C ognifive

Devefop~e~f,

12, 463-476

~7YY7~~ex

(1997)

~bfis~;ng

Ali rights of re production

ISSN 0885-2~7 4

CHILDREN’S UNDERSTANDIN G AND DISC O NFIRMIN G VISUAL

reserve d.

OF BELIEF EVIDEN CE Josef Perner

University

of Salzburg,

Austria

Alejandro Ldpez Max Planck Institute Research,

for Psychological Munich,

Germany

Data from 3- to 5year-old children show that even the youngest have an understanding that a story character cannot rhink an object is in a box when the character looks into the empty box. This understanding is, however, limited to the case where children share the same visual perspective with the character. When visual perspectives are different the relevant knowledge develops shortly after 4 years at about the time that children master the false belief task. These results are interpreted as young children using an implicit constraint that “think” cannot be applied in view of disconfirming visual evidence. This implicit knowledge becomes explicit and, thus, perspective independent with some delay.

A central question

about our acquisition of a “theory of mind” is when children can understand belief. Experimental investigation of this question has relied heavily on the false belief test (Wimmer & Perner, 1983). Children are told a short story in which the protagonist puts an object into one of two locations. In his absence the object is unexpectedly transferred. On his return children are asked where he will look for his object. Children below the age of 4 years tend to answer wrongly that he will look where the object actually is, while most children older than 4 answer correctly with the object’s original location.

The authors thank Alarcos Cieza, Anna-Maria

Ludescher, Christine Neudert, and Miriam Steffens

for their help in collecting the data. The research was carried out while the first author was a visiting professor from the Laboratory of Experimental

Psychology, University of Sussex at the Max Planck

institute for Psychological Research, Munich, and the second author held a Research Fellowship from the Max Planck Society.

Direct all correspondence nerstrasse 34, A-5020

to: Josef Perner, lnstitut fiir Psychologie, Universitat Salzburg, Hellbrun-

Salzburg, Austria.

Manuscript received April 30, 1996; revision accepted March 4, 1997

463

464

J. Perner and A. Lhpez

In contrast to the experimental evidence, observation of children’s use of the word “think” in everyday contexts (Shatz, Weltman & Silber, 1983; Bartsch & Wellman, 1995) suggests that belief is understood by most children at their third birthday. Although these authors carefully distinguish between different uses of the word “think,” even when an instance of “think” is classified as referring to belief, there is no guarantee that children have really grasped the critical, defining features of the concept of belief. Children use many words seemingly with competence. Only critical tests show that they lack understanding of some defining features as Keil (1989) has amply documented. So, for instance, children use the word “uncle” quite proficiently. Only a critical test pitting defining features against typical, characteristic features, reveals gaps in their grasp of this concept. Not until they are 7 or 8 years old do they appreciate the defining feature that an uncle must be related. In the case of belief Perner (1988) has argued that a defining feature of belief is that it is taken as true by its hotder. Only this feature differentiates belief (f~i~~~~g that a state of affairs is the case) from closely related concepts, such as, thinking @some state of affairs (Perner, 1991) or pretending a state of affairs is the case. Hence, instead of having a proper concept of belief, the young 3-year old may have a concept of “prelief’ which captures all defining features of belief except that one that differentiates it from the mental state underlying pretense (Pemer, Baker & Hutton, 1994; Pemer, 1995). Bartsch and Wellman (1995, p. 40) give one example of the use of “think” where the child seems to emphasize this particular defining feature of belief: “He thinks dis a monster. He thinks it’s a real one, doesn’t he?‘. However, since this child was 4 years old, the example does not provide good evidence for the early understanding of belief at 3 years. Much of the alleged evidence for earlier understanding of “think” as belief in Bartsch and Wellman’s corpus of data does not allow us to conclude that these children use “think” to mean belief rather than prelief Unexpected evidence for young 3-year-olds use of “think” to mean “prelief’ comes from a study by Moses and Flavell (1990) who investigated the claim by Bartsch and Wellman (1989) that children are first able to explain action with belief before they can predict action on the basis of belief as required in the false belief task. Moses and Flavell were able to find a superio~ty of explanation over prediction but at the expense of children making errors on a control question. That is, children correctly explained Mary’s futile trip to the empty bandaids box by saying that she thought there were bandaids in it. However, when asked what Mary rzow thought was in the box after she had, to her surprise, discovered that the box was empty, most of these children again said that Mary still thought there were bandaids in the box. This unexpected wrong response on the follow up question can be explained by the prelief hypothesis (Pemer, 1995). Prefief is a mental state (like belief) that predisposes people to act as-if something were the case. Since Mary acts-as-if the

465

Belief and Visual Evidence

bandaids were in the empty box children infer that she is “thinking” (i.e., prelieving) the bandaids are in the box. However, unlike belief and pretence, prelief does not commit the child to distinguish between Mary acting-as-if the bandaids were in the box because she believes they are there versus because she knows they are not there but pretends they are. So, once the child has formed the idea that Mary prelieves that there are bandaids in the box, then it is perfectly consistent for the child to assume that Mary keeps prelieving that there are bandaids in the box even after she had just looked inside the empty box. So, one critical difference between belief proper and the less developed concept of “prelief’ is that belief prescribes a definite relationship to visual evidence. In particular, disconfinning visual evidence (the person sees the box empty) must lead to abandoning the belief (that the object is in that box). The finding by Moses and Flavell indicates that young 3-year-olds do not fully understand this constraint. However, some incidental observation during a pilot study indicated that under certain circumstances children may obey this constraint. This observation was made while trying to explore the results by Wellman and Bartsch (1988) who found that 3-year-olds when directly told where a person thinks that an object is understand that this person will look for the object there. We had the ill-fated idea to tell children that the person thinks that a white cupboard (in plain view of person and child) is brown. That gained us little understanding on part of the children but a puzzled glare of incomprehension. It felt as if the words “he thinks this cupboard is brown” make no sense in open view of a white cupboard. The first objective of our present investigation is to explore whether the observation of our past pilot studies can be demonstrated in a controlled experiment. That is, we investigate when children understand the relationship between visual evidence and belief (commonly expressed as “thinking”) and whether this understanding emerges before the mastery of the false belief task. EXPERIMENT

1

To assess understanding of the impact of visual evidence on belief, children are told two short sketches involving two toy boxes (e.g., one blue one green) and a Playmobil@ character. Children are shown that both boxes are empty (an idea gleaned from Sheffield, Sosa, & Hudson, 1993, to avoid problems of a reality bias). The character appears and announces that he thinks that his briefcase is in, for example, the blue box. As he stands in front of the two boxes the door of the blue one happens to open so that he can clearly see that it is empty (CE: counter evidence condition). The children are then asked, where the character thinks the briefcase is. The correct answer is that he now thinks the briefcase is in the other, green box as the only logical alternative having seen the blue box to be empty. In the other story (NCE: no counter evidence) events unfold exactly as in the first story except the green box (instead of the blue one) opens. In that case the cor-

466

1. Perner and A.

16pez

reel answer to the question is that the character maintains his belief that the object is in the blue box. Children are scored as understanding the relationship between visual evidence and belief if they give correct answers in both stories. Since we are also interested in whether this underst~d~ng develops before or with the mastery of the false belief task, a traditional false belief task was administered. Method ~~~ci~~~~s. Twenty-eight middle-class German preschoolers (14 female, 14 male) from two kinderg~ens in Munich p~ici~ated in this experiment. They ranged in age from 3;2 to 4; 11, with a mean age of 4 years 3 months. Design. This experiment had two tests: (1) a test of understanding False Belief (FB) and (2) a test for underst~ding the dependence of belief on Visual Evidence (VE). The VE test consisted of two parts, the counter-evidence condition(CE) and the no-counter-evidence condition (NCE) which were presented in counterbalanced order. The EB test was presented between these two parts of the VE test. ~~~r~~~ and Procedure. Children were individually tested for about 15 min each in a separate room of their preschool. They were first familiarized with the experimental situation in a pretest that assessed their understanding of when a story character can and cannot see inside a box. The child faced a small box (12cm x 4cm x l&m) with a closed door containing a piece of chocolate. Then the following interaction ensued: Experimenter: ‘“Look, here is a box with a door. Open the door, and look inside the box. Child: Experimenter Child:

What is inside?” “a piece ofchocolate." (If the child answers wrongly or does not provide an answer at alt, the experimenter always provides the right answer.) (closes the door, and then places a 7cm tall Playmobil@ doll behind the box) “Here comes a man. Can the man see what is inside the box?” “ n o .”

~xp~~menter: (opens the door) “Can the man see what is inside the box now?’ Child: Experimenter: Child: Experimenter: Chiid:

“ I l O .”

(closes the door, and moves the man to the front side of the box) “Can the man see what is inside the box now?’ “ S l O .”

(opens the door) “Can the man see what is inside the box now?” “yes.”

The pretest was followed by the first part of the Visual Evidence Test (CE or NCE task) involving two plastic dolls (7cm tall) and four colored cardboard boxes (9cm x 5cm x 13cm) with a door on the front side. One of the dolls and two of the boxes were used for each of the two tasks.

Belief and Visual

Evid enc e

467

The first task used the blue and the green box (both empty) facing the child with doors closed and a Playmobil@ figure of a doctor with stethoscope (Doctor scenario). The following dialogue ensued: Experimenter: ChiId: Experimenter

“Look, here are two boxes with doors.’ (Open the doors, and look inside the boxes). “What is inside?” “nothing.” (Experimenter supplies the right answer if the child gets it wrong.) (closes both doors and places a small doll in front of the closed boxes) “Here comes the doctor. He is looking for his briefcase.” (Pointing to the blue box.) “He thinks that his briefcase is in the blue box.” (note: “think” is for this context an adequate translation of the German word “glauben”).

CE-condition (the door of the blue box opens): “Look, the door of the blue box opens up all by itself.” (Child and story character look into the empty blue box.) NCE-condition (the door of the green box opens): “Look, the door of the green box opens up all by itself.” (Child and story character look into the empty green box.) Test question:

“Where does the doctor think that his briefcase is now?’

After the intervening False Belief test the second part of the Visual Evidence test followed (Peter scenario). It was modeled after the doctor scenario of the first part but red and yellow boxes were used and a Playmobil~ boy represented Peter looking for his pencil. If the doctor scenario was enacted in its CE version then the Peter scenario was enacted in its NCE version, and the other way round. For the False Belief test an A4 (21cm x 29Scm) sized cardboard with a drawing of two doghouses was used. And the following story was enacted with paper cut outs of a bone, a dog, and a man: “Look, here are two doghouses. They both belong to Belle. Here comes Belle with a bone.” (Dog figure put with his bone in front of the doghouses.) Bello wants to put the bone into one of the doghouses. Bello puts the bone into this doghouse (doghouse-l), and then goes away. (The bone is slipped under doghouse-l so that it cannot be seen anymore. The dog is removed.) While Belle is away, its master comes and finds the bone. (The man figure placed in front of the doghouses.) He takes the bone out of this doghouse (doghouse-l), puts the bone into this other doghouse (doghouse-2), and then goes away. (The bone is slipped under doghouse-2 so that it cannot be seen anymore. The man is removed.) Here comes Bello back. Bello wants to have the bone again. (The dog is placed in front of the doghouses.)” Test question:

“Where does Bello think that the bone is?’

468

J. Perner and A. 16pez

Results and Discussion. All children successfully performed the pretest task as be expected from literature Flavell, & Flavell, 1977; Flavell, Everett, Croft, Flavell, 198 Children’s on the False Belief test were scored as correct if they answered with doghouse- 1. All their wrong responses consisted of pointing to doghouse-2 where the bone really was. Table 1 shows the percent children giving correct answers within three age groups of approximately equal size. The result replicates previous findings that around 4 years most children give correct answers. On the Visual Evidence test children were scored as passing if they gave the correct answer in the counter-evidence condition (e.g., pointing to the green box) and in the no-counter-evidence condition (e.g., pointing to the blue box). Errors consisted of erroneously pointing to the open box in the no-counter-evidence condition (1 child), in the counter-evidence condition (2 children) or in both tasks (2 children). Table 1 shows that there is no clear improvement over the age range tested comparable to that on the false belief test, since even most of the youngest children passed the Visual Evidence test. Clearly more children passed the visual evidence test than the false belief test (McNemar’s x2 (1, N = 28) = 8.33, p < .Ol). This result supports the contention by Wellman (1990; Bartsch & Wellman, 1995) and our pilot observations that, in their use of the word “think,” young children understand something about belief before they can pass the false belief test. These children understand that looking inside an empty box is incompatible with “thinking” that the briefcase is in that box. So, with few exceptions almost all children always chose the closed box as an answer to the question about where he thinks the briefcase is. If “thinking” had been interpreted by the younger children as “prelief’ (thinking about the briefcase being in the blue box, or pretending, as Pemer et al., 1994, contended) then there would have been no pressure on these children always to choose the closed box. For, one can keep “thinking of the briefcase in the blue box” or “pretending that it is there” regardless of the visual evidence that the box is empty. An important next question is how general the child’s grasp of this constraint on the use of the word “think” (as belief) is. For, in Experiment 1 the conditions were

Table 1. Percent Children who Pass False Belief Test and Visual Evidence Test Experiment 1. Age Group 3;2-4;0

4;1-4;3

4;4-4; 11

All ages

(II= 9)

(n = 9)

(n = 10)

(N = 28)

False Belief

22%

33%

80%

46.4%

Visual Evidence

78%

78%

90%

82.1%

Test

469

Belief and Visual Evid enc e

very favourable for demonstrating early competence. When the box opens and the child and the story character are staring into an empty box the situation is similar to our ill-fated pilot study where children seemed to consider the instruction, “he thinks the (white) box is brown” a nonsensical sentence. It seems as if a direct linguistic constraint prevents us from using the word “think” in this context. The same holds for looking with a person into an empty box where it feels nonsensical to say that the person thinks there is something in the box. It may be for this reason that children refrain from giving wrong responses under these conditions. That children make their use of “think” dependent on these types of situation is not implausible, since these types play an important role in communication as those situations that directly support mutual knowledge (Lewis, 1969, Schiffer, 1972) or mutual manifestness (Sperber & Wilson, 1986) of the target fact. In these situations no questions tend to arise about any participant not knowing that the other one knows the target fact, or about any higher-order knowledge state of one participant knowing that the other one knows that he (the first one) knows, etc. It also can explain why children did not obey this constraint in Moses and Flavell’s (1990) experiment. Although at the time of asking the critical question the child knew that the box did not contain bandaids, the child did not share an open view of the interior of the box with the story character. The box was shown in side view on a video still (Moses, personal communication). In our next experiment we investigate children’s performance when they do not look with the story character into the box but observe the character’s viewing from a different vantage point. Since this is not the typical mutual knowledge situation (Perner & Garnham, 1988) we expect that children may now persist in saying that the story character still thinks the briefcase is in the box that just opened. EXPERIMENT

2

In order to separate the child’s visual perspective from that of the story character the two boxes were replaced with two doll house rooms that were open at the top. Hence, the children, looking down upon the scene, were in constant view of the emptiness of the rooms. In contrast, the story character who was much smaller than the walls of the room could see inside only when the door to the room was opened. Thus, children had information about whether the other could or could not see the interior of the room but their visual access was not the same as that of the story character. Children’s understanding of the character’s perspective was again established in a pretest, using the same protocol as in Experiment 1. Method. Participants. Forty-seven middle-class German preschoolers (22 female, 25 male) from four different kindergartens in Munich participated in this experiment. They ranged in age from 3;2 to 4;11, with a mean age of 4 years and 3 months. None of the children had participated in the previous experiment.

1. Perner and A. Liipez

470

Design and Procedure. as Experiment 1.

This experiment

had the same design and procedure

Materials. The only difference between this and the previous experiment consisted in the materials for the Visual Evidence tasks. A cardboard dollhouse 30cm x 22cm x 13cm(high) with no roof was subdivided into two equal sized rooms. The walls were 13cm high while the story characters (Playmobil@ dolls) were only 7cm tall. Both rooms had large doors on the same side of the house. The doors were practically as wide as each room so that it was clear that with a door open the story character could see that the room was empty. The doors had the same colors as the boxes in the previous experiment. In the pretest children’s understanding of the character’s perspective was tested with a 9cm x 5cm x 13cm(high) topless box, which had a door on one of the sides. Results and Discussion All children successfully performed the pretest task with the topless box. Their performance on the False Belief test is shown in Table 2. The age trend is reminiscent of the previous experiment with an increase in the proportion of children giving correct responses at about 4 years. Results on the Visual Evidence test, however, were quite different from Experiment 1. Children found this version of the VE test as difficult as the false belief test. Their errors consisted of changing rooms in the NCE task or saying “nowhere” (6 children), not changing rooms or saying “nowhere” in the CE task (9 children), changing rooms in NCE but not in CE task (8 children), and saying “don’t know” in one or both tasks (2 children). This version of the Visual Evidence test was not only of equal difficulty to the False Belief test, but performance on these two tasks also correlated significantly: o = 0.32, x2( 1, N = 47) = 4.85, p < .05. Moreover, this relationship between tasks goes beyond a common improvement with age. Logistic regression (using SPSS) on the proportion of children passing the false belief test showed a significant relationship with the visual evidence test after children’s age had been pa~ialled out: Wald Statistic = 3.91, p q.05.

Table 2. Percent Children who Pass False Belief Test and Visual Evidence Test Experiment 2. Age Group -. Test

3;2-4;l

4;2-4;s

(n = 16)

False Belief

38%

Visual Evidence

19%

--

4;6-4;ll

All ages

(n = 16)

(n = 15)

56%

60%

51.1%

50%

73%

46.6%

--- (N = 47)

Belief and Visual Evidence

471

An across-experiment comparison of the two versions of the visual evidence test shows that the second version (different visual perspectives) where the child saw the other look through the door (46.8% correct) was clearly more difficult than the earlier version (matching visual perspectives) where the child looked through the door with the other character (82.1% correct). This difference was significant (&I, N = 75) = 9.12, p < .Ol) even though children in the second experiment tended to do better on the false belief test than those in the first experiment. This result suggests that children’s good performance in Experiment 1 was due to the particular situation used, in which the emptiness of the box was visually evident (mutu~ly manifest) to child and story character alike. And under the changed conditions the correct use of “think” does not precede mastery of the false belief task. In the following experiment we check whether this result was not due to the peculiar task demands of this experiment (e.g., the aerial view of the rooms) and can be obtained again with different perspectives but using the same materials as in Experiment 1. EXPERIMENT

3

In order to obtain a difference in visual perspectives between child and story characters using the same boxes as in Experiment 1 the no-counter-evidence (NCE) condition was changed. As in the counter-evidence (CE) condition the door of the blue box opened (where the character said he thought the object was) but the character approached the box from behind so that, despite the open door, he could not see inside. Thus, in response to the question where the character thinks the object is children should not change to the other box. Method ~~~~c~~u~~~. Fifteen middle-class German preschoolers (6 female, 9 male) from a single kindergarten in Munich participated in this experiment. They ranged in age from 3;5 to 4; 11, with a mean age of 4 years and 2 months. None of the children had participated in either of the previous two experiments. lksign and ~~~e~~. Experiment 1.

This experiment

had the same design and materials as

Procedure. The procedure was the same as in Experiment 1 with the one difference, which was the procedure for the NCE task exemplified here for the Peter scenario using the red and yellow boxes: Expe~menter: Child:

“Look, here are two boxes with doors.” (Open the doors, and look inside the boxes.) “What is inside?” “nothing.” (The experimenter provides the right answer if the child gets it wrong.)

472

J. Perner and A. 16pez

Experimenter:

Test question:

(closes bath doors and places the boy doll hehind the boxes): “Here comes Peter. He is looking for his pencil.” (Pointing to the red box.) “He believes that his pencii is in the red box.” (Opening the door of the red box.) “Look, the door of the red box opens up all by itself.” (The child, but not Peter, can see inside the red box.)

“Where does Peter think that his pencil is now?’

Results As Table 3 shows, despite use of the original material of Experiment 1, the difficulty of the visual evidence test is similar to the version used in Experiment 2. That is, even though performance on the false belief test (46.7% correct) is practically the same as in Experiment 1 (46.4%) performance on the Visual Evidence tasks (33.3%) is much below that of Experiment 1 (82.1%): x2( 1, N= 43) = 10.24, 12< .oOl. Children’s errors on the visual evidence test consisted of changing boxes in the NCE task (7 children). not changing boxes in the CE task (31, and combining both errors (2). As in Experiment 2, performance on the visual evidence test does not exceed that on the false belief test. If anything, it stays somewhat below it, though not significantly so (Binomial Test: N = 8, x = 2, p(one-tail) r .14f. This result confirms that when children do not look into the boxes with the story character correct use of the word “think” does not develop before mastery of the false belief task. However, the results also suggest that children’s use of “think” is not tied to situations of mutual manifestness. Rather, it is dominated by their own visual experience as their predominant error in the NCE condition suggests. They switch in their answers to the other box because they themselves look into the box that is being opened even though the story character cannot see inside that box. GENERAL

DISCUSSION

The important finding is that children showed some early understanding of “think” expressing belief (well before they passed the false belief test) when their and the story character’s visual perspective matched so that the child looked into

Table 3.

Percent Children who Pass False Belief Test and Visual Evidence Test

Experiment 3. Age Group 3;5-4;2 Test

Belief Visual Evidence False

(n =

7)

43% 14%

4;3-4;ll ____-

(n = 8)

All ages -.______(N =

15)

50%

46.7%

50%

33.3%

Belief and Visual Evid enc e

473

the box with the story character, However, when perspectives differed, so that the child could see from a different vantage point whether the character had or did not have visual access, this understanding was not apparent until the time of passing the false belief test. One tempting explanation for these results comes from simulation theory. Simulation theory assumes that belief attributions are made by taking the believer’s perspective and then ascribing one’s relevant experiences within that perspective as beliefs to the other person. Harris (1991) explained developmental difficulties by the complexity of the perspective shifts involved. The more complex the required perspective shift the later a task is mastered. This assumption naturally predicts our observation that children do better in the Visual Evidence tasks of Experiment 1 than in the versions used in Experiments 2 and 3. In all versions the child’s cognitive perspective (the child knew that both boxes were empty) differs from that of the story character. However, in Experiment 1 children obtained support for the necessary perspective shift by sharing the same visual perspective with the story protagonist, while in the other versions and in the false belief tasks no such support was fo~hcoming. Unfortunately, this elegant explanation runs into a problem when we consider different versions of the false belief test as Peter Mitchell (personal communication) has pointed out. In deceptive container tasks (Gopnik & Astington, 1988; Perner et al., 1987) children have to predict what another person thinks is in a typical container (e.g., Smarties). In this task the child, although knowing the real contents, is given the same misleading visual clue (same visual perspective) as the person in question. Yet, children are no better (Jenkins & Astington, 1996) or even worse (Doherty, 1994; Holmes, Black, & Miller, 1996; Schneider, Perner, Bullock, Stefanek, & Ziegler, in press) in describing that person’s false belief in this situation than in the more traditional task (that was used in this series of experiments) where no visual perspective support was available. This fact certainly poses a problem for the traditional “introspe~tionist” version of simulation theory (Harris, 1992; Goldman, 1993) where it is presumed that children have a pre-existing well-formed concept of belief, which is applicable to their own mental states. The young child merely suffers from performance problems with the complex perspective shifts that are necessary to apply the concept to other people. From that theoretical perspective the concept of belief does not develop, only sophistication in its application. The problem for this position is the fact that correct use of “think’ is not guaranteed by just any visual perspective support. The critical difference between the helpful situation in Experiment 1 and the deceptive container false belief task is that in our experiments direct visual information about the contents of a box was provided, while in the deceptive container tasks merely a visual clue is given to what might be inside. Considering this difference, simulation theory may fare better if one considers Gordon’s “non-introspectionist” alternative. In order to avoid Wittgenstein’s criticism of the possibility of introspection, Gordon (1992, 1995) replaced introspec-

474

J. Perner and A. Lbpez

tion by ascent routines. An ascent routine is a gratuitously correct (uninformative) attribution within a particular perspective. For instance, every fact X within a person’s perspective is a fact that that person believes (otherwise it would not be a fact within that perspective). Hence by taking that person’s perspective one can “ascend” (within that perspective) from “X” to “I believe X.” If we now assume that the prime basis for such ascension is the direct visual evidence of facts then we can see why children are helped in the situations of Experiment 1 but not in the deceptive container tasks since, there, no direct visual evidence of the fact is provided. Of course, correct application of this ascent routine does not constitute a proper unde~tanding of belief. ~nde~tanding of belief only emerges with the ability to combine the ascent routine with a shift in perspective. That is, after applying the routine within the other’s perspective one has to shift back to one’s own perspective and change the belief attributions from oneself to the other person, i.e., to “He believes X.” If we further make the plausible assumption that these ascent routines are closely tied to the use of the word “think” then we can see an affinity to our original intuition that it is a linguistic constraint that prevents children (and us) from using “think” in view of openly contradictory visual evidence. Whichever of these two ways we look at children’s early competence in Experiment 1 their knowledge is at best implicit. It is implicit if it is embedded in the hnguistic procedures, as Karmiloff Smith (1992) has demonstrated for many other areas, and it is implicit in ascent routines before these routines can be co-ordinated with differences in perspective (Perner, 1996). This interpretation of our findings provides an interesting compromise between the opposed intuitions about the level of 3-year-olds’ understanding of belief. The data of Experiment I support those whose intuition is based on natural language use (Shatz et al., 1983; Bartsch & Wellman, 1995) that the early use of “think” shows more of an understanding of belief proper rather than just an undifferentiated concept of “prelief’ covering belief and pretence, as Pemer et al. (1994; Perner, 1995) have argued. In particular, children appear to understand “think” with the meaning of “thinking that,” i.e., that the believer takes the belief as true of the real state of affairs, since they understand that “think” cannot be used in the face of disconfirming visual evidence. However, since this early understanding of “think” as belief is only implicit in the child’s own linguistic constraints a compromise is possible. Explicit understanding of this constraint, which is necessary for applying it independently of the child’s own visual perspective, develops later, roughly at the time the false belief test is mastered as the “pre.lief’ hypothesis would have it (Pemer et al., 1994). REFERENCES Bartsch, K., & Wellman, H.M. (1989). Young children’s attribution of action to beliefs and desires. Child Devekpnent, 60, 946-964.

Belief and Visual Evidence

475

Bartsch, K., & Wellman, H.M. (1995). Children talk about the mind. New York: Oxford University Press. Doherty, M. (1994). Metulinguistic understanding and theory of mind. Unpublished doctoral dissertation, Laboratory of Experimental Psychology, University of Sussex. Flavell, J.H., Everett, B.A., Croft, K., & Flavell, E.R. (198 1). Young children’s knowledge about visual perception: Further evidence for the Level l-Level 2 distinction. L&vetopmental P.~ycho~ogy, 17, 99-103. Goldman, AI. (1993). The psychology of folk psychology. Behavioral and Brain Sciences, 16, 15-28. Gopnik, A.I., & Astington, J.W. (1988). Children’s understanding of representational change and its relation to the understanding of false belief and the appearance-reality distinction. Child Development, 59, 26-37. Gordon, R.M. (1992). The simulation theory: Objections and misconceptions. Mind & Ianguage, 7, 11-34. Gordon, R.M. (1995). Simulation without introspection or inference from me to you. In M. Davies Br T. Stone (Eds.), mental simulation: revaluations and applications (pp. 5367). Oxford: Blackwell Publishers. Harris, P.L. (1991). The work of the imagination. In A. Whiten (Ed.), Natural theories of mind: The evolution, development and simulation of everyday mindreading (pp. 283-304). Oxford: Blackwell. Harris, P.L. (1992). From simulation to folk psychology: The case for development. Mind & Language, 7, 120-144. Holmes, H.A., Black, C., & Miller, S.A. (1996). Task comparison of false belief understanding in a Head Start population. Journal of Experimental Child Psyc~logy, 63, 263-285. Jenkins, J.M. & Astington, J.W. (1996). Cognitive factors and family structure associated with theory of mind development in young children. Developmental Psychology, 32, 70-78. Karmiloff-Smith, A. (1992). Beyond modularity: A developmental perspective on cognitive science. Cambridge, MA: MIT Press. Keil, EC. (1989). Concepts, kinds, and cognitive development. Cambridge, MA: MIT Press. Lempers, J.D., Flavell, E.R., & Flavell, J.H. (1977). The development in very young children of tacit knowledge concerning visual perception. Genetic Psychology Monographs, 95, 3-53. Lewis, D. (1969). Convention: A philosophical study. Cambridge, MA: Harvard University Press. Moses, L.J., & Flavell, J.H. (1990). Inferring false beliefs from actions and reactions. Child Development, 61, 929-945. Pemer, J. (1988). Developing semantics for theories of mind: From propositional attitudes to mental representation. In J.W. Ast~ngton, PL. Harris, & D.R. Olson (Eds.), Developing theories of mind (pp. 141-172). New York: Cambridge University Press. Pemer, J. (199 1). On representing that: The asymmetry between belief and intention in children’s theory of mind. In D. Frye & C. Moore (Eds.), Children’s theories o f mind (pp. 139-l 55). Hillsdale, NJ: Erlbaum. Pemer, J. (1995). The many faces of belief: Reflections on Fodor’s and the child’s theory of mind. Cognition, 57, 241-269.

476

J. Perner and A. Ld p ez

Pemer, J. (1996). Simulation as explicitation of predication-implicit knowledge about the mind: Arguments for a simulation-theory mix. In P. Carruthers & P.K. Smith (Eds.), Theon’es oftheories ofmind (pp. 90-I 04). Cambridge: Cambridge University Press. Pemer, J., Baker, S., & Hutton, D. (1994). Prelief: The conceptual origins of belief and pretence. In C. Lewis & P. Mitchell (Eds.), Children S early understanding of mind: Origins and development (pp. 261-286). Hove, U.K.: Erlbaum. Pemer, J., & Garnham, A. (1988). Conditions for mutuality. Journal of Semantics, 6, 369385. Pemer, 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. Schiffer, S. (1972). Meaning. Oxford, U.K.: Oxford University Press. Schneider, W., Perner, J., Bullock, M., Stefanek, J., & Ziegler, A. (in press). The development of intelligence and thinking. In F. E. Weinert (Ed.), The Munich longitudinal study of the genesis of individual competencies (LOGIC). Cambridge, U.K.: Cambridge University Press. Shatz, M., Wellman, H.M., & Silber, S. (1983). The acquisition of mental verbs: A systematic investigation of the first reference to mental state. Cognition, 14, 301 321. Sheffield, E.G., Sosa, B.B., & Hudson, J.A. (1993, March). Narrative complexity and 2and 3-year olds’ understanding offalse helie$ Paper presented at the 60th Anniversary Meeting of the Society for Research in Child Development, New Orleans. Sperber, D., & Wilson, D. (1986). R e 1evance: Communication and cognition. Oxford: Blackwell. Wellman, H.M. (1990). The child’s theory of mind. Cambridge, MA: MIT Press. Wellman, H.M., & Bartsch, K. (1988). Young children’s reasoning about beliefs. Cognition, 30, 239-277. Wimmer, H., & Pemer, J. (1983). Beliefs about beliefs: Representation and constraining function of wrong beliefs in young children’s understanding of deception. Cognition, 13, 103-128.