Perner, J., Leekam, S.R., Myers, D., Davis, S. and Odgers, N. Misrepresentation and Referential Confusion: Children's Difficulty with False Beliefs and Outdated Photographs Unpublished manuscript

Misrepresentation and Referential Confusion: Children's Difficulty with False Beliefs and Outdated Photographs

Josef Perner
Susan R. Leekam1
Deborah Myers
Shalini Davis
Nicola Odgers

Laboratory of Experimental Psychology
University of Sussex
East Sussex


1Sue Leekam is now at the Institute of Social and Applied Psychology, University of Kent at Canterbury.


Three and 4-year-old children were tested on matched versions of Zaitchik's (1990) photo task and Wimmer and Perner's (1983) false belief task. Although replicating Zaitchik's finding that false belief and photo task are of equal difficulty, this applied only to mean performance across subjects and no substantial correlation between the two tasks was found. This suggests that the two tasks tap different intellectual abilities. It was further discovered that children's performance can be improved by drawing their attention to the back of the photo but not by drawing attention to the person holding the false belief. Results are interpreted as showing that children's difficulty with the photo task is due to referential confusion about which scene the question refers to (the picture or reality) while the hurdle in the false belief task is to understand that the believer misrepresents reality.

Zaitchik (1990) presented evidence for an important change around the age of 4 years in children's understanding of how photographs relate to what they depict. In her photo (or camera) task a picture is taken of a state of affairs (person 1 lying on the mat). Then, while the photograph develops, a change occurs in the world (person 2 takes the original person's place on the mat). Children are then asked to predict who will be lying on the mat in the picture. Most 4-year olds correctly answered that person 1 will appear in the photograph. In contrast, most 3-year olds answered wrongly that person 2 (the one currently lying on the mat) will be in the photo.

Zaitchik also found that this change concurs with a change in children's understanding of false belief as measured by a version of the false belief test developed by Wimmer and Perner (1983). Because of this concurrence between understanding how photos relate to reality and how beliefs relate to reality Zaitchik's finding was taken as important evidence in favour of the claim that the observed changes on 'theory of mind' tasks around 4 years reflect the emergence of (Forguson & Gopnik, 1988; Perner, 1988) or important changes in (Flavell, 1988; Wellman, 1990) children's understanding of representations (ability to metarepresent). Zaitchik's finding boosted this theoretical position in favour of other views, for instance, the position that the observed changes are but manifestations of the gradually improving performance (e.g., due to growing information processing capacities) of deeper competencies that developed earlier at the end of infancy (Chandler, 1988; Leslie, 1993), or the position that children at this age figure out the causation of mental states (Leslie, 1988; Wimmer, Hogrefe & Sodian, 1988).

The objective of our first experiment is twofold. One is to replicate Zaitchik's basic finding that 3- and 4-year old children have as much difficulty with the photo task as with the false belief task. Part of this objective is to check that the two tasks are not just of comparable difficulty when averaged across subjects but that it is the same children who have dificulty with each task.

The second objective is to establish Zaitchik's interpretation that young children's difficulty with the task is due to the fact that photos are representations. Alternative explanations cannot be ruled out. For instance, one alternative is that the parallel performance on the two tasks is due to similar information processing demands. This is particularly plausible, since Zaitchik took great care to make the structure of the photo task resemble that of the false belief task. Another alternative is that performance on both tasks reflects the growing understanding of the causation of representational states and not the understanding of representation as such. For, it may well be that children's understanding of how, e.g., looking leads to belief develops at the same time as their understanding how the pointing of a camera at a scene leads to a photo of that scene.

To establish Zaitchik's claim that photo and false belief task are difficult because they require understanding of representation, we designed a new Colour Transmission Task that used the same causal mechanism as the photo task and was of the same complexity as photo and belief task, but fell short of requiring an understanding of representation. This was achieved by describing the camera as a machine that turned a piece of paper inside it into the same colour as the screen it is pointing at. Unlike the photo of a doll on a mat produced in the photo task, the resulting piece of glossy, e.g., blueish paper produced in the colour transmission task is not, in any obviously recognizable way, a representation of some external state of affairs. When simply handed the piece of paper and asked what it is, one could only say that it is a glossy piece of blue paper. One would not spontaneously identify it as a picture of something blue. Hence, if children perform substantially better on this new task than on the photo task, then this will support Zaitchik's claim at least for the photo task.

In order to match our new task as closely as possible to the photo task we had to change Zaitchik's original version of it. Instead of a person lying on a mat we pointed the camera at a doll named Judy who was dressed all in one colour, say blue, when the photograph was being taken. While the photograph developed, Judy changed into a different colour dress. After this change children were asked what colour the doll will be in the picture.



Subjects. Eighty children from a nursery school in Brighton, Sussex took part. Of these, 4 missed the second testing session and 9 did not pass the preliminary colour naming test, leaving 32 3-year-olds (20 boys, 12 girls ranging from 3 years; 4 months to 3;11, median age = 3;8) and 35 4- year-olds (20 boys, 16 girls ranging from 4;1 to 4;9, median age =4;6). For one 4-year old boy the data for false belief were incomplete.

Materials. The photograph task and the false belief task both involved a 19 cm high wooden jointed doll and a set of different coloured dresses. For the Photo task these were red, green, yellow, and blue. For the False Belief task these were black and white. For the colour transmission task a 81cm x 53cm large, self-supporting cardboard screen was used which could be covered by sheets of cloth in the same colours as used for the photo task.

Photographs were taken with a Polaroid camera.

Design. A within subjects design was used for the main experimental manipulation. The three tasks falsebelief (FB), photo (P) and colour transmission (CT) were administered in the following 4 different sequences: FB-P--CT, FB-CT--P, P-CT--FB, and CT-P--FB. Each child was seen on two occasions. The child was given 2 tasks in the first session and the third task in the second session about a week apart. For the false-belief task only colours black and white were used. The sequence in which these colours were used (i.e., whether the change of dress was from white to black or vice versa) was counterbalanced within age groups. The use of red-green or blue-yellow and the sequence of use within each colour pair were counterbalanced, as far as possible, for age and photo/colour-transmission tasks.

Two different wordings of the test question were used (Form of Question) which was varied between subjects, counterbalanced with age and order of tasks. The form of question modelled after Zaitchik's procedure was: "In the picture, what colour is Judy?"(Form A) and the new version was: "What colour is Judy in the picture?" (Form B). Similarly for the colour transmission task we asked "In my hand, what colour is the paper?" (Form A) and "What colour is the paper in my hand?" (Form B).

Procedure. Each subject was first given a colour labelling task in which the experimenter pointed to each of the coloured dresses and asked subjects "What colour is this?" This was a qualifying condition for acceptance on the experimental tasks. Testing proper started with one of the three experimental tasks. The experimenter followed a routine script as shown in Table 1. Subjects responses were recorded on the script and the test was also recorded on videotape.

Table 1. Scripts for False-Belief, Photograph, and Colour-Transmission Tasks in Experiment 1.

False Belief

Colour Transmission
This is Judy. She has 2 dresses. They are different colours. Let's put this BLACK dress on her.
Here is Ann. She has 2 dresses. They are different colours. Let's put this YELLOW dress on her.
Here are different colours. Let;s put the RED one over here [point to screen].
Now here comes Susan. She wants to see what colour Judy is wearing. So she has a look.
This is a machine with a piece of paper in it. If you press this button it makes a "click". When it goes click it makes ...

... a picture of Judy wearing this colour.
... the paper inside the same colour as this.
PROMPT: What colour does Susan see?
PROMPT: What colour is Judy?
PROMPT: What colour is this?
Susan says to Judy: "Ah. I see you are in BLACK today."


Now Susan goes to fetch the belt for Judy's dress. Now we have to wait until her friend comes back.
Now we have to wait. Can you remember what happened when we worked the machine? You pressed the button and it made a click. So what's going to happen? Yes, it's going to make ...

... a picture of Judy wearing this colour.
... the paper inside the same colour as this.
While we are waiting let's change her dress from BLACK to WHITE.
While we are waiting let's change her dress from YELLOW to BLUE.
While we are waiting let's change this colour from RED to GREEN.
REALITY QUESTION: What colour is Judy now?
REALITY QUESTION: What colour is Judy now?
REALITY QUESTION: What colour is here now? [point]
MEMORY QUESTION: What colour was Judy when Susan came round to see her?
MEMORY QUESTION: What colour was Judy when you pressed the button?
MEMORY QUESTION: What was the colour when you pressed the button?
TEST QUESTION: What colour does Susan think Judy is?
TEST QUESTION: What colour is Judy in the picture?
TEST QUESTION: What colour is the paper in my hand?

All three tasks ended with a series of questions. First a Reality question asking what the colour is now. Next a Memory question was asked about the original colour before the change took place. Finally, this question was followed by the critical Test Question. This question was asked in two different forms, Form A and Form B. The standard forms of False Belief Question was used "What colour does Susan think Judy is?".


Inspection of the data showed no discernible effects for wording of test questions, the order in which tasks were administered or which colours were used. Answers to test questions consisted of either naming the correct or incorrect colour. Some answers (9 in photo task and 7 in colour- transmission task) were "don't know" answers. They were scored as incorrect. The percentages of children giving correct answers are shown in Table 2 for the 3 tasks independently.

Table 2. Percentage Children Giving Correct Answers to Test Questions in Experiment 1.

Study and Age

Present Study (n=66)
Zaitchik (1990)
Experiments 1+5 (n=70)

False Belief
Colour Transmission

The first two rows in this table also show results by Zaitchik (1990) averaged over her Experiments 1 and 5 which both included 3- and 4-year-old children. Our results closely replicate Zaitchik's findings in several respects.

There is clear improvement on the false belief task between 3 and 4 years (c2 (1, N = 66) = 11.5, p < .001) and our percentages for each age group mirror those reported by Zaitchik (Table 2) and of most other studies of this kind, for instance, Wimmer and Perner (1983): 15% and 76%, Hogrefe, Wimmer and Perner (1986): 17% and 56%, Perner, Leekam and Wimmer (1987): 40% and 87%, Zaitchik (1990, Experiments 1 and 5 combined): 40% and 84%.

Also our percentages of correct answers in the Photo task closely match those reported by Zaitchik (Table 2). Moreover, we found that performance on the false belief task tends to be better than on the photo task as Zaitchik had noted. However, when we look at the contingency between the two tasks (which Zaitchik did not report for her data) in Table 3 we find a surprisingly weak and statistically nonsignificant correlation (PHI = .15; c2(1, N = 66) = 1.53, p > .20). Although unexpected, this lack of a correlation is very important. As subsequent experiments will demonstrate, it is the first sign that performance on the false belief and the photo task are dissociable and not based on a common difficulty in understanding representation.

Table 3. Correct-Response Contingency Between False-Belief and Photo Task in Experiment 1.


Number Correct
3 years
4 years
All Ages
False Belief Only
Photo Only

The bottom two rows of Table 2 show that performance on the colour transmission task was better than performance on the photo task (McNemar's c2 (1, N = 24) = 8.17, p < .01). The difference is small, but provides prima facie evidence for Zaitchik's claim that the difficulty with the photo task is that it requires understanding of representation (which the colour transmission task does not). However, Zaitchik also thought that the same difficulty underlies the false belief task. In view of the minuscule correlation between belief and photo task our support for Zaitchik's position remains tentative.

Control Questions. All children gave correct answers to the Reality Question about what colour dress the doll was wearing now. Responses to the Memory Question about the colour of the dress at the time the camera was used were less reliable. Several children (9 3-year and 7 4-year olds) mentioned a colour not used on this particular trial. Some of these children changed their answers spontaneously. Others, when prompted by the experimenter to consider only the two colours used on this trial, gave an appropriate and correct answer. We let these children pass as having answered the question correctly. However, 11 children (10 3-year olds and 1 4-year old) had a more serious problem remembering the doll's original colour in at least one of the 3 tasks. Yet, even if we exclude all of these 11 children from our analysis the pattern of results remains essentially the same as shown in Table 2. In particular, the main result of a higher frequency of correct responses on the colour transmission than on the photo task was still statistically reliable: McNemar's c2(1, N = 19) = 6.36, p < .02.


The somewhat better performance on the colour transmission task than on the photo task provides some of the expected support for Zaitchik's suggestion that children find her task difficult because it involves an understanding of representation. However, the low correlation between belief and photo task makes it difficult to accept her wider claim that the same understanding underlies performance on the false belief task.

Moreover, performance on the colour transmission task was not as impressive as it should have been if understanding of representation were the problem for children on the photo task. The disappointing performance has two plausible explanations. One is that the camera and the process of photographic development is such an unfamiliar and obscure process that even the older children remain uncertain when a photograph's content may or may not change. Another explanation is that despite our efforts to describe the colour transmission in nonrepresentational language, because of the use of a camera, many children still conceived of the piece of paper as a representation of the coloured screen, thus defying our experimental intentions.

The next experiment overcomes these shortcomings of the colour transmission task by switching from photographs to drawings and to a new copy control condition. It also explores a fortuitous discovery in pilot work that may explain the lack of correlation between photo and belief task, a point more fully investigated in the last experiment.


The original aim of this experiment was to introduce a better non- representational control condition than the colour transmission task of Experiment 1. This was made possible by switching from photographs to drawings. The general scenario was one with a doll named Curly who used a stencil to print, say, a bicycle on his washable tee shirt. In the picture condition, a drawing was made of Curly including the print on his tee shirt. The drawing was then turned face down (analogous to waiting until the photo develops) and Curly decided to change his print to a different shape. Children were asked what shape was on Curly's tee shirt in the picture.

The new copy condition followed the same steps except that instead of the shape on Curly's tee shirt being duplicated onto a drawing of Curly it is put on the tee shirt of Curly's friend Blondy. Unlike the colour transmission condition of Experiment 1, the copy condition does not involve a representational relationship and the contrast to the picture condition is clearer. In the picture condition the shape on the tee shirt in the drawing of Curly represents the shape on Curly's tee shirt, whereas the shape on Blondy's tee shirt does not represent the shape on Curly's tee shirt. It is merely a copy of what is on Curly's shirt due to the fact that it was produced by the same stencil.

In addition to the contrast between picture and copy conditions, another manipulation was introduced. In pilot work it was observed that children, when asked about the picture, gave only a cursory glance at the back of the picture and then looked at the experimenter or the real scene and proceeded to describe how things were in the real scene. When children were helped to keep their attention on the picture, by making them look at its back, then almost all their answers were correct. The effectiveness of this helpful method was explored systematically in this experiment. On one trial of each task (picture and copy task) children's attention was left undirected, while on another trial their attention was directed to the picture.


Subjects. Seventeen children (6 girls and 11 boys) from "Brightstart" nursery in Brighton participated in this study. Due to this nursery's highly selective admissions policy children were very bright and socially precocious. Their age ranged from 2 years and 2 months to three years and 8 months (median age 3 years and 1 month).

Materials. Two 35 cm tall cardboard cut-outs were used as dolls. One was a boy with curly black hair called "Curly" and the other had blond hair called "Blondy". They wore tee shirts made of acetate-type material so that poster paints could be applied and then wiped off with a damp cloth. The tee shirts could thus be given ornamental prints with sponge stencils which came in two different shapes: a bicycle and a butterfly. Two colours, green and blue were used.

Procedure. Subjects sat next to the experimenter and helped with printing of tee shirts. In the picture condition Curly was placed in front of the child and the child was told that he wanted a shape put onto his plain tee shirt. One of the stencils was chosen at random and dipped into one of the two colours (green or blue) and then pressed onto Curly's tee shirt. The child was then asked: "What shape is on Curly's tee shirt?". All children could identify the shapes. They were then told that Curly was very proud of this new print and wanted a picture made of him wearing it. So a large picture was drawn with the shape on Curly's tee shirt stencilled in. The picture was then turned face down and put sideways behind the experimenter.

It was then explained that Curly was bored with his tee shirt and would like a new shape put on it. Curly's tee shirt was washed clean with a damp cloth and the other stencil was used to put a new print on Curly's tee shirt. The child was then asked the following questions:

1. Reality question: "What's Curly got on his tee shirt now?
2. Memory question: "What did we wash off Curly's tee shirt?
3. Test question:"What's Curly got on his tee shirt in the picture?"

At the end each child was shown the picture.

In the copy condition the same scenario unfolded except that, instead of a picture of Curly being made, his friend Blondy wanted the same shape as Curly on his tee shirt. After Blondy had received the desired print he was treated like the picture of Curly in the other condition: placed face down behind the experimenter. After Curly's changing the shape on his tee shirt, children were asked the same 3 questions as in the picture condition with a necessary change in one question:

Test Question: "What's Blondy got on his tee shirt?"

In the attention-directed condition the experimenter pointed to the back of the picture (or the back of Blondy) and made sure the child looked at the picture (or Blondy) while answering the test question. In the attention-undirected condition the experimenter did not point and was not concerned about where the child happened to look.

Design. Each subject was tested on all four conditions (picture with undirected gaze, picture with directed gaze, copy with undirected and copy with directed gaze) over two sessions between 3 days to 1 week apart. About half the subjects were first tested with eye gaze directed, the other half with eye gaze undirected. Within each session the order of picture and copy tasks were counterbalanced as far as possible.


All children gave correct answers to all Reality questions. The success rate on the memory question was 86.8%. The few errors were committed in the following conditions: 4 errors in the copy condition with gaze directed, 2 in the copy condition with gaze undirected, 3 in the picture condition with gaze directed and no error in the picture condition with gaze undirected. Except for one case these errors occurred when correct answers to the subsequent test question were given. It appears that these errors on the memory question were haphazard gaps in attention rather than an indication that the sequence of events was not attended to.

As Table 4 shows, practically all children gave correct answers to the test question in all conditions except the picture condition with undirected gaze.[1] Hence, on the one hand these results confirm the finding from Experiment 1 that the nonrepresentational colour transmission is easier than the representational photo task, because the representational picture condition is more difficult than the nonrepresentational copy condition (Binomial Test: N = 7, x = 0, p = .016 two tailed).

On the other hand, the results also show that by directing children's gaze to the back of the picture their difficulty is significantly reduced (Binomial Test: N = 9, x = 1, p = .04 two tailed).

Table 4. Correct Responses to Test Question in Experiment 2.


Picture undirected
Picture directed
Copy undirected
Copy directed


The results from this experiment establish two points. The first is that data from the undirected attention conditions show the nonrepresentational copy condition as being substantially easier than the representational picture condition. This confirms and improves on the finding in Experiment 1 where the nonrepresentational colour transmission task was somewhat easier than the representational photo task. It provides further evidence that the difficulty with the photo task resides in the fact that it involves representation.

The second point established is that the difficulty with the picture task can be largely alleviated by the simple technique of directing children's attention to the back of the picture. As a next step we need to confirm that the same technique works not only for pictures but also for Zaitchik's original photos.

Nevertheless, the finding opens the interesting prospect that we have discovered a means of not only making the picture and perhaps photo task much easier, but also the false belief task. This might be successfully done by directing the child's attention to the believer when the belief question is asked. The next experiment, therefore, investigates whether the beneficial effects of directing the child's attention generalise to Zaitchik's original photo task and to the false belief task.


The main objective of this experiment is to investigate whether the technique of directing children's attention to the back of pictures will also improve their responses in Zaitchik's original photo task and in the false belief task. In both tasks children are asked about an object (mat) or person (Judy) which appears in two different situations. In Zaitchik's photo task the critical question pertains to the mat and who is sitting on it. The answer depends on which situation the question asker has in mind. Is it the situation in the photo where person 1 sits on the mat, or the real situation where person 2 is sitting on it? That the experimenter is talking about the pictorial situation is made clear by the context marker: "in the picture." For young children this may not be strong enough to keep their attention from straying to the real situation. Young children, one could say, suffer from referential confusion, as to which situation the question refers. Making their eyes focus on the back of the picture, however, helps make clear that the question is about the situation in the picture and not reality.

Could referential confusion also apply to the belief task? Thsi is possible given that the child has to realize that the question being asked refers to the believed situation and not to the real situation. Making children focus on the believer in the belief task should, therefore, have a similar effect to making them focus on the photo in the photo task.

In contrast, such referential confusion does not arise in the colour transmission task of Experiment 1 and in the copy condition of Experiment 2. For instance when asked, "What's Blondy got on his tee shirt?" there is no referential confusion possible since Blondy only ever had one shape on his shirt. So we see that referential confusion can not only explain the effectiveness of attention directing in Experiment 2, but also why the nonrepresentational conditions (colour transmission, copy) were easier than the representational ones (photo, picture).

Referential confusion might also play a role for the memory questions. For instance, when asked in Experiment 1, "What colour was Judy when you pressed the button?" the child has to decide whether the question refers to Judy and the colour of her dress in the past situation or in the present situation. The decision has to be made on the basis of the context marker "when you pressed the button", helped by the use of the past tense form "was". In contrast, the question form used in Experiment 2, "What did we wash off Curly's tee shirt?" does not raise the same problem since only one shape was ever washed off Curly's shirt. Results from the two experiments give some weak support to this contention since in Experiment 1 children had some problem with the memory question 18% of the time while only 12% in Experiment 2. These comparisons are, however, very tenuous since the sample of Experiment 2 was a very unusual one. A proper comparison between these two question forms is incorporated in the present experiment.

To summarize, if referential confusion is the main difficulty in the photo and belief tasks we might expect parallel results on these two tasks. The low correlation between belief and photo task in Experiment 1 casts some doubt on this prediction. This experiment enables us to determine whether performance on the belief task and on the photo task do indeed correlate just minimally, or whether the result in Experiment 1 was an unfortunate statistical accident.

There are also theoretical grounds on which different performance in the photo and belief task might be expected. The belief task requires understanding of misrepresentation which the photo task does not. The belief is a false one because it is about the present situation but misrepresents it. In contrast the photograph is not false but only outdated. If understanding misrepresentation, in addition to any referential confusion problems, is what makes the false belief task difficult for young children, then directing children's attention to the believer will not alleviate that difficulty.


Subjects. Thirty-six children (21 girls and 15 boys) from five different nursery schools participated. Their ages ranged from 3 years 2 months to 4 years 10 months with a mean age of 4 years.

Materials. Two different scenarios were used: A doll scenar modelled after Experiment 2, and a mat scenario modelled after Zaitchik's original material.

In the doll scenario two cardboard cut-out dolls, one boy and one girl doll, about 30 cm tall were used. Cardboard squares with either a house, duck, flower, or a fish on them could be attached to the boy's tee shirt indicating that he was wearing a different print on his shirt. The girl doll was only used in the false-belief task as the observer of the boy's appearance.

In the mat scenario two soft toys, a rabbit and a bear, and a piece of paper symbolizing a mat were used. in both scenarios a 40 cm high wall was used behind which the observer in the belief tasks disappeared. For the photo conditions the Polaroid camera of Experiment 1 was used.

Design. Each child was tested in 4 conditions: directed attention in photograph task (dPH), undirected photograph (uPH), directed false belief (dFB), and undirected false belief (uFB). These conditions were administered in 4 different orders (dPH-dFB-uPH-uFB, dFB-dPH-uFB-uPH, uPH- uFB-dPH-dFB, and uFB-uPH-dFB-dPH) which were allocated at random. The first and third task always used the doll, and the second and fourth test the mat scenario.

Half the subjects, randomly allocated, were asked Type A questions, the other half Type B questions. Type A questions take the typical form of the belief question, "What does Mary think is on Peter's tee shirt?" and equate the form of the photo question to this by the use of the word 'show': "What does the photo show is on Peter's tee shirt?" Type B questions take the usual form of the photo question as a starting point, "In the photo, what is on Peter's tee shirt?" and give the belief question the same form: "In Mary's head, what is on Peter's tee shirt?". This question also assesses whether children think of beliefs as something like pictures in the head.[2]

Procedure. Each child was taken into a quiet corner of the nurs room and was seated at a small table opposite the experimenter. The cardboard wall was erected on the table between child and experimenter dividing the table into a left and right half from the child's point of view.

In the false belief task the child was introduced to Peter, the male cardboard doll, and was told that Peter wanted to put on a tee shirt. A picture-square with either duck or house on it was then put on Peter's shirt. The child was asked to confirm what picture was on Peter's shirt to establish that the child could identify the picture correctly. All children could do this. Mary, the girl doll was then introduced and it was explained that she wanted to see what was on Peter's tee shirt. The child was asked to confirm what Mary could see on Peter's shirt. Mary was then put on the other side of the cardboard wall and it was explained that she could not see Peter. The child was asked to confirm this fact. If necessary the reason for this fact was explained to the child. Then Peter, tired of his shirt, changed it, which was symbolized by removing the picture square on his chest with the other one.

The child was then asked the reality question: "What picture is on Peter's tee shirt?" and the child was reminded of the fact that Mary had not been able to see Peter change his shirt. Finally the child was asked the following questions (half the children were given type A the other half type B):

Belief Question:

Type A:"What does Mary think is on Peter's shirt?"
Type B: "In Mary's head, what is on Peter' shirt?"

Memory Question:[3]

Type A:"Remember, we removed a picture from Peter's shirt, what was that?"
Type B:"Remember when Mary came to look, what was on Peter's shirt then?"

At the beginning of the first photo task each child was familiarized with the camera. Every child took one practice photo of a doll. It was explained to them that they must look through the "peephole" and when they pressed the button a picture would be made of everything they could see through the viewer.

For the task proper, Peter was introduced and it was explained that he was deciding which tee shirt to put on. A picture-square of either a fish or a flower was put onto the front of his shirt. The child was asked what picture was on his tee shirt and told that Peter was so proud of his shirt that he wanted someone to take a photo of it. The child then took a photo of Peter. When the photo emerged from the camera it was placed face down on the other side of the cardboard wall. It was then explained that Peter was bored with his shirt and wanted to change. The picture-square was removed and replaced by another one. The child was then asked the same reality question as in the belief condition and the following test questions:

Photo Question:

Type A:"What does the photo show is on Peter's shirt?"
Type B: "In the photo, what is on Peter' shirt?"

Memory Question:

Type A:"Remember, we removed a picture from Peter's shirt, what was that?"
Type B:"Remember when the camera went click and we made a photo, what was on Peter's shirt then?"

In the conditions where the child's attention was directed the child was told to look at Mary (belief task) or the back of the photo (photo task) when the Belief/ Photo question was asked. If the child's gaze strayed (typically to the experimenter or over to Peter) the child's attention was made to refocus on Mary/ the back of the photo. In the condition where the attention was undirected the child was free to look anywhere when the critical question was asked. On the two occasions when the "mat" scenario with the two soft toys was used the belief and photo task closely followed the format for the "doll" scenarios just described.


All children gave correct answers to the reality question in all 4 stories.

Belief and Photo Questions. Inspection of the data showed that there was no systematic effect of children's sex or story order on the mean correct responses to the Belief and Photo questions.

Responses to the Belief question were influenced by the wording of the question as Table 5 shows. Making the belief question similar to the Photo question made it much more difficult: c2(1, N = 48) = 8.00, p < .01 (collapsing numbers of children with one and two correct answers). No similar difference was found for the photo task: c2(2, N = 48) = 1.41, p > .40.

Table 5. The Effect of Question Type and Direction of Attention on Percent Subjects Answering False-Belief and Photo Question Correctly in Experiment 3.

Question Type

"in photo/head?"

Table 5 also shows that directing the child's attention to the back of the photo when answering the photo question produced a sizable improvement of over 30% due to the fact that 13 children answered correctly when their attention was directed and wrongly when it was left undirected, while only 2 children showed the opposite pattern: MCNemar's c2(1, N = 15) = 8.06 < .01. In contrast, directing children's attention to the story protagonist's friend had little effect on how children answered the question about that friend's belief: Binomial Test, N = 2, x = 0, p > .25.

To test whether the effect for directing attention was reliably stronger in the photo than in the belief condition we scored children's responses in the following way. If they gave a correct response when their attention was directed but an incorrect one when left undirected they were assigned +1, -1 if they showed the opposite response pattern and 0 if there was no difference between attention conditions. The score computed in this way for the two belief tasks was then deducted from the score for the two photo tasks. 12 children obtained an overall positive score (indicating that directing of attention was more effective in the photo than on the belief tasks), whereas only 3 children obtained a negative score. This is significant on a Sign test: c2(1, N = 15) = 5.14, p < .05. To avoid criticism that this interaction between question type and belief vs. photo tasks was unduly helped by the fact that performance on the type B belief question (In Mary's head,...?") was near floor, one should point out that the same computation yielded a significant result for the reduced sample of 18 children tested on type A questions: c2(1, N = 10) = 5.33, p < .05.

To enable a fair comparison between performance in both Experiment 1 and Zaitchik's study and performance in the present study, we need to look at those conditions in this experiment that are most similar to these other experiments. These are the undirected attention conditions with the type A ("think") Belief question and the type B ("In the picture") Photo question. For the Belief question we get 50% and for the Photo question 44.4% correct responses, which is fairly close to our data in Experiment 1 and Zaitchik's data as summarized in Table 2.

In Table 2 we also looked at the correlation between false belief and photo task and found a surprisingly low correlation of Phi = .15. We can do the same for this Experiment if we use the data from the type A ("show") Photo question, which yielded very much the same level of performance as the type A Photo question. On these measures the correlation between false belief task and photo task was again very low (PHI = .12) and non- significantly different from zero (c2(1, N = 18) = .28, p > .50).

Memory Questions. Inspection of the data showed that children's and story order had no systematic effect on answers to these questions. Also, there was no systematic difference between the four stories. This was to be expected, since the temporal context and memory problems that these questions may involve were the same for the photo and the belief tasks. We thus analyzed the number of correct answers to the four stories by analysis of variance with age (younger half versus older half) and question type as independent variables. Both main effects were significant but not their interaction.

Factor age (F(1, 32) = 8.86, p < .01) reflects an improvement from 52.8% correct responses among 3-year olds to 77.8% correct among 4-year olds. More importantly, question type (F(1, 32) = 4.65, p < .05) reflected the following difference: If the question made a unique reference to that picture which was removed from Peter's tee shirt ("Remember, we removed a picture from Peter's tee shirt,...?") then children gave 73.6% correct answers. If, however, reference was made to a temporal context (e.g., Remember when the camera went click,...?") then children gave only 56.9% correct answers.


Three main findings emerged from the three experiments reported. Firstly, we replicated Zaitchik's finding that young children find it difficult to predict what a photo shows when the photographed scene has changed since the picture had been taken and that this task is about as difficult as the false belief task. Furthermore, we were also able to show that the difficulty has something to do with the fact that photos (or drawings) are representations, because closely matched nonrepresentational tasks (colour transmission, copy task) were substantially easier. These results were originally expected on grounds of Zaitchik's conjecture that children's difficulty with belief and photo task resides in children's failure to understand representations. However, our eventual interpretation of these results is different due to other, less expected findings that emerged during the course of the investigation.

Our second main finding was the discovery that by directing attention to the back of the photo (drawing) children's performance improved markedly. That this simple manipulation had such a strong effect can explain discrepant results between studies. For instance, Charman and Baron-Cohen (in press) working with drawings reported that only 40% of 3- to 4-year old children gave at least 2 correct answers on 3 problems, while Lillard and Cho (1991) found that 84% of the same age group met this criterion. This discrepancy can plausibly be explained by the experimenter being unaware of the effectiveness of attention directing cues. These cues, therefore, could exert their important influence in an uncontrolled manner to a different degree in each study.

The ease with which performance on the photo task can be improved also speaks to "theory-of-mind" proposals trying to explain children's performance on Zaitchik's task. For instance, Wellman (1990, pp. 265-266) proposed that young children fail the task because they have a 'copy theory' of representation, i.e., they understand representations like photos or pictures as "direct copies of reality and hence believe the content of the photograph faithfully depicts the real situation." (p. 266). This proposal is difficult to square with our results from Experiment 3, since it would be difficult to explain why children would relinquish their copy theory, simply because they are induced to look at the back of the picture when answering questions about the picture.

Perner (1991, p. 100) pointed out that although young children would not approach Zaitchik's task the same way as adults do, according to his theory they should be able to give the correct answer by being able to establish the sameness between the external situation at the time that the picture was taken and the situation in the photograph. Our results are compatible with this theory. However, to explain children's difficulty on Zaitchik's task Perner (pp. 100-101) also adduced some additional assumptions about children's use of photographs in contrast to their use of verbal statements. His claim is that children have no experience in relating photos to reality and so, caught without any definite expectations about the photo's content, answer by default in terms of reality. If children have no expectations of what to find in the photo, then how could looking at the back of it generate such expectations?

The natural explanation for the effects of directing attention is that it helps overcome children's referential confusion. Such confusion occurs when a question refers to an entity (mat, Peter's tee shirt) that occurs in two different situations and the child has to rely on context markers (e.g., "in the picture", "when the camera went click") to know which situation is meant. Directing the child's attention to the intended situation (e.g., by making them look at the back of the photograph) helps the child answer the question correctly.

Lewis and Freeman (1992) discovered an alternative method to help children in Zaitchik's photo task by making two photographs, one of the original situation (person 1 on the mat) and another of the current situation (person 2 on the mat). The introduction of the second photograph helped children give more correct answers to the question about the first photograph (71%) than when the regular Zaitchik task with only one photograph was used (41% correct). The authors explained their data with a specific version of referential confusion. They argued that children may misinterpret Zaitchik's test question "In the picture, who is lying on the mat?" as "who is (now in reality) lying on the mat (which is shown in the picture)?". By making a contrast between two photographs and asking, "in this picture," it becomes clearer that the question is about how things stand in that particular photo rather than in the other photo or in reality.

Referential confusion can also explain why children did much better on our nonrepresentational tasks (colour transmission, copy task). In these tasks the confusion does not arise because the entity the question refers to occurs in only one situation. For instance, in the copy condition of Experiment 2 the question, "What's on Blondy's tee shirt?", can only refer to the particular shape that Blondy has, and ever had on his shirt (it was Curly who first had one and then another shape on his shirt).

Referential confusion occurs not only with pictorial (and other representational) contexts but also with temporal contexts. This was demonstrated by manipulating the form of the memory question in Experiment 3. When the question (type B) needed disambiguation as to which shape on Peter's shirt was meant by specifying temporal context markers ("when Mary came to look/ when the camera went click") children had relative difficulty (only 56.9% correct). In contrast when the question (type A) referred directly and uniquely to what was removed from Peter's tee shirt, children fared much better (73.6% correct).

This finding highlights the fact that referential confusion is not the same as Zaitchik's explanation for the photo task, namely that children have difficulty understanding how representations like photos relate to what they represent. As the example of temporal contexts shows, referential confusion can also occur with nonrepresentational contexts.

The third important finding from this series of experiments is that directing attention has no comparable effect on the false belief task and that the belief task correlates only minimally with the photo task. The fact that directing attention has no effect on false belief performance parallels the findings by Lewis and Freeman (1992). Although the contrast between the outdated photograph and a more updated version improved performance from 41% to 71% percent correct answers, no such effect was found for false beliefs. When a contrast was created between one person holding a false belief and another person believing the truth, performance of 3-year-olds remained at a low 31% correct as in the traditional task.

The ineffectiveness of directing attention in the belief task suggests, as Leekam and Perner (1991) emphasised, that children's real problem with false beliefs is the beliefs' falseness, that is, the fact that the belief misrepresents the current real situation. Perner (1991) pointed out that photos can be treated as situations which are independent of the real situation in a way that beliefs cannot. A belief (in particular a false one) must be understood as making a claim about (being about) the real situation or else it would not be a belief but just a thought (like a picture or image) in the head. Children's failure on the type B belief question in Experiment 3 ("In Mary's head, what is on Peter's shirt?") shows that children do not conceive of beliefs like pictures in the head. Hence, if children fail to understand misrepresentation then making them focus on the believer's head cannot help them as the data show.

An alternative, but in our view less convincing, explanation for the ineffectiveness of directing attention in the belief task is that beliefs, unlike photos, are intangible and abstract entities which are not seen as located in the believer. Hence directing attention to the believer has no effect.

Recent research by Parkin (1993) on misleading direction signs helps distinguish between the two alternative explanations. He established that when children do not know where an object is they can use a direction sign to locate it and correctly answer the question "Where does this sign show the object is?". However when the sign points to location 1 and children know the object is in location 2 then the younger children answer this question wrongly with "location 2". Performance on this task correlates strongly with children's ability to answer a question about a person's false belief, "Where does this person think the object is?". This correlation suggests that it is not the intangibility of beliefs that makes them different from Zaitchik's photos but their falseness, since they only share the falseness with the equally difficult misleading signs but not their tangibility.

In sum, there is converging evidence from several studies that Zaitchik's photo task and the false belief task do not tap the same underlying intellectual development. The false belief task assesses children's understanding of misrepresentation (understanding that something is about a situation external to itself and misrepresents that situation) while the problem in the photo task consists of referential confusion created through multiple contexts involving the same entities. This dissociation of photo from belief task is underlined by the consistently low correlations between these tasks: PHI = .15 in Experiment 1, .12 in Experiment 3, and .09 in Experiment 3 by Lewis and Freeman.


Chandler, M. J. (1988). Doubt and developing theories of mind. In J. W. Astington, P. L. Harris, and D. R. Olson (Eds.), Developing theories of mind, 387-413. New York: Cambridge University Press.

Charman, T. & Baron-Cohen, S.(1992). Understanding drawings and beliefs: A further test of the metarepresentation theory of autism. Journal of Child Psychology and Psychiatry, 33, 1105-1112.

Flavell, J. H. (1988). The development of children's knowledge about the mind: From cognitive connections to mental representations. In J. Astington, P. L. Harris, & D. R. Olson (eds.), Developing theories of mind (pp. 244-267). New York, NY: Cambridge University Press.

Forguson, L. & Gopnik, A. (1988). The ontogeny of common sense. In J. Astington, P. L. Harris, & D. R. Olson (Eds.), Developing theories of mind (pp. 226-243). New York, NY: Cambridge University Press.

Hogrefe, G. J., Wimmer, H., and Perner, J. (1986). Ignorance versus false belief: A developmental lag in attribution of epistemic states. Child Development, 57, 567P582.

Leekam, S. R., and Perner, J. (1991). Does the autistic child have a metarepresentational deficit? Cognition, 40,203-218.

Leslie, A. M. (1988). Some implications of pretense for mechanisms underlying the child's theory of mind. In J. W. Astington, P. L. Harris, and D. R. Olson (Eds.), Developing theories of mind, 19P46. New York: Cambridge University Press.

Leslie, A. M. (1993). The theory of mind deficit in autism: The metarepresentation theory. In Baron-Cohen, S., Tager-Flusberg, H., & Cohen, D. (Eds.), Understanding other minds: Perspectives on the theory of mind hypothesis of autism. Oxford: University of Oxford Press.

Leslie, A. M., & Thaiss, L. (in press). Domain specificity in conceptual development: Evidence from autism. Cognition.

Lewis, C. N., & Freeman, N. H. (1992). Dissociation of inferences about beliefs and pictures in preschoolers. Unpublished manuscript, Departments of Psychology, Universities of Lancaster and Bristol.

Lillard, A. & Cho, V. Y. (1991). Young children's understanding of representations. Poster presented at the Annual Convention of the American Psychological Association, San Francisco, August, 1991.

Parkin, L. J. (1993). Young children's understanding of false beliefs, outdated photos and misleading signs. Paper presented at the Annual Meeting of the Developmental Section of the British Psychological Society, Birmingham, September, 1993.

Perner, J. (1988). Developing semantics for theories of mind: From propositional attitudes to mental representation. In J. Astington, P. L. Harris, & D. R. Olson (Eds.), Developing theories of mind, (pp. 141-172). New York, NY: Cambridge University Press.

Perner, J. (1991). Understanding the representational mind. Cambridge, MA: MIT Press/ Bradford Books.

Perner, J., Leekam, S. R., & Wimmer, H. (1987). Three-year olds' difficulty with false belief: The case for a conceptual deficit. British Journal of Developmental Psychology, 5, 125-137.

Pollack, J. (1990). children's understanding of "false" drawings. Unpublished manuscript, Department of Experimental Psychology, University of Oxford.

Wellman, H. M. (1990). The child's theory of mind. Cambridge, MA: MIT Press/ Bradford Books.

Wimmer, H., Hogrefe, G.-J., and Sodian, B. (1988). A second stage in children's conception of mental life: Understanding sources of information. In J. W. Astington, P. L. Harris, and D. R. Olson (Eds.), Developing theories of mind, 173P192. New York: Cambridge University Press.

Wimmer, H. & Perner, J. (1983). Beliefs about beliefs: Representation and constraining function of wrong beliefs in young children's understanding of deception. Cognition, 13, 103-128.

Zaitchik, D. (1990). When representations conflict with reality: The preschooler's problem with false beliefs and "false" photographs. Cognition, 35, 41-68.


[1]. It should be noted that children in this study performed extremely well, which made it necessary to include so unusually young subjects. It is unlikely that this good performance was due to the fact that we used drawings instead of photos since other studies that have used drawings (Charman & Baron-Cohen, 1992; Pollack, 1990; Lillard & Cho, 1991, including our own pilot work on a different group of children) reported fewer but still similar difficulties as in Zaitchik's original photo study. So the most likely explanation for this precocious performance in this experiment is Brighstart nursery's admission policy of admitting only bright and socially aware infants.

[2].Originally we wanted to use the expression "In Mary's mind,..." but pilot work showed that children found "In the head,..." easier to understand.

[3]. Memory questions were asked after the Belief (or Photo) question to avoid their different forms (type A vs. B) influence answers to the Belief question.

Author Notes

The reported research was financially supported by research grants G8720952N and G9026071N from the Medical Research Council. The authors would also like to thank the head, teachers and pupils of Tarnerland Nursery School, Fiveways Playgroup and the East Sussex Education Office for their friendly cooperation. Requests for reprints should be directed to the first author at Laboratory of Experimental Psychology,University of Sussex, Brighton, East Sussex, BN1 9QG, England. E-mail: