Beliefs about beliefs : Representation and constraircing function of wrong bekfs in young children ’ s understanding of deception HEINZ

Understanding of another person’s wrong belief requires explicit represer;tation of the wrongness of this person’s belief in relation to one’s own knowledge. Three to nine year old children’s understanding of two sketches was tested. In each sketch subjects observed how a protagonist put an object into a location x and then witnessed that in the absence of the protagonist the object was transferred from x to location v. Since this transfer came as a surprise they had to assume that the protagontst still believed that the object was in x. Subjects had to indicate where the protagonist will look for the object at his return. None of the 3-4-year old, 57% of 4-6-year old, and 86% of 6-g-year old children pointed correctly ,to location x in both sketches. Of the many cases where J-6-year olds made an error the]; failed *This project was financially supported by a research grant from the Stiftung Volkswagenwerk to Professor Erwin Roth, University of Salzbuq. The international cooperation of the two authors was supported by the Academic Link Interchange Scheme of The British Council. The authors wi-h to thank Silvia Gruber for expert advice on procedures and for her help in data collection. Gratefully we acknowledge the friendly cooperation of the ‘Amt fur Kindergarten und Horte der Salzburger Landesregiexung (Frau Kfchdorfer)‘, of the ‘Stadtjugendamt Salzburg (Herr Artner)‘, and staff and children in various kindergartens in Salzburg. Requests for reprints should be directed to either Heinz Wimmen, Institut fur Psychologie der Universitiit Salzburg, A-5020 Salzburg, Akademiestrasse 22, Austria, or Josef Peaner, Experimental Psychology, University of Sussex, Brighton BN19QG, England. OOlO-0277/83/010103-26/$08.10 0 Eisevier Sequoia/Printed in The Netherlands 104 H. Wirnmer and J. Perner in only about 2O70 to remember the initial location correctl.v. As a test of the stability of children’s representation of the protagonist’s wrong belief the sketches continued with a statement about the protagonist S intention to either deceive an antagonist or truthfur& inform a friend about the object’s Ix&ion. independent of age, of those children who correctly thought that the protagonist would search in x, 85% of the time they also correctly thought that he would direct his antagonist to location y and his friend to location x. This shows that once children can represent a person’s beliefs they can constrain their interpretation of this person’s stated intentions to the l*erson’s beliefs. In a more story-like situation another group of children had to infer a deceptive plan from the depiction of a goal conflict between two story characters and one character’s expedient utterance. At the age of 4-S years children correctly judged this utterance as a lie only 28% of the time while S-6-year olds did so 94% of the time. These results sugg~t that around the ages of 4 to 6 years the ability to repiasent the re&ionship between two or more person’s epistemic srates emerges and becomes firmly established. Premack and Woodruff (1978) argued that one could demonstrate that chimpanzees have a ‘theory of mind’, i.e., an ability to impute mental states to themselves and others. In Pylyshyn’s (1978) explicatiori this means that somebody who has a theory of mind does not only have a representation about a state of affairs (x) and stands in certain relationships to these representations (e.g., wanting x, believing x, etc.) but also represents these relationships explicitly. Pylyshyn refers to this ability as an ability for ‘metarepresentation’. This ability looms large in philosophical discussion as a necessary precondition for moral responsibility, self-consciousness, and social interaction (e.g., Dennett, 1978a). It was thought that the best method of demonstrating an ability for metarepresentation in non-linguistic chimpanzees is to teach these animals to deceive a competitor. Deceptive action is a good way of demonstrating the presence of a theory of mind, because on the one hand, flexible application of a deception strategy ought to have high adaptive value, while on the other hand it requires the conceptualization of the deceived person’s wrong belief as a subgoal in one’s planning strategy. Woodruff and Premack (1979) succeeded in teaching deceptive pointing to 2 out of 4 chimpanzees but only a.fter an extensive training effort of 5 months. This difficulty in teaching chimpanzees to deceive is comparable to the difficulties encountered in teaching them language. Thus it seems that the natural acquisition of language and of a ‘theory of mind’ is a human characteristic. Beliefs about beliefs 105 Although the natural occurrence of deception is a convincing indicator of meta-representational activity it is a very conservative measure because it requires many additional complex planning steps. Since ch.ildren are capable of language there is alternative access to their meta-representational ability. One obvious indicator of this ability is children’s use of meta-representational langiiage, i.e., their reference to people’s ‘wants’, ‘beliefs’, etc. In an observational study Bretherton and Beeghly (in press) found that at the age of 2 l/2 years the majority of children spontaneously used a substantial vocabulary about perception, volition, major emotions, and knowledge. The most popular of these words were used equally frequently for self and others. Hood and Bloom (1979) reported that at 3 years children referred quite frequently to their own and others’ intentions in response to ‘why’questions. Experimental studies tested the correct use and understanding of these words at an early age. For instance, Shultz et al. (1989) demonstrated that 3 to 5-year olds distinguished correctly betwe?en intended acts and unintended behaviour, such as mistakes, reflexes, and passive movement. There is also evidence that 4-year olds start to understand correctly some, although not all aspects of ‘know’ and ‘guess’ (Johnson and Maratsos, 1977; Miscione et al., 1978), of ‘remember’ and ‘forget’ (Johnson and Wellman, 1980) and start to understand the presuppositions of these words (Macnamara et al., 1976). The studies cited above demonstrate that very young children already explicitly represent and hence are able to verbalize the relation in which they and others stand to their representation of propositional content. A more complicated meta-reRresentationa1 problem arises when one has to explicitly represent the difference between one’s own anl;l somebody else’s relation to the same propositional conten.t. This problem arises when one has to account for the lack of knowledge in another person. In a series of studies on epistemie egocentrism subjects had information about some event and also had information about the fact that another person had no such information. Chandler and Greenspan (1972) and Flavell et al. (1968) generated this difference in knowledge by introducing a late arriving bystander. Marvin ef al. (1976) and Mossler et al. (1976) used a different technique based on selective presentation of sense specific information. For instance, in one study children were shown a televisilon sketch in which a boy was entering a house and asked his sister for a biscuit. Then they watched the same sketch over again in the presence of their mother but with the sound track turned off, so that it was clear that the mother did n& hear the auditorily conveyed information. Sixty percent of 4-year and 85% of S-year olds answered correctly that the mother knew that the boy entered 106 H. Wimmer and J. Perner the house and that sh,e did not know that he asked his sister ror a biscuit. These results show that from 4 years onwards children are atile to differentiate between their own knowledge and the absence of this knol-,vledge in the other person. in contrast to these studies which investigated children’? ability to represent the absence of knowledge in another person the present series of experiments investigated children’s competence in representing another person’s definite belief which differs from what the subject knows to be true This problem has not been investigated developmentally. It can, however, be conveniently studied in the following formal paradigm which has been outlined by Bennett (1978), Dennett (19786) and Harman (1978) independently in their comments on the paper by Premack and Woodruff (1978): The subject is aware that he/she and another person observe a certain state of affairs x. Then, in the absence of the other person the subject witnesses an unexpected change in the state of affairs from x to y. The subject now knows that y is the case and also knows that the other person still believes that x is the case. In order to test subjects’ comprehension of the other person’s wrong belief, stories like the following were constructed: A story character, Maxi, puts chocolate into a cupboard x. In his absence his mother displaces the chocolate from x into cupboard ‘y. Subjects have to indicate the box where Maxi will look for the chocolate when he returns. Only when they are able to represent Maxi’s wrong belief #(‘Chocolate is in x’) apart from what they themselves know to be the case (‘Chocolate is in y’) will they be able to point correctly to box x. This procedure tests whether subjects have an explicit and defmite representation of the other’s wrong belief. Yet, there is neither a problem in framing the test question by using mental verbs (e.g., ‘What does Maxi believe?‘) nor are !;ubjects required to verbalize their knowledge about other’s beliefs since a mere pointing gesture suffices. The practical importance of representing another person’s wrong beliefs consisis in the use of this representation as a frame of reference for interpreting or anticipating the other person’s actions. That is, the interpretations and anticipations have to be constrained to the realm of the other person’s beliefs. In order for such a constraining function to occur, the representation of the other person’s beliefs has to be frnnly established. As a test of the firmness of children’s representation of wrong beliefs, the ‘wrong belief paradigm as outlined above was extended by describing Maxi as wanting to tell something wrong or something true. fn a competitive story version Maxi’s brother was introduced who, too, wants the chocolate and asks Maxi where it is. Maxi (who wrongly believes that the chocolate is in x) decides to tell his brother something wrong, SubBeliefs aiiou t beliefs 107 jects are then asked ‘Where will Maxi tell his brother the chocolate is?‘. A correct answer to this question depends on the correct interpretation of Maxi’s intention to tell something wrong. Maxi’s message to his brother has to be ‘wrong’ in relation to ‘Maxi’s wrong belief’. Ironically, this correct interpretation can lead to the answer t+!at Maxi will try to mislead his brother by telling him: ‘The chocolate is in y’, where it actually is. Hence there is a temptation for subjects to make the typically ‘egocentric’, incorrect interpretation. Maxi’s message, whick he intends to be wrong, could easily be construed as actually wrong, i.e.. It is interpreted as being ‘wrong’ in relation to the subject’s own knowledge. In a cooperative story version Maxi asks his Grandpa for help and has the explicit intention to tell the truth about the location of the chocolate. Again, there are two possible interpretation; of ‘true’: the correct intergretation relates the ‘true’ statement to Maxi’s wrong belief, while the incorrect one relates it to the subject’s own knowledge of the truth. The correct interpretation leads to the answer that Maxi tells his Grandpa: ‘The chocolate is in x’, where the subject knows that it is not. Therefore, the correct interpretation of the instruction to make Maxi say something wrong can lead to a statement that is actually true, while the correct interpretation of the instruction to make him say something true should lead to an utterance that is actually wrong. One can empathise with the subject’s temptation to take the wrong interpretation: when asked to make Maxi say something (which he thinks is) wrong or true, it is but natural to make him say something that is actually wrong or true. This temptation will meet little opposition if subjects are still uncertain about the distinction between their own knowledge and the protagonist’s beliefs. Hence, we can expect that correct interpretations will only occur once this distinction has been firmly established in the child’s mind. The representation of Maxi’s belief has to be firm enough so that the interpretation of his intention to say something wrong or true can be constrained to the realm of his beliefs.