic people are unable to make the distinction between appearance and reality: to them the sponge looks like a rock, therefore, it must be a rock, regardless of the fact that they can scrub themselves in the bath with it. Simon Baron-Cohen says, 'If, as appears to be the case, most children with autism really are unaware of the appearance-reality distinction, as well as being blind to their own past thoughts and to other people's possibly different thoughts, their world must be largely dominated by current perceptions and sensations.'

Norman Freeman and Hazel Lacohée, from Bristol University, wanted to find out whether children's own false beliefs were lost and gone for ever, or whether they were buried in their brains and could be accessed. As they point out, the beauty of the Smarties test is that a handful of Smarties is a vivid reminder of their earlier belief. 'It would be a tribute to "amnesia" if children would, through a mouthful of Smarties, say that they had thought that the Smarties box contained pencils all along,' says Freeman sardonically.

What Freeman and Lacohée discovered was that children as young as three were able to remember their own false belief about the Smarties, but only during certain conditions. If they were given a handful of Smarties to eat when they first saw the Smarties tube, and the experimenter had a brief conversation with them about the sweets, and then, when they were shown the real contents of the tube, they were asked, 'When you were eating those sweets I gave you, what did you think was in this tube?' the children were able to access their past false belief. As Freeman says, 'Giving the children Smarties to put in their mouths can be categorised as getting the children literally to test a sample of reality.' Simply stuffing a few Smarties at the children and asking them what they thought was in the tube without any elaborate prompting did not work.

Giving the children a picture of whatever they thought was in the Smarties tube or in an egg box and asking them to post the picture through a slit in a shoebox also helped. The children were much better at recalling their past beliefs when they were given a picture of the object rather than the actual object to post into the shoe box. Presumably, this is because the picture reminds them of the egg that will be in the egg box, whereas, if they post an egg they are confused as, in their minds, the egg cannot now be in the egg box. These tests don't suddenly turned three year olds into the mental equivalent of a four or five year old. Children of this age can compute a false belief, but can't deal with two seemingly conflicting versions of the truth without clues to help them recall what they'd previously thought.

Even at the tender age of three, children are beginning to see glimmerings of the mental cogs whirring away in other people. Alan Leslie and Daniel Roth, from the University of Tel Aviv, tried to make the Sally-Ann test easier for young children by framing the story as a great deception. Sally has been given a bar of chocolate which Ann hides in her box when Sally goes out of the room. When Sally returns and cannot find the chocolate, she asks Ann where she's put it. Ann tells a lie and says it's in the dog's kennel. The children were asked where Ann thinks the chocolate is and then where Sally thinks it is. Although the three year olds still couldn't understand a false belief (where Sally thinks the chocolate is), they could say that Sally would think that Ann was lying about the whereabouts of the chocolate. This, Leslie and Roth say, indicates that three year olds have some inkling of the characters' mental states. Autistic children, however, always answered without recourse to mental states.

In fact, children as young as two years and eleven months old have an implicit understanding of false beliefs, even though they cannot verbally answer questions about a person's false belief until they are nearly five. Josef Perner, along with Wendy Clements, who is based at Sussex University, built an underground mouse house which had two exits on ground level. At each exit there was a coloured box, one blue, the other red. The two puppet mice protagonists were called Katie and Sam and Sam had a piece of cheese. The experiment was essentially the same as the Sally-Ann task, except that the children were filmed. Sam hides his cheese in one box, and when he's not looking Katie moves it to the other box. When the children were asked where Sam will look for his cheese, they correctly looked at the box where Sam had hidden it, but until they were four years and five months old, they always said that he would look in the box where Katie had hidden the cheese. The onset of an implicit understanding of Theory of Mind is very abrupt, as is the onset of an explicit and hence communicable Theory of Mind.

Understanding a false belief is quite different from understanding that the world has altered. This was aptly demonstrated by Alan Leslie and Laila Thaiss, currently at McGill University in Montreal, who showed normal four-year-old children and twelve-year-old autistic children how to use a Polaroid camera. They took a photo of a toy cat sitting on a chair, then laid the photograph on the table, face down and removed the cat. Leslie asked the children, 'In the photo, where is the cat?' This ingenious experiment is the equivalent of the false belief task: the Polaroid is 'out-of-date', just as Sally has an 'out-of-date' belief. Strikingly, all the autistic children were able to do this test, yet four year olds could not, even those who could pass false belief tasks. The reason may be simple though: autistic children who are given psychology tests tend to be older than four and so they may be better at the test because they have more experience with cameras and they are more familiar with the concept of pictures mismatching reality.

Leslie recognised this problem, so he repeated the experiment, but this time he made a map of the toy room and placed a sticker on it where the toy was before removing the toy from the room. The map is now 'out of date'. All the children found this task more difficult, but the results stayed the same: the autistic children who passed the map test were still unable to answer a false belief task properly and the normal children who passed the false belief test, could not do the map task. Leslie says, 'Autistic children can pass photograph and map tasks because they have a good command over the general cognitive processing demands of the tasks and over the concepts of "photographs" and "maps". However, these children tend to fail false belief tasks because they do not have easy access to the concept "belief". Three year olds and some four year olds tend to fail both sets of tasks because they cannot meet the general processing demands made by both sets of tasks (false belief and photographs). Older four year olds pass both sets of tasks because they have access to all the required concepts and enough of the processing resources required.' This still does not explain why the four and five year olds who can pass a false belief task sometimes fail the photo test. Leslie explains, 'They have greater access to the concept "belief" than they do to the concepts "photograph" and "map". This is so despite the fact that beliefs are private, unobservable and thoroughly "theoretical", while photos are public and observable entities. Beliefs ought to be much harder to learn about. For this reason, I don't think the normal mechanisms of learning will account for these findings. I'd argue that normally developing children possess a sort of "instinct" for attending to the invisible mental states of other people - a special intelligence.' Thus we are born with the potential to understand mental interactions well before we can understand concepts from the 'real' world.

John Swettenham, from The University of London, tried to see if he could teach children to pass false belief tasks. He designed a computer game based on the Sally-Ann test. In the program, Sally puts her ball in either a red box or a blue box. Sally leaves and Ann hides the ball. She then asks the child where Sally thinks her ball is. If the child correctly indicates where Sally thinks her ball is, music plays, and a flashing message, which changes colour, appears saying, 'Yes, well done!' Sally then searches for her ball and says, 'The ball is not there any more,' and waves her arms and stamps her feet in disappointment – this acts as a reinforcement, showing that the child has understood that the ball is not where Sally left it. If the child chooses incorrectly, Sally tells her the answer by saying, 'I think the ball is in the blue box where I left it.’

Normal three-year-old children, children with Down's syndrome and autistic children were then given the non-electronic version of the Sally-Ann task and three other belief tasks such as the Smarties test. Both the three year olds and the children with Down's syndrome were able to pass these other tests. Swettenham says, 'I think at three years normal children are pretty close to passing false belief tasks and the cognitive mechanisms for passing are there. Perhaps they just need the input - repeated examples and explanations - in order to get things working.' However, even though the autistics were almost eleven years old, they could pass only the Sally-Ann task they had practised and none of the other tests. Swettenham says they were probably using a rule of thumb, such as 'Pick the box where the ball isn't', so they were completely floored by other tests. When he visited their school three months later they recited the entire text of the computer game out loud and from memory!

Since autistic children are good at learning rules, Swettenham wondered whether he could come up with a rule that he could explicitly teach children to help them solve all false belief tasks. Initially he thought about using the analogy of the computer as a person with its own mind and teaching the children that computers make decisions on the basis of the information they have, thus using the computer to bridge the gap for the children. In this way, the children could learn about the concepts required for Theory of Mind so that they could use them to understand people. Eventually he, Simon Baron-Cohen and colleagues decided to capitalise on the idea of a photo in the head, which, he thought, was visually easier to grasp than the idea of a computer as a person. He said to the autistic children that the eye is like a camera and people have something similar to pictures in their heads: just as pressing the shutter on a camera captures an image, looking at a scene and blinking is a way of imprinting a picture on your mind. As Baron-Cohen says about autistic children, 'They understand about cameras so we've already got to first base.'

It is unlikely that normal children use a photo-in-the-head model for mental states (we've already seen how difficult it is for them to understand an out-of-date photograph), but for adults photos serve as a reasonable metaphor for what beliefs are - they're about things, and they endure even when reality changes. Swettenham says, 'For children with autism, they may be the closest approximation to what beliefs are actually like, if they cannot conceive of beliefs themselves.' For an hour a day for five days, the children were taught a Theory of the Mind-as-a-camera. They and the experimenter took Polaroid photographs of Sally's 'beliefs'. The photo was slotted into a manikin’s head, so that Sally, the manikin, literally had a picture in her head. The children were then taught to predict where Sally would look for her ball on the basis of the picture in her head.

After the training, the autistic children could not spontaneously use photos to understand false beliefs, nor explicitly link mental states to photos, but they could connect the photos to the character's actions. Before the teaching exercise, they all failed false belief tasks, and did not understand that seeing leads to knowing. After their training, most of them passed both the Sally-Ann test and a seeing-knowing test and about a third of them could do the Smarties task. Despite this explicit coaching, the children were still not using the photos to infer mental states, they were operating on a behaviouristic principle. 'I don't think they understood any more about mental states,' says Swettenham. "But that's okay. We're saying, "Let's teach them some behavioural rules or strategies to help them solve social situations - at least be able to predict peoples' actions a little better - with their own 'picture- based theory of mind."' He thinks that the results are promising and that a larger scale and more intensive study could prove fruitful. Baron-Cohen, pointing out that each child was only taught for five hours, says, 'You could argue that there should be a special curriculum for mindreading!' This does not mean that the autistic children would eventually become like normal ones, Swettenham cautions. 'I doubt they would eventually recognise mental states in the same way that we do, but I think they would develop their own way of explaining and understanding other people's behaviour using this strategy.'

Normal children don't need this special teaching, but, as Norman Freeman and John Swettenham have both shown, they can be helped to understand false beliefs. Freeman and a college, Charlie Lewis, told false belief stories to children, but did not read to the end of the stories. They then asked the children to retell the stories in their own words. They then read the same stories, but in full, and the children were tested for their understanding of the characters’ false beliefs. 'The kids cracked it,' says Freeman, 'even young three year olds. They'd got an understanding of false beliefs but they needed to consolidate the episode in their minds before they were able to switch to a mental framework.'

Freeman argues that these three crucial experiments - his and Charlie Lewis's where children rehearsed part of the narrative of the test, John Swettenham's study in which children were taught how to do a false belief test using a computer game, and the Perner and Clements experiment indicating that young children had an implicit understanding of false beliefs - all point in one direction. 'These three tasks show that three year olds have the computational base for understanding false beliefs, but they don't know when to use it. They can understand false beliefs if they are prompted whereas autistic people have to be taught.'

Although Theory of Mind is universal, the age at which children are able to pass a false belief test varies. Henry Wellman says, 'The main thing is not the exact age at which children pass or fail but the sequence of acquisitions, and it's the sequences and patterns with age that are most important.' Children begin by understanding joint attention, then comprehending fulfilled desires, followed by unfulfilled desires. They then develop an understanding of other people’s beliefs before finally comprehending false beliefs. However, a number of factors affect the age at which an individual child will acquire the various skills that constitute Theory of Mind. High verbal IQ, a large number of siblings and, most importantly, a great deal of co-operation and affection between the children and good maternal verbal fluency help a child acquire a Theory of Mind at an earlier age. Family discourse allows children to enquire, argue and reflect on why people behave in the ways they do. For instance, parents often sort out disagreements between siblings by explaining how the other child's mistaken beliefs governed his or her actions: 'He thought you had finished yours'; 'She didn't know I had promised it to you'; 'He thought it was his turn'. Girls usually acquire Theory of Mind slightly earlier than boys do.

As I have mentioned, there is some evidence that chimpanzees can understand the link between seeing and knowing and that they can tell the difference between mental states such as knowing and guessing. But can animals pass this crucial test for Theory of Mind? Can they understand the beliefs of others and more importantly, can they understand false beliefs?

Dorothy Cheney and Robert Seyfarth, a husband and wife team from the University of Pennsylvania, spent eleven years looking at the social behaviour of East African vervet monkeys to find out how monkeys see the world and how much they know about what goes on in another monkey's head. They discovered that the monkeys have three different alarm calls which are related to specific predators. They give a barking alarm call in response to leopards and wild cats such as caracals; monkeys hearing this alarm call react by running into the trees. They also have an alarm call for crowned and martial eagles, which elicits a quick scan of the sky and a dash into the nearest bush; and a warning call for snakes, whereupon the monkeys stand on their hind legs and peer into the grass. The monkeys aren't simply acting like this because they have spotted the predator and are frightened, rather they react specifically to the alarm calls a Cheney and Seyfarth showed. They played back the calls of the monkeys to them, hiding loudspeakers in the bushes, and no matter whether they altered the calls so they were louder or softer, or longer or shorter than normal, the monkeys always responded appropriately.

The key question, as Cheney and Seyfarth say, is this: 'What do monkeys mean when they vocalise to each other? Can we actually define "leopard alarm", or "grunt to a dominant" in the same way we define words like anarchist, bordello, or sycophant?' The listening monkeys clearly treat the alarm calls as if they're conveying some kind of information, but do the calling monkeys intend the others to understand what they're saying?

Steven Pinker, author of The Language Instinct, says, 'Human communication is not just a transfer of information like two fax machines connected by a wire; it is a series of alternating displays of behaviour by sensitive, scheming, second-guessing social animals.' Genuine communication where symbols, words or vocalisations have a meaning only occurs when the speaker intends listeners to understand the meaning of the word as the speaker understands it. In Through the Looking-Glass, Humpty Dumpty says, 'When I use a word, it means just what I choose it to mean - neither more nor less.'

'The question is,' said Alice, 'whether you can make words mean different things.'

'The question is,' said Humpty Dumpty, 'which is to be master - that's all.'

In human society, there are conventions about what words mean, but they can mean several different things at the same time, or have different meanings in differing circumstances.

Humpty Dumpty says, '"Impenetrability" That's what I say!'

When Alice asks what he means, he replies, 'I meant that we've had enough of that subject and it would be just as well if you'd mention what you mean to do next, as I suppose you don't intend to stop here for the rest of your life.'

'That's a great deal to make one word mean,' Alice said in a thoughtful tone.

'When I make a word do a lot of work like that,' said Humpty Dumpty, 'I always pay it extra.'

Because autistic people cannot scheme or second-guess they don't understand the intention behind words. When one autistic child was asked, 'What should you do if you cut yourself?', the reply was, 'Bleed'; a standard response to a request to pass the salt would be, 'Yes.'

Daniel Dennett defined several levels of intentionality which I discussed in the introduction, and which we, because we know what we mean and we mean what we say, use unconsciously. In the film, The Lion in Winter, Peter O'Toole plays Henry II, and Katherine Hepburn, his estranged wife, Eleanor of Aquitaine. The two of them are plotting against each other as to which of their three sons should inherit the throne. Henry says of Eleanor, 'She knows I want John on the throne and I know she wants Richard. We're very frank about it.' Which leaves the third son, Jeff, who is equally frank. In a brilliant exposition of levels of intentionality, Jeff says, 'I know. You know I know. I know you know I know. We know Henry knows and Henry knows we know it. We're a very knowledgeable family.' After Jeff has left the scene, Eleanor pithily sums him up, 'He'll sell us all you know. But only if he thinks we think he won't.'

To return to the vervets, what level of intentionality are they capable of understanding? If they are using zero order, then they are reacting because they are frightened and we have already established that that is not the case. If they are using first- order intentionality, the monkeys have beliefs and desires, but they have no beliefs about their beliefs. When a monkey calls, it is because he believes that there is a leopard nearby and he wants the other monkeys to run into the trees. He need have no conception of his audience's state of mind, nor does he need to recognise the distinction between his own and another animal's beliefs. If the vervets were capable of using second- or third-order intentionality, they would have some idea about another monkey's state of mind. A second- order communication would be one in which the caller wants the rest of the group to believe that there is a leopard in the vicinity. Using third-order intentionality, both the signaller’s and the audience's state of mind are considered. A calling monkey would give a warning because he wanted the other monkeys to believe that he wanted them to run into the trees to escape from the leopard.

It is unlikely, though, that the monkeys have such a sophisticated understanding. Baby vervets give eagle alarm calls to pigeons and falling leaves, but the adults never correct them, or 'praise' them when they call properly. As Cheney says, 'They learn by observation - those that don't get eaten.' There is no vervet word for a larger category of meaning, such as 'danger' or 'family'. The adults don't give calls to warn baboons of danger, even when the predator poses no risk to themselves, and when it is highly likely that the baboons' babies will be killed. They are, says Cheney, sublimely egoistic. Both she and her husband believe that 'Monkeys have a kind of laser beam intelligence - extraordinarily powerful when focused in a single domain, but much less well developed outside that narrow sphere. Although they solve social problems with little difficulty or training, they often flounder when confronted with the same problems outside the social domain. They do not always generalise their social abilities to other species or to inanimate objects and in this sense their skills seem relatively restricted. Apparently, the animals do not know what they know and cannot apply their knowledge to problems in other domains.'

For instance, the monkeys don't give alarm calls if they see a python's tracks disappearing into a bush, neither would they give a leopard call if they saw a carcass hanging in a tree – an indication that a leopard is almost certainly nearby. They don't react to alarm calls given by a starling or an impala and they don't act surprised to hear a hippo or a stilt's calls when the monkeys are nowhere near water.

The two researchers went on to study baboons. The females give contact calls to one another when they're travelling in a group through the woods. What they wanted to know was whether the females were deliberately calling to share information or to maintain contact if they were on the periphery of the group? Do baboons intend to communicate? Cheney and Seyfarth recorded contact calls and played them back to the group. They discovered that only females who were related would respond to calls. In other words, if one baboon's calls were played, only her sisters, her aunts and her cousins would reply. This type of 'altruism' has evolved in many animals since those who are related to one another share the same genes. The 'selfish gene' theory states that animals react to safeguard their own genes - whether those genes are in their own body, or if copies are in their relatives' bodies. Cheney and Seyfarth also discovered that female relatives only responded to the playbacks of the contact calls when they themselves were on the periphery of the group, right at the back of the troop or when there were no other females around. They concluded that the calls are used to keep in touch, but there is no mental understanding on the part of either the callers or the listeners. 'Monkeys [including baboons],' say Cheney and Seyfarth, 'see the world in terms of things that act, not things that think and feel.'

If monkeys do not have Theory of Mind, how do chimpanzees fare? Sarah, the original subject of Theory of Mind tasks, was taught to push a button that controlled the lock to a cabinet mounted on a wall just outside her cage. The cabinet was divided into two. One side contained pastries which her favourite trainer would share with her, the other side contained repulsive items such as rotting rubber and faeces. Every time the trainer came to see her, she pushed the button to release the lock to the good side of the cabinet. One day a villain wearing a mask entered her room and, while she watched, removed all the goodies and replaced them with the repulsive stuff from the other side of the cabinet. Sarah responded aggressively, hurling things at him from her cage. A few minutes later, the good trainer entered, yet Sarah showed no change to her normal behaviour. She made no obvious attempt to warn the trainer, and neither did she hesitate before pushing the button. She acted as if she didn't recognise the discrepancy between her knowledge and the trainer's.

In contrast, Kanzi seems to have understood this distinction. Sue Savage-Rumbaugh told him he couldn't have any more M&Ms (sweets similar to Smarties). He then asked if he could go and play in the next room. He knew, although Savage-Rumbaugh did not, that there were M&Ms hidden in one of the cupboards. He ran into the other room, stole the sweets and returned.

At least one of three chimpanzees tested by Andy Whiten appear to understand someone else's state of mind. The chimps' keeper would place food in one of two boxes, padlock them and hang the keys above the boxes. Whiten would then come in and unlock which ever box the chimps pointed to and, if they were correct, give them the food. Whiten was conducting this experiment with the symbol-using chimpanzees, Sherman, Austin and Panzee. He adopted an experimental procedure which went as follows: he asked Panzee what kind of food she would like to eat. He then put her choice of food in one of the boxes and locked both of them. He told her that he was just going out of the room and when he returned, he would then give her the treat. In the meantime, the keeper came in and hid the key. When Whiten came back in, Panzee immediately pointed to where the key had been hidden and not to the box with food in as she had been trained to do. In other words, she seemed to understand that Whiten knew where the food was, but didn't know where the key was. In another version of this test, Whiten opened the boxes and hid the key. The keeper came in, placed food in one box and padlocked them. When Whiten came back, Panzee pointed to the box rather than the food - as if this time she knew that Whiten knew where the key was, but not the food. Austin and Sherman did less well on these tasks - they were able to point appropriately only after several trials. 'You could say that they had simply learnt whether it was correct to point to either the key or the box to get me to give them the food,' says Whiten, 'You could also say that Panzee is an exceptional chimpanzee.'

An experiment which I have conducted on four chimpanzees does seem to show that the brightest chimps can understand a false belief. I built a huge box which was so big that when a person crouched down behind it, a chimpanzee or a child facing the front of the box could not see the person. It had four drawers at the front, and a fan belt above the drawers. I used a large clothes peg, one of the old wooden ones, which I had painted bright red, as a signal to the chimps. The test involved a training phase prior to the experimental phase. In the training phase, each chimp had to learn that if I hooked the peg on to the fan belt above one of the drawers, there would be a sweet inside that drawer (after putting the peg above a drawer, I would go round the back of the box, and place a sweet in the drawer; the box was pushed forward to the wire of the cage, and the chimp reached through to pull the drawer handles and take out the reward). Once all the chimps had learnt this simple association, we progressed to the experimental part. This time, I again put the peg above one of the drawers but someone else moved the peg so that it was above another drawer. During half the trials the peg was moved while I was behind the box and could not see what was going on. The rest of the time I saw where the peg had been moved to. When I didn’t see that the peg had been moved I put the sweet in the drawer below it. I was alternately in a state of Knowledge and Ignorance. In Knowledge, I know where the peg is, but in Ignorance, I do not know that the peg has been moved and thus I do not know where the peg is: I have a false belief about the peg's location, and hence I bait the drawer where I wrongly think the peg is. To get this test right, the chimp has to understand my state of mind. Of course, the chimp could pick the drawer where the peg is and would be correct half the time - but this would obviously show that the chimp had learnt an association between the peg and the drawer. If the chimp was able to choose the correct drawer throughout the course of the experiment, she would have an understanding of my state of mind, knowing both when I knew and when I did not know. One of the chimps was quite clearly not up to the task. He was much older than the others and had had little mental stimulation for years. He was given three times the number of trials in the training phase as the others. The other three chimps were young, between six and eight. Overall, they responded better than four-year-old children to this test, but not as well as five and six year olds. One female did exceptionally well and was able correctly to pick the drawer where the sweet was above chance levels in both the Ignorance and the Knowledge condition.

The children whom I tested with this non-verbal false belief task were able to pass it at the same age as they passed a verbal false belief task, the Smarties test. Autistic people could not do it at all. They consistently chose the drawer beneath the peg, no matter what the state of mind of the experimenter. One young man was able to do the task at first. We asked him how he knew which drawer to look in. The staff who cared for the autistic people had not wanted us to give them sweets, so we hid a bean instead. Mark said he could hear the slight click of the dried bean as we put it into the drawer and that gave him the clue!

A new Theory of Mind task has just been carried out on sign language-using bonobos and shows that bonobos with symbol use, at least, are capable of understanding a person's false belief. Panbanisha watched as a trainer hid some sweets in a box. When the trainer was out of the room, a second person substituted pine needles for the sweets. The trainer returned and attempted to open the box, Savage-Rumbaugh asked the bonobo what she was looking for. Panbanisha replied that the trainer was looking for sweets.

We've seen that normal children from four to five year onwards can understand false beliefs, as can Down’s syndrome people. Most autistic people are unable to pass a false belief task, although a minority are able to give the correct answers. These false belief tasks that we've been using are all second order - the child knows that Sally does not know where her chocolate has been hidden. Josef Perner and Heinz Wimmer have gone a stage beyond this to discover at what age a third-order intentionality false belief task can be solved. They invented the following story, which is told with dolls to act out the events:

This is Mary and this is John. Today they are in the park. Along comes the ice cream van. John wants to buy an ice-cream, but he has left his money at home. He'll have to go home first and get his money before he can buy an ice-cream. The ice-cream man tells John, 'It's all right John, I'll be here in the park all day, so you can go and get your money and come back and buy your ice-cream. I'll still be here.' John runs off home to get his money.

But when John has gone, the ice-cream man changes his mind. He decides he won't stay in the park all day, instead he'll go and sell ice-cream outside the church. He tells Mary, 'I won't stay in the park, like I said. I'm going to the church.'

So in the afternoon, Mary goes home and the ice-cream man sets off for the church. But on his way he meets John. He tells John, 'I changed my mind, I won't be in the park, I'm going to sell ice-cream outside the church this afternoon.' The ice cream man then drives to the church.

In the afternoon, Mary goes over to John's house and knocks on the door. John's mother answer the door and says, 'Oh, I'm sorry Mary, John's gone out. He's gone to buy an ice cream.'

The false belief test is one question: 'Where does Mary think John has gone to buy an ice cream?' The child has to think about one character's false belief about another character's belief. The child who passes the test knows that Mary thinks John doesn't know that the ice cream van is at the church. Normal children are able to answer correctly at around five to seven years old.

About a fifth of the autistic children that Simon Baron-Cohen tested could give the correct answer to the Sally-Ann task. Although these children had a verbal age of between seven and seventeen, none of them could understand Perner’s and Wimmer’s more complex belief scenario. Some of the people he tested had Asperger's syndrome and they were able to understand what Mary thought about John; this is something we'll discuss in chapter eight.

People with autism and Asperger's syndrome are not the only ones to find Theory of Mind tasks difficult. Chris Frith, from the Wellcome Department of Cognitive Neurology, London, has shown that schizophrenics are also unable to understand false beliefs. In fact, the term autistic was initially invented in 1908 to describe people with schizophrenia. Although there are similarities between schizophrenics and autistics the two disorders are quite different. Frith suggests that the inability of schizophrenics to represent the thoughts, beliefs and intentions of other people underlies their delusions of persecution. 'Some schizophrenic patients have a symptom in which they hear voices talking about them in the third person, for example, saying "He is stupid,"' says Frith. 'I think that this might arise as follows: the patient believes that other people think that he, the patient, is stupid. This belief is represented (in his brain) and is normally labelled as a belief. The labelling goes wrong so the information is no longer experienced as a belief, but a perception. Rather than believing that other people think he is stupid, he hears other people saying that he is stupid.'

‘Spider’ is schizophrenic. He believes that his father has killed his mother and a whore has moved in to their house. Spider’s doctor describes him as 'hallucinating floridly in the visual, auditory and olfactory spheres,' and says that he has 'ideas of persecution and thought injection.' Discharged from a secure unit, he now believes his caretaker is his 'father's tart'. He writes in his diary, 'Despite further layers of brown paper taped to my torso, despite the layers of vests and shirts and jerseys on top, the smell of gas was with me until dawn [. . .] A new strategy from the creatures in the attic: I kept my light on all night, of course, and the bulb crackled at me as it usually did, and I paid it no attention - until, that is, the crackling grew suddenly louder and the voices were producing a sort of chant that came out of the bulb, and the chant went: KILL her kill her kill her kill her KILL her...' This is from Patrick McGrath's chilling but intriguing novel, Spider.

As Frith says, 'Paranoid patients make incorrect inferences about the mental states of others - believing that certain people are against you is an example of such an inference.' He has shown that people with schizophrenia have problems understanding the meaning behind words - just as autistic people do. For example, if you say, 'It's very cold in here,' whilst looking pointedly at the window, they won't understand that you actually mean, 'Is it all right if we close the window?' Frith and Rhiannon Corcoran, from the Psychology Department at University College London, showed that both second- and third-order intentionality Theory of Mind tasks are especially difficult for schizophrenic patients with paranoia.

In this chapter we have seen that by the time a child is seven years old, her ability to understand the mental states of others is almost fully formed. She has a very sophisticated understanding of other people’s desires and beliefs, a complex grasp of states of knowledge and how knowledge can be acquired, she can make fine distinctions between intentional and unintentional acts and can tell whether the consequences of a person's act could be foreseen or not. A child of seven, like an adult, is capable of passing a third-order false belief task; she, like many adults, will have trouble solving problems that require higher levels of intentionality. But her grasp of an other person's mental states, her knowledge of what motivates a person and her ability to imagine what it is like to be them means that she is capable of showing an advanced degree of empathic understanding. In the next chapter we will explore the concept of empathy, compassion and the role of the imagination.